\nDate: <date> Attendees: <names>\nPurpose: <one line — why are we here?>\n\nDECISIONS\n-\n\nACTION ITEMS\n- [ ] <task> — <owner> — <due date>\n\nOPEN QUESTIONS\n-\n\nNOTES / CONTEXT\n-\n\nNext meeting / follow-up: <date or trigger>","outro":"That is it. Prep with it, fill it during the call, clean it up for five minutes after, and send it out. The structure does most of the work for you."},"faq":{"title":"Frequently Asked Questions"},"faqs":[{"question":"How do you take good meeting notes?","answer":"Prepare a simple template before the meeting, then capture only what matters: decisions, action items with an owner and a due date, and open questions. Don't try to transcribe every word — listen for signals like \"we've decided\", \"I'll take that\", or \"by Friday\". Clean the notes up within a few minutes of the meeting ending, then share them where your team already works."},{"question":"What should meeting notes include?","answer":"Good meeting notes include the date, attendees, and purpose at the top, followed by four sections: decisions made, action items (each with an owner and a deadline), open questions, and any background context worth keeping. Everything else — small talk, thinking out loud, anything already in the deck — can be left out."},{"question":"What is the best format for meeting notes?","answer":"A consistent, reusable format beats any clever one. The most reliable structure is four buckets: Decisions, Action items, Open questions, and Notes. Using the same template every time means you spend your attention listening instead of deciding how to lay the page out, and it makes notes from different meetings easy to scan and search later."},{"question":"What is the difference between meeting notes and meeting minutes?","answer":"Meeting minutes are a formal, often required record of attendance, motions, and votes — common for boards and legal settings. Meeting notes are a lighter, practical summary aimed at helping the team remember decisions and follow through on action items. Most working meetings need good notes, not formal minutes."},{"question":"How do you take notes and pay attention at the same time?","answer":"Stop trying to write everything down. When you only capture decisions, owners, and deadlines, your hands are free most of the time and you can actually listen. A meeting note taker that records and transcribes the call for you removes the conflict entirely: you stay present during the meeting and edit a ready-made draft afterward."},{"question":"Should you record meetings to take notes?","answer":"Recording lets you stay present and get an accurate transcript instead of typing live, but always tell participants first — many places require consent. The bigger question is where the audio goes. Most cloud note takers upload your meeting to their servers. For sensitive conversations, a local-first tool like OpenWhispr records, transcribes, and labels speakers entirely on your own device, so nothing is uploaded."},{"question":"Is it legal to record a meeting for notes?","answer":"It depends on where the participants are. Some regions allow recording with just one person's consent, while others require everyone on the call to agree, and rules like the EU's GDPR treat a recorded voice as personal data. This isn't legal advice, but the safe habit everywhere is the same: tell people you're recording and get their okay before you start."},{"question":"Where does my audio go when I use an AI meeting note taker?","answer":"With most cloud note takers, your meeting audio and transcript are uploaded to the company's servers to be processed and stored, and some use that data to train their models. With a local-first tool like OpenWhispr, the recording, transcript, and speaker labels stay on your own device and are never uploaded — which matters most for sensitive meetings like HR, legal, or anything involving customer data."},{"question":"How do you take notes in a meeting you are leading?","answer":"When you are running the meeting you can't also type. Either assign a notetaker, or let a meeting note taker capture the transcript so you can focus on the room. If you must self-serve, jot only decisions and action items in short fragments, then expand them in the five minutes after the call while everything is still fresh."},{"question":"What is the best app for taking meeting notes?","answer":"The best meeting notes app is the one that captures the conversation accurately and fits your privacy needs. Cloud tools like Otter, Fireflies, and Granola are convenient but send your audio to their servers. OpenWhispr is a free, open source meeting note taker that transcribes and identifies speakers locally, so your audio never leaves your device."}],"cta":{"title":"Take Meeting Notes Without Typing a Word","body":"OpenWhispr is a free, open source meeting note taker that records, transcribes, and labels who said what — entirely on your own device. Stay present in the meeting and edit a ready-made draft afterward. Works on macOS, Windows, Linux and iOS.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":[{"href":"/blog/local-speaker-diarization","label":"Private meeting note taker"},{"href":"/blog/local-vs-cloud-transcription","label":"Local vs cloud transcription"},{"href":"/blog/best-dictation-tools-windows-2026","label":"Best dictation tools for Windows"},{"href":"/blog","label":"All blog posts"}]},"blogHowWhisper":{"metadata":{"title":"How Whisper AI Works: A Complete Guide | OpenWhispr","description":"How Whisper AI works explained: transformer architecture, 680K training hours, model sizes, and accuracy benchmarks. Open-source, runs on-device.","keywords":["how whisper AI works","whisper speech recognition","openai whisper explained","whisper model architecture","whisper AI accuracy","whisper transformer model","automatic speech recognition","whisper vs other ASR","whisper.cpp explained"]},"schema":{"title":"How Whisper AI Works: A Complete Guide","description":"A technical deep dive into OpenAI's Whisper speech recognition model — architecture, training data, accuracy benchmarks, and practical usage."},"breadcrumb":{"home":"Home","blog":"Blog","article":"How Whisper AI Works"},"hero":{"title":"How Whisper AI Works: A Complete Guide","subtitle":"A technical deep dive into OpenAI's open-source speech recognition model — from mel spectrograms to transformer decoding.","date":"February 3, 2026","tag":"Technical"},"sections":[{"id":"what-is-whisper","title":"What Is Whisper?"},{"id":"architecture","title":"Architecture"},{"id":"training-data","title":"Training Data"},{"id":"model-sizes","title":"Model Sizes"},{"id":"accuracy","title":"Accuracy"},{"id":"whisper-cpp","title":"whisper.cpp"},{"id":"comparison","title":"ASR Comparison"},{"id":"applications","title":"Applications"},{"id":"sources","title":"Sources"}],"intro":{"paragraph":"Whisper is an automatic speech recognition (ASR) model developed by OpenAI. Released in September 2022, it is a general-purpose speech recognition system trained on 680,000 hours of multilingual audio data collected from the internet. Whisper uses an encoder-decoder Transformer architecture that converts audio into log-mel spectrograms, processes them through a neural network encoder, and autoregressively decodes text tokens. It supports transcription in 99 languages, translation to English, language identification, and timestamp prediction — all within a single unified model. Unlike most commercial ASR services, Whisper's weights are <openSourceLink>fully open source</openSourceLink> under the MIT license, which has enabled a thriving ecosystem of ports, optimizations, and applications — including <whisperCppLink>whisper.cpp</whisperCppLink>, the C/C++ implementation that makes Whisper practical for real-time, on-device speech recognition.","lastUpdated":"Last updated: February 17, 2026. Quantitative claims in this article are cross-checked against at least two primary sources."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","items":[{"label":"Whisper release, training scale, and license:","body":"OpenAI released Whisper in September 2022, trained on 680,000 hours, under MIT."},{"label":"Task scope and language support:","body":"Whisper supports transcription, translation-to-English, language identification, and timestamp tokens across 99 languages."},{"label":"English benchmark context:","body":"around 3% WER is a benchmark figure on LibriSpeech test-clean for English-focused evaluation, not a universal real-world rate."},{"label":"Recent model updates:","body":"large-v3 reports lower errors than large-v2 on many languages, and turbo focuses on lower latency with smaller decoder compute."},{"label":"OpenWhispr relevance:","body":"OpenWhispr uses Whisper via whisper.cpp for local, offline-first dictation across desktop platforms."}],"sources":["<a1>arXiv paper</a1> · <a2>OpenAI Whisper repo</a2>","<a1>Whisper model card</a1> · <a2>Whisper README</a2>","<a1>large-v2 model card</a1> · <a2>benchmark details</a2>","<a1>large-v3 model card</a1> · <a2>turbo release discussion</a2>","<a1>whisper.cpp</a1> · <a2>OpenWhispr</a2>"],"sourceUrls":[["https://arxiv.org/abs/2212.04356","https://github.com/openai/whisper"],["https://github.com/openai/whisper/blob/main/model-card.md","https://github.com/openai/whisper"],["https://huggingface.co/openai/whisper-large-v2","https://github.com/openai/whisper/blob/main/model-card.md"],["https://huggingface.co/openai/whisper-large-v3","https://github.com/openai/whisper/discussions/2363"],["https://github.com/ggml-org/whisper.cpp","https://openwhispr.com"]]},"whatIs":{"title":"What Is Whisper?","paragraph1":"Whisper was introduced in the paper <paperLink>\"Robust Speech Recognition via Large-Scale Weak Supervision\"</paperLink> by Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever at OpenAI. The code and model weights were published on <githubLink>GitHub</githubLink> in September 2022 under the MIT license.","paragraph2":"Before Whisper, state-of-the-art speech recognition systems were typically trained on curated, human-labeled datasets — often limited to specific languages, accents, or domains. Whisper broke from this approach by training on a massive, diverse dataset of 680,000 hours of audio paired with transcripts scraped from the internet. The key insight was that this \"weakly supervised\" approach — using imperfect, machine-generated or user-uploaded transcripts rather than hand-labeled data — could produce a model with remarkable robustness across languages, accents, background noise, and technical vocabulary.","paragraph3":"The result is a single model that can handle multiple speech processing tasks: transcription (speech to text in the same language), translation (speech in any language to English text), language identification, and voice activity detection with timestamps. This multitask capability is baked into the model's architecture through a system of special tokens that specify which task to perform.","paragraph4":"Whisper's open-source release was significant. It democratized access to high-quality ASR, allowing developers, researchers, and companies to use a model competitive with commercial APIs — without per-minute pricing, without sending audio to the cloud, and without vendor lock-in. This catalyzed the development of tools like <whisperCppLink>whisper.cpp</whisperCppLink>, <fasterWhisperLink>Faster Whisper</fasterWhisperLink>, and dozens of applications built on top of them — including <openWhisprLink>OpenWhispr</openWhisprLink>."},"architecture":{"title":"Architecture Deep Dive","intro":"Whisper uses an encoder-decoder Transformer architecture — the same fundamental design behind models like the original Transformer (Vaswani et al., 2017) and many machine translation systems. The encoder processes audio, the decoder generates text. Here is how each stage works.","preprocessing":{"title":"Audio Preprocessing","intro":"Before any neural network processing, raw audio is converted into a visual representation of sound called a log-mel spectrogram. The process works as follows:","steps":["Audio is resampled to 16,000 Hz (16 kHz mono).","A Short-Time Fourier Transform (STFT) is computed using 25-millisecond windows with a 10-millisecond stride (hop length).","The frequency spectrum is projected onto 80 mel-frequency filter banks (128 for large-v3), which approximate human auditory perception by spacing frequency bands logarithmically.","A logarithmic scaling is applied, producing the final log-mel spectrogram.","The spectrogram is divided into 30-second chunks. Audio shorter than 30 seconds is zero-padded; longer audio is processed in sequential chunks."],"outro":"The result is a 2D representation with shape <code>(80, 3000)</code> — 80 mel channels by 3,000 time steps (30 seconds at 10ms stride). This is analogous to an image, and the encoder processes it much like a vision model processes pixel data."},"encoder":{"title":"The Encoder","intro":"The encoder converts the mel spectrogram into a sequence of learned audio representations. It consists of:","steps":["Two 1D convolution layers — The first convolution has a kernel size of 3 and padding of 1, followed by a GELU activation. The second convolution has a kernel size of 3, stride of 2, and padding of 1, which halves the time dimension. This downsamples the 3,000 time steps to 1,500 positions.","Sinusoidal positional embeddings — Unlike the decoder, the encoder uses fixed sinusoidal embeddings (not learned) to encode the position of each time step.","Transformer blocks — A stack of standard Transformer encoder layers, each containing multi-head self-attention and a feed-forward network with GELU activation, connected by residual connections and layer normalization."],"outro":"The output is a sequence of 1,500 contextualized audio feature vectors — one for every 20 milliseconds of input audio — that the decoder will attend to during text generation."},"decoder":{"title":"The Decoder","intro":"The decoder generates text tokens autoregressively — one token at a time, each conditioned on the encoded audio and all previously generated tokens. It consists of:","steps":["Token embedding layer — Converts token IDs into dense vector representations.","Learned positional embeddings — Unlike the encoder's sinusoidal embeddings, the decoder uses learned position embeddings that are trained alongside the model.","Transformer blocks with cross-attention — Each decoder layer has three sub-layers: masked self-attention (preventing the model from looking at future tokens), cross-attention to the encoder output (allowing the model to \"listen\" to the audio), and a feed-forward network."],"outro":"The final decoder output is projected onto the vocabulary to produce token probabilities. Decoding strategies include greedy search, beam search, and temperature-based sampling."},"tokens":{"title":"Special Tokens and Multitask Training","intro":"Whisper performs multiple tasks with a single model through a clever system of special tokens that serve as instructions. The decoder input follows a structured format:","startOfTranscript":"— Signals the start of the output sequence.","language":"— Specifies the language (one of 99 language tokens). Can be auto-detected or manually set.","taskJoin":"or","task":"— Specifies whether to transcribe in the original language or translate to English.","timestampsPart1":"— Controls whether to produce timestamp tokens. When timestamps are enabled, the model generates time offset tokens like","timestampsJoin":"and","timestampsPart2":"that align text to audio.","outro":"This unified token format means a single model can perform transcription, translation, language detection, and timestamped transcription — all without separate model heads or fine-tuning. The task is determined entirely by the token sequence provided to the decoder."}},"trainingData":{"title":"Training Data","paragraph1":"The original Whisper models (tiny through large-v2) were trained on 680,000 hours of audio paired with transcripts collected from the internet. This dataset is not publicly available. The key characteristic is that it is weakly supervised — the transcript labels were not hand-verified by human annotators. Instead, they came from sources like subtitles, closed captions, and other user-generated text paired with audio.","compositionTitle":"Data Composition","compositionIntro":"According to the <modelCardLink>official model card</modelCardLink>, the 680,000-hour training set breaks down as follows:","headers":{"segment":"Segment","hours":"Hours","share":"Share","description":"Description"},"rows":[{"segment":"English ASR","hours":"438,000","share":"65%","description":"English audio with English transcripts"},{"segment":"Translation (X-to-English)","hours":"126,000","share":"18%","description":"Non-English audio with English transcripts"},{"segment":"Multilingual ASR","hours":"117,000","share":"17%","description":"Non-English audio with native-language transcripts (covering 98 languages)"}],"paragraph2":"The heavy English bias (65% of training data) explains why Whisper performs significantly better on English than on lower-resource languages. For the large-v3 model released in November 2023, OpenAI expanded the training set to 1 million hours of weakly labeled audio plus an additional 4 million hours of pseudo-labeled audio generated by running Whisper large-v2 over unlabeled data — a form of self-training or knowledge distillation.","weakTitle":"Why \"Weakly Supervised\" Matters","weakParagraph1":"Most prior ASR systems trained on datasets like LibriSpeech (960 hours) or Common Voice (a few thousand hours per language) — carefully curated, human-verified datasets. Whisper's approach traded label quality for sheer scale: 680,000 hours versus the hundreds or low thousands of hours used by conventional systems. The paper demonstrated that this trade-off was worthwhile. The diversity of the internet-sourced data gave Whisper exceptional robustness to real-world conditions: background noise, overlapping speech, accents, domain-specific vocabulary, and acoustic environments that would challenge models trained on studio-quality read speech.","weakParagraph2":"However, the weak supervision also introduces a known limitation: hallucination. Because the training transcripts are imperfect, the model sometimes generates plausible-sounding text that was not actually spoken — particularly during silence or very quiet passages. This is a trade-off inherent to the approach and remains an active area of improvement."},"modelSizes":{"title":"Model Sizes","intro":"Whisper is available in multiple sizes, ranging from 39 million to 1.55 billion parameters. Smaller models are faster but less accurate; larger models are more accurate but require more compute and memory. The <code>.en</code> variants are English-only models that tend to perform better for English, especially at smaller sizes. No English-only variants exist for the large models.","headers":{"model":"Model","parameters":"Parameters","englishOnly":"English-Only","vram":"VRAM","speed":"Relative Speed"},"rows":[{"model":"tiny","parameters":"39 M","englishOnly":"tiny.en","vram":"~1 GB","speed":"~10x"},{"model":"base","parameters":"74 M","englishOnly":"base.en","vram":"~1 GB","speed":"~7x"},{"model":"small","parameters":"244 M","englishOnly":"small.en","vram":"~2 GB","speed":"~4x"},{"model":"medium","parameters":"769 M","englishOnly":"medium.en","vram":"~5 GB","speed":"~2x"},{"model":"large (v1, v2, v3)","parameters":"1,550 M","englishOnly":"--","vram":"~10 GB","speed":"1x"},{"model":"turbo (large-v3-turbo)","parameters":"809 M","englishOnly":"--","vram":"~6 GB","speed":"~8x"}],"tableNote":"Speed is relative to the large model. VRAM requirements are for GPU inference using fp16 precision via the original Python implementation. <whisperCppLink>whisper.cpp</whisperCppLink> uses significantly less memory (e.g., the large model requires ~3.9 GB RAM instead of ~10 GB VRAM).","turbo":"The turbo model (released October 2024) is a distilled version of large-v3 with a significantly smaller decoder. It retains most of large-v3's accuracy while running approximately 8x faster — making it a strong choice when speed matters. Note that turbo does not support the translation task.","large":"The large model has gone through three iterations: large-v1 (September 2022), large-v2 (December 2022), and large-v3 (November 2023). Each version improved accuracy, with large-v3 introducing 128 mel-frequency bins (up from 80) and training on a much larger dataset. For a detailed breakdown of each size and how to choose, see our <guideLink>guide to Whisper model sizes</guideLink>."},"accuracy":{"title":"How Accurate Is Whisper?","intro":"Whisper's accuracy is typically measured using Word Error Rate (WER) — the percentage of words that are incorrectly transcribed (insertions, deletions, and substitutions combined). Lower WER is better.","englishTitle":"English Benchmarks","englishParagraph":"On LibriSpeech test-clean — the most widely used English ASR benchmark, consisting of clean read speech from audiobooks — Whisper large-v2 achieves a WER of approximately 3.0%<ref1>[1]</ref1><ref2>[2]</ref2>. This is competitive with commercial ASR systems and purpose-trained models, though specialized models fine-tuned on LibriSpeech can achieve lower WER on that specific benchmark. Whisper's strength lies not in beating narrow benchmarks but in robustness — performing consistently well across a wide range of real-world conditions.","largeV3Title":"Large-v3 Improvements","largeV3Paragraph":"According to OpenAI's evaluation, Whisper large-v3 shows a 10-20% reduction in errors compared to large-v2 across languages where the model achieves below 60% error rate on the Common Voice 15 and Fleurs evaluation datasets<ref2>[2]</ref2><ref3>[3]</ref3>. The improvements are particularly notable for non-English languages, where the expanded training data (5 million total hours) makes the biggest difference.","compareTitle":"How Whisper Compares (February 2026)","compareParagraph":"Since Whisper's release, newer open-source ASR models have surpassed it on clean-speech benchmarks. NVIDIA's Parakeet TDT (0.6B parameters) achieves 1.69% WER and their Canary model reaches 1.6% WER on LibriSpeech test-clean — both significantly below Whisper large-v3's ~2.7%. However, Whisper remains the most widely deployed open-source ASR model due to its ecosystem maturity, language coverage, and the availability of optimized runtimes like whisper.cpp.","chartNote":"Note: Commercial cloud APIs (Deepgram Nova-3, Google Chirp, AssemblyAI Universal-2) do not publish LibriSpeech WER figures, so they cannot be directly compared in this chart. Their benchmarks use proprietary real-world test sets. OpenAI also released GPT-4o-transcribe in March 2025 as an API-only model (not open-source, not Whisper-based) with reportedly lower error rates — but it is cloud-only and not available for local inference.","excelsTitle":"Where Whisper Excels","excels":[{"title":"Robustness to accents and dialects","body":"— Trained on diverse internet audio, Whisper handles regional accents better than systems trained on read speech."},{"title":"Background noise tolerance","body":"— The model maintains reasonable accuracy even with music, other speakers, or ambient noise present."},{"title":"Technical vocabulary","body":"— Due to the breadth of its training data, Whisper handles domain-specific terms (medical, legal, technical) better than many general-purpose ASR systems."},{"title":"Multilingual capability","body":"— A single model supporting 99 languages, with no per-language fine-tuning required."}],"strugglesTitle":"Where Whisper Struggles","struggles":[{"title":"Hallucination","body":"— The most commonly cited issue. During silence or very quiet passages, Whisper can generate text that was never spoken. This is a consequence of the weakly supervised training approach."},{"title":"Low-resource languages","body":"— Languages with little training data (much of the 680K hours is English) have significantly higher error rates."},{"title":"Repetitive text generation","body":"— The autoregressive decoder can get \"stuck\" in loops, producing the same phrase repeatedly. Beam search with specific parameters can mitigate but not eliminate this issue."},{"title":"Not real-time by default","body":"— Whisper processes 30-second chunks, so there is inherent latency. Streaming solutions (like whisper.cpp's real-time mode) work around this but add complexity."}]},"whisperCpp":{"title":"whisper.cpp and Local Inference","paragraph1":"The original Whisper implementation requires Python, PyTorch, and a CUDA-capable GPU for reasonable performance. This makes it impractical for desktop applications, mobile devices, and edge deployments. <whisperCppLink>whisper.cpp</whisperCppLink>, created by Georgi Gerganov, solves this problem.","paragraph2":"whisper.cpp is a complete reimplementation of Whisper inference in plain C/C++ with no dependencies. It reads the same model weights but runs without Python, without PyTorch, and without a GPU (though it benefits enormously from one). Key features include:","features":[{"title":"Zero runtime memory allocations","body":"— All memory is pre-allocated, making performance predictable and suitable for embedded systems."},{"title":"Apple Silicon optimization","body":"— On Macs with M-series chips, the encoder can run on the Apple Neural Engine via Core ML, delivering more than 3x faster inference compared to CPU-only execution. Metal GPU acceleration is also supported for full GPU inference."},{"title":"Quantization support","body":"— Models can be quantized to 4-bit, 5-bit, or 8-bit integer precision, dramatically reducing memory usage and increasing speed with minimal accuracy loss."},{"title":"GPU acceleration","body":"— Supports NVIDIA CUDA (via cuBLAS), Vulkan (cross-vendor), OpenVINO (Intel), and OpenBLAS for CPU acceleration."},{"title":"Voice Activity Detection (VAD)","body":"— Built-in VAD to skip silent segments, improving both speed and transcript quality."}],"memoryTitle":"Memory Requirements (whisper.cpp)","memoryHeaders":{"model":"Model","disk":"Disk Size","ram":"RAM Required"},"memoryRows":[{"model":"tiny","disk":"75 MB","ram":"~273 MB"},{"model":"base","disk":"142 MB","ram":"~388 MB"},{"model":"small","disk":"466 MB","ram":"~852 MB"},{"model":"medium","disk":"1.5 GB","ram":"~2.1 GB"},{"model":"large","disk":"2.9 GB","ram":"~3.9 GB"}],"paragraph3":"whisper.cpp runs on macOS (Intel and Apple Silicon), Linux, Windows, iOS, Android, FreeBSD, Raspberry Pi, and even in the browser via WebAssembly. It achieves faster-than-real-time transcription on modest consumer hardware — including devices as small as the Raspberry Pi 4 and iPhone 13.","paragraph4":"This portability is what makes local, private speech recognition practical. Applications like <openWhisprLink>OpenWhispr</openWhisprLink> use whisper.cpp under the hood to provide real-time push-to-talk dictation that runs entirely on the user's device — no audio ever leaves the machine."},"comparison":{"title":"Whisper vs Other ASR Systems","intro":"Whisper exists in a broader landscape of speech recognition systems. Here is how it compares to the most common alternatives.","cards":[{"title":"Google Cloud Speech-to-Text","body":"Google's cloud ASR service. Excellent accuracy, especially on English, with real-time streaming support and speaker diarization. The primary differences: it is proprietary, requires sending audio to Google's servers, and is priced per minute of audio processed. Whisper offers comparable accuracy for general use, runs locally, and is free — but lacks built-in streaming and diarization.","tags":["Cloud-only","Proprietary","Pay-per-use"]},{"title":"Amazon Transcribe (AWS)","body":"Amazon's cloud ASR, tightly integrated with the AWS ecosystem. Offers medical transcription, call analytics, and custom vocabulary. Similar trade-offs to Google: cloud dependency, per-minute pricing, and proprietary. Whisper is a better fit for applications that need offline operation, open-source licensing, or no vendor lock-in.","tags":["Cloud-only","Proprietary","AWS integration"]},{"title":"Mozilla DeepSpeech","body":"Mozilla's open-source ASR engine based on Baidu's Deep Speech research. Was one of the first high-quality open ASR models, but Mozilla discontinued active development in 2021. Whisper has effectively superseded DeepSpeech in accuracy, multilingual support, and community momentum. DeepSpeech remains useful as a lightweight, well-understood baseline but is no longer recommended for new projects.","tags":["Open source","Discontinued","English-focused"]},{"title":"Vosk","body":"An open-source offline speech recognition toolkit that supports 20+ languages with small, fast models. Vosk is excellent for resource-constrained environments — its models are a fraction of Whisper's size and run well on low-power hardware. The trade-off is accuracy: Vosk is noticeably less accurate than Whisper, particularly on accented speech, noisy audio, and technical vocabulary.","tags":["Open source","Lightweight","Lower accuracy"]},{"title":"Faster Whisper (CTranslate2)","body":"A reimplementation of Whisper using SYSTRAN's CTranslate2 inference engine. It runs the same Whisper models with the same accuracy but is up to 4x faster than the original OpenAI implementation with lower memory usage. Faster Whisper is Python-based (unlike whisper.cpp's C/C++) and is a strong choice for server-side batch transcription. It supports CUDA GPUs and CPU inference with INT8 quantization.","tags":["Open source","Python / CTranslate2","Same Whisper accuracy"]}]},"applications":{"title":"Practical Applications","intro":"Whisper's combination of accuracy, multilingual support, and open-source availability has made it the backbone of a wide range of applications:","cards":[{"title":"Desktop Dictation","body":"Apps like <openWhisprLink>OpenWhispr</openWhisprLink> use whisper.cpp to provide system-wide push-to-talk dictation on macOS, Windows, and Linux. Audio is processed locally — nothing is sent to the cloud."},{"title":"Subtitle Generation","body":"Whisper's timestamp prediction makes it well-suited for generating subtitles and closed captions. Tools like <stableTsLink>stable-ts</stableTsLink> and <autoSubtitleLink>auto-subtitle</autoSubtitleLink> automate this workflow."},{"title":"Podcast and Meeting Transcription","body":"Whisper handles long-form audio transcription well, especially with the chunked processing approach. Many podcast editors and meeting note tools use Whisper (or Faster Whisper) for batch transcription of recordings."},{"title":"Translation and Localization","body":"Whisper's built-in translation capability (any language to English) is used for cross-lingual content access, media localization, and real-time translation prototypes."},{"title":"Accessibility","body":"Real-time captioning for deaf and hard-of-hearing users, voice interfaces for users with motor disabilities, and audio description transcription are all active areas where Whisper is deployed."},{"title":"Data Annotation and Research","body":"Researchers use Whisper to create pseudo-labeled datasets for training other models — the same approach OpenAI used to create large-v3's training data. It is also used extensively in linguistic research and corpus creation."}]},"sources":{"title":"Sources and Further Reading","items":[{"label":"[Paper]","text":"Radford, A., Kim, J. W., Xu, T., Brockman, G., McLeavey, C., & Sutskever, I. (2022). Robust Speech Recognition via Large-Scale Weak Supervision. arXiv:2212.04356.","linkText":"arxiv.org/abs/2212.04356"},{"label":"[Code]","text":"OpenAI Whisper GitHub repository.","linkText":"github.com/openai/whisper"},{"label":"[Code]","text":"whisper.cpp — C/C++ port by Georgi Gerganov.","linkText":"github.com/ggml-org/whisper.cpp"},{"label":"[Model Card]","text":"Whisper large-v2 on Hugging Face (WER benchmark: ~3.0% on LibriSpeech test-clean).","linkText":"huggingface.co/openai/whisper-large-v2"},{"label":"[Model Card]","text":"Whisper large-v3 on Hugging Face (10-20% error reduction over v2).","linkText":"huggingface.co/openai/whisper-large-v3"},{"label":"[Model Card]","text":"Whisper large-v3-turbo on Hugging Face (latency-focused distilled variant).","linkText":"huggingface.co/openai/whisper-large-v3-turbo"},{"label":"[Model Card]","text":"OpenAI Whisper model card (training data composition breakdown).","linkText":"github.com/openai/whisper/blob/main/model-card.md"},{"label":"[Release Notes]","text":"OpenAI large-v3 announcement and evaluation notes.","linkText":"github.com/openai/whisper/discussions/1762"},{"label":"[Release Notes]","text":"OpenAI turbo release discussion and speed-focused tradeoffs.","linkText":"github.com/openai/whisper/discussions/2363"},{"label":"[Code]","text":"Faster Whisper — CTranslate2-based reimplementation by SYSTRAN.","linkText":"github.com/SYSTRAN/faster-whisper"},{"label":"[Product]","text":"OpenWhispr uses Whisper through whisper.cpp for local dictation workflows.","linkText":"openwhispr.com"}],"links":["https://arxiv.org/abs/2212.04356","https://github.com/openai/whisper","https://github.com/ggml-org/whisper.cpp","https://huggingface.co/openai/whisper-large-v2","https://huggingface.co/openai/whisper-large-v3","https://huggingface.co/openai/whisper-large-v3-turbo","https://github.com/openai/whisper/blob/main/model-card.md","https://github.com/openai/whisper/discussions/1762","https://github.com/openai/whisper/discussions/2363","https://github.com/SYSTRAN/faster-whisper","https://openwhispr.com"]},"cta":{"title":"Try Whisper Locally with OpenWhispr","body":"OpenWhispr uses Whisper (via whisper.cpp) for local, private dictation on macOS, Windows, and Linux. Push-to-talk, 99+ languages, no cloud required. Try it free.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":[{"href":"/blog/whisper-model-sizes-explained","label":"Whisper model sizes compared"},{"href":"/blog/local-vs-cloud-transcription","label":"Local vs cloud transcription"},{"href":"/blog","label":"All blog posts"},{"href":"/use-cases/developers","label":"OpenWhispr for developers"},{"href":"/compare","label":"Compare dictation tools"}]},"blogVoicePrivacy":{"metadata":{"title":"Voice Data Privacy: Keep Your Audio On-Device | OpenWhispr","description":"Cloud speech tools log your voice. See why local AI transcription with OpenWhispr keeps audio on-device, private, and fully offline — free and open source.","keywords":["local AI privacy","on-device AI","voice data privacy","local speech recognition","private AI processing","edge AI voice","local LLM","AI without cloud","voice data security","nvidia parakeet"]},"schema":{"title":"The Case for Local AI: Why Your Voice Data Shouldn't Leave Your Device","description":"AI is powerful enough to run on your laptop. Here's why local processing is better for privacy, security, and control."},"breadcrumb":{"home":"Home","blog":"Blog","article":"The Case for Local AI"},"layout":{"title":"The Case for Local AI: Why Your Voice Data Shouldn't Leave Your Device","subtitle":"The hardware in your pocket is powerful enough to run speech recognition that rivals cloud services. So why are we still sending voice data to servers?","date":"February 12, 2026","tag":"Blog"},"sections":[{"id":"opening","title":"Overview"},{"id":"fact-check","title":"Fact-Check Snapshot"},{"id":"local-ai-revolution","title":"The Local AI Revolution"},{"id":"voice-sensitivity","title":"Voice Data Sensitivity"},{"id":"voice-in-the-cloud","title":"Voice in the Cloud"},{"id":"performance-gap","title":"Performance Gap"},{"id":"regulatory-landscape","title":"Regulatory Landscape"},{"id":"cloud-makes-sense","title":"When Cloud Makes Sense"},{"id":"future-of-local-ai","title":"Future of Local AI"},{"id":"what-you-can-do","title":"What You Can Do"},{"id":"sources","title":"Sources"}],"opening":{"lead":"Local AI processes data entirely on your device — no internet connection, no data leaving your machine, no trusting a third party with your information.","body1":"With models like","whisperLink":"OpenAI Whisper running efficiently via whisper.cpp","body2":"and NVIDIA Parakeet, consumer laptops can now transcribe speech with accuracy that matches or approaches cloud services. This article makes the case for why voice data, specifically, should stay on your device — and why the technology to make that happen is already here.","lastUpdated":"Last updated: February 17, 2026. Quantitative and legal claims are cross-checked against at least two primary sources."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","item1":{"lead":"Local ASR is practical on consumer hardware:","body":"Whisper + whisper.cpp and Parakeet have public model/runtime docs with hardware guidance."},"item2":{"lead":"Cloud provider logging behavior is configurable, not uniform:","body":"data handling depends on vendor defaults and opt-ins/opt-outs.","link1":"Google data logging","link2":"AWS AI opt-out policy"},"item3":{"lead":"Voice data can trigger strict compliance obligations:","body":"biometric and AI-system risk rules are codified in EU law.","link1":"GDPR official text","link2":"EU AI Act official text"},"item4":{"lead":"Breach impact is financially material:","body":"breach costs remain high, which raises the stakes of centralizing sensitive audio.","link1":"IBM data breach report","link2":"FTC voice-cloning risk alert"},"item5":{"lead":"OpenWhispr context:","body":"OpenWhispr is designed for local-first transcription with optional cloud usage.","link2":"whisper.cpp backend"}},"revolution":{"title":"The Local AI Revolution: What Changed","p1":"Five years ago, running a capable AI model on a laptop was impractical. The models were too large, the hardware too slow, and the software ecosystem barely existed. That has changed dramatically.","p2a":"The catalyst was","p2b":", Georgi Gerganov's C/C++ port of OpenAI Whisper speech recognition model. By rewriting the model inference in plain C++ with zero runtime dependencies, Gerganov made it possible to run Whisper on everything from a MacBook to a Raspberry Pi. The project supports models from the 75 MB \"tiny\" variant (needing roughly 273 MB of RAM) to the full \"large-v3\" model at 2.9 GB on disk — with integer quantization to reduce memory usage further. It runs on Mac (Intel and Apple Silicon), Linux, Windows, Android, iOS, and even in WebAssembly.","p3a":"The same approach exploded across the AI landscape. Gerganov's","p3b":"did for large language models what whisper.cpp did for speech — it brought LLaMA, Mistral, Qwen, DeepSeek, and dozens of other models to consumer hardware, with quantization from 8-bit down to as low as 1.5-bit. Tools like","p3c":"and","p3d":"wrapped these capabilities in user-friendly interfaces.","p4a":"Hardware caught up, too. Apple's M-series chips, starting with M1 in 2020, put a unified memory architecture and a dedicated Neural Engine into every Mac — making local AI inference fast enough for real-time use. Apple leaned into this with","p4b":", which processes AI tasks on-device by default and only routes to Apple's Private Cloud Compute servers when necessary — with strong guarantees that data is never stored and is used only for the immediate request.","p5":"The trend is clear: models are getting smaller and better simultaneously. Techniques like knowledge distillation (training small models to mimic large ones) and quantization (reducing numerical precision without destroying accuracy) mean that what required a data center GPU in 2023 can run on a laptop in 2026. Whisper's \"large-v3-turbo\" model, for example, is significantly faster than its predecessor while maintaining comparable accuracy."},"sensitivity":{"title":"Why Voice Data Is Uniquely Sensitive","p1":"Not all data is equally sensitive. Text can be anonymized. Browsing history can be cleared. But voice is different — it is a biometric identifier. Your voice is as unique to you as your fingerprint.","p2":"A recording of your voice can reveal far more than the words you spoke. Research in speech analysis has demonstrated that voice patterns can indicate:","identity":{"title":"Identity markers","item1":"Your unique identity (voice is used for biometric authentication by banks and security systems)","item2":"Gender, approximate age, and regional accent","item3":"Native language and socioeconomic background"},"health":{"title":"Health and emotional state","item1":"Emotional state — stress, anxiety, fatigue, and depression can be detected from vocal patterns","item2":"Neurological conditions — voice biomarkers have been studied for early detection of Parkinson's disease","item3":"Respiratory health — conditions like COVID-19 and asthma affect vocal characteristics"},"p3":"This is not speculative. Companies are actively building products that analyze voice for health screening, insurance risk assessment, and hiring decisions. The biometric richness of voice data is precisely what makes it valuable — and precisely what makes it dangerous when it leaves your control.","p4a":"Unlike a password, you cannot change your voice if it is compromised. Once a recording is on a server, you have permanently lost exclusive control over that biometric data. And data breaches are not hypothetical — IBM's 2025 Cost of a Data Breach Report found the global average cost of a breach reached","breachCost1":"$$4.88 million","p4b":"in 2024 and","breachCost2":"$$4.44 million","p4c":"in 2025","ibmCite":"IBM, 2025","p4d":"The question is not whether breaches happen, but when."},"cloud":{"title":"What Happens to Your Voice in the Cloud","p1":"When you use a cloud-based speech recognition service, your audio is transmitted to remote servers for processing. What happens after that varies by provider — and the details matter.","google":{"a":"Google's documentation states that","bold1":"by default, Cloud Speech-to-Text does not log customer audio data or transcripts","b":". However, Google offers a voluntary data logging program where users can opt in to share audio data in exchange for discounted pricing. When opted in, Google uses the data to \"improve service quality.\" Notably, logged data","bold2":"is not deleted when you delete your project","c":"— you must submit a separate deletion request","cite":"Google Cloud Docs"},"aws":{"a":"AWS AI services, including Amazon Transcribe,","bold1":"may use customer content to develop and improve AWS services","b":"unless users explicitly opt out via an organization-level policy. AWS updated this policy in 2023 after public pressure, but the default in many regions still allows data usage for model improvement","cite":"AWS Transcribe Docs"},"azure":{"a":"Azure's Speech Service","bold1":"processes audio in real-time and does not persist customer audio data by default","b":". However, users who opt in to custom model training will have their data stored. Microsoft's data handling is governed by the","cite":"Data Protection Addendum","c":", which varies by service tier and region."},"p2a":"Beyond the providers themselves, consider the infrastructure chain. Your audio may traverse multiple network hops, be processed in a data center in a different country, and potentially be accessible to subprocessors or contractors. Even with strong provider policies, data stored on US-based servers can be subject to government access under","effLink":"National Security Letters and FISA Section 702","p2b":", which do not require notification to the data subject.","callout":{"lead":"To be clear:","body":"these providers generally have strong security practices and legitimate reasons for their policies. The point is not that cloud services are malicious — it is that sending data to any third party inherently means trusting their policies, their security, and their jurisdiction. Local processing eliminates that trust requirement entirely."}},"performance":{"title":"The Performance Gap Is Closing","p1":"The traditional argument for cloud transcription was simple: cloud models were dramatically more accurate. That gap has narrowed significantly.","p2":"OpenAI Whisper large-v3, released in late 2023, achieves word error rates competitive with commercial cloud APIs across most languages. NVIDIA's Parakeet models have pushed accuracy even further, achieving some of the lowest WERs on standard English benchmarks. Independent benchmarks on English speech show these open models achieving word error rates (WER) in the 3-5% range on clean speech — comparable to or better than Google's and AWS's commercial offerings. The newer \"large-v3-turbo\" variant is significantly faster while maintaining similar accuracy, making real-time local transcription practical on modern hardware.","stat1":"Whisper model sizes (tiny to large)","stat2Value":"99+ languages","stat2":"Supported by Whisper models","stat3Value":"Real-time","stat3":"On Apple Silicon and modern x86 CPUs","p3":"Hardware acceleration has made a major difference. whisper.cpp supports ARM NEON and Apple's Accelerate framework and Metal GPU on Mac, CUDA on NVIDIA GPUs, Vulkan for cross-platform GPU inference, and even specialized backends for Ascend NPUs and Intel's OpenVINO. The GGUF model format, developed alongside llama.cpp, enables efficient quantization — reducing a model's memory footprint by 50-75% with minimal accuracy loss.","p4":"For dictation specifically — where recordings are typically 5-30 seconds long — local inference is effectively instant on any machine made in the last three years. The \"small\" Whisper model (466 MB, ~852 MB RAM) offers excellent accuracy for most languages and runs comfortably on machines with 8 GB of memory. You do not need a high-end GPU. You do not need a gaming PC. You need a reasonably modern laptop.","p5":"Apple recognized this trend early. Their on-device speech recognition, built into iOS and macOS, improved substantially with each OS release — supporting dictation without an internet connection for dozens of languages. The fact that Apple, a company with vast cloud infrastructure, is moving speech recognition on-device should tell you something about where this technology is heading."},"regulatory":{"title":"The Regulatory Landscape","p1":"Privacy regulations worldwide are catching up to the reality that voice data is sensitive. For organizations handling voice data, local processing is not just a privacy preference — it is increasingly a compliance advantage.","gdpr":{"title":"GDPR (European Union)","a":"Under the General Data Protection Regulation, voice data is classified as","bold1":"biometric data","b":"when used to uniquely identify a person — placing it in the \"special categories\" of personal data under","link":"Article 9","c":"Processing biometric data requires explicit consent or one of a narrow set of legal bases. Even when not used for identification, voice recordings are personal data subject to GDPR's data minimization principle. Local processing avoids many GDPR obligations entirely — if data never leaves the device, there is no data transfer, no third-party processing, and no cross-border compliance issues."},"aiAct":{"title":"EU AI Act","a":"The EU AI Act, which entered into force in August 2024 with provisions being phased in through 2026, classifies AI systems that process biometric data for identification purposes as","bold1":"high-risk","link":"EU AI Act","b":"High-risk systems face extensive requirements including conformity assessments, documentation obligations, and ongoing monitoring. Real-time biometric identification in public spaces is outright prohibited with narrow exceptions. While standard speech-to-text may not trigger the high-risk classification by itself, any system that could be used to identify speakers from voice — even inadvertently — faces scrutiny under this framework."},"hipaa":{"title":"HIPAA and Professional Privilege","a":"In healthcare contexts, voice recordings that contain patient information are protected under","b":"in the United States. Sending such recordings to a cloud service requires a Business Associate Agreement (BAA) and compliance with the Security Rule. Similarly, attorneys dictating notes about client matters face","bold2":"attorney-client privilege","c":"considerations — sending privileged communications to a third-party server introduces risk. Local processing sidesteps these issues: if the audio never leaves the device, there is no third-party data handler to worry about."},"ccpa":{"title":"CCPA / CPRA (California)","a":"The California Consumer Privacy Act, as amended by the California Privacy Rights Act, treats voice and audio data as regulated personal information, with additional rules when data is used for biometric identification","link1":"California AG","link2":"CPPA FAQ","b":"Businesses collecting voice data must provide additional disclosures and respect consumer rights to deletion and opt-out. Multiple other US states — including Illinois (BIPA), Texas, and Washington — have enacted their own biometric privacy laws with varying requirements and enforcement mechanisms."},"p2":"The regulatory direction is clear: voice data is being treated with increasing seriousness worldwide. For organizations that handle voice data — whether in healthcare, legal, finance, or any regulated industry — local processing offers a structurally simpler compliance posture. No data transfers to audit. No subprocessors to vet. No cross-border data flows to justify."},"cloudSense":{"title":"When Cloud Still Makes Sense","p1":"Intellectual honesty matters. Local AI is not always the right choice, and pretending otherwise would undermine the legitimate points above. Here are situations where cloud processing remains the better option:","case1":{"title":"Massive-scale real-time streaming","body":"Call centers processing thousands of concurrent audio streams need cloud infrastructure. No single device can handle that throughput."},"case2":{"title":"Speaker diarization at scale","body":"Identifying \"who said what\" in multi-speaker recordings is computationally expensive. Cloud APIs from Google and AWS handle this more reliably than most local solutions today."},"case3":{"title":"Underrepresented languages","body":"While Whisper and Parakeet support many languages, accuracy varies. Specialized cloud providers may offer better models for specific languages or dialects that Whisper handles poorly."},"case4":{"title":"Very long recordings on limited hardware","body":"Transcribing a 4-hour meeting recording on a low-end laptop may be painfully slow. Cloud APIs process hours of audio in minutes."},"p2":"The ideal approach is often hybrid: local by default, cloud by choice. Process your everyday dictation locally. Keep sensitive content on-device. And when you genuinely need cloud-scale compute or specialized features, make that a conscious decision — not an invisible default."},"future":{"title":"The Future of Local AI","p1":"The hardware trajectory strongly favors local AI. Modern chips are being designed specifically to run neural networks efficiently:","npu1":{"title":"Apple Neural Engine","body":"Apple's dedicated neural processing unit, present in every M-series and A-series chip, handles up to 38 TOPS (trillion operations per second) on M4. Apple Intelligence relies heavily on this hardware for on-device processing."},"npu2":{"title":"Qualcomm Hexagon NPU","body":"The Snapdragon X Elite series, powering Windows laptops since 2024, includes a 45 TOPS NPU designed explicitly for local AI workloads. Microsoft's Copilot+ PC initiative requires NPU capability."},"npu3":{"title":"Intel AI Boost NPU","body":"Intel's Meteor Lake Core Ultra processors introduced dedicated NPUs at around 11 TOPS, while the newer Lunar Lake generation reaches 48 TOPS. Intel has explicitly positioned these as enabling \"AI PC\" experiences."},"p2a":"On the model side, research in distillation, pruning, and efficient architectures continues to push the boundary of what runs locally. Projects like","p2b":"demonstrate that purpose-built smaller models can achieve 99% of a large model's accuracy at a fraction of the compute cost.","p3":"The convergence is unmistakable: every major chip manufacturer is adding dedicated AI hardware. Every major OS vendor is building on-device AI features. Every major model lab is publishing smaller, more efficient models. On-device AI processing is not an alternative approach — it is becoming the default. Cloud processing will remain important for heavy workloads, but for personal computing tasks like dictation, the future is local."},"actions":{"title":"What You Can Do Today","intro":"You do not need to wait for the future. Local AI is practical right now. Here are concrete steps you can take:","step1":{"title":"Use local speech recognition for dictation","a":"provides push-to-talk dictation powered by OpenAI Whisper and NVIDIA Parakeet with a full GUI. If you prefer the command line,","b":"can be used directly. Both run entirely on your machine."},"step2":{"title":"Run LLMs locally for sensitive work","a":"Tools like","b":"and","c":"let you run capable language models locally. For queries involving personal, financial, or confidential information, a local model eliminates the data exposure entirely."},"step3":{"title":"Audit your current tools","body":"Check the privacy policies of every voice-enabled service you use. Does your dictation app send audio to a server? Does your virtual assistant store recordings? Does your meeting transcription tool train on your content? The answers may surprise you."},"step4":{"title":"Prefer apps that offer local processing","body":"When choosing new tools, favor those that offer on-device processing. The capability exists — demand it. Every user who chooses local processing sends a market signal that privacy matters."}},"sources":{"title":"Sources and Further Reading","s1":"C/C++ port of OpenAI Whisper. Model sizes, hardware requirements, and supported platforms.","s2":"Original ASR model release, training methodology, and evaluation context.","s3Title":"NVIDIA Parakeet model card","s3":"Open ASR metrics and benchmark details for local speech recognition.","s4":"C/C++ LLM inference. Supports 1.5-bit to 8-bit quantization across dozens of model architectures.","s5Title":"Google Cloud Speech-to-Text Data Logging","s5":"Default no-logging policy and opt-in data program details.","s6Title":"AWS Transcribe Security Documentation","s6":"Data handling, opt-out policies, and encryption practices.","s7Title":"IBM Cost of a Data Breach Report 2024","s7":"Global breach costs, average records exposed, and industry analysis.","s8Title":"GDPR Article 9","s8":"Processing of special categories of personal data, including biometric data.","s9Title":"EU AI Act","s9":"Classification of AI systems processing biometric data as high-risk.","s10":"California privacy guidance for personal and sensitive data handling.","s11Title":"Azure Speech data privacy","s11":"Microsoft's service-level speech data handling documentation.","s12":"On-device processing and Private Cloud Compute architecture.","s13Title":"Voice biomarkers for depression detection","s13":"Low, Bentley, Ghosh (2020). Automated assessment of psychiatric disorders using speech.","s14Title":"EFF on FISA and National Security Letters","s14":"Government access to data stored by US companies."},"cta":{"title":"Your Voice Should Stay Yours","body":"OpenWhispr processes everything locally by default. No audio ever leaves your device unless you explicitly choose cloud processing.","subtext":"Open source. Powered by OpenAI Whisper and NVIDIA Parakeet. Available on macOS, Windows, Linux and iOS.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"links":{"openSource":"Why open source matters","localVsCloud":"Local vs Cloud Transcription","allArticles":"All articles","offline":"Offline use cases","medical":"Medical dictation"}},"blogDiarization":{"metadata":{"title":"Speaker Diarization: 100% Local & Private | OpenWhispr","description":"See how OpenWhispr runs speaker diarization locally — pyannote, CAM++ embeddings, sherpa-onnx. Private by design. Open source and free.","keywords":["local speaker diarization","on-device speaker diarization","private meeting transcription","private meeting note taker","local meeting notes","offline meeting transcription","open source meeting assistant","open source meeting note taker","open source AI meeting notes","audio never leaves device","private AI meeting assistant","self-hosted meeting transcription","speaker identification privacy","pyannote segmentation","cam++ speaker embedding","sherpa-onnx diarization","voice fingerprint local","who said what transcript","otter alternative privacy","otter open source alternative","granola open source alternative","fireflies alternative on-device"]},"schema":{"title":"Local Speaker Diarization: How We Built a 100% Private Meeting Note Taker","description":"The full technical breakdown of OpenWhispr's local speaker diarization — pyannote, CAM++, sherpa-onnx. Your audio never leaves your device."},"breadcrumb":{"home":"Home","blog":"Blog","article":"Local Speaker Diarization"},"hero":{"title":"Local Speaker Diarization: How We Built a 100% Private Meeting Note Taker","subtitle":"OpenWhispr is the open source meeting assistant where your audio never leaves your device. Every voice fingerprint, every speaker label, every embedding — processed and stored on your own machine.","date":"April 15, 2026","tag":"Technical"},"sections":[{"id":"what-is-diarization","title":"What Speaker Diarization Is"},{"id":"pipeline","title":"The Four-Stage Pipeline"},{"id":"vad","title":"Stage 1: Voice Activity Detection"},{"id":"segmentation","title":"Stage 2: Segmentation"},{"id":"embeddings","title":"Stage 3: Speaker Embeddings"},{"id":"clustering","title":"Stage 4: Clustering"},{"id":"live","title":"Live Diarization"},{"id":"profiles","title":"Voice Profiles"},{"id":"privacy","title":"Privacy"},{"id":"limitations","title":"Honest Limitations"},{"id":"why-sherpa","title":"Why Sherpa-ONNX"},{"id":"faq","title":"FAQ"}],"opening":{"lead":"Local speaker diarization is the process of labeling who spoke when in a meeting recording, run entirely on the user's device with no audio, embeddings, or transcripts transmitted to any external server.","body":"Meeting transcripts have a “who said what” problem. Zoom gives you a wall of text. Most AI note takers — Otter, Fireflies, Read, Granola — solve it by uploading your meeting audio to their servers, running a speaker model there, and sending back labeled segments. Your voice, and your colleagues' voices, end up as vectors in someone else's database. We didn't want that, and neither do our users. OpenWhispr is the open source meeting note taker built on the opposite assumption: audio never leaves your device, speaker fingerprints live in a local SQLite file on your disk, and nothing about a meeting gets uploaded unless you explicitly choose to. This article is the technical breakdown of the diarization side of that promise — what it is, which models we chose, how they fit together, and exactly where every byte of data lives at every step.","footnote":"Last updated: April 15, 2026. Implementation details reference the speaker diarization feature merged into the OpenWhispr desktop app on April 14, 2026. Every technical claim below is sourced from primary references or directly inspectable in the open source repository."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","items":[{"lead":"Speaker diarization answers “who spoke when.”","body":"It is evaluated against community benchmarks like NIST Rich Transcription and the pyannote.audio reference pipeline.","sources":[{"href":"https://www.nist.gov/itl/iad/mig/rich-transcription-evaluation","label":"NIST RT evaluation"},{"href":"https://arxiv.org/abs/1911.01255","label":"pyannote.audio paper (Bredin et al., 2020)"}]},{"lead":"We use pyannote-segmentation-3.0 and CAM++ from the 3D-Speaker project.","body":"Both are open-source, both are ONNX-portable, and both are directly downloaded from the sherpa-onnx model releases.","sources":[{"href":"https://huggingface.co/pyannote/segmentation-3.0","label":"pyannote-segmentation-3.0 model card"},{"href":"https://arxiv.org/abs/2303.00332","label":"CAM++ / 3D-Speaker paper (Wang et al., 2023)"}]},{"lead":"All inference runs through sherpa-onnx, a native ONNX binary — no Python runtime, no PyTorch, no GPU required.","body":"We version-pin the binary and its dependencies at build time.","sources":[{"href":"https://github.com/k2-fsa/sherpa-onnx","label":"sherpa-onnx on GitHub"},{"href":"https://github.com/k2-fsa/sherpa-onnx/releases","label":"sherpa-onnx releases"}]},{"lead":"Voice fingerprints are 512-dimensional float32 vectors stored as local SQLite BLOBs,","body":"roughly 2KB per speaker. They never leave the device.","sources":[{"href":"https://www.sqlite.org/datatype3.html","label":"SQLite datatype docs"},{"href":"https://onnxruntime.ai/","label":"ONNX Runtime"}]},{"lead":"No audio, no embeddings, and no transcripts are transmitted off-device during diarization.","body":"The only network activity in the entire feature is a one-time model download on first launch, which you can inspect yourself in the open source repo.","sources":[{"href":"https://openwhispr.com/privacy","label":"OpenWhispr privacy policy"},{"href":"https://github.com/OpenWhispr","label":"OpenWhispr on GitHub"}]}]},"whatIs":{"title":"What Speaker Diarization Actually Is","lead":"Speaker diarization is the process of labeling who spoke when in a recording — not what they said.","paragraph1":"Transcription answers what; diarization answers who. The two are complementary: a good meeting note taker needs both, and getting them working together is harder than either one alone.","paragraph2":"Diarization is genuinely hard for a handful of stubborn reasons. Two voices can overlap on the same call. Background noise bleeds in. Short utterances (“yeah,” “okay”) carry almost no speaker information. New speakers show up halfway through a meeting. And unlike transcription, diarization has to make global decisions — it cannot label segment seven without comparing it to segment two. That global context is why most commercial solutions do diarization server-side on the full recording, then ship the labeled transcript back.","paragraph3":"Transcription alone is not enough. A ten-person meeting with forty-five minutes of cross-talk, rendered as one undifferentiated wall of text, is unreadable. Meeting notes — the actual deliverable people want — require speaker attribution, because action items, decisions, and commitments are all attached to people. “Alice will ship the migration by Friday” is a note. “Will ship the migration by Friday” is noise.","calloutTitle":"Diarization is not the same as speaker recognition","calloutBody":"Diarization says “these three chunks are Speaker A; these two are Speaker B.” It does not know who Speaker A is. Speaker recognition puts a name on Speaker A by matching a voice fingerprint to a stored profile. OpenWhispr does both — diarization runs on every meeting, recognition kicks in once you have labeled someone by name. Both stay local."},"pipeline":{"title":"The Four-Stage Local Pipeline","lead":"OpenWhispr's speaker diarization pipeline has four stages: voice activity detection, segmentation, speaker embedding, and clustering — followed by a merge step that reconciles speaker segments with the transcript.","paragraph1":"Each stage is a separate ONNX model or native binary. No Python runtime, no PyTorch, no CUDA. The total on-disk footprint of the models is about 45MB.","paragraph2":"Audio capture is dual-stream from day one. The microphone is definitionally “you,” so we never spend compute trying to diarize your own voice — every mic segment gets labeled you by source, not by voice matching. System audio (everyone else on the call) is the only thing that actually needs speaker-splitting, which immediately halves the work.","paragraph3":"Once audio is captured, the four stages run in sequence. VAD drops silence so the expensive embedding stage never wastes cycles on empty frames. Segmentation cuts continuous speech into single-speaker chunks. The embedding model turns each chunk into a 512-dimensional voice fingerprint. Agglomerative clustering groups fingerprints that look alike, producing a final set of speaker IDs. A last pass merges those IDs back into the transcript by timestamp overlap, so every word ends up tagged with a speaker.","paragraph4Prefix":"The whole thing runs inside a single child process spawned from the app — no server, no listening port, no IPC to anything outside Electron. You can confirm that by reading","paragraph4Suffix":"in the open source repo."},"vad":{"title":"Stage 1: Voice Activity Detection (Silero)","lead":"Voice activity detection (VAD) is the cheap filter that separates speech from silence before the expensive stages run.","paragraph1":"Running a speaker embedding model on silence is wasted CPU and produces garbage vectors. VAD is what keeps the rest of the pipeline fast.","paragraph2Prefix":"We use","sileroLinkLabel":"Silero VAD","paragraph2Suffix":": a 2MB ONNX model, MIT-licensed, effectively the open-source industry default. Silero returns a per-frame probability that the current 32-millisecond window contains speech. On a modern CPU, the cost is negligible — about 0.1 milliseconds per 32-millisecond chunk, or roughly a third of a percent of one core.","paragraph3":"The live pipeline runs Silero continuously on system audio during a call. The thresholds we actually ship with are tuned for real meeting audio, not a clean benchmark:","list":[{"label":"Window size:","value":"512 samples (32ms at 16kHz)"},{"label":"Speech threshold:","value":"0.15 — deliberately aggressive so quiet or distant speakers are not missed"},{"label":"Silence threshold:","value":"0.08"},{"label":"Segment ends after:","value":"16 consecutive silent windows (~512ms)"},{"label":"Minimum segment for embedding:","value":"0.8 seconds"},{"label":"Live identification cadence:","value":"every 1 second once ≥1.6 seconds of speech have accumulated"}],"calloutTitle":"Honest trade-off","calloutBody":"An aggressive speech threshold of 0.15 catches soft talkers on distant mics, but it also occasionally false-triggers on loud keystrokes or room noise. The 0.8-second minimum-segment filter discards most of those false positives before they reach the embedding stage, so downstream accuracy is protected."},"segmentation":{"title":"Stage 2: Segmentation (pyannote 3.0)","lead":"Segmentation takes continuous speech and cuts it into chunks that each contain only one speaker, detecting both speaker-change points and overlap regions.","paragraph1":"It is the stage that decides where the speaker boundaries are. Everything after it assumes those boundaries are correct.","paragraph2Prefix":"We use","paragraph2Mid":", the current version of the model from the pyannote.audio team, ONNX-exported by the","paragraph2Suffix":"project. Pyannote is the de-facto academic baseline: the pyannote.audio pipeline achieves diarization error rates in the 12–15% range on standard benchmarks like AMI and CALLHOME — the same numbers commercial cloud providers like to quote.","paragraph3Prefix":"Why the ONNX port specifically? Because shipping PyTorch inside an Electron app is a non-starter — the runtime alone is roughly 1.5GB and effectively requires a GPU to be fast. The ONNX export is a single 6.6MB file that runs on CPU via","paragraph3Suffix":". Same model, tiny fraction of the weight.","paragraph4Prefix":"We invoke it through sherpa-onnx's offline diarization binary with a handful of flags. The minimum-duration settings come from the pyannote paper's recommended defaults: short speech bursts get merged into adjacent silence, and short silences do not break a continuous speaker segment. Output is a list of start/end time pairs tagged with provisional speaker IDs (","paragraph4Suffix":", etc.)."},"embeddings":{"title":"Stage 3: Speaker Embeddings (CAM++)","lead":"A speaker embedding is a fixed-length list of numbers that represents a voice the way a hash represents a file.","paragraph1":"Two recordings of the same voice produce embeddings that are close in vector space; different voices produce embeddings that are far apart. “Close” is measured by cosine similarity — a single number between -1 and 1. This is the mathematical heart of everything else in the pipeline.","paragraph2Prefix":"Our embedding model is CAM++ from the 3D-Speaker project at Alibaba's DAMO Academy, trained on the","paragraph2Mid":"speaker-verification benchmark. The specific file we ship is","paragraph2Mid2":", a roughly 28MB ONNX export that produces a 512-dimensional float32 vector per segment. The paper is","camppPaperLabel":"arXiv:2303.00332 (Wang et al., 2023)","paragraph2Suffix":".","paragraph3Prefix":"We picked CAM++ over the obvious alternative,","paragraph3Suffix":", for three concrete reasons: roughly half the parameters, lower equal-error rate on VoxCeleb1-O, and faster CPU inference. NVIDIA NeMo's MSDD model has no first-party ONNX export and leans heavily on PyTorch. Picovoice Falcon carries a commercial per-seat license. Apple's CoreML Speech framework is macOS-only. CAM++ was the only option that satisfied our constraints — cross-platform, ONNX, open license, CPU-fast.","paragraph4Prefix":"To feed CAM++ we ship our own log-mel filterbank feature extractor in","paragraph4Mid":": 80 mel bands, 25-millisecond windows, 10-millisecond hops, standard parameters for this class of model. Features flow into the ONNX model via","paragraph4Suffix":", running in the Electron main process. Out the other side comes a 512-entry Float32Array, which we write into SQLite as a 2,048-byte BLOB.","paragraph5":"The matching step itself is the simplest possible thing: cosine similarity between two vectors. Here is the literal code that decides whether two meeting segments are the same speaker:","calloutTitle":"Why 512 dimensions?","calloutBody":"512 is the current sweet spot on the VoxCeleb speaker-verification leaderboard between accuracy and storage cost. 256 dimensions start losing discriminative power on close-but-different voices; 1024 dimensions add storage and compute with diminishing returns. 512 float32s × 2KB × a thousand meetings of ten speakers is 20MB on disk. That fits comfortably in a local SQLite file."},"clustering":{"title":"Stage 4: Clustering — the “how many speakers?” problem","lead":"After embedding every segment you have N vectors. Clustering groups them into K speakers — but the hard part is that you usually don't know K up front.","paragraph1":"A sales call might have 2 speakers. An all-hands might have 8. Fixed-K clustering breaks at both ends.","paragraph2":"We use agglomerative clustering with a cosine-similarity threshold of 0.5: start with every segment as its own cluster, merge the two clusters with the highest similarity, repeat until no remaining pair exceeds the threshold. The threshold, not a target count, decides where to stop. This scales naturally from two speakers to ten without any configuration.","paragraph3Prefix":"The actual invocation lives in","paragraph3Suffix":". It is a spawn of the sherpa-onnx binary with a handful of CLI flags:","paragraph4Prefix":"Output is plain text: one line per segment, formatted as","paragraph4Mid":". We parse that into an array of objects and then walk the transcript, matching each transcript segment to the diarization segment with maximum time overlap. Mic-sourced segments are always labeled","paragraph4Suffix":" — source beats voice, every time.","paragraph5Prefix":"The 0.5 threshold is not a guess. We tuned it against our own eval harness (","paragraph5Suffix":") on real meeting recordings. Higher thresholds under-cluster — one speaker gets split into two. Lower thresholds over-cluster — two similar voices get merged. 0.5 is the sweet spot we converged on."},"live":{"title":"Live Diarization: Labels As You Speak","lead":"Batch-only diarization is accurate but invisible during a call. Live-only is immediate but noisy. OpenWhispr runs both: live labels appear during the meeting, and a full batch pass refines them after.","paragraph1Prefix":"The live path lives in","paragraph1Mid":". It taps the system audio stream at 16kHz, feeds every 32-millisecond frame into Silero VAD, and accumulates speech frames until a segment either closes (after ~512ms of silence) or hits the live-identification cadence (every 1 second once ≥1.6 seconds have accumulated). When it triggers, it extracts a CAM++ embedding, compares it against the in-memory map of active-call embeddings and stored speaker profiles, and fires an IPC event —","paragraph1Suffix":" — with the best match. The React side picks that up and updates the transcript bubble in place.","paragraph2Prefix":"Once a human has manually set a speaker name in the UI, that assignment is","lockedWord":"locked","paragraph2Mid":". The post-meeting batch pass is not allowed to overwrite it, even if its own analysis disagrees. This lock behavior lives in","paragraph2Suffix":", and it matters: the worst possible UX is letting an automated pass silently overwrite a user's correction.","paragraph3":"When the recording stops, we run the full offline diarization on the complete WAV file. Batch has full context — it sees the whole meeting, not just the last 1.6 seconds — and produces a cleaner final labeling. The batch result gets reconciled with live labels, respects every user lock, and replaces provisional guesses with the more accurate grouping.","calloutTitle":"Why hybrid beats either alone","calloutBody":"Live-only is fast but noisy on short segments. Batch-only is accurate but silent until the meeting ends. Hybrid gives you both — instant feedback during the call, and a final transcript that reflects the full-context analysis. The cost is one extra reconciliation step, which runs in the background once and is effectively free."},"profiles":{"title":"Voice Profiles: Remembering Speakers Across Meetings","lead":"Once you label a speaker “Alice” in one meeting, OpenWhispr labels her automatically in every future meeting — without you doing anything, and without her voice fingerprint ever leaving your device.","paragraph1":"The mechanism is straightforward: every speaker's centroid embedding gets stored in your local SQLite database. When new embeddings appear on a future call, we compare them by cosine similarity against every stored profile. A match above threshold auto-assigns the name. Nothing about this touches a server. The three tables that power it are plain SQLite:","matchingIntro":"Matching is three-tier, to balance confidence against false positives:","matchingTiers":[{"label":"Cosine ≥ 0.70:","body":"auto-confirm. The label appears immediately with no prompt."},{"label":"0.55 ≤ Cosine < 0.70:","body":"suggest. The UI shows “Is this Alice?” with confirm / dismiss controls, and waits for your input."},{"label":"Cosine < 0.55:","body":"stay anonymous. The segment remains Speaker N until you name them manually."}],"paragraph2Prefix":"When you confirm a match, the stored centroid updates via running mean rather than a raw overwrite. The formula is simple:","paragraph2Suffix":". This handles voice drift — a cold morning, a different mic, a bad cable — without breaking future matches. The more samples of Alice's voice the system has seen, the more stable her profile becomes.","paragraph3Prefix":"New profiles also trigger","retroactiveWord":"retroactive relabeling","paragraph3Suffix":". When you name a speaker for the first time, a background task walks every past meeting's stored embeddings, finds matches, and updates previously-unnamed speakers in historical transcripts. User-locked mappings are never touched. The result: naming Alice once retroactively labels her in every prior call she attended.","paragraph4Prefix":"If Google Calendar is connected and the meeting has attendees, the speaker-picker UI pre-populates with their names and emails. One click maps","speakerTwo":"Speaker 2","paragraph4Suffix":" and persists forever — attendees and voice profiles stay linked, so the right names show up even across devices."},"privacy":{"title":"Where Your Data Actually Lives","lead":"Audio never leaves your device. That is not a marketing line — it is a mechanical property of the code, and because OpenWhispr is open source you can verify it yourself.","intro":"Here is exactly what happens to each piece of data at each step.","tableHeaderWhat":"What","tableHeaderWhere":"Where it lives","tableHeaderLeaves":"Leaves your device?","tableRows":[{"what":"Raw meeting audio","where":"OS temp directory, deleted after diarization","leaves":"No — ever","leavesEmphasis":true},{"what":"Diarization ONNX models","where":"~/.cache/openwhispr/diarization-models/","leaves":"Downloaded once on first launch, then never again"},{"what":"Speaker embeddings (voice fingerprints)","where":"Local SQLite BLOB","leaves":"No — ever","leavesEmphasis":true},{"what":"Speaker-to-name mappings","where":"Local SQLite table","leaves":"No — ever","leavesEmphasis":true},{"what":"Transcript text","where":"Local SQLite (optional cloud sync if you turn it on)","leaves":"Only if you opt in"}],"paragraph1Prefix":"The raw PCM audio from a meeting is written to a file in your OS temp directory, handed to the locally-spawned sherpa-onnx binary, and deleted in a","paragraph1Mid":"block when diarization completes. There is no code path in OpenWhispr's diarization pipeline that opens a network socket to upload audio. Because the code is open source, you can grep","paragraph1Mid2":", and","paragraph1Mid3":"for","paragraph1Mid4":", or","paragraph1Suffix":"and confirm that yourself.","paragraph2Prefix":"IPC between Electron's main and renderer processes uses the built-in","paragraph2Suffix":"— a kernel-level pipe inside a single process, not a network socket. Speaker embeddings and voice profiles are stored as SQLite BLOBs in the same database file as your notes, with the same OS-level file permissions.","calloutTitle":"One exception, transparently","calloutBody":"The only network activity in this feature is the one-time model download on first launch — three ONNX files fetched from the sherpa-onnx GitHub releases CDN. We document this explicitly because “one network call during setup, zero after” is a very different privacy story from “always-on cloud inference,” and as an open source meeting assistant we would rather tell you than have you discover it."},"limitations":{"title":"Honest Limitations","lead":"Local diarization is good, not perfect. Here are the cases where it genuinely struggles, and what we do about each one.","items":[{"label":"Overlapping speakers.","body":"Two people talking at once is inherently lossy for single-label attribution. Pyannote 3.0 detects the overlap region, but a single speaker_id per segment is not enough to represent “both people talking.” We mark those segments as provisional and let you correct them in the UI."},{"label":"Cold-start accuracy.","body":"The first meeting with a new colleague uses only the generic VoxCeleb-trained features. Once you label them once and the profile has a few samples, subsequent meetings are noticeably more accurate."},{"label":"Short utterances.","body":"“Yeah,” “okay,” a quick laugh — anything under ~0.8 seconds cannot produce a reliable embedding. We drop those segments from the embedding stage and fall back to label propagation from the surrounding context."},{"label":"Far-field or low-SNR audio.","body":"Someone on a phone speakerphone in a noisy room will degrade gracefully but visibly. Automatic gain control and VAD tuning help, but no amount of model sophistication completely overcomes bad physics."},{"label":"CPU-only inference.","body":"A GPU would be faster, but we deliberately don't require one. In practice, throughput on a modern CPU is fine: ~100 milliseconds per embedding, and about 30 seconds of batch diarization for a 45-minute meeting on an M1 Mac."}],"outroPrefix":"We publish our eval harness (","outroSuffix":") in the open source repo so you can measure any of this yourself, on your own recordings — not our cherry-picked demo clips. That feels more honest than a glossy number in a marketing deck."},"whySherpa":{"title":"Why We Chose Sherpa-ONNX","lead":"We needed a diarization runtime that was cross-platform, Python-free, CPU-fast, and open source. Sherpa-ONNX was the only option that hit all four.","paragraph1":"Our requirements were concrete. It had to run on macOS (both Intel and Apple Silicon), Windows x64, and Linux x64 without separate code paths. It could not bundle PyTorch — the runtime alone is about 1.5GB and essentially requires a GPU to be usable at interactive latencies. It could not require CUDA. And the license had to be permissive enough to ship in a commercial-but-open-source desktop app.","evaluatedIntro":"Here is what we actually evaluated:","evaluated":[{"label":"pyannote.audio (Python):","body":"ruled out. Bundling PyTorch in an Electron app is a non-starter."},{"label":"NVIDIA NeMo MSDD:","body":"PyTorch-dependent, no first-party ONNX export, GPU-centric."},{"label":"Picovoice Falcon:","body":"commercial per-active-user license plus cloud-activation on first run."},{"label":"Apple CoreML + Speech framework:","body":"macOS-only, no cross-platform parity."},{"label":"sherpa-onnx","body":"Apache 2.0, native ONNX binary, ~35MB of models, spawn-and-read-stdout interface, actively maintained by k2-fsa. Same toolchain powers production ASR in other shipping products."}],"outroPrefix":"One honest caveat: sherpa-onnx's offline speaker diarization binary is relatively new. We version-pin to","outroSuffix":" and run our full eval harness against every update before bumping the pin. That's the right trade-off for something this load-bearing."},"faqHeading":"Frequently Asked Questions","faqs":[{"question":"Is there a private meeting note taker that doesn't upload my audio to the cloud?","answer":"Yes. OpenWhispr is a free, open source meeting note taker where your audio never leaves your device. Transcription runs locally with OpenAI Whisper or NVIDIA Parakeet, speaker diarization runs locally with pyannote segmentation and CAM++ embeddings, and voice fingerprints are stored in a local SQLite file on your own machine — not on our servers."},{"question":"Can Otter, Fireflies, or Granola see my meetings?","answer":"Yes. Cloud-based AI note takers upload your meeting audio to their servers in order to transcribe and diarize it. Speaker labels, transcripts, and voice embeddings are computed in their infrastructure and stored in their databases. If you need an alternative that keeps every byte of audio on your device, you need a local-first tool like OpenWhispr."},{"question":"What is speaker diarization?","answer":"Speaker diarization is the process of labeling who spoke when in a recording, as distinct from transcription, which answers what was said. A diarized transcript shows each utterance tagged with a speaker label — Alice, Bob, Speaker 3 — instead of one undifferentiated wall of text."},{"question":"Does local speaker diarization actually work, or is it a downgrade from the cloud?","answer":"It works. OpenWhispr's pipeline — pyannote-segmentation-3.0 plus CAM++ embeddings from the 3D-Speaker project, run via sherpa-onnx — achieves diarization error rates in the same ballpark as commercial cloud APIs on standard benchmarks. The main honest trade-offs are overlapping-speaker handling and cold-start accuracy on voices the system has never heard before."},{"question":"How much disk space do the local diarization models use?","answer":"About 45MB in total: pyannote segmentation (6.6MB) plus the CAM++ speaker embedding model (28MB) plus Silero VAD (2MB). All three are downloaded once on first launch and cached at ~/.cache/openwhispr/diarization-models/. After that they stay on disk and never re-download."},{"question":"Do I need a GPU to run OpenWhispr's local diarization?","answer":"No. Everything runs on CPU via ONNX Runtime. Modern Apple Silicon and recent x86 CPUs handle it comfortably — in practice, about 30 seconds of batch diarization for a 45-minute meeting on an M1 Mac, with live speaker labels appearing within a second or two during the call itself."},{"question":"Is OpenWhispr actually open source?","answer":"Yes. The diarization code, the transcription pipeline, and the rest of the desktop app are open source on GitHub. The privacy claims in this article are directly verifiable: grep the diarization helpers for fetch, axios, or http and you will not find any network calls in the diarization path."},{"question":"What happens when two people talk over each other on a call?","answer":"Pyannote 3.0 explicitly detects and annotates overlap regions, but single-label attribution during overlap is inherently lossy. We mark those segments as provisional in the transcript and let you correct them by clicking the speaker label. Your manual corrections are locked and are never overwritten by later automated passes."}],"cta":{"title":"Try the Private Meeting Note Taker","body":"OpenWhispr is a free, open source meeting assistant where your audio never leaves your device. Local diarization, local transcription, local voice profiles. Works on macOS, Windows, and Linux.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":[{"href":"/blog/local-vs-cloud-transcription","label":"Local vs cloud transcription"},{"href":"/blog/how-whisper-ai-works","label":"How Whisper AI works"},{"href":"/blog/local-ai-voice-data-privacy","label":"The case for local AI"},{"href":"/blog/why-open-source-voice-tools","label":"Why open source voice tools"},{"href":"/blog","label":"All blog posts"}]},"blogLocalVsCloud":{"metadata":{"title":"Local vs Cloud Speech-to-Text: Privacy, Speed & Cost","description":"Compare local vs cloud speech-to-text on privacy, accuracy, speed, and cost. See why on-device transcription keeps your audio private and free.","keywords":["local vs cloud transcription","local speech to text","cloud transcription privacy","offline transcription","local voice recognition","private speech to text","on-device transcription","cloud ASR vs local ASR","nvidia parakeet","whisper vs parakeet"]},"schema":{"title":"Local vs Cloud Transcription: Privacy, Speed, and Accuracy Compared","description":"An honest comparison of local and cloud-based speech-to-text — privacy, latency, accuracy, cost, and offline capability."},"breadcrumb":{"home":"Home","blog":"Blog","article":"Local vs Cloud Transcription"},"sections":{"factCheck":"Fact-Check Snapshot","howCloudWorks":"Cloud Transcription","howLocalWorks":"Local Transcription","privacy":"Privacy","accuracy":"Accuracy","speed":"Speed & Latency","cost":"Cost","offline":"Offline Capability","comparison":"Comparison Table","recommendations":"Recommendations","sources":"Sources"},"hero":{"title":"Local vs Cloud Transcription: Privacy, Speed, and Accuracy Compared","subtitle":"Your voice data is valuable. Here is what happens to it depending on where you process it.","date":"February 8, 2026","tag":"Comparison"},"intro":{"lead":"Local transcription processes your audio entirely on your device","body1":"using models like OpenAI Whisper and NVIDIA Parakeet, running via optimized runtimes such as whisper.cpp. No audio ever leaves your machine.","cloudLead":"Cloud transcription sends your audio to remote servers","body2":"operated by providers like Google, AWS, Microsoft, Deepgram, or AssemblyAI, where large GPU clusters process it and return a text transcript. The key tradeoffs between the two approaches come down to five factors: privacy, latency, accuracy, cost, and offline capability. This article examines each one with real numbers and real policies, so you can make an informed decision.","lastUpdated":"Last updated: February 17, 2026. Quantitative and policy claims are cross-checked against at least two primary sources."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","item1":{"lead":"Open local models are production-capable:","body":"Whisper and Parakeet have public model cards and open implementations suitable for on-device inference.","source1":"OpenAI Whisper","source2":"NVIDIA Parakeet"},"item2":{"lead":"Cloud STT is usage-priced:","body":"major providers meter by audio volume and model tier.","source1":"Google pricing","source2":"AWS pricing"},"item3":{"lead":"Provider data handling differs:","body":"logging/retention settings and opt-outs are service-specific and must be configured.","source1":"Google data logging","source2":"AWS AI opt-out policy"},"item4":{"lead":"Regulatory context matters:","body":"biometric and sensitive-data obligations vary by jurisdiction and use case.","source1":"GDPR official text","source2":"EU AI Act official text"},"item5":{"lead":"OpenWhispr deployment model:","body":"OpenWhispr defaults to local processing and also supports BYOK cloud routing when users need it.","source1":"OpenWhispr","source2":"whisper.cpp runtime"}},"howCloudWorks":{"title":"How Cloud Transcription Works","paragraph1":"When you use a cloud-based speech-to-text service, your audio follows a multi-step journey. First, your device captures raw audio from your microphone and compresses it into a format suitable for transmission — typically Opus, FLAC, or linear PCM. This compressed audio is then sent over the internet to the provider's API endpoint, usually via HTTPS or a WebSocket connection for real-time streaming.","paragraph2":"On the server side, the provider runs your audio through large neural network models hosted on GPU clusters. These models are typically trained on hundreds of thousands of hours of labeled speech data, far more than what any individual could assemble. The resulting transcript is sent back to your device.","providersIntro":"The major cloud transcription providers include:","providers":{"google":"— Broad language coverage with batch and real-time streaming options.","aws":"— Amazon's offering with automatic language identification, custom vocabulary support, and PII redaction.","azure":"— Deep integration with the Microsoft ecosystem, custom speech models, and real-time transcription.","deepgram":"— Known for speed and developer experience, with their Nova-3 model offering strong accuracy at competitive pricing.","assemblyai":"— Focused on accuracy and developer tooling, with built-in speaker diarization and content moderation."},"paragraph3":"Cloud latency is variable because it includes capture, upload, server-side inference, and response delivery. In well-connected environments it can feel fast, while unstable networks or high-latency regions can make the experience noticeably slower. Batch transcription can take from seconds to minutes depending on file length and queueing."},"howLocalWorks":{"title":"How Local Transcription Works","paragraph1":"Local transcription runs the speech recognition model directly on your device. There is no network involved — audio goes from your microphone straight into a model running on your CPU or GPU, and the text comes out the other side. The entire process happens in memory on your machine.","paragraph2a":"The breakthrough that made high-quality local transcription practical was OpenAI's release of Whisper in September 2022 — an open-source model trained on 680,000 hours of multilingual audio. More recently, NVIDIA released its Parakeet family of ASR models, which achieve state-of-the-art accuracy on English benchmarks and are available under open licenses. Shortly after, Georgi Gerganov created","paragraph2b":", a C/C++ reimplementation optimized for CPU inference, which brought Whisper's accuracy to everyday hardware without requiring a dedicated GPU.","modelTable":{"intro":"Whisper comes in several model sizes, each trading accuracy for speed:","headerModel":"Model","headerParameters":"Parameters","headerVram":"VRAM","headerSpeed":"Relative Speed","rows":[{"model":"Tiny","params":"39M","vram":"~1 GB","speed":"~10x"},{"model":"Base","params":"74M","vram":"~1 GB","speed":"~7x"},{"model":"Small","params":"244M","vram":"~2 GB","speed":"~4x"},{"model":"Medium","params":"769M","vram":"~5 GB","speed":"~2x"},{"model":"Large-v3","params":"1550M","vram":"~10 GB","speed":"1x (baseline)"},{"model":"Turbo","params":"809M","vram":"~6 GB","speed":"~8x"}]},"paragraph3":"On Apple Silicon (M1 through M4), whisper.cpp runs inference fully on the GPU via Metal, achieving real-time or faster-than-real-time speeds even with the medium model. The Turbo model — an optimized variant of large-v3 with minimal accuracy loss — runs at high throughput on modern MacBooks, which is why it is commonly used for low-latency local dictation workflows.","paragraph4a":"Other local transcription engines include","paragraph4b":"(lightweight, lower accuracy) and","paragraph4c":"(next-gen Kaldi with ONNX runtime support). However, OpenAI Whisper via whisper.cpp and NVIDIA Parakeet remain the gold standard for local transcription quality."},"privacy":{"title":"Privacy: The Core Tradeoff","paragraph1":"Privacy is the single biggest differentiator between local and cloud transcription. With local processing, the question of \"what happens to my audio data?\" has a simple answer: nothing. It stays on your device, in your RAM, and is discarded after transcription. There is no third party to trust, no policy to read, no data breach to worry about.","paragraph2":"Cloud transcription requires you to trust your provider. Here is what the major providers actually say about how they handle your audio:","google":{"body1":"Google's data logging for Speech-to-Text is off by default. When disabled, Google states that audio data is processed in memory, used only to return the transcription result, and not stored on servers. However, if you opt into data logging (or use certain features that require it), your audio and transcriptions may be stored and used to improve Google's models. Google's","link":"data logging documentation","body2":"details these distinctions.","footnote":"Data processing location can be specified by region, which matters for GDPR compliance."},"aws":{"body1":"By default, AWS may use content processed by Amazon Transcribe to develop and improve AWS AI/ML services. However, you can opt out of this by using","link":"AWS AI service opt-out policies","body2":". When opted out, AWS states that your content is not stored or used for service improvement. Audio data is encrypted in transit and at rest.","footnote":"Important: the opt-out is not the default. You have to actively configure it."},"azure":{"body1":"Azure does not use customer data to improve its base models by default. Their","link":"data privacy documentation","body2":"states that audio sent to the Speech service is processed and immediately deleted from their servers. If you create a custom model, the training data you provide is stored in the same region as the resource until you explicitly delete it.","footnote":"Azure offers disconnected container deployment for fully on-premises processing."},"regulatory":{"title":"Regulatory Considerations","gdprLead":"GDPR:","gdprBody":"Voice recordings are considered biometric data under EU law. Sending audio to US-based cloud providers raises data transfer concerns under the GDPR, even with Standard Contractual Clauses (SCCs) or the EU-US Data Privacy Framework. Local processing avoids this entirely by keeping biometric data within the data subject's own device.","hipaaLead":"HIPAA:","hipaaBody":"Healthcare providers using cloud transcription must ensure their provider offers a Business Associate Agreement (BAA). Google, AWS, and Azure all offer BAAs, but the configuration requirements are strict. Local transcription sidesteps HIPAA data handling rules because protected health information never leaves the covered entity's control."},"localAdvantage":{"title":"The Local Advantage","body":"With local transcription, there is no privacy policy to parse, no data retention schedule to trust, and no opt-out toggle to find. Your audio is processed in RAM and never written to disk (unless you choose to save a recording). This is a fundamentally different trust model — you do not need to trust anyone, because the data never leaves your machine."}},"accuracy":{"title":"Accuracy: How Close Is Local?","paragraph1":"Let us be honest: the top cloud providers still have an edge in raw accuracy for certain use cases. They train on vast proprietary datasets, offer real-time model updates, and can leverage context-specific custom vocabularies. For challenging audio — heavy accents, noisy environments, domain-specific jargon — cloud services like Google's Chirp 2 or Deepgram's Nova-3 tend to perform well.","paragraph2":"That said, the gap has narrowed dramatically. OpenAI Whisper large-v3, trained on 680,000 hours of multilingual audio, achieves competitive Word Error Rates (WER) across most common languages. NVIDIA's Parakeet models have pushed the bar even further, achieving some of the lowest WERs on standard English benchmarks. Independent benchmarks consistently show these open models performing within a few percentage points of commercial cloud services for English, and sometimes outperforming them for specific languages in the long tail.","paragraph3":"The Turbo model — a distilled variant of large-v3 with 809M parameters instead of 1.55B — retains most of the accuracy while running roughly 8x faster. For real-time dictation where you are speaking clearly into a decent microphone, the difference between Turbo and a cloud API is negligible for most users.","cloudAdvantages":{"title":"Cloud Accuracy Advantages","item1":"Custom vocabulary and domain adaptation","item2":"Continuous model improvements server-side","item3":"Better speaker diarization (who said what)","item4":"Automatic punctuation and formatting tuned per language"},"localAdvantages":{"title":"Local Accuracy Advantages","item1":"No audio compression artifacts from network transmission","item2":"Consistent, deterministic performance — no server variability","item3":"Full uncompressed 16kHz audio fed directly to the model","item4":"Whisper large-v3 and Parakeet match cloud for clear-speech dictation"},"bottomLine":"Bottom line: if you are dictating in a quiet environment with a decent microphone, models like Whisper Turbo, Whisper large-v3, or Parakeet running locally will give you results indistinguishable from cloud services for most practical purposes. Cloud pulls ahead for noisy multi-speaker scenarios, niche languages, and domain-specific vocabularies."},"speed":{"title":"Speed and Latency","paragraph1":"Latency matters most for real-time dictation, where you want to see your words appear immediately after speaking. Here, local and cloud transcription have fundamentally different latency profiles.","cloudBreakdown":{"title":"Cloud Latency Breakdown","item1":"Audio capture + encoding","item2Lead":"Network upload and download","item2Rest":"(variable by route quality)","item3Lead":"Server-side queueing + inference","item3Rest":"(variable by provider load)","summary":"User-perceived latency can vary significantly by network and region."},"localBreakdown":{"title":"Local Latency (Apple Silicon)","item1Lead":"Tiny/base:","item1Rest":"fastest response, lower quality ceiling","item2Lead":"Small:","item2Rest":"balanced speed and quality","item3Lead":"Turbo:","item3Rest":"high-quality real-time dictation candidate","item4Lead":"Large models:","item4Rest":"highest quality, more compute-heavy","summary":"No network dependency. Consistent regardless of connection."},"paragraph2":"For real-time streaming, cloud providers can return partial hypotheses quickly when connectivity is strong. But this still requires a stable low-latency connection. On a plane, in congested public Wi-Fi, or behind restrictive VPN routes, perceived latency can degrade quickly.","paragraph3":"Local transcription with whisper.cpp is not real-time streaming in the traditional sense. Most local implementations use a \"push-to-talk\" pattern: you speak, the audio is buffered, and then the model processes the complete segment. With the Turbo model on an Apple M-series chip, this processing step is fast enough that the delay feels negligible for dictation — typically under a second for short utterances.","footnote":"For long-form transcription of pre-recorded files, cloud services can parallelize across GPU clusters and often finish large batches quickly. Local throughput is bounded by your machine and model choice, so runtime can be slower for very long recordings."},"cost":{"title":"Cost: Free vs. Pay-Per-Minute","paragraph1":"Local transcription costs nothing beyond the hardware you already own. There is no API key, no billing dashboard, no per-minute charge. You download a model file (ranging from 75MB for Tiny to 3GB for Large-v3), and you are done. Forever.","paragraph2":"Cloud providers charge for usage (per-minute or per-hour, depending on vendor and model). Plans and discounts change frequently, so verify current rates before budgeting:","table":{"headerProvider":"Provider","headerBatch":"Batch / Pre-recorded","headerStreaming":"Streaming / Real-time","headerFreeTier":"Free Tier","rows":[{"provider":"Google Speech-to-Text","batch":"Usage-priced (tiered by model)","streaming":"Usage-priced (tiered by model)","free":"Trial/free quotas vary by account"},{"provider":"AWS Transcribe","batch":"Usage-priced","streaming":"Usage-priced","free":"Intro free tier (time-limited)"},{"provider":"Azure Speech","batch":"Varies by region/model","streaming":"Varies by region/model","free":"Limited free quota"},{"provider":"Deepgram","batch":"Usage-priced by model tier","streaming":"Usage-priced by model tier","free":"Credit-based trial"},{"provider":"AssemblyAI","batch":"Usage-priced by model tier","streaming":"Usage-priced by model tier","free":"Credit-based trial"},{"provider":"Local (Whisper / Parakeet)","batch":"Free","streaming":"Free","free":"Unlimited, forever"}]},"estimate":{"title":"How to Estimate Your Cloud Spend","body1a":"Use this simple formula:","formula":"annual minutes of audio x vendor rate","body1b":"If your team dictates heavily, total cost scales linearly with usage and seat count.","body2":"Local inference (Whisper/Parakeet) avoids ongoing API charges, which is why many teams keep local as the default and only route special cases to cloud."},"footnote":"Note: cloud pricing often includes additional charges for features like speaker diarization, PII redaction, or custom vocabularies. The base prices above are for standard transcription only."},"offline":{"title":"Offline Capability: When Local Wins Definitively","paragraph1":"This one is binary: cloud transcription requires internet, local transcription does not. For many people, this is the deciding factor.","scenariosIntro":"Scenarios where offline capability is not optional:","travel":{"title":"Travel and Remote Work","body":"Flights, trains, rural areas, developing regions — anywhere with spotty or nonexistent Wi-Fi. Local transcription works exactly the same at 35,000 feet as it does at your desk."},"airGapped":{"title":"Air-Gapped Environments","body":"Government facilities, defense contractors, secure research labs, and financial institutions often operate on air-gapped networks where outbound internet access is blocked entirely. Local transcription is the only option."},"fieldWork":{"title":"Field Work","body":"Journalists in conflict zones, researchers in remote field stations, healthcare workers in rural clinics — these professionals need reliable transcription in environments where cell service may be unreliable or unavailable."},"reliability":{"title":"Reliability","body":"Even in well-connected offices, cloud services experience outages. AWS, Google Cloud, and Azure have all had multi-hour incidents that affected speech APIs. Local transcription has no external dependency that can go down."}},"comparisonTable":{"title":"Side-by-Side Comparison","headerFactor":"Factor","headerLocal":"Local","headerCloud":"Cloud","headerNotes":"Notes","rows":[{"factor":"Privacy","local":"full","cloud":"none","notes":"Local: data never leaves device. Cloud: audio sent to third-party servers."},{"factor":"Accuracy (English)","local":"partial","cloud":"full","notes":"Cloud models are marginally better; Whisper large-v3 and NVIDIA Parakeet close the gap significantly."},{"factor":"Accuracy (Multilingual)","local":"partial","cloud":"full","notes":"Cloud providers have broader language coverage with custom-trained models."},{"factor":"Latency","local":"full","cloud":"partial","notes":"Local avoids network round-trips. Cloud latency depends on connection quality and region."},{"factor":"Cost","local":"full","cloud":"none","notes":"Local has no per-minute API billing. Cloud services use usage-based pricing."},{"factor":"Offline Use","local":"full","cloud":"none","notes":"Local works anywhere. Cloud requires internet."},{"factor":"Setup Ease","local":"partial","cloud":"full","notes":"Cloud is an API call. Local requires model downloads and hardware consideration."},{"factor":"Speaker Diarization","local":"none","cloud":"full","notes":"Multi-speaker identification is a strong cloud advantage."},{"factor":"Scalability","local":"none","cloud":"full","notes":"Cloud scales infinitely. Local is bound by your hardware."}]},"recommendations":{"title":"When to Choose What","intro":"Neither approach is universally better. The right choice depends on what you prioritize.","chooseLocal":{"title":"Choose Local If...","item1":"Privacy is non-negotiable (medical, legal, personal journals)","item2":"You need offline or air-gapped operation","item3":"You want zero ongoing costs for transcription","item4":"Your primary use case is dictation (single speaker, clear audio)","item5":"You are subject to GDPR, HIPAA, or similar regulations"},"chooseCloud":{"title":"Choose Cloud If...","item1":"You need speaker diarization for multi-person meetings","item2":"Maximum accuracy for difficult audio conditions","item3":"You are building an application that needs to scale to many users","item4":"You need real-time streaming transcription","item5":"Your hardware is limited and cannot run large models"},"hybrid":{"title":"The Hybrid Approach","body":"You do not have to choose one exclusively. Some applications — including OpenWhispr — support both modes: use local models like Whisper and Parakeet by default for privacy and zero cost, and optionally bring your own cloud API key (OpenAI, Deepgram, etc.) when you need the extra accuracy or features that cloud provides. This \"BYOK\" (Bring Your Own Key) model gives you the best of both worlds without locking you into either approach."}},"sourcesSection":{"title":"Sources and Further Reading","whisper":"— Original model repository with model sizes, benchmarks, and documentation.","whisperCpp":"— C/C++ port of Whisper optimized for CPU and Apple Silicon GPU inference.","parakeetLabel":"NVIDIA Parakeet RNNT model card","parakeet":"— Open ASR model metrics and usage guidance.","googlePricingLabel":"Google Cloud Speech-to-Text Pricing","googlePricing":"— Current pricing tiers for Google's speech API.","awsPricingLabel":"AWS Transcribe Pricing","awsPricing":"— Amazon's current usage-based pricing for batch and streaming.","azurePricingLabel":"Azure Speech Services Pricing","azurePricing":"— Microsoft's pricing for real-time and batch speech-to-text.","deepgramPricingLabel":"Deepgram Pricing","deepgramPricing":"— Deepgram model-tier pricing details.","assemblyaiPricingLabel":"AssemblyAI Pricing","assemblyaiPricing":"— AssemblyAI usage and model-tier pricing details.","googleLoggingLabel":"Google Speech-to-Text Data Logging","googleLogging":"— Google's documentation on how audio data is handled and stored.","awsOptOutLabel":"AWS AI Service Opt-Out Policies","awsOptOut":"— How to prevent AWS from using your content to improve their AI services.","azurePrivacyLabel":"Azure Speech Data Privacy","azurePrivacy":"— Microsoft's data privacy and security documentation for Speech Services.","whisperPaperLabel":"Robust Speech Recognition via Large-Scale Weak Supervision","whisperPaper":"— The original Whisper paper by Radford et al. (2022), detailing training methodology and WER benchmarks.","gdprLabel":"GDPR (official EU text)","gdpr":"— Legal baseline for personal/biometric data handling.","aiActLabel":"EU AI Act (official EU text)","aiAct":"— AI risk-tier obligations relevant to biometric and high-risk systems.","openwhispr":"— Example of a local-first + optional BYOK cloud workflow in one app."},"cta":{"title":"Try Both Approaches in One App","body":"OpenWhispr supports both local models (OpenAI Whisper and NVIDIA Parakeet) for privacy and zero cost, and bring-your-own cloud API keys when you need maximum accuracy. Open source, free forever.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":{"localAi":"The case for local AI","offline":"Offline transcription","allBlog":"All blog posts","allComparisons":"All comparisons"}},"blogModelSizes":{"metadata":{"title":"Whisper Model Sizes: Complete Guide | OpenWhispr","description":"Every Whisper model size compared: parameters, speed, accuracy, and RAM. Choose between tiny, small, large-v3, and turbo for local transcription.","keywords":["whisper model sizes","whisper tiny vs large","whisper model comparison","which whisper model to use","whisper large v3","whisper model accuracy","whisper model speed","whisper model parameters","whisper.cpp model size"]},"schema":{"title":"Whisper Model Sizes Explained: Tiny vs Base vs Small vs Medium vs Large","description":"A practical guide to choosing the right Whisper model. Parameters, speed, accuracy, and hardware requirements compared."},"breadcrumb":{"home":"Home","blog":"Blog","article":"Whisper Model Sizes Explained"},"layout":{"title":"Whisper Model Sizes Explained: Tiny vs Base vs Small vs Medium vs Large","subtitle":"Every Whisper model, compared. Parameters, speed, accuracy, and which one you should actually use.","date":"February 5, 2026","tag":"Blog"},"sectionsToc":{"factCheck":"Fact-Check Snapshot","comparisonTable":"Model Comparison","tiny":"Tiny (39M)","base":"Base (74M)","small":"Small (244M)","medium":"Medium (769M)","large":"Large (1.55B)","turbo":"Turbo (809M)","englishVsMultilingual":"English vs Multilingual","hardware":"Hardware Requirements","recommendations":"Recommendations","openwhispr":"OpenWhispr","sources":"Sources"},"cards":{"resources":"Resources","speed":"Speed","accuracy":"Accuracy","bestFor":"Best for:"},"intro":{"lead":"OpenAI's Whisper comes in 9 model variants across 5 size categories:","body1":"tiny (39M parameters), base (74M), small (244M), medium (769M), and large (1.55B). Each size has a multilingual version and an English-only version (except large and turbo, which are multilingual only). There is also a turbo model (809M) — a distilled version of large-v3 that is nearly as accurate but dramatically faster. Larger models are more accurate but slower and require more memory.","body2":"For the whisper.cpp quantized (GGML) format used by most local transcription apps, model files range from 75 MB (tiny) to 2.9 GB (large). Memory usage at runtime ranges from around 273 MB for tiny to 3.9 GB for large. The right model depends on your hardware, your language, and how much accuracy you need.","lastUpdated":"Last updated: February 17, 2026. Parameter, memory, and benchmark claims are cross-checked against at least two primary sources."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","item1":{"lead":"Official model families and parameter counts:","body":"tiny/base/small/medium/large plus turbo are documented by OpenAI.","link1":"OpenAI Whisper README","link2":"Whisper model card"},"item2":{"lead":"GGML/GGUF disk and runtime memory context:","body":"local deployment footprints come from whisper.cpp conversion/runtime docs.","link1":"whisper.cpp","link2":"whisper.cpp memory notes"},"item3":{"lead":"Benchmark caveat:","body":"WER varies heavily by dataset and language; LibriSpeech figures are not a universal quality score.","link1":"Whisper paper","link2":"large-v2 model card"},"item4":{"lead":"Turbo model positioning:","body":"turbo is a speed-focused variant with accuracy tradeoffs documented in release notes/model cards.","link1":"turbo model card","link2":"turbo release discussion"},"item5":{"lead":"OpenWhispr usage context:","body":"OpenWhispr exposes multiple Whisper sizes so users can tune speed/accuracy on-device.","link1":"OpenWhispr","link2":"whisper.cpp backend"}},"table":{"title":"The Complete Model Comparison","intro":"This table covers every official Whisper model. \"Relative Speed\" is relative to large (1x baseline). VRAM figures are for PyTorch inference on GPU. The GGML columns show file sizes and RAM usage for whisper.cpp, which is what most desktop apps use.","headers":{"model":"Model","params":"Params","englishOnly":"English-Only","vram":"VRAM (GPU)","ggmlDisk":"GGML Disk","ram":"RAM (whisper.cpp)","speed":"Speed","englishWer":"English WER","multilingualWer":"Multilingual WER"},"notAvailable":"N/A","notesLead":"Notes:","notesBody":"WER (Word Error Rate) figures are approximate averages from the Whisper paper, HuggingFace model cards (LibriSpeech test sets for English), and OpenAI's published benchmarks. Exact WER varies significantly by dataset, language, audio quality, and evaluation methodology. Lower is better. Speed is relative to large (1x = slowest)."},"tiny":{"title":"Tiny (39M Parameters)","para1":"The tiny model is the smallest and fastest Whisper variant. At just 39 million parameters and a GGML file size of 75 MB, it can run comfortably on almost any hardware — including Raspberry Pi, older laptops, and low-end mobile devices. It transcribes at roughly 10x real-time speed relative to the large model, making it the best choice when latency matters more than perfect accuracy.","resources":{"l1":"GGML: 75 MiB on disk","l2":"RAM: ~273 MB runtime","l3":"VRAM: ~1 GB (GPU mode)"},"speed":{"l1":"~10x faster than large","l2":"Near-instant on Apple Silicon","l3":"Real-time on most CPUs"},"accuracy":{"l1":"English WER: ~7.6%","l2":"Multilingual WER: ~12%","l3":"Struggles with accents/noise"},"para2":"On the tiny.en variant (English-only), LibriSpeech test-clean WER is around 5.6%, rising to ~14.9% on test-other (noisy/accented audio). The multilingual tiny model performs slightly worse on English but supports all 99 Whisper languages.","para3":"Tiny is ideal for quick notes, low-stakes dictation, live captioning prototypes, and embedded or edge devices where compute is limited. It is the default starting model for many whisper.cpp-based tools because it downloads in seconds and runs instantly.","bestFor":"Fastest possible transcription, low-resource hardware, Raspberry Pi, real-time applications"},"base":{"title":"Base (74M Parameters)","para1":"The base model doubles the parameter count of tiny (74M vs 39M) while remaining very lightweight. At 142 MiB (GGML) and ~388 MB of RAM at runtime, it still fits comfortably on virtually any modern device. The accuracy improvement over tiny is noticeable — especially on noisy audio and accented speech.","resources":{"l1":"GGML: 142 MiB on disk","l2":"RAM: ~388 MB runtime","l3":"VRAM: ~1 GB (GPU mode)"},"speed":{"l1":"~7x faster than large","l2":"Still very fast on CPU","l3":"Sub-second on Apple Silicon"},"accuracy":{"l1":"English WER: ~5.0%","l2":"Multilingual WER: ~10%","l3":"Better on noisy audio than tiny"},"para2":"The base.en variant achieves around 4.3% WER on LibriSpeech test-clean and ~12.8% on test-other. This is a meaningful improvement over tiny.en (5.6% and 14.9% respectively), particularly on challenging audio.","para3":"Base is a good choice for low-power devices where tiny is not quite accurate enough. It is also popular for real-time transcription pipelines where you want a balance of speed and quality without going above 500 MB of RAM.","bestFor":"Low-power devices that need better accuracy than tiny, real-time streaming, budget hardware"},"small":{"title":"Small (244M Parameters)","para1":"Small is where Whisper starts to feel genuinely accurate. At 244M parameters, it represents a 3.3x jump from base and delivers a significant accuracy improvement — especially for non-English languages and noisy recordings. The GGML file is 466 MiB, and runtime RAM usage is around 852 MB. It runs at about 4x the speed of large.","resources":{"l1":"GGML: 466 MiB on disk","l2":"RAM: ~852 MB runtime","l3":"VRAM: ~2 GB (GPU mode)"},"speed":{"l1":"~4x faster than large","l2":"Fast on modern laptops","l3":"Real-time on Apple Silicon"},"accuracy":{"l1":"English WER: ~3.4%","l2":"Multilingual WER: ~7%","l3":"Handles accents well"},"para2":"The small.en variant reaches about 3.0% WER on LibriSpeech test-clean — a remarkable result for a model that fits in under 500 MB. For many English-only use cases, small.en provides accuracy that is close to medium at a fraction of the resource cost.","para3":"Small is often recommended as the default model for most users. It runs well on any modern laptop (including MacBook Air M1 with 8 GB RAM), delivers good accuracy across languages, and transcribes fast enough for real-time dictation workflows.","recommendationLead":"Our recommendation:","recommendationBody":"If you are not sure which model to start with, start with small. It is the best balance of speed, accuracy, and resource usage for the majority of hardware and use cases.","bestFor":"Most users on modern laptops, default recommended model, good multilingual support"},"medium":{"title":"Medium (769M Parameters)","para1":"Medium is where diminishing returns begin to set in — but the accuracy gains over small are still meaningful, especially for multilingual transcription, technical vocabulary, and challenging audio conditions. At 769M parameters, it requires 1.5 GiB of disk space (GGML) and about 2.1 GB of RAM at runtime.","resources":{"l1":"GGML: 1.5 GiB on disk","l2":"RAM: ~2.1 GB runtime","l3":"VRAM: ~5 GB (GPU mode)"},"speed":{"l1":"~2x faster than large","l2":"Comfortable on 8 GB+ machines","l3":"Needs decent hardware"},"accuracy":{"l1":"English WER: ~2.9%","l2":"Multilingual WER: ~5%","l3":"Strong on technical content"},"para2":"The medium.en model achieves around 3.0% WER on LibriSpeech test-clean and approximately 7.5% on test-other. However, across broader benchmarks — particularly those with diverse accents, background noise, and domain-specific vocabulary — medium consistently outperforms small.","para3":"Medium is a strong choice for professional transcription workflows where accuracy matters but you do not have the hardware (or patience) for large. It runs well on machines with 8 GB or more of RAM and benefits significantly from GPU acceleration.","bestFor":"Professional use, multilingual transcription, technical content, machines with 8 GB+ RAM"},"large":{"title":"Large (1,550M Parameters)","para1":{"part1":"The large model is the most accurate Whisper variant, with 1.55 billion parameters. There are three versions:","part2":"(the original release),","part3":"(trained for 2.5x more epochs with added regularization), and","part4":"(trained on more data with 128 Mel frequency bins instead of 80). Large-v3 is the current recommended version for maximum accuracy."},"resources":{"l1":"GGML: 2.9 GiB on disk","l2":"RAM: ~3.9 GB runtime","l3":"VRAM: ~10 GB (GPU mode)"},"speed":{"l1":"1x (baseline — slowest)","l2":"GPU strongly recommended","l3":"CPU: may be slower than real-time"},"accuracy":{"l1":"English WER: ~2.4% (v3)","l2":"Multilingual WER: ~3.5% (v3)","l3":"Best across all languages"},"versionsTitle":"large-v2 vs large-v3","v2":{"l1":"Trained for 2.5x more epochs than large-v1","l2":"Added regularization for better generalization","l3":"LibriSpeech test-clean WER: ~3.0%","l4":"More stable on some audio types"},"v3":{"l1":"128 Mel bins (vs 80) — finer frequency resolution","l2":"Trained on 1M hours labeled + 4M hours pseudo-labeled audio","l3":"10-20% error reduction over large-v2 across languages","l4":"Added Cantonese language support"},"whichLead":"Which large model to use:","whichBody":"For new projects, use large-v3. It is strictly better than large-v2 on average across languages. However, some users have reported that large-v2 produces fewer hallucinations (generating text not present in the audio) in specific edge cases with very quiet or silent audio segments. If you encounter hallucination issues with large-v3, try large-v2 as a fallback.","bestFor":"Maximum accuracy, professional/medical/legal transcription, low-resource languages, batch processing"},"turbo":{"title":"Turbo (809M Parameters)","para1":"The turbo model is a distilled version of large-v3 that dramatically reduces inference time while retaining most of the accuracy. OpenAI achieved this by pruning the decoder from 32 layers down to just 4 layers — cutting the parameter count from 1,550M to 809M. The encoder (which does the heavy lifting) remains identical to large-v3.","resources":{"l1":"GGML: ~1.6 GiB on disk","l2":"RAM: ~2.3 GB runtime","l3":"VRAM: ~6 GB (GPU mode)"},"speed":{"l1":"~8x faster than large","l2":"Near-large accuracy, near-tiny speed","l3":"Great on GPU with torch.compile"},"accuracy":{"l1":"English WER: ~2.5%","l2":"Multilingual WER: ~3.7%","l3":"Minor quality loss vs large-v3"},"para2":"Turbo is the best choice when you want near-large-v3 accuracy but cannot afford the latency of the full large model. On GPU with optimizations like torch.compile, turbo can achieve up to 4.5x speed improvement over standard inference, making it extremely fast.","limitationLead":"Important limitation:","limitationBody":"The turbo model was not trained for translation tasks. If you need to translate speech from one language to English, use medium or large-v3 instead.","bestFor":"High-accuracy real-time transcription, latency-sensitive apps, when large-v3 is too slow"},"englishVsMultilingual":{"title":"English-Only vs Multilingual Models","para1":{"part1":"For the tiny, base, small, and medium sizes, Whisper offers two variants: a multilingual model and an English-only (","part2":") model. The large and turbo models are multilingual only — there are no English-only versions."},"enTitle":"When to use .en (English-only)","en":{"l1":"You only transcribe English audio","l2":"Better English accuracy at tiny/base size levels","l3":"Slightly faster — no language detection overhead","l4":"Same file size as multilingual equivalent"},"multiTitle":"When to use multilingual","multi":{"l1":"You transcribe non-English audio","l2":"Your audio contains mixed languages","l3":"You need speech translation (to English)","l4":"Required if using large or turbo (no .en variant exists)"},"para2":"The performance gap between .en and multilingual models is most significant at smaller sizes. For tiny and base, the .en variants are noticeably better at English. By the time you reach small and medium, the difference narrows considerably. At the large level, there is only a multilingual model — and it performs excellently on English regardless.","ruleLead":"Rule of thumb:","ruleBody":"If you exclusively use English and are choosing tiny or base, pick the .en variant. For small or larger, the multilingual version works great for English and gives you flexibility for other languages."},"hardware":{"title":"Hardware Requirements: What Runs Where","intro":"The hardware you need depends heavily on the model size and whether you are using whisper.cpp (CPU/Metal/CUDA) or the PyTorch implementation (GPU-focused). Here is what to expect on common hardware classes.","m1Air":{"title":"MacBook Air M1 (8 GB RAM)","r1Label":"tiny/base:","r1":"Instant. Faster than real-time.","r2Label":"small:","r2":"Fast. Real-time or better with Metal.","r3Label":"medium:","r3":"Usable. May be slightly slower than real-time on CPU. Metal helps significantly.","r4Label":"large:","r4":"Possible but tight on RAM. Expect slower than real-time.","r5Label":"turbo:","r5":"Good fit. Near real-time with Metal acceleration."},"mbpPro":{"title":"MacBook Pro M2/M3 (16-36 GB RAM)","r1Label":"tiny/base/small:","r1":"Instant. All well within hardware limits.","r2Label":"medium:","r2":"Fast. Comfortable with Metal.","r3Label":"large:","r3":"Good. Real-time or near it with Metal.","r4Label":"turbo:","r4":"Excellent. Fast real-time transcription."},"midRange":{"title":"Mid-Range Windows/Linux Laptop (8 GB RAM, integrated GPU)","r1Label":"tiny/base:","r1":"Fast. CPU-only is fine.","r2Label":"small:","r2":"Good. Comfortable on most modern CPUs.","r3Label":"medium:","r3":"Workable. May be 2-3x slower than real-time on CPU.","r4Label":"large:","r4":"Challenging. CPU-only will be slow.","r5Label":"turbo:","r5":"Workable. Benefits from Vulkan if available."},"nvidia":{"title":"Desktop with NVIDIA GPU (RTX 3060+, 8 GB+ VRAM)","r1Label":"All models:","r1":"Fast. GPU handles everything comfortably.","r2Label":"large:","r2":"Needs 10 GB VRAM (RTX 3060 12GB or better).","r3Label":"turbo:","r3":"Fastest option with CUDA. Significantly faster than large.","r4":"Use CUDA backend in whisper.cpp for best performance."},"backendsTitle":"whisper.cpp Acceleration Backends","backendsIntro":"whisper.cpp — the C/C++ port that most desktop apps use — supports several hardware acceleration backends that can dramatically improve performance:","metal":{"title":"Metal (Apple Silicon)","body":"Full GPU inference on M1/M2/M3/M4 Macs. Runs the entire model on the GPU with zero memory copies. This is the fastest backend for macOS users and what OpenWhispr uses on Mac."},"coreml":{"title":"Core ML (Apple)","body":"Uses Apple's neural engine for inference. Can be faster than Metal for some model sizes on newer chips, and more power-efficient. Requires converting models to Core ML format first."},"cuda":{"title":"CUDA (NVIDIA)","body":"GPU acceleration for NVIDIA cards. Fastest backend for users with dedicated NVIDIA GPUs. Requires CUDA toolkit. Best choice for large and turbo models on desktop."},"vulkan":{"title":"Vulkan (Cross-vendor GPU)","body":"Works with Intel, AMD, and NVIDIA GPUs. Good fallback for non-NVIDIA systems. Available on Linux and Windows. Performance varies by GPU vendor and driver."},"speedTitle":"Approximate Transcription Speed","speedIntro":"Real-time factor (RTF) for a 60-second audio clip. RTF < 1.0 means faster than real-time. These are rough estimates — actual performance depends on audio content, model quantization, and system load.","speedTable":{"model":"Model","m1":"M1 MacBook Air","m3":"M3 Pro MacBook","intel":"Intel i7 (CPU)","rtx":"RTX 3060 (CUDA)"},"readingLead":"Reading the table:","readingBody":"0.1x means transcribing 60 seconds of audio takes about 6 seconds. 1.0x = real-time. Values above 1.0x mean slower than real-time. All figures use whisper.cpp with default settings and the Metal (Mac) or CUDA (NVIDIA) backend where applicable."},"recommendations":{"title":"Which Model Should You Use?","intro":"Here are concrete recommendations based on common scenarios. When in doubt, start with small and move up or down based on your experience.","fastest":{"q":"\"I want the fastest possible\"","part1":"Use","part2":"or","part3":". Transcribes in milliseconds on modern hardware. Accuracy is rough but usable for quick notes and low-stakes dictation."},"laptop":{"q":"\"I want good accuracy on a laptop\"","part1":"Use","part2":". Best balance of speed and accuracy for modern laptops. If your machine has 16 GB+ RAM,","part3":"is also an excellent choice.","tag":"small or medium"},"accuracy":{"q":"\"I want the best accuracy\"","part1":"Use","part2":". It is the most accurate Whisper model overall. If speed matters too,","part3":"gets you 95% of the accuracy at 8x the speed.","tag":"large-v3 or turbo"},"englishOnly":{"q":"\"I only use English\"","part1":"Use the","part2":"variant of your chosen size for the best English accuracy. For tiny and base, the .en models are noticeably better. For small+, the difference is minimal.","tag":"small.en or medium.en"},"multiLang":{"q":"\"I use multiple languages\"","part1":"Use the multilingual variant.","part2":"or","part3":"for balanced performance.","part4":"for best multilingual accuracy.","tag":"small, medium, or large-v3"},"limited":{"q":"\"I have very limited hardware\"","part1":"Use","part2":"or","part3":". Both run on 1 GB of RAM and even work on Raspberry Pi and other single-board computers. Base offers a meaningful accuracy bump over tiny with minimal added cost.","tag":"tiny or base"}},"openwhispr":{"title":"How OpenWhispr Handles Model Selection","intro":"is an open-source desktop dictation app that uses whisper.cpp under the hood. It lets you download and switch between any Whisper model directly from the settings — no command-line required. Models are downloaded once and stored locally. You can change models at any time to balance speed and accuracy for your hardware.","c1":{"title":"Download any model","body":"Pick from tiny through large-v3 in settings. OpenWhispr downloads the GGML version automatically."},"c2":{"title":"Switch instantly","body":"Change models at any time. No restart needed. Test different sizes to find what works best on your hardware."},"c3":{"title":"Hardware optimized","body":"Uses Metal on Apple Silicon, CUDA on NVIDIA, and optimized CPU inference elsewhere. All local, all private."},"suggestionLead":"Our suggestion:","suggestionPart1":"Start with","suggestionPart2":". If it feels slow, drop down to base or tiny. If you want better accuracy and your hardware can handle it, move up to medium or large-v3. OpenWhispr makes it easy to experiment."},"sources":{"title":"Sources","s1":{"link":"Radford et al. \"Robust Speech Recognition via Large-Scale Weak Supervision\" (2022)","desc":"— The original Whisper paper with full benchmark results."},"s2":{"link":"OpenAI Whisper GitHub Repository","desc":"— Official model table, parameter counts, VRAM requirements, and relative speed figures."},"s3":{"link":"OpenAI Whisper Model Card","desc":"— Official training data composition and evaluation caveats."},"s4":{"link":"whisper.cpp GitHub Repository","desc":"— GGML model sizes, RAM usage, and hardware acceleration details."},"s5":{"link":"Whisper large-v2 Model Card (HuggingFace)","desc":"— LibriSpeech benchmark context used in many WER comparisons."},"s6":{"link":"Whisper large-v3 Model Card (HuggingFace)","desc":"— Training details, architecture changes from large-v2, and multilingual improvements."},"s7":{"link":"Whisper large-v3-turbo Model Card (HuggingFace)","desc":"— Turbo architecture details, decoder pruning specifics, and performance tradeoffs."},"s8":{"link":"OpenAI Whisper Turbo Announcement (GitHub Discussion #2363)","desc":"— Official turbo release notes and cross-language performance comparisons."},"s9":{"link":"OpenWhispr","desc":"— Practical local deployment context for choosing model sizes."}},"cta":{"title":"Try Different Whisper Models in OpenWhispr","body":"Open source, local-first dictation. Download any Whisper model, switch between sizes with one click, and find the perfect balance for your hardware.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":{"howWhisperWorks":"How Whisper AI works","linux":"Dictation for Linux","allPosts":"All blog posts","developers":"OpenWhispr for developers","localVsCloud":"Local vs Cloud Transcription"}},"blogOpenSource":{"metadata":{"title":"Why Open Source Matters for Voice Tools | OpenWhispr","description":"Voice tools process some of the most personal data imaginable — your actual voice. Here's why open source is the only approach that earns real trust.","keywords":["open source voice recognition","open source dictation","open source speech to text","why open source matters","voice privacy open source","open source AI tools","transparent voice software","open source whisper","nvidia parakeet"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"Why Open Source Matters"},"schema":{"title":"Why Open Source Matters for Voice Tools","description":"Voice tools process some of the most personal data imaginable. Here's why open source is the only approach that earns real trust."},"layout":{"title":"Why Open Source Matters for Voice Tools","subtitle":"Your voice is biometric data. The software that processes it should be auditable.","date":"February 10, 2026","tag":"Blog"},"sections":[{"id":"fact-check-snapshot","title":"Fact-Check Snapshot"},{"id":"voice-data-is-different","title":"Voice Data Is Different"},{"id":"the-trust-problem","title":"The Trust Problem"},{"id":"what-open-source-gives-you","title":"What Open Source Gives You"},{"id":"the-open-model-effect","title":"The Open Model Effect"},{"id":"open-source-polish","title":"Open Source Polish"},{"id":"business-model","title":"Business Model"},{"id":"what-to-look-for","title":"What to Look For"},{"id":"sources","title":"Sources"}],"opening":{"lead":"Open source voice tools let you verify exactly what happens to your audio.","body":"With proprietary voice software, you are trusting a company's privacy policy. With open source, you can read the code, build it yourself, and confirm that your voice data never leaves your device. This distinction matters more for voice tools than almost any other category of software, because voice data is uniquely personal — it is biometric, it is hard to anonymize, and once it is recorded, it cannot be unheard.","lastUpdated":"Last updated: February 17, 2026. Incident and policy claims are cross-checked against at least two primary sources."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","voiceData":{"lead":"Voice data can be identifying and high-risk:","body":"legal and policy frameworks increasingly treat biometric processing as sensitive.","source1":"GDPR official text","source2":"EU AI Act official text"},"assistants":{"lead":"Major assistants have faced voice-data controversies:","body":"Amazon, Google, and Apple incidents are documented by regulators and/or major news outlets.","source1":"FTC Amazon case","source2":"Reuters Siri settlement"},"cloning":{"lead":"Voice-cloning abuse is an active consumer risk:","body":"US regulators and cybersecurity guidance both flag impersonation risk.","source1":"FTC alert","source2":"IBM risk context"},"openModel":{"lead":"Open model ecosystem is real:","body":"Whisper and downstream runtimes enable auditable local deployments.","source1":"OpenAI Whisper","source2":"whisper.cpp"},"positioning":{"lead":"OpenWhispr positioning:","body":"OpenWhispr is built as an open, local-first voice workflow on top of open ASR.","source1":"OpenWhispr","source2":"Whisper upstream"}},"voiceData":{"title":"Voice Data Is Different","paragraph1":"Text is text. You can strip metadata, anonymize names, and aggregate it into statistics. Voice is none of those things.","paragraph2":"Your voice is a biometric identifier. Like a fingerprint, it is unique to you — voice authentication systems used by banks and government agencies depend on this fact. But that same uniqueness makes voice recordings inherently identifying. A leaked database of text transcriptions is a privacy incident. A leaked database of voice recordings is a biometric breach.","reveal":{"title":"What voice recordings can reveal","items":["Identity — voice biometrics can uniquely identify a speaker","Health conditions — vocal biomarkers can indicate Parkinson's, depression, respiratory illness","Emotional state — stress, anger, fatigue are detectable in vocal patterns","Demographics — age, gender, accent, native language, regional origin"]},"anonymize":{"title":"Why anonymization is harder","items":["You can redact words from a transcript, but you cannot redact a voice from a recording without destroying it","Voice cloning technology means leaked audio can be used to generate deepfakes of your voice","The EU AI Act classifies biometric identification systems — including voice — as high-risk AI, subject to strict compliance requirements"]},"paragraph3a":"Voice cloning has gone from a research curiosity to a commodity service. With just a few seconds of audio, modern voice synthesis models can produce convincing replicas of a person's voice. This is not a hypothetical risk — the","ftcLink":"FTC has warned about AI-powered voice cloning scams","paragraph3b":"since 2023. When the software processing your voice is a black box, you have no way to know whether recordings are stored, transmitted, or used for model training."},"trustProblem":{"title":"The Trust Problem with Proprietary Voice Tools","intro":"The argument for open source is not theoretical. Every major voice assistant has been caught handling audio in ways their users did not expect.","alexa":{"title":"Amazon Alexa","body1":"In April 2019, Bloomberg","reportedLink":"reported","body2":"that Amazon employed thousands of workers worldwide to listen to Alexa voice recordings captured in customers' homes and offices. Workers transcribed, annotated, and reviewed audio clips to improve the speech recognition system. Some reviewers reported hearing recordings they found distressing, including what sounded like a sexual assault. Amazon's response acknowledged the practice but emphasized it applied to \"an extremely small sample.\" The company later added opt-out controls, but the recordings had already been collected under a privacy policy most users never read closely. In 2023, Amazon","settlementLink":"paid $25 million","body3":"to settle FTC charges that it violated children's privacy by retaining kids' Alexa voice recordings indefinitely."},"google":{"title":"Google Assistant","body1":"In July 2019, Belgian broadcaster VRT NWS","obtainedLink":"obtained over 1,000 Google Assistant recordings","body2":"from a contractor. The recordings included bedroom conversations, business calls, and interactions with children — many captured by accidental activations when no one had said \"OK Google.\" Google confirmed that human reviewers listened to about 0.2% of all voice recordings and suspended the practice in Europe after the leak. A subsequent investigation by Hamburg's data protection authority led to a temporary EU-wide ban on the practice."},"siri":{"title":"Apple Siri","body1":"In July 2019, a whistleblower revealed to The Guardian that Apple contractors","heardLink":"regularly heard confidential medical information, drug deals, and sexual encounters","body2":"while grading Siri recordings. Apple had not disclosed the human review program in its privacy documentation. The company apologized, suspended the program globally, and made it opt-in — but the recordings that had already been reviewed were collected without informed consent. Apple later settled a","lawsuitLink":"$$95 million class-action lawsuit","body3":"over Siri privacy violations in January 2025."},"pattern":"The pattern is consistent: a company says \"we take your privacy seriously,\" while the actual data handling practices are only discovered through leaks, whistleblowers, or regulatory action. These were not small companies cutting corners. They were Apple, Google, and Amazon — three of the most technically sophisticated and well-resourced organizations on the planet.","quote":"\"We don't store your data\" is a claim that requires trust. Open source is a claim that requires only reading comprehension."},"whatOpenSource":{"title":"What Open Source Actually Gives You","intro":"Open source is not a marketing label. It is a set of concrete properties that change the trust model between you and the software you use.","properties":[{"title":"Auditability","body":"Anyone can read the source code and verify exactly how audio data is handled. Does the app send recordings to a server? Does it store audio on disk? Does it phone home with telemetry? You do not need to trust a privacy policy — you can check."},{"title":"Reproducibility","body":"You can build the application from source and compare it to the distributed binary. This closes the gap between \"we published the code\" and \"the app you downloaded actually runs this code.\" Reproducible builds are the gold standard of software trust."},{"title":"Community Review","body":"Security researchers, privacy advocates, and experienced developers can independently review the code. Vulnerabilities are found and patched in the open. This is not a theoretical benefit — it is the reason Linux, OpenSSL, and the entire internet infrastructure run on open source."},{"title":"Fork Rights","body":"If the maintainers make decisions you disagree with — adding telemetry, removing features, selling to an acquirer — the community can fork the project and continue development independently. This is not just insurance; it is a structural incentive for maintainers to act in users' interests."},{"title":"No Lock-in","body":"Your workflows are not hostage to a company's business decisions. If a proprietary dictation tool gets acquired, pivots, or shuts down, your investment in learning and configuring it is lost. Open source tools persist as long as anyone cares enough to run them."},{"title":"Longevity","body":"Companies disappear. Open source projects endure. Apache HTTP Server has been running since 1995. PostgreSQL since 1996. GCC since 1987. Dragon NaturallySpeaking was the leading dictation product for decades — then Nuance was acquired by Microsoft, and the standalone product was effectively discontinued. Open source does not have this failure mode."}]},"openModel":{"title":"The Open Model Effect: From Whisper to Parakeet","paragraph1a":"In September 2022, OpenAI","releasedLink":"released Whisper","paragraph1b":"— a general-purpose speech recognition model trained on 680,000 hours of multilingual audio — and open-sourced it under the MIT license. It was a watershed moment for voice tools.","paragraph2":"Before Whisper, high-quality automatic speech recognition (ASR) was expensive and proprietary. Google Cloud Speech-to-Text, Amazon Transcribe, and Microsoft Azure Speech Services all charged per minute of audio and required sending recordings to their servers. The best offline options — CMU Sphinx, Kaldi, VOSK — were functional but noticeably less accurate than the cloud APIs.","paragraph3":"OpenAI Whisper changed that equation overnight. For the first time, a model that matched or exceeded commercial API accuracy was available for anyone to download and run locally, at no cost, with no data leaving their machine. NVIDIA followed a similar path with its Parakeet family of ASR models — achieving state-of-the-art English accuracy and releasing them under open licenses. The trend is clear: the best speech recognition models are increasingly open source.","ecosystem":{"title":"The ecosystem that followed","items":[{"name":"whisper.cpp","desc":"— Georgi Gerganov's C/C++ port, optimized for CPU inference. Runs Whisper on everything from laptops to Raspberry Pis. Over 46,000 GitHub stars."},{"name":"faster-whisper","desc":"— CTranslate2-based reimplementation that achieves up to 4x faster inference than the original with the same accuracy."},{"name":"WhisperX","desc":"— Adds forced alignment and speaker diarization to Whisper, enabling word-level timestamps and multi-speaker transcription."},{"name":"Apps built on open models","desc":"— Dozens of desktop, mobile, and web applications now use OpenAI Whisper or NVIDIA Parakeet as their speech recognition engine, including OpenWhispr, MacWhisper, Buzz, and Vibe."}]},"paragraph4":"The release of OpenAI Whisper is one of the clearest demonstrations of how open source creates ecosystems. A single open model spawned optimized runtimes, new capabilities, and dozens of products that would never have existed if the model remained behind an API paywall. NVIDIA Parakeet continued this pattern, proving that multiple organizations see value in open-sourcing high-quality ASR. Today, any developer can build a voice tool with state-of-the-art accuracy without paying per minute, without sending audio to a third party, and without asking permission."},"polish":{"title":"Open Source Does Not Mean Less Polished","intro":"There is a persistent assumption that open source means rougher edges. That you trade polish for freedom. This was arguably true in 2005. It is not true in 2026.","apps":[{"name":"Firefox","desc":"A mainstream browser used by hundreds of millions"},{"name":"VS Code","desc":"The most popular code editor in the world"},{"name":"Signal","desc":"End-to-end encrypted messaging with a polished UX"},{"name":"Blender","desc":"Oscar-winning 3D animation and VFX tool"},{"name":"VLC","desc":"Plays everything, runs everywhere, no ads"},{"name":"Linux","desc":"Runs the majority of the world's servers and smartphones"}],"paragraph1":"The quality of open source software has improved dramatically because the incentive structures have changed. Companies now build businesses on open source cores (Red Hat, Elastic, GitLab). Well-funded teams maintain projects full-time. Community contributions catch bugs and add features that small teams would never have bandwidth for.","paragraph2":"To be fair about the tradeoffs: open source voice tools sometimes have smaller teams and slower release cycles than well-funded proprietary competitors. Documentation can be uneven. But these are practical challenges, not inherent limitations. And the gap has been closing quickly — especially in the voice tool space, where open models like OpenAI Whisper and NVIDIA Parakeet give open source projects access to the same foundational model quality as any commercial product."},"businessModel":{"title":"How Open Source Voice Tools Sustain Themselves","intro":"A reasonable question: if the software is free, how does anyone get paid to work on it?","paragraph1a":"The answer is that \"open source\" and \"free\" are not synonyms. The most sustainable open source projects separate the core engine (free, open, auditable) from convenience features that justify a paid tier. This is called the","openCoreModel":"open core model","paragraph1b":", and it works because it aligns incentives: the company earns revenue by making the product better, not by locking users in or monetizing their data.","openCore":{"title":"Why open core works well for privacy-focused tools","points":["The core product works locally and offline — no data collection required for the free tier","Cloud features (like managed transcription APIs) provide genuine convenience worth paying for","Community contributions improve the product for everyone, including paying customers","Users who cannot pay still benefit, and they contribute bug reports, translations, and code"]},"paragraph2":"This is the model OpenWhispr uses: the open source desktop app handles transcription locally using models like OpenAI Whisper and NVIDIA Parakeet at no cost. An optional Pro plan adds cloud-powered transcription and AI text cleanup for users who want faster results or do not want to run models on their own hardware. The important part is that the choice is yours — the local, private, fully functional version is always free."},"whatToLookFor":{"title":"What to Look For in an Open Source Voice Tool","intro":"Not all \"open source\" claims are equal. Here is a practical checklist for evaluating voice tools.","checklist":[{"question":"Is the full source code available?","detail":"Some projects open-source a library but keep the application proprietary. Check that the actual product you use — the desktop app, the mobile app — has its source published."},{"question":"Can you build from source?","detail":"Published source code that does not compile is not meaningfully open. Look for build instructions, CI pipelines, and community reports of successful builds."},{"question":"What is the license? (MIT, Apache, GPL, AGPL?)","detail":"Permissive licenses (MIT, Apache 2.0) allow maximum flexibility. Copyleft licenses (GPL, AGPL) ensure derivatives stay open. Both are legitimate — just know what you are getting."},{"question":"Does it work offline and locally?","detail":"Open source code that requires a cloud connection to function still sends your data to a server. For voice tools, offline-capable local processing is the privacy baseline."},{"question":"Is it actively maintained?","detail":"Check the commit history, issue tracker, and release cadence. An abandoned open source project may have unpatched security vulnerabilities."},{"question":"Is there a community? (Issues, PRs, discussions)","detail":"A healthy open source project has more than one contributor. Look for community engagement — pull requests from external contributors, issue discussions, and responsive maintainers."}]},"sources":{"title":"Sources and Further Reading","bloomberg":"Bloomberg: \"Is Anyone Listening to You on Alexa? A Global Team Reviews Audio\" (April 2019)","vrt":"VRT NWS: \"Google employees are eavesdropping, even in Flemish living rooms\" (July 2019)","guardian":"The Guardian: \"Apple contractors 'regularly hear confidential details' on Siri recordings\" (July 2019)","reuters":"Reuters: \"Apple agrees to pay $95 million to settle Siri privacy lawsuit\" (January 2025)","ftcAmazon":"FTC: Amazon charged with violating children's privacy law by keeping kids' Alexa recordings (May 2023)","ftcCloning":"FTC: Scammers use AI voice cloning to enhance family emergency schemes (March 2023)","openaiWhisper":"OpenAI: Introducing Whisper (September 2022)","whisperGithub":"OpenAI Whisper GitHub repository (models, license, docs)","radford":"Radford et al. (2022): Robust Speech Recognition via Large-Scale Weak Supervision","gdpr":"GDPR official text (EU 2016/679)","aiAct":"EU AI Act official text (EU 2024/1689)","whisperCpp":"whisper.cpp — C/C++ port of OpenAI Whisper","fasterWhisper":"faster-whisper — CTranslate2-based Whisper implementation","openwhispr":"OpenWhispr — open, local-first dictation product context"},"cta":{"title":"OpenWhispr Is Open Source","body":"We built OpenWhispr in the open because we believe voice software should be transparent. Don't take our word for it — check the code yourself.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · MIT-licensed · Open source forever"},"internalLinks":{"localAi":"The case for local AI","localVsCloud":"Local vs cloud transcription","allPosts":"All blog posts","compare":"Compare OpenWhispr","useCases":"Use cases"}},"blogGoogleDocsVoiceTyping":{"metadata":{"title":"Google Docs Voice Typing: The Complete Guide | OpenWhispr","description":"How to turn on voice typing in Google Docs, every voice command, why it stops working, and how to dictate in any app, even offline.","keywords":["google docs voice typing","voice typing google docs","voice to text google docs","google docs dictation","turn on voice typing google docs","google docs voice commands","google docs voice typing not working","google docs voice typing languages"]},"schema":{"title":"Google Docs Voice Typing: The Complete Guide","description":"How to turn on voice typing in Google Docs, every voice command, why it stops working, and how to dictate in any app, even offline."},"breadcrumb":{"home":"Home","blog":"Blog","article":"Google Docs Voice Typing"},"layout":{"title":"Google Docs Voice Typing: The Complete Guide","subtitle":"How to turn on voice typing in Google Docs, the main voice commands, why it sometimes stops working, and how to dictate everywhere else on your computer.","date":"June 26, 2026","tag":"Guide"},"sections":[{"id":"quick","title":"Quick Answer"},{"id":"enable","title":"How to Turn It On"},{"id":"commands","title":"Voice Commands"},{"id":"languages","title":"Languages"},{"id":"mobile","title":"On Phone and Tablet"},{"id":"troubleshooting","title":"Not Working?"},{"id":"limits","title":"The Limitations"},{"id":"alternative","title":"Voice Typing Anywhere"},{"id":"faq","title":"FAQ"}],"intro":{"lead":"To turn on voice typing in Google Docs, open a document in Chrome, go to Tools > Voice typing, or press Ctrl+Shift+S on Windows or Cmd+Shift+S on a Mac, then click the microphone and start speaking.","body":"It is free, built in, and useful for a quick first draft. The trade-off is that it only runs in Chrome on desktop, needs an internet connection, and only types inside Google Docs.","summary":"Below you will find setup steps, the command categories that matter, supported language notes, troubleshooting fixes, privacy limitations, and a practical option when you need dictation outside Docs."},"enable":{"title":"How to Turn On Voice Typing in Google Docs","intro":"Voice typing is part of Google Docs on desktop and is designed for Google Chrome. Here is the full process:","steps":["Open a document at docs.google.com in Chrome.","In the menu bar, click Tools > Voice typing. Shortcut: Ctrl+Shift+S on Windows, Cmd+Shift+S on Mac.","A microphone box appears on the left side of the page. Click it to start.","The first time, your browser asks for microphone permission. Click Allow.","Speak naturally. Your words appear at the cursor as you talk.","Click the microphone again when you want to stop."],"tipLead":"Tip:","tipBody":"Voice typing also works in the speaker notes field of Google Slides with the same Tools menu and microphone control."},"commands":{"title":"Voice Commands: Punctuation, Formatting, and Editing","intro":"You can speak punctuation, line breaks, and editing commands. The important catch: formatting and editing commands only work when the document language is set to English.","groups":[{"title":"Punctuation","items":["\"period\"","\"comma\"","\"question mark\"","\"exclamation point\"","\"new line\" / \"new paragraph\""]},{"title":"Formatting","items":["\"bold\" / \"italics\"","\"underline\"","\"apply heading 1\"","\"create bulleted list\"","\"align center\""]},{"title":"Editing and moving","items":["\"select [word]\"","\"delete last word\"","\"go to end of paragraph\"","\"next paragraph\"","\"stop listening\""]}]},"languages":{"title":"Which Languages Are Supported?","paragraph1":"Voice typing handles dictation in about a hundred languages and dialects. Pick yours from the dropdown on the microphone box. Basic punctuation works broadly across languages.","paragraph2":"Spoken formatting and editing commands are the exception. Commands like bold, select paragraph, or apply heading work only in English, so non-English dictation usually means formatting with the keyboard or mouse."},"mobile":{"title":"Voice Typing on Phone and Tablet","intro":"The Google Docs mobile app does not include the desktop Voice typing tool. On phones and tablets, dictate the same way you would in any app: tap the microphone on your keyboard.","items":["Android: tap the microphone on Gboard or your preferred keyboard while the cursor is in the document.","iPhone and iPad: tap the dictation key on the iOS keyboard."],"outro":"That is your device dictation, not the Google Docs desktop feature, so accuracy, commands, and behavior will not match the Chrome version."},"troubleshooting":{"title":"Voice Typing Not Working? Try This","items":[{"problem":"The Voice typing button is missing or grayed out","fix":"Use Google Chrome on desktop. Voice typing is not available in the mobile app, and other browsers are not officially supported."},{"problem":"It will not pick up my voice","fix":"Check that docs.google.com has microphone permission in Chrome, and make sure another app is not holding the microphone."},{"problem":"It keeps stopping on its own","fix":"Voice typing can stop after a pause or when you click away from the tab. Click the microphone again and keep the Docs tab focused."},{"problem":"Nothing happens at all","fix":"You may be offline. Voice typing is cloud-based and needs an internet connection. Reconnect and reload the document."}]},"limits":{"title":"What Google Docs Voice Typing Cannot Do","intro":"It is a genuinely useful free feature, but it is built for one job: dictating inside a Google Doc. These limits matter if you dictate often:","items":[{"title":"Chrome only.","body":"No reliable support in Safari, Firefox, or Edge."},{"title":"Online only.","body":"Audio is processed in Google's cloud, so no connection means no dictation."},{"title":"Docs only.","body":"It will not type into Gmail, Slack, your browser bar, a code editor, or any other app."},{"title":"Cloud privacy.","body":"Your voice is sent to Google for transcription, which is not ideal for sensitive material."}]},"alternative":{"title":"Want Voice Typing in Every App? Use OpenWhispr","introBefore":"If Docs is the only place you dictate, voice typing does the job. But when you want to talk into an email, a Slack message, a code comment, a form, or your notes app,","introAfter":"covers what Google Docs does not:","points":["One push-to-talk hotkey types into any app or text field, system-wide","Runs 100% on your device with OpenAI Whisper and NVIDIA Parakeet, fully offline","Your audio never leaves your computer","Works in any browser and in desktop apps, not just Chrome","Open source, so anyone can inspect what it does","Agent mode can clean up and reformat what you dictated"],"outro":"Think of it as Google Docs voice typing for your whole computer, without the cloud dependency.","viewOnGitHub":"View on GitHub"},"faqTitle":"Frequently Asked Questions","faqs":[{"question":"How do I turn on voice typing in Google Docs?","answer":"Open a document in Chrome, then go to Tools > Voice typing, or press Ctrl+Shift+S on Windows or Cmd+Shift+S on Mac. Click the microphone box, allow microphone access when prompted, and start speaking."},{"question":"Why is my Google Docs voice typing not working?","answer":"The usual causes are using a browser other than Chrome, missing microphone permission, another app holding the microphone, a paused or unfocused Docs tab, or no internet connection."},{"question":"Does Google Docs voice typing work offline?","answer":"No. Google Docs voice typing is cloud-based and needs an internet connection. If you need offline dictation, use a local tool such as OpenWhispr."},{"question":"What languages does Google Docs voice typing support?","answer":"It supports dictation in about a hundred languages and dialects, but spoken formatting and editing commands work only when the document language is English."},{"question":"Does Google Docs voice typing work on iPhone or Android?","answer":"Not as the desktop Voice typing tool. In the mobile app, use your phone keyboard microphone instead, such as Gboard on Android or the iOS dictation key."},{"question":"Is Google Docs voice typing private?","answer":"Audio is sent to Google for transcription, so it is convenient but not local-only. For confidential notes, legal work, patient information, or unreleased business content, local dictation is a better fit."},{"question":"Can I use voice typing outside of Google Docs?","answer":"Google's built-in voice typing only works inside Google Docs and speaker notes in Google Slides. OpenWhispr provides a system-wide hotkey that dictates into any app."}],"links":{"items":[{"href":"/use-cases","label":"All use cases"},{"href":"/compare/apple-dictation","label":"vs Apple Dictation"},{"href":"/pricing","label":"Pricing"}]}},"useCaseMedical":{"breadcrumb":{"home":"Home","useCases":"Use Cases","medical":"For Clinicians"},"metadata":{"title":"Free Medical Dictation Software For Clinicians - OpenWhispr","description":"OpenWhispr is free, open-source medical dictation software. Dictate the visit, get the SOAP note drafted, and keep patient audio on your machine.","keywords":["medical dictation software","clinical dictation software","physician dictation","HIPAA compliant dictation software","dictation software for doctors","SOAP note dictation","offline medical transcription"]},"schema":{"name":"OpenWhispr","applicationCategory":"Medical dictation software","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source medical dictation software. Patient audio is transcribed on the clinician's own device and never leaves it. Free for unlimited local dictation.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Medical dictation software for doctors & clinicians","title":"Dictate the visit once. Get the note drafted.","body":"OpenWhispr drafts the SOAP note, the referral letter or the discharge summary from what you dictate, on your own computer. Free for every clinician, and it keeps working on a locked-down hospital network.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Clinician walking through a sunlit hospital corridor"}},"trustBar":{"label":"Dictate into:","names":["Epic","Oracle Health","athenahealth","eClinicalWorks","NextGen","Meditech"]},"pain":{"label":"Cloud or local. Choose for the visit.","title":"Cloud for speed. Local for sensitive or offline work.","subtitle":"Cloud transcription provides top accuracy and speed without a local model download. Local models process entirely on-device and need no internet.","cards":[{"title":"The whole encounter, uploaded.","body":"An ambient scribe can send the whole encounter to the cloud, including things the patient asked you not to put in the chart. OpenWhispr processes what you dictate on your device."},{"title":"Pajama-time charting.","body":"A full clinic, then an hour of notes at the kitchen table. OpenWhispr drafts the note from your dictation while the visit is fresh, so you review and sign instead of type."},{"title":"One more vendor for compliance to vet.","body":"Each cloud scribe adds a BAA, a risk assessment and another data flow to map. In local mode, no transcription vendor sits in the data path."},{"title":"\"Trust us\" is not a compliance posture.","body":"OpenWhispr's code is public on GitHub. Give your security team the repository, then choose the cloud or local path that fits your review."}]},"features":{"label":"What you get","title":"Faster charting. Same confidentiality.","subtitle":"Medical dictation software often trades privacy for speed or bills per provider every month. OpenWhispr keeps the speed without sending patient audio away or charging per clinician for local mode.","items":[{"image":{"label":"Clinician speaking beside a laptop in a medical practice"},"title":"Your mic, or the telehealth call, transcribed on-device","body":"Dictate the visit into your mic, or record a telehealth consult straight from your speakers on macOS 14.2 or later, Windows and most Linux desktops. It's transcribed locally and never uploaded. You decide what's kept and when it's deleted."},{"image":{"label":"Clinician reviewing a printed document at a hospital workstation"},"title":"SOAP notes and letters, drafted for you","body":"Save your own SOAP, referral and discharge prompts as one-click actions, then run whichever one the visit needs. You review every word before it's signed, then paste it into the chart."},{"image":{"label":"Doctor speaking while working on a laptop in an examination room"},"title":"Dictate into the chart field you're already in","body":"Press a hotkey and dictate wherever the cursor already is: the HPI box, a reply to a portal message, a refill note in your inbox. Clean, punctuated text, processed locally."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Patient audio sent to us","value":"None","footnote":"in local mode, transcription never leaves your device"},{"label":"Cost per clinician","value":"$$0","footnote":"unlimited local dictation, the whole practice, forever"},{"label":"Local dictation is","value":"Free","footnote":"unlimited on every plan"}]},"howItWorks":{"label":"How it works","title":"The note drafted before the next patient is roomed.","subtitle":"Most medical dictation is a per-provider cloud subscription or an outsourced transcription service. OpenWhispr lives on your machine, with no account or credit card required.","image":{"label":"Doctor reaching for a closed laptop in an examination room"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model during setup."},{"title":"Dictate the visit.","body":"Speak your findings, plan and follow-up, and it's transcribed on your device."},{"title":"Review, sign, delete.","body":"Edit the draft note, paste it into the chart, and delete the audio. You control all of it."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons clinicians choose OpenWhispr","subtitle":"Choosing medical dictation software is a patient-confidentiality decision before it's anything else. Here are the facts, and your compliance officer can verify every one in the repo or at trust.openwhispr.com.","items":[{"title":"Audio never leaves your device","body":"Dictation, and note drafts with a local model, are processed on your machine. The open-source code shows there's no upload step."},{"title":"Open source, auditable by anyone","body":"Your IT or compliance team can read every line on GitHub. No faith required."},{"title":"Retention is yours, not a vendor's","body":"Audio and transcripts are files on your machine, not rows in someone else's retention schedule. Delete the audio the moment the note is signed."},{"title":"Optional Corti clinical transcription","body":"For specialist terminology, Corti Transcribe is available as a clinical-grade medical transcription provider. Choose a local model for note drafting that stays on-device."},{"title":"No bot joins the telehealth call","body":"Nothing appears in the participant list, so you never have to explain a third party to the patient before you can begin."},{"title":"Works fully offline","body":"Ward round, rural clinic, a locked-down hospital network. No signal required, so it never stalls mid-dictation."},{"title":"Dictate into any EHR","body":"A system-wide hotkey drops clean text wherever your cursor is: Epic, athenahealth, a referral letter, an email."},{"title":"Free for the whole practice","body":"Unlimited local dictation at $0, every clinician in the group, forever. No per-provider licence to renew, and you can test it on your own clinic list before anyone signs anything."},{"title":"SOC 2 Type 2 audit complete. HIPAA attestation. ISO 27001-aligned.","body":"Your security review can start at trust.openwhispr.com. Enterprise includes custom data processing agreements and compliance reviews."},{"title":"100+ languages","body":"Dictate the history in the language the visit actually happened in, transcribed locally."}]},"faq":{"label":"Before you download","title":"The questions clinicians ask first.","subtitle":"Straight answers on privacy, compliance and accuracy. If yours isn't here, ask us in Discord.","items":[{"question":"Does patient audio ever leave my device?","answer":"Not in local mode. Pick a local model during setup and transcription runs on your machine; the open-source code shows there is no upload step. Cloud transcription remains optional."},{"question":"Is it HIPAA compliant?","answer":"Compliance depends on your workflow, not the tool alone. OpenWhispr's SOC 2 Type 2 audit is complete, with the final report available under NDA once issued; it has a HIPAA attestation and an ISO 27001-aligned information-security programme. For PHI, local-only and bring-your-own-key modes route content around OpenWhispr Cloud. Cloud voice audio is processed in real time and not persisted after transcription. Enterprise includes custom data processing agreements and compliance reviews."},{"question":"Can it handle medical terminology?","answer":"Yes. It runs Whisper-class models trained on a wide range of accents and 100+ languages. The custom dictionary feeds the words that trip everything else up to both the transcriber and the AI cleanup: metoprolol succinate ER, cholecystectomy, the abbreviations your specialty writes and nobody else reads."},{"question":"How is this different from an ambient AI scribe?","answer":"An ambient scribe listens to the whole encounter, sends it to a vendor's cloud, and bills per provider every month. Medical dictation software is the other shape: you say what belongs in the note, a local model drafts it from your own prompt, and it stays on your machine at $0. For a telehealth consult it captures your speakers directly, so nothing joins the call and nothing is uploaded."},{"question":"Does it draft the note, or just transcribe?","answer":"Both. Run a local model (Llama, Qwen, Gemma) or connect your own AI key, then save your own SOAP, referral and discharge prompts as one-click actions and run whichever one the visit needs. A local model keeps the PHI on your machine; your own key means the model and data path are ones you already trust."},{"question":"Is it actually free?","answer":"Yes. Unlimited local dictation, 100+ languages, all three desktop platforms: $0 forever, for every clinician. Pro ($8/month) adds more cloud transcription and mobile dictation if you want them."}]},"reviews":{"label":"What users say","title":"Built by people who don't want to be recorded either.","subtitle":"From our Discord, GitHub and inboxes, the same instinct that matters in the exam room.","score":{"value":"0","label":"Patient recordings on a server","body":"In local mode, audio stays on your device and is never stored on a server."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for doctors & clinicians","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Clinicians who want free, private, offline dictation and are fine with a quick local setup"}]},{"parts":[{"text":"Clinicians who need unlimited cloud transcription, cross-device sync and mobile access"}]},{"parts":[{"text":"Practices and teams that need unlimited meeting recordings, speaker labels and Agent mode"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Free offline dictation, with patient audio staying on your device"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models or keys"}]},{"parts":[{"text":"Unlimited meeting recordings, speaker labels and workflow features for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"Some conversations stay in the exam room. We built for those.","paragraphs":["OpenWhispr started with a simple objection: what a patient tells you in the exam room shouldn't have to pass through a stranger's server to become a note.","Our code is public on GitHub. In local mode, transcription runs on the device with no internet required.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Sign the last note before you leave the building.","body":"Download OpenWhispr. Dictate the visit, draft the note, then paste it into the chart while it is still fresh. Choose local models when processing needs to stay on your computer.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free unlimited local dictation on Mac, Windows & Linux."}}},"useCaseMac":{"metadata":{"title":"Mac Dictation — Offline AI Voice-to-Text | OpenWhispr","description":"Mac dictation powered by Whisper & Parakeet — runs locally on Apple Silicon, works offline, open source, and free. The Dragon Dictate replacement.","keywords":["mac dictation","best dictation software for mac","dragon dictate for mac","apple dictation alternative","apple silicon dictation","mac speech to text"]},"breadcrumb":{"home":"Home","useCases":"Use Cases","mac":"Mac"},"hero":{"badge":"Mac Dictation","title":"The Modern AI Mac Dictation App","subtitle":"A real upgrade over built-in Apple Dictation and a replacement for discontinued Dragon Dictate for Mac. Whisper and Parakeet accuracy runs locally on Apple Silicon, with offline privacy and open-source transparency.","viewOnGitHub":"View on GitHub"},"problem":{"title":"Mac Dictation Has a Gap, and OpenWhispr Fills It","paragraph1":"For years, serious dictation on a Mac meant Dragon Dictate. Then Nuance discontinued it, leaving Mac users with built-in Apple Dictation: convenient, but limited in accuracy, formatting, and control.","paragraph2":"AI speech recognition has improved dramatically. OpenAI Whisper and NVIDIA Parakeet now run well on Apple Silicon, and OpenWhispr brings those models to a native desktop workflow.","callout":"The result is modern Mac speech-to-text with a system-wide hotkey, stronger accuracy, local processing, and no cloud dependency for core transcription."},"benefits":{"title":"Why OpenWhispr Is the Best Dictation Software for Mac","items":[{"title":"Whisper and Parakeet Accuracy","body":"Choose from multiple local speech models. Larger models handle accents, technical vocabulary, and natural speech better than basic built-in dictation."},{"title":"Built for Apple Silicon","body":"OpenWhispr runs well on M1, M2, M3, and M4 Macs. Intel Macs are supported too; smaller models are best for speed there."},{"title":"System-Wide Push-to-Talk","body":"Hold a global hotkey to dictate anywhere on macOS: Notes, Mail, Messages, Slack, Safari, Chrome, Pages, Word, Xcode, or your terminal."},{"title":"Agent Mode AI Cleanup","body":"Convert rough dictation into polished text, bullet points, emails, or structured notes using prompts you control."},{"title":"Private and Offline","body":"Audio is transcribed on-device. Disconnect from Wi-Fi and core dictation still works."},{"title":"Open Source and Auditable","body":"The full codebase is public on GitHub, so you can inspect how the app handles data."}]},"appleComparison":{"title":"OpenWhispr vs Apple Dictation","intro":"Apple Dictation is fine for quick, casual use. If you dictate often, the differences add up.","columns":[{"title":"OpenWhispr","items":["Whisper and Parakeet models with size you choose","Agent mode reformats dictation into clean text","Always local and works fully offline","Open source and auditable","Local transcription history"]},{"title":"Apple Dictation","items":["Single built-in engine with no model choice","No structured AI reformatting","Convenient but weaker on hard speech","Closed source","No dedicated transcription history"]}]},"scenarios":{"title":"Real-World Mac Dictation Scenarios","items":[{"title":"Replacing Dragon Dictate for Mac","body":"If you relied on Dragon for years, OpenWhispr gives you a modern push-to-talk flow that works in any app, while running locally and staying actively maintained."},{"title":"Writing Email and Notes Faster","body":"Draft a reply in Mail, capture thoughts in Notes, or send a Slack message by voice. Agent mode can turn a rough first pass into a clean message."},{"title":"Coding with Voice in Xcode and VS Code","body":"Dictate comments, documentation, and commit messages without leaving your development flow."}]},"setup":{"title":"Quick Setup and Which Model Suits Your Mac","stepsTitle":"Getting started in minutes","steps":["Download OpenWhispr for macOS and open it.","Grant microphone and accessibility permissions when prompted.","Pick a speech model; it downloads and runs locally.","Set your push-to-talk hotkey, then dictate in any app."],"modelTitle":"Choosing a model","modelTips":["Apple Silicon M1-M4: run larger Whisper or Parakeet models for top accuracy with good speed.","Older or Intel Macs: choose a smaller model for fast, lightweight transcription.","Need near-instant results: smaller models return text faster; bigger models trade speed for accuracy."],"noteLead":"Privacy note:","noteBody":"Core transcription always runs locally on your Mac. Agent-mode cleanup is optional and only uses a cloud LLM if you configure one."},"toolComparison":{"title":"How It Compares to Other Mac Dictation Tools","items":[{"title":"vs MacWhisper","body":"MacWhisper is excellent for transcribing existing audio files. OpenWhispr is built for live, system-wide dictation into any app."},{"title":"vs SuperWhisper","body":"OpenWhispr stands out with both Whisper and Parakeet models, agent-mode cleanup, cross-platform support, and a fully open-source codebase."},{"title":"vs Apple Dictation","body":"OpenWhispr adds model choice, stronger AI transcription, structured cleanup, local history, and open-source transparency."}]},"faqTitle":"Mac Dictation FAQ","faqs":[{"question":"What is the best dictation software for Mac?","answer":"For accuracy and privacy, OpenWhispr is a strong choice. It runs Whisper and Parakeet locally, works system-wide, and keeps core transcription on your Mac."},{"question":"Is there a Dragon Dictate replacement for Mac?","answer":"Yes. OpenWhispr fills much of the gap left by Dragon Dictate for Mac with modern local AI speech recognition and a push-to-talk workflow."},{"question":"How is OpenWhispr better than Apple Dictation?","answer":"It offers stronger model-based transcription, agent-mode cleanup, model-size choice, local history, and open-source auditability."},{"question":"Does Mac dictation work offline with OpenWhispr?","answer":"Yes. Core transcription runs on your Mac with no internet connection."},{"question":"Does OpenWhispr run on Apple Silicon?","answer":"Yes. It runs well on M1, M2, M3, and M4 Macs, and also supports Intel Macs."},{"question":"Does it work in any app?","answer":"Yes. Hold the global hotkey and your transcribed text appears wherever the cursor is focused."}],"links":{"items":[{"href":"/pricing","label":"Pricing"},{"href":"/use-cases","label":"All use cases"},{"href":"/use-cases/windows","label":"Windows dictation"},{"href":"/compare/apple-dictation","label":"Compare Apple Dictation"},{"href":"/compare/macwhisper","label":"Compare MacWhisper"}]},"cta":{"title":"Get the Best Dictation App for Mac","description":"Download OpenWhispr and start dictating on your Mac in minutes. Runs on Apple Silicon, works offline, and is open source.","footer":"Free to start · Modern macOS · No account required"}},"authGate":{"providers":{"google":"Google","microsoft":"Microsoft","apple":"Apple"},"continueWithProvider":"Continue with {provider}","or":"or","fields":{"email":"Email address","password":"Password","fullName":"Full name","passwordHint":"At least 8 characters"},"actions":{"checking":"Checking...","continueWithEmail":"Continue with Email","back":"Back","signingIn":"Signing in...","signIn":"Sign In","createAccount":"Create account","tryAgain":"Try again","sending":"Sending...","sendResetLink":"Send reset link","creatingAccount":"Creating account...","resendVerificationEmail":"Resend verification email","useDifferentEmail":"Use a different email","useSso":"Use single sign-on (SSO)"},"status":{"completingSignIn":"Completing sign in...","redirecting":"Redirecting..."},"errors":{"generic":"Something went wrong. Please try again.","authenticationTimedOut":"Authentication timed out. Please try again.","sessionExpired":"Session expired. Please try signing up again.","waitingForVerification":"Still waiting for verification. Check your inbox or resend the email.","unableToCheckEmail":"Unable to check this email. Please try again.","network":"Network error. Please check your connection.","networkRetry":"Network error. Please check your connection and try again.","incorrectPassword":"Incorrect password. Please try again.","accountExists":"An account with this email already exists. Please sign in.","createAccountFailed":"Failed to create account. Please try again.","resendFailed":"Failed to resend. Please try again."},"signin":{"title":"Welcome back","forgotPassword":"Forgot password?","newHere":"New here?"},"forgot":{"title":"Reset your password","description":"Enter your email and we'll send you a reset link","emailSentTitle":"Check your email","emailSentBody":"We sent a password reset link to","didNotReceive":"Didn't receive the email? Check your spam folder or try again."},"signup":{"title":"Create your account","haveAccount":"Have an account?"},"verifyEmail":{"title":"Check your email","message":"We sent a verification link to"},"legal":{"line":"By continuing, you agree to our <termsLink>Terms of Service</termsLink> and <privacyLink>Privacy Policy</privacyLink>."}},"getStarted":{"title":"Get started","subtitle":"Download OpenWhispr, free and open source.","processing":"Starting your download...","skip":"Continue without an account","verificationMessage":"Verify your email to start your download. We sent a link to","metadata":{"title":"Get Started | OpenWhispr","description":"Create your OpenWhispr account and get started with private AI voice dictation."}},"checkout":{"title":"Get started with OpenWhispr","processing":"Redirecting to checkout...","period":{"month":"/mo","year":"/yr"},"plan":{"proMonthly":"Pro · Monthly","proAnnual":"Pro · Annual","businessMonthly":"Business · Monthly","businessAnnual":"Business · Annual"},"badge":{"twoMonthsFree":"2 months free"},"metadata":{"title":"Checkout | OpenWhispr","description":"Complete your OpenWhispr subscription purchase securely."}},"checkoutSuccess":{"appRedirect":{"title":"Redirecting to OpenWhispr...","body":"You can close this tab","fallback":"Tap here if the app didn't open"},"success":{"title":"Welcome to Pro!","body":"Your subscription is now active"},"download":{"autoNotice":"Your download should start automatically...","fallback":"Click here if it didn't start","open":"Download OpenWhispr","all":"See all downloads"},"linuxOptions":[{"label":"Linux AppImage (x86_64)","meta":"AppImage"},{"label":"Linux DEB (amd64)","meta":"Debian package"},{"label":"Linux RPM (x86_64)","meta":"RPM package"},{"label":"Linux tar.gz (x64)","meta":"Portable tarball"}],"metadata":{"title":"Checkout Successful | OpenWhispr","description":"Your OpenWhispr checkout is complete and your subscription is ready."}},"upgrade":{"expiredTitle":"Link expired","redirecting":"Redirecting to checkout...","errors":{"missingToken":"Missing upgrade token. Please check your email link.","invalidLink":"This link is no longer valid. You can upgrade from within the app.","generic":"Something went wrong. Please try again or upgrade from within the app."},"metadata":{"title":"Upgrade | OpenWhispr","description":"Upgrade your OpenWhispr plan to unlock more voice dictation."}},"authCallback":{"failedTitle":"Sign-in failed","redirecting":"Redirecting to OpenWhispr","closeTab":"You can close this tab","desktopFallback":"Click here if OpenWhispr didn't open","mobileFallback":"Tap here if the app didn't open","backToSignIn":"Back to sign-in","errors":{"account_not_linked":"An account with this email already exists with a different sign-in method.","state_mismatch":"Sign-in expired before completing. Please try again.","please_restart_the_process":"Sign-in expired before completing. Please try again.","invalid_code":"Sign-in failed. Please try again.","unable_to_get_user_info":"Couldn't read your profile from the provider. Please try again.","noSession":"Could not retrieve session. Please try signing in again.","default":"Sign-in failed. Please try again.","sso_required":"This workspace requires single sign-on. Sign in with your company SSO instead.","account_mismatch":"You were signed in as a different account than the one you entered."},"metadata":{"desktopCallback":{"title":"Redirecting... | OpenWhispr","description":"Completing your OpenWhispr desktop sign-in and redirecting you to the app."},"mobileCallback":{"title":"Redirecting... | OpenWhispr","description":"Completing your OpenWhispr mobile sign-in and redirecting you to the app."},"error":{"title":"Sign-in error | OpenWhispr","description":"We couldn't complete your OpenWhispr sign-in. Review the error and try again."}},"mismatch":{"switchAccount":"Sign in as a different account","continueAnyway":"Continue with the signed-in account"}},"notFound":{"title":"Oops! Page not found","subtitle":"Looks like this page got lost in transcription!","quip":"Don't worry, we think faster than we type...","description":"The page you're looking for might have been moved, deleted, or is temporarily unavailable.","home":"Take me home","back":"Go back","autoRedirectLabel":"Auto-redirect:","autoRedirectBody":"Taking you home in {seconds} seconds..."},"verifyEmail":{"loading":"Verifying your email...","errors":{"noToken":"No verification token found. Please check your email link.","failed":"Verification failed. The link may have expired.","network":"Network error. Please check your connection and try again."},"resend":{"signInFirst":"Please sign in first, then request a new link.","failed":"Could not resend. Please try again.","sent":"Verification email sent! Check your inbox.","network":"Network error. Please try again."},"success":{"title":"Email verified!","desktop":"If you're not redirected automatically, click below.","browser":"You can close this tab and return to your signup to continue.","openApp":"Open OpenWhispr"},"actions":{"sending":"Sending...","requestNewLink":"Request a new link"},"metadata":{"title":"Verify Email | OpenWhispr","description":"Verify your email address to finish setting up your OpenWhispr account."}},"resetPassword":{"loading":"Loading...","errors":{"passwordMismatch":"Passwords do not match.","passwordTooShort":"Password must be at least 8 characters.","network":"Network error. Please check your connection and try again."},"invalid":{"title":"Invalid reset link","description":"This password reset link is invalid or has expired.","cta":"Go to Get Started"},"success":{"title":"Password updated","description":"You can now sign in with your new password.","cta":"Continue to sign in"},"form":{"title":"Create new password","description":"Enter your new password below","newPassword":"New password","confirmPassword":"Confirm new password","passwordHint":"At least 8 characters","resetting":"Resetting...","submit":"Reset password"},"metadata":{"title":"Reset Password | OpenWhispr","description":"Choose a new password for your OpenWhispr account."}},"consent":{"banner":{"ariaLabel":"Cookie consent","title":"We value your privacy","eu":"We use cookies to measure usage and our marketing. Analytics and marketing cookies stay off until you allow them.","us":"We and our partners use cookies for analytics and advertising. You can opt out anytime.","row":"We use cookies to understand usage and improve OpenWhispr.","gpc":"We detected a Global Privacy Control signal and have opted you out of ad cookies.","policyLine":"See our <cookieLink>Cookie Policy</cookieLink> and <privacyLink>Privacy Policy</privacyLink>.","rejectAll":"Reject all","customize":"Customize","acceptAll":"Accept all","privacyChoices":"Your privacy choices","cookieSettings":"Cookie settings","gotIt":"Got it","updated":"We've updated our cookie policy since your last visit. Your existing choices remain in effect until you change them."},"settings":{"title":"Cookie preferences","description":"Choose which cookies OpenWhispr may use. You can change this anytime from the footer.","necessary":{"title":"Strictly necessary","description":"Required for sign-in and security. Always active.","ariaLabel":"Strictly necessary cookies (always on)"},"analytics":{"title":"Analytics","description":"Helps us understand usage (Google Analytics 4).","ariaLabel":"Analytics cookies"},"marketing":{"title":"Marketing","description":"Measures our ads and campaigns (Google Ads, Meta).","gpcDescription":"Off — your browser's Global Privacy Control signal opts you out.","ariaLabel":"Marketing cookies"},"rejectAll":"Reject all","acceptAll":"Accept all","save":"Save preferences"},"trigger":{"privacyChoices":"Your Privacy Choices","cookieSettings":"Cookie Settings"}},"referral":{"someone":"Someone","loading":"Loading...","invitedBy":"Invited by {name}","invited":"You're invited","title":"Unlock 1 free month of OpenWhispr Pro","claimPrompt":"Sign up to claim your free month of Pro","continueWithGoogle":"Continue with Google","or":"or","processing":"Setting up your account...","alreadyHaveAccount":"Already have an account?","downloadOpenWhispr":"Download OpenWhispr","fields":{"email":"Email address","password":"Password","fullName":"Full name","passwordMin":"Password (min. 8 characters)"},"actions":{"checking":"Checking...","continueWithEmail":"Continue with Email","signingIn":"Signing in...","signInClaim":"Sign in & claim offer","useDifferentEmail":"Use a different email","creatingAccount":"Creating account...","createClaim":"Create account & claim offer","useSso":"Use single sign-on (SSO)"},"signin":{"as":"Signing in as"},"signup":{"for":"Creating account for"},"legal":{"line":"By continuing, you agree to our <termsLink>Terms</termsLink> and <privacyLink>Privacy Policy</privacyLink>."},"errors":{"network":"Network error. Please try again.","incorrectPassword":"Incorrect password. Please try again.","accountExists":"Account already exists. Please sign in.","createAccountFailed":"Failed to create account.","signInTimedOut":"Sign-in timed out. Please try again.","signInFailed":"Failed to sign in.","ssoEmailRequired":"Enter your work email above to use single sign-on.","ssoFailed":"Single sign-on failed."},"spectrogram":{"friend":"a friend","signalId":"Signal ID","scanAriaLabel":"Scan audio signature","scan":"Scan","stop":"Stop","replay":"Replay","unlocked":"Unlocked","freeMonth":"1 free month of Pro","signature":"{name}'s audio signature","scanning":"Scanning..."},"continueWithProvider":"Continue with {provider}","providers":{"google":"Google","microsoft":"Microsoft","apple":"Apple"},"metadata":{"title":"Referral | OpenWhispr","description":"Join OpenWhispr through a referral link and start using private AI voice dictation."}},"demo":{"ariaLabel":"Product demo","tabpanelFallback":"Demo","tabs":[{"label":"Speech-to-Text"},{"label":"Meeting Notes"},{"label":"AI Chat"},{"label":"Audio Upload"}],"panelLabels":["Speech-to-Text demo","Meeting Notes demo","AI Chat demo","Audio Upload demo","Custom Dictionary demo","Integrations demo","History demo"],"sidebar":{"search":"Search","settings":"Settings","userName":"Alex Chen","userEmail":"alex@company.com","nav":{"home":"Home","notes":"Notes","upload":"Upload","dictionary":"Dictionary","integrations":"Integrations"}},"speech":{"dictationText":"Hey team, I just finished reviewing the Q2 roadmap and I think we should prioritize the voice pipeline before the new integrations.","withoutOpenWhispr":"Without OpenWhispr","slowWpm":"~40 WPM","messageApp":"Slack — #product","placeholder":"Type a message...","withOpenWhispr":"With OpenWhispr","fastWpm":"~150 WPM","faster":"3x faster"},"history":{"transcriptions":{"title":"Transcriptions","items":[{"time":"9:15 AM","text":"The quarterly metrics look promising. Revenue is up 12% month over month and user retention has improved significantly."},{"time":"4:30 PM","text":"Let's schedule the design review for next week. I want the whole team to see the new components before we ship."}]},"upcomingMeetings":{"title":"Upcoming Meetings","items":[{"time":"11:00 AM - 11:30 AM","title":"Design Review"},{"time":"2:00 PM - 2:45 PM","title":"Sprint Planning"},{"time":"4:00 PM - 4:30 PM","title":"1:1 with Sarah"}]}},"notes":{"rawNotes":"- ship voice pipeline by march\n- talk to design team re: new UI\n- budget Q: ask finance","meetingTitle":"Sprint Planning","title":"Product Sync","recording":"Recording","you":"You","date":"Mar 11, 2026","foldersLabel":"Folders","notesLabel":"Notes","viewTabs":{"transcript":"Transcript","notes":"Notes","enhanced":"Enhanced"},"sections":[{"heading":"Decisions","items":["Voice pipeline is the top priority — ship by end of March before any other workstream","UI redesign deferred until pipeline lands, don't want to split focus","Dictionary feature approved: custom words for medical, legal, and technical terms"]},{"heading":"Action Items","items":["Sarah: frontend migration to new component library by end of sprint","Alex: API refactor + latency benchmarks (target sub-200ms, currently ~280ms)","Gabriel: follow up with finance on Q2 budget, need headcount approval by Friday","Design review for new editor UI — next Tuesday 2pm, full team"]},{"heading":"Open Questions","items":["Real-time collaboration on notes — CRDT vs OT approach? No timeline yet","Third-party dependency audit needed before public open source release","How to handle offline sync for mobile — local-first or queue-based?"]}],"done":"Done","primaryAction":"Sync Template","enhanceNotes":"Enhance Notes","actions":[{"name":"Sync Template","description":"Format using team template"},{"name":"Meeting Summary","description":"Summarize decisions & actions"},{"name":"Email Draft","description":"Draft follow-up email"}],"folders":[{"name":"Personal","count":5,"isActive":true},{"name":"Meetings","count":3,"isActive":false},{"name":"Work","count":2,"isActive":false}],"noteList":[{"title":"Product Sync","preview":"Action items from weekly sync...","time":"10:30 AM"},{"title":"Voice pipeline design","preview":"Architecture notes for v2...","time":"9:15 AM"},{"title":"Weekly standup","preview":"Sprint progress and blockers...","time":"Yesterday"}]},"aiChat":{"title":"AI Chat","listening":"Listening...","askAnything":"Ask anything about your notes","speakQuestion":"Speak your question","voicePoweredQa":"Voice-powered Q&A over your data","userMessage":"Summarize what we discussed in today's meetings","response":"Based on today's discussions:\n\n• Voice pipeline — ship by end of March\n• UI redesign deferred post-pipeline\n• Follow up with finance on Q2 budget","releaseToFinish":"Release {hotkey} to finish","holdToSpeak":"Hold {hotkey} to speak"},"upload":{"noteTitle":"Design System Migration Planning","transcript":"The team agreed to move forward with the new design system. Sarah will lead the frontend migration and Alex will handle the API refactor by end of sprint.","now":"Now","preview":"Team aligned on migrating to the new design system...","dateMeta":"Mar 12, 2026 · 847 words","structuredIntro":"Team aligned on migrating to the new component-based design system. Frontend migration begins this sprint with Sarah leading the component audit. Alex will refactor the API layer in parallel.","sections":[{"heading":"Key Decisions","items":["Adopt new design tokens across all product surfaces","Frontend migration handled component-by-component","API refactor scoped to auth and data-fetching layers"]},{"heading":"Action Items","items":["Sarah — Component audit and migration plan (due Fri)","Alex — Scope API refactor; draft RFC by Wednesday","Jamie — Update CI pipeline for new token format","Team — Review Figma library v2 before next standup"]},{"heading":"Timeline","items":["Sprint 14: Audit + migration plan + API RFC","Sprint 15: Component migration, design tokens","Sprint 16: Integration testing, deprecate legacy"]}],"dropPrompt":"Drop audio file or browse","formats":"MP3, WAV, M4A, WebM, OGG, FLAC, AAC","fileName":"meeting-recording.mp3","fileMeta":"12:04 duration · 18.3 MB","folder":"Personal","transcribe":"Transcribe","transcribing":"Transcribing locally...","complete":"Transcription Complete","openNote":"Open Note","newUpload":"New Upload"},"dictionary":{"title":"Custom Dictionary","clearAll":"Clear all","placeholder":"Add words or phrases...","hint":"Words are sent as hints to improve transcription accuracy. Separate multiple words with commas."},"integrations":{"title":"Integrations","subtitle":"Connect your favorite tools","calendarDescription":"Auto-join meetings and transcribe in real time","connected":"Connected","comingSoon":"More integrations coming soon"}},"blogComponents":{"author":{"defaultRole":"Engineering","updated":"Updated {date}"},"layout":{"defaultTag":"Blog","tableOfContents":"Table of contents","onThisPage":"On this page"},"diarizationPipeline":{"title":"The Local Diarization Pipeline","steps":[{"label":"Audio Capture","sub":"Mic + system, 16kHz"},{"label":"VAD","sub":"Silero, 2MB"},{"label":"Segmentation","sub":"pyannote 3.0"},{"label":"Embedding","sub":"CAM++, 512-dim"},{"label":"Clustering","sub":"Agglomerative, 0.5"},{"label":"Labels","sub":"Match profiles in SQLite"}],"whyLocal":"Why it stays local:","bullets":["Every stage runs on your device via a single sherpa-onnx binary — no Python, no cloud.","~45MB of ONNX models downloaded once, then cached at ~/.cache/openwhispr/diarization-models/.","Embeddings are 512-dim float32 vectors stored as SQLite BLOBs on your local disk."]},"linuxSpectrum":{"title":"Linux Dictation Landscape (2026)","left":"Polished UX","right":"DIY control"},"windowsSpectrum":{"title":"Windows Dictation Landscape (2026)","left":"AI-powered","right":"Specialized"},"localVsCloud":{"title":"Local vs Cloud: Key Tradeoffs","local":{"title":"Local (OpenWhispr)","points":[{"label":"Privacy","detail":"Audio stays on device"},{"label":"Latency","detail":"No network round trip"},{"label":"Cost","detail":"No per-minute API billing"},{"label":"Offline","detail":"Works without internet"},{"label":"Control","detail":"Deterministic local behavior"}]},"cloud":{"title":"Cloud APIs","points":[{"label":"Privacy","detail":"Policy + config dependent"},{"label":"Latency","detail":"Network and region sensitive"},{"label":"Cost","detail":"Usage-based billing"},{"label":"Offline","detail":"Unavailable"},{"label":"Scale","detail":"Easy for large batch workloads"}]},"recommendation":"Best practical pattern: local-first, route edge cases to cloud only when needed."},"openSourceTrust":{"title":"Why Open Source Improves Voice-Data Trust","layers":[{"title":"Auditability","detail":"Inspect data flows directly in code"},{"title":"Reproducibility","detail":"Build from source and verify binaries"},{"title":"Fork rights","detail":"Users keep control if maintainers change direction"},{"title":"No hard lock-in","detail":"Migrate stacks without losing workflows"}]},"voicePrivacyFlow":{"title":"Voice Data Path: Local vs Cloud Risk Surface","microphone":"Microphone","localInference":"Local Inference","localText":"Local Text","zeroExposure":"Zero exposure","cloudApi":"Cloud API","remoteText":"Remote Text","networkExposure":"Network exposure","defaultPath":"OpenWhispr default: local path first, optional cloud only when explicitly configured."},"whisperArchitecture":{"title":"How Whisper Processes Audio","steps":["Audio Input","Log-Mel Features","Encoder","Decoder","Text"],"whyItMatters":"Why it matters for OpenWhispr:","bullets":["Runs locally via whisper.cpp — raw audio stays on-device.","Model size controls the speed/accuracy tradeoff.","Same pipeline powers dictation across macOS, Windows, and Linux."]},"werChart":{"title":"Word Error Rate: Whisper vs Open-Source Competitors","subtitle":"LibriSpeech test-clean (English read speech) — lower is better","werLabel":"WER","turboNote":"8x faster","otherOpenSource":"Other open-source","sources":"Sources: Hugging Face model cards, NVIDIA model cards, Open ASR Leaderboard (Feb 2026). English-clean read speech only — real-world accuracy varies by audio quality, accent, and domain. Commercial APIs omitted as they do not publish LibriSpeech benchmarks."},"modelTradeoff":{"title":"Whisper Model Sizes: Speed vs Accuracy","subtitle":"Bubble size reflects relative parameter count","ariaLabel":"Scatter chart of Whisper model speed vs accuracy tradeoffs","fasterAxis":"Faster →","accuracyAxis":"More accurate →","recommendation":"Recommendation: start with small, adjust based on hardware."},"openAsrWerChart":{"averageWer":"average WER","title":"English Accuracy: Average Word Error Rate","caption":"Multi-dataset averages from the Hugging Face Open ASR Leaderboard benchmarks (lower is better). Nemotron shown in its streaming configuration.","sources":"Sources: NVIDIA model cards on Hugging Face and the Open ASR Leaderboard, July 2026. WER = percentage of words transcribed incorrectly."},"streamingAsrFlow":{"title":"The Live Streaming Path in OpenWhispr"}},"compareFathom":{"metadata":{"title":"Fathom Alternative — Local, Private & Open Source","description":"Looking for a Fathom alternative? OpenWhispr runs voice-to-text and meeting notes locally — no bot, no cloud, no account needed. Free and open source.","keywords":["fathom alternative","fathom ai alternative","private fathom alternative","open source fathom alternative","local meeting notes","no bot meeting notes"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","fathom":"vs Fathom"},"hero":{"title":"The Private, Open-Source Fathom Alternative","description":"OpenWhispr is a Fathom alternative for people who want <voice>voice-to-text</voice> and <meetings>meeting notes</meetings> that run 100% on their own device. No bot joins your calls, your audio never leaves your machine, and the whole app is open source.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","fathomIntro":"is a polished cloud meeting notetaker with a famously generous free tier. Its bot joins your Zoom, Teams, and Meet calls, records and transcribes them, and produces clean summaries, highlights, and action items with very little setup. For automated meeting recaps, it is one of the easiest tools to start with.","openWhisprIntro":"optimizes for something Fathom does not: keeping your audio on your own machine. It is a voice-to-text tool plus on-device meeting notes built on OpenAI Whisper and NVIDIA Parakeet. No bot joins the call, no account is required, and nothing is uploaded. Press a hotkey and your words appear in any app — or record a meeting locally and transcribe it on-device.","chooseFathom":"Choose Fathom if: you want the easiest automated meeting summaries with a strong free tier and don't mind cloud processing.","chooseOpenWhispr":"Choose OpenWhispr if: you want private, local voice-to-text and meeting notes that never upload your audio to anyone."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"100% local — meeting audio stays on your device"},"competitor":{"status":"none","details":"Cloud-based — recordings stored on servers"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"MIT licensed, auditable on GitHub"},"competitor":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Works Offline","openWhispr":{"status":"full","details":"Local models run with no internet"},"competitor":{"status":"none","details":"Requires an internet connection"}},{"feature":"Meeting Capture","openWhispr":{"status":"partial","details":"Records system audio locally — no bot joins"},"competitor":{"status":"full","details":"A bot auto-joins Zoom, Teams, and Meet calls"}},{"feature":"General Dictation","openWhispr":{"status":"full","details":"System-wide voice typing in any app"},"competitor":{"status":"none","details":"Meeting-only; not a dictation tool"}},{"feature":"Transcription & Languages","openWhispr":{"status":"full","details":"Whisper + Parakeet, 100+ languages"},"competitor":{"status":"full","details":"Cloud transcription with AI summaries"}},{"feature":"Free Tier","openWhispr":{"status":"full","details":"Unlimited local dictation, free forever"},"competitor":{"status":"full","details":"Generous free plan for individuals"}},{"feature":"Team & CRM Features","openWhispr":{"status":"partial","details":"API/MCP access; basic team on Business"},"competitor":{"status":"full","details":"CRM sync and team features on paid plans"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","items":[{"title":"Local by default:","body":"Transcription runs on your machine. No bot in the participant list, no audio uploaded, nothing stored on a third-party server."},{"title":"Dictation everywhere:","body":"A system-wide hotkey types your voice into any app — email, docs, your editor, an AI chat. Fathom is meeting-only."},{"title":"Open source:","body":"MIT licensed and auditable — you can verify exactly how your audio is handled."}]},"whyFathom":{"title":"Why Fathom?","items":[{"title":"Effortless meeting recaps:","body":"Its bot joins and summarizes calls automatically, with highlights and action items ready the moment the meeting ends."},{"title":"Generous free tier:","body":"One of the most capable free plans among cloud notetakers for individual users."},{"title":"Polished, low-setup UX:","body":"Clean summaries and CRM sync that most teams can adopt with almost no configuration."}]},"faq":{"title":"Frequently Asked Questions"},"faqs":[{"question":"What is the best Fathom alternative?","answer":"It depends on what you need. If you want a private, open-source alternative for everyday voice-to-text and on-device meeting notes, OpenWhispr is a strong choice — it runs speech-to-text 100% locally using OpenAI Whisper and NVIDIA Parakeet, works system-wide via a push-to-talk hotkey, and never sends your audio to the cloud. If you specifically want the easiest automated meeting summaries with a generous free tier, Fathom itself is excellent at that."},{"question":"Is there a free Fathom alternative?","answer":"Yes. OpenWhispr is free and open source. Unlimited local dictation in 100+ languages works offline at no cost on macOS, Windows, and Linux. An optional Pro plan ($8/mo) adds unlimited cloud transcription, more meeting recording hours, mobile, and device sync — but you never need it for core local voice-to-text."},{"question":"Is there a local or private alternative to Fathom?","answer":"OpenWhispr is built for exactly this. Core transcription happens on your own machine — your audio never leaves the device and works fully offline — and because the app is open source, you can audit precisely how your data is handled. There is no bot in your calls and nothing is uploaded, which matters most for sensitive meetings like HR, legal, or anything with customer data."},{"question":"Does OpenWhispr join meetings with a bot like Fathom?","answer":"No — and that is the point. Fathom sends a bot that appears in your call and uploads the audio to its cloud. OpenWhispr instead captures the meeting's system audio directly on your device, so no bot joins the participant list and nothing is uploaded. You still disclose that you are recording, but the conversation stays entirely on your machine."},{"question":"How is OpenWhispr different from Fathom?","answer":"Fathom is a cloud meeting notetaker focused on effortless, automated summaries with a strong free tier. OpenWhispr is a privacy-first voice-to-text tool plus on-device meeting notes: press a hotkey and speak into any app, or record and transcribe a meeting locally. They solve different problems — many people use a cloud notetaker for routine calls while using OpenWhispr for private, everyday dictation and sensitive meetings."}],"cta":{"title":"A Fathom alternative that never uploads your audio","description":"Private, open-source voice-to-text and on-device meeting notes. Free on macOS, Windows, and Linux.","viewOnGitHub":"View on GitHub"}},"featureDictationSoftware":{"metadata":{"title":"Dictation Software (Download for Free) | OpenWhispr","description":"OpenWhispr is free dictation software that turns speech into text in any app. Works offline, runs locally and keeps your voice on your device.","keywords":["dictation software","voice dictation software","free dictation software","speech to text software","voice typing","dictation app","offline dictation","local dictation software"]},"breadcrumb":{"home":"Home","dictationSoftware":"Dictation Software"},"structuredData":{"description":"Open-source dictation software. Turn speech into text in any app — run it privately on your device or on OpenWhispr Cloud for top accuracy — in 100+ languages.","offerDescription":"Free to start. Unlimited local dictation, with optional Pro and Business plans.","featureList":["System-wide push-to-talk dictation in any app","Local, offline speech to text plus OpenWhispr Cloud for top accuracy","100+ languages with built-in translation","OpenAI Whisper and NVIDIA Parakeet models","Custom dictionary for names, jargon and acronyms","Built-in AI writing assistant to reformat and clean up dictation","Open source on macOS, Windows and Linux"]},"hero":{"badge":"Free to start · Open source","title":"Free Dictation Software for Every App You Use","description":"Press a hotkey, speak, and clean, punctuated text appears anywhere — email, docs, Slack, your editor or an AI chat. Run it privately on your device, or switch on OpenWhispr Cloud for top accuracy on any machine. Either way, your audio is never stored.","github":"View on GitHub","platforms":"Free to start on macOS, Windows & Linux · No card, no account"},"trustBar":{"title":"Dictate into every app you already use"},"painPoints":{"eyebrow":"Why people switch","title":"Most dictation software makes you trade something away.","description":"Speed for privacy, privacy for accuracy, or your whole workflow for one company’s cloud. OpenWhispr is built so you don’t have to choose.","items":[{"title":"Your ideas outrun your fingers.","body":"You speak several times faster than you type. OpenWhispr turns speech into clean text at the speed you think, so email, prompts and drafts stop getting squeezed through a keyboard."},{"title":"You can’t see what most voice tools do with your voice.","body":"Most dictation software is closed and cloud-only. OpenWhispr is open source, and whether you run it locally or on OpenWhispr Cloud, your audio is transcribed and never stored, sold, or used for training — verifiable in the code."},{"title":"Your setup shouldn’t be locked to one app or OS.","body":"Closed tools trap your workflow in their cloud and their platform. OpenWhispr runs on Mac, Windows and Linux, works in every app, and lets you bring your own AI keys — your workflow stays yours."},{"title":"Built-in dictation isn’t enough.","body":"Apple Dictation and Windows Speech are inaccurate, online-dependent and locked to one OS. OpenWhispr runs Whisper-class voice recognition on all three platforms, with a custom dictionary for your names and jargon."}]},"features":{"eyebrow":"What you get","title":"One hotkey between your voice and every app.","description":"Dictation software that works wherever your cursor is — dictate emails, prompt AI, and capture ideas with no destination app to copy and paste from.","items":[{"title":"Dictate into any app","body":"Press the hotkey, speak, release. Punctuated text appears wherever your cursor is — email, docs, Slack, your code editor. Nothing to copy and paste."},{"title":"Talk to AI instead of typing at it","body":"Speak long, context-rich prompts into ChatGPT, Claude or Cursor. Dictation is the fastest way to give AI the detail it needs for a better answer."},{"title":"Accurate in your words","body":"Teach OpenWhispr your names, jargon and acronyms with a custom dictionary, then pick the engine that fits — run locally, or switch on OpenWhispr Cloud for top accuracy on any machine."}]},"stats":{"items":[{"value":"100+","label":"Dictate in","foot":"languages, with translation built in"},{"value":"0","label":"Recordings stored","foot":"local audio never leaves your device; cloud audio is transcribed, then discarded"},{"value":"5 min","label":"Setup takes","foot":"from download to your first dictated sentence"}]},"howItWorks":{"eyebrow":"How it works","title":"Talk to your computer in three steps.","description":"No cloud account, no signup. Just software on your machine, running in about five minutes.","items":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model, or use OpenWhispr Cloud — no account, no card."},{"title":"Set your hotkey.","body":"One shortcut — hold-to-record or push-to-toggle — working system-wide across every app."},{"title":"Press. Speak. Done.","body":"Clean, punctuated text lands wherever your cursor is, in 100+ languages."}]},"reasons":{"eyebrow":"No convincing needed","title":"10 reasons people make OpenWhispr their dictation software","description":"Switching is a trust decision. Every claim here is verifiable in the open-source repo.","items":[{"title":"Works in every app","body":"A system-wide hotkey. If you can place a cursor there, you can dictate there."},{"title":"Private by design, and auditable","body":"Open source with zero data retention — local audio stays on your device, and cloud audio is transcribed then discarded."},{"title":"Local or cloud — your call","body":"Run on-device for full privacy, or switch on OpenWhispr Cloud for top accuracy and speed on any machine."},{"title":"Whisper-class and Parakeet accuracy","body":"The leading open speech-to-text models, plus a cloud option that keeps up on modest hardware."},{"title":"A custom dictionary for your words","body":"Teach it names, jargon and acronyms so your transcripts come out right the first time."},{"title":"100+ languages plus translation","body":"Speak one language, output another — set a language or auto-detect on the fly."},{"title":"Bring your own AI keys","body":"Use your own OpenAI, Anthropic, Gemini or Groq keys for unlimited cloud transcription, no markup."},{"title":"Clean up as you speak","body":"A built-in AI writing assistant reformats and polishes raw dictation — locally or in the cloud."},{"title":"Mac, Windows & Linux","body":"One of the only dictation apps that runs natively on all three, with the same features everywhere."},{"title":"Easier on your hands","body":"If typing is slow, tiring or painful — RSI, fatigue, long days — your voice does the work instead."}]},"faq":{"eyebrow":"Before you download","title":"Common questions about OpenWhispr dictation software","items":[{"question":"Does OpenWhispr work in every app?","answer":"Yes. Unlike voice-to-text software that lives in one app, the hotkey works system-wide — email, docs, Slack, browsers, code editors and AI chat windows. Wherever you can type, you can dictate."},{"question":"How accurate is the speech to text, accents included?","answer":"It runs Whisper-class models and NVIDIA Parakeet, trained on a huge range of accents and 100+ languages. Pick a larger local model, add a custom dictionary for your jargon, or switch on OpenWhispr Cloud for top accuracy on any machine."},{"question":"What is the difference between local and cloud dictation?","answer":"Local dictation runs on your own device — fully private, works offline, and unlimited. OpenWhispr Cloud runs on our servers for top accuracy and speed on any hardware, with no models to download, and audio is transcribed then never stored. Free includes 2,000 cloud words a week; Pro makes it unlimited."},{"question":"Is it free? What do I pay for?","answer":"The app is open source and free — unlimited local dictation in 100+ languages, a custom dictionary, plus 2,000 OpenWhispr Cloud words a week and 5 hours of meeting recording. Pro ($8/month) unlocks unlimited cloud transcription, 20 meeting hours, mobile and device sync."},{"question":"Does my voice ever leave my device?","answer":"In local mode, never — and the open-source code proves there is no upload step. If you use OpenWhispr Cloud or your own API keys, audio is sent only to transcribe it, then discarded. We never store, sell, or train on your data."},{"question":"Does it work as dictation software for Mac, Windows and PC?","answer":"Yes. OpenWhispr is one of the only open-source dictation apps that runs natively on macOS, Windows and Linux. Mac dictation, Windows dictation and Linux all use the same engine and the same features."},{"question":"Can it learn my vocabulary and jargon?","answer":"Yes. The custom dictionary teaches OpenWhispr your names, technical terms and acronyms so specialized language transcribes correctly the first time — included on every plan, local or cloud."},{"question":"Is this AI dictation? Can it format and clean up text?","answer":"Yes. Beyond raw transcription, a built-in AI writing assistant punctuates, reformats and rewrites what you said — using local models or the cloud, with agent mode to automate actions on your transcripts."},{"question":"How is it different from voice recognition software like Dragon or built-in dictation?","answer":"Dragon is powerful but expensive and Windows-first; Apple and Windows dictation are online-dependent and locked to one OS. OpenWhispr runs Whisper-class voice recognition on all three platforms, is open source, and lets you choose local or cloud."},{"question":"What do I need to run it, and does it work offline?","answer":"A reasonably modern Mac, Windows or Linux machine. Smaller local models run on modest hardware, and once installed, local dictation works fully offline. Prefer not to manage models? OpenWhispr Cloud needs nothing to download."}]},"testimonials":{"eyebrow":"What users say","title":"People who stopped typing."},"pricing":{"eyebrow":"Pricing","title":"Start free. Upgrade for cloud speed and more.","description":"Unlimited local dictation is free forever. Pro and Business add unlimited OpenWhispr Cloud, meeting hours, mobile, and team features."},"openSource":{"eyebrow":"Built in the open","title":"Your voice, handled the way you would handle it.","description":"OpenWhispr started with a simple objection: most dictation software listens to everything you say from behind closed source. So we built ours in the open — every release and every line of code is on GitHub, where 5,000+ engineers have starred, forked and audited the project. Local or cloud, we never store, sell, or train on your data, and you can prove it by reading the code.","github":"Star us on GitHub"},"finalCta":{"title":"Your next sentence doesn’t need a keyboard.","description":"Download OpenWhispr and dictate into any app on your computer — in 100+ languages, private by design, five minutes from now.","footer":"Free to start on Mac, Windows & Linux · No card, no account"},"demo":{"dictationText":"Hey team, I just finished reviewing the Q2 roadmap and I think we should prioritize the voice pipeline before the new integrations.","without":"Without OpenWhispr","typingSpeed":"~40 WPM","windowTitle":"Slack — #product","messagePlaceholder":"Type a message…","with":"With OpenWhispr","dictationSpeed":"~150 WPM","faster":"3x faster","transcribing":"Transcribing…","listening":"Listening…","hotkeyPrompt":"Press your hotkey and speak…","indicator":"Hold the hotkey, speak, release — text lands in the active app."}},"compareFireflies":{"metadata":{"title":"Fireflies.ai Alternative — Private, Local, Open Source","description":"A private, open-source Fireflies.ai alternative. Voice-to-text and meeting notes run 100% locally — no bot joins your calls, no audio uploaded.","keywords":["fireflies alternative","fireflies.ai alternative","private fireflies alternative","open source fireflies alternative","local meeting notes","no bot meeting notes"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","fireflies":"vs Fireflies.ai"},"hero":{"title":"The Private, Open-Source Fireflies.ai Alternative","description":"OpenWhispr is a Fireflies.ai alternative for people who want <voice>voice-to-text</voice> and <meetings>meeting notes</meetings> that run 100% on their own device. No bot joins your calls, your audio never leaves your machine, and the whole app is open source.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","firefliesIntro":"is a cloud meeting assistant. Its bot joins your Zoom, Teams, and Meet calls, records and transcribes them, generates summaries and action items, and layers on conversation intelligence and CRM integrations for sales and revenue teams. It is genuinely good at that job.","openWhisprIntro":"is built around a different priority: privacy and ownership. It is a voice-to-text tool plus on-device meeting notes that runs entirely on your computer using OpenAI Whisper and NVIDIA Parakeet. No bot joins the call, no account is required, and nothing is uploaded. Press a hotkey and your words appear in any app — or record a meeting's audio locally and transcribe it on-device.","chooseFireflies":"Choose Fireflies.ai if: you want an automated bot that joins calls with team analytics and CRM sync.","chooseOpenWhispr":"Choose OpenWhispr if: you want private, local voice-to-text and meeting notes that never upload your audio to anyone."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"100% local — meeting audio stays on your device"},"competitor":{"status":"none","details":"Cloud-based — audio and transcripts stored on servers"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"MIT licensed, auditable on GitHub"},"competitor":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Works Offline","openWhispr":{"status":"full","details":"Local models run with no internet"},"competitor":{"status":"none","details":"Requires an internet connection"}},{"feature":"Meeting Capture","openWhispr":{"status":"partial","details":"Records system audio locally — no bot joins"},"competitor":{"status":"full","details":"A bot auto-joins Zoom, Teams, and Meet calls"}},{"feature":"General Dictation","openWhispr":{"status":"full","details":"System-wide voice typing in any app"},"competitor":{"status":"none","details":"Meeting-focused; not a dictation tool"}},{"feature":"Transcription & Languages","openWhispr":{"status":"full","details":"Whisper + Parakeet, 100+ languages"},"competitor":{"status":"full","details":"Cloud transcription with AI search"}},{"feature":"CRM & Team Intelligence","openWhispr":{"status":"partial","details":"API/MCP access; basic team on Business"},"competitor":{"status":"full","details":"Salesforce/HubSpot sync, conversation analytics"}},{"feature":"Free Tier","openWhispr":{"status":"full","details":"Unlimited local dictation, free forever"},"competitor":{"status":"partial","details":"Free tier with limits — see their site"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","items":[{"title":"Local by default:","body":"Transcription runs on your machine. No bot in the participant list, no audio uploaded, nothing stored on a third-party server."},{"title":"Dictation everywhere:","body":"A system-wide hotkey types your voice into any app — email, docs, your editor, an AI chat. Fireflies is meeting-only."},{"title":"Open source:","body":"MIT licensed and auditable — you can verify exactly how your audio is handled."}]},"whyFireflies":{"title":"Why Fireflies.ai?","items":[{"title":"Automated call-joining:","body":"Its bot joins scheduled calls on its own and captures them without you lifting a finger."},{"title":"CRM & revenue intelligence:","body":"Deep Salesforce/HubSpot sync and conversation analytics built for sales teams."},{"title":"Shared team workspace:","body":"Searchable, collaborative transcripts across an organization out of the box."}]},"faqs":[{"question":"What is the best Fireflies.ai alternative?","answer":"It depends on what you need. If you want a private, open-source alternative for everyday voice-to-text and on-device meeting notes, OpenWhispr is a strong choice — it runs speech-to-text 100% locally using OpenAI Whisper and NVIDIA Parakeet, works system-wide via a push-to-talk hotkey, and never sends your audio to the cloud. If you specifically need an automated bot that joins calls with team analytics and CRM integrations, Fireflies.ai itself remains excellent at that."},{"question":"Is there a free Fireflies.ai alternative?","answer":"Yes. OpenWhispr is free and open source. Unlimited local dictation in 100+ languages works offline at no cost on macOS, Windows, and Linux. An optional Pro plan ($8/mo) adds unlimited cloud transcription, more meeting recording hours, mobile, and device sync — but you never need it for core local voice-to-text."},{"question":"Is there a local or private alternative to Fireflies.ai?","answer":"OpenWhispr is built for exactly this. Core transcription happens on your own machine — your audio never leaves the device and works fully offline — and because the app is open source, you can audit precisely how your data is handled. There is no bot in your calls and nothing is uploaded, which matters most for sensitive meetings like HR, legal, or anything with customer data."},{"question":"Does OpenWhispr join meetings with a bot like Fireflies?","answer":"No — and that is the point. Fireflies sends a bot that appears in your call and uploads the audio to its cloud. OpenWhispr instead captures the meeting's system audio directly on your device, so no bot joins the participant list and nothing is uploaded. You still disclose that you are recording, but the conversation stays entirely on your machine."},{"question":"How is OpenWhispr different from Fireflies.ai?","answer":"Fireflies.ai is a cloud meeting assistant focused on automated call-joining, conversation intelligence, and CRM sync for sales and revenue teams. OpenWhispr is a privacy-first voice-to-text tool plus on-device meeting notes: press a hotkey and speak into any app, or record and transcribe a meeting locally. They solve different problems — many people keep a bot tool for collaborative sales calls while using OpenWhispr for private, everyday dictation and sensitive meetings."}],"faq":{"title":"Frequently Asked Questions"},"cta":{"title":"A Fireflies.ai alternative that never uploads your audio","description":"Private, open-source voice-to-text and on-device meeting notes. Free on macOS, Windows, and Linux.","viewOnGitHub":"View on GitHub"}},"unsubscribe":{"metadata":{"title":"Unsubscribe — OpenWhispr","description":"Manage your OpenWhispr marketing email subscription.","keywords":["OpenWhispr unsubscribe","email preferences","marketing email"]},"breadcrumb":{"home":"Home","unsubscribe":"Unsubscribe"},"invalid":{"title":"This link isn't valid","description":"The unsubscribe link is missing or damaged. Try the link from a newer email, or ask us to remove you directly.","contact":"Email <email>support@openwhispr.com</email> and we'll take care of it."},"success":{"title":"You're unsubscribed","description":"<email>{emailAddress}</email> won't receive marketing email from OpenWhispr anymore. This took effect immediately.","serviceMessages":"You'll still get service messages — receipts, security alerts, and password resets — because those are part of your account."},"confirm":{"title":"Unsubscribe from marketing email?","description":"<email>{emailAddress}</email> will stop receiving product news and offers from OpenWhispr. Service messages about your account aren't affected.","button":"Unsubscribe"}},"legalDpa":{"metadata":{"title":"Data Processing Addendum | OpenWhispr","description":"OpenWhispr's Data Processing Addendum (Gizmo Labs Inc. as processor) — self-executing, with a no-AI-training commitment, SCC and UK Addendum transfer clauses, and a product subprocessor annex."},"breadcrumb":{"home":"Home","dpa":"Data Processing Addendum"},"authoritativeNotice":"This translation is provided for convenience only. If it differs from the English version, the English version governs.","header":{"title":"Data Processing Addendum","subtitle":"How Gizmo Labs Inc. (OpenWhispr) processes personal data on behalf of its customers. Customer content is never used to train AI models (Section 3).","effective":"Effective July 17, 2026 · Version 1.2, updated August 21, 2026"},"preamble":{"title":"About this Addendum","body":"$16"},"definitions":{"title":"1. Definitions","body":"Capitalised terms not defined here have the meaning given in Data Protection Law. \"Data Protection Law\" means all laws applicable to the processing of Personal Data under the Agreement, including the EU GDPR (Regulation (EU) 2016/679), the UK GDPR and Data Protection Act 2018, the Swiss FADP, and applicable US state privacy laws (including the CCPA/CPRA). \"SCCs\" means the Standard Contractual Clauses in Commission Implementing Decision (EU) 2021/914. \"UK Addendum\" means the ICO's International Data Transfer Addendum to the EU SCCs. \"Restricted Transfer\" means a transfer of Personal Data to a third country lacking an adequacy decision. \"Sub-processor\" means a third party engaged by OpenWhispr to process Personal Data on the Controller's behalf."},"scope":{"title":"2. Scope and roles","body":"$17"},"noAiTraining":{"title":"3. No AI training on Customer Data","body":"OpenWhispr will not use Customer Personal Data — including voice audio, transcriptions, notes, prompts and AI-agent conversations — to train, retrain or improve any artificial intelligence or machine learning model, whether OpenWhispr's own or a third party's, and does not permit its Sub-processors to do so. The model providers listed in Annex 3 process such content solely to provide the Services, with training disabled or opted out for OpenWhispr's accounts. This restriction survives termination of the Agreement."},"processorObligations":{"title":"4. Processor obligations","intro":"OpenWhispr will:","items":["process Personal Data only on documented instructions from the Customer (the Agreement, this DPA, and the Customer's configuration of the Services constitute such instructions), including for transfers, unless required by law to which OpenWhispr is subject — in which case OpenWhispr informs the Customer of that legal requirement before processing, unless legally prohibited;","ensure persons authorised to process Personal Data are bound by confidentiality obligations;","implement the technical and organisational security measures set out in Annex 2 and keep them under periodic review;","assist the Customer, taking into account the nature of processing, with data-subject requests, security, breach notification, data protection impact assessments and prior consultation (Sections 6–9);","at the Customer's choice, delete or return all Personal Data at the end of the Services and delete existing copies within 30 days, unless law to which OpenWhispr is subject requires retention;","make available information reasonably necessary to demonstrate compliance and allow and contribute to audits per Section 10; and","immediately inform the Customer if, in OpenWhispr's opinion, an instruction infringes Data Protection Law."]},"subprocessing":{"title":"5. Sub-processing","body":"The Customer grants OpenWhispr general written authorisation to engage Sub-processors. The current list is set out in Annex 3 and published on the OpenWhispr trust center at <trustLink>trust.openwhispr.com</trustLink> (Subprocessors section). OpenWhispr imposes data protection obligations on each Sub-processor equivalent to those in this DPA (including transfer obligations) and remains liable for their performance. OpenWhispr gives the Customer at least 30 days'' prior written notice of intended additions or replacements (by email to the Customer's account owner, and by updating the published list), during which the Customer may object on reasonable, documented data-protection grounds. If the parties cannot resolve an objection, the Customer may terminate the affected Services and receive a pro-rata refund of any prepaid fees for the terminated Services."},"security":{"title":"6. Security","body":"OpenWhispr maintains the measures in Annex 2, appropriate to the risk, taking account of the state of the art and the nature of the Personal Data. Voice audio processed through the cloud transcription path is processed in real time and not persisted by OpenWhispr after transcription."},"breach":{"title":"7. Personal data breach","body":"OpenWhispr notifies the Customer without undue delay, and in any event within 72 hours, of becoming aware of a Personal Data breach affecting the Customer's Personal Data, providing the information the Customer reasonably needs to meet its own notification obligations, and cooperates on remediation."},"dataSubjectRequests":{"title":"8. Data-subject requests","body":"Taking into account the nature of the processing, OpenWhispr assists the Customer by appropriate technical and organisational measures in responding to requests to exercise data-subject rights (access, rectification, erasure, restriction, portability, objection). End users can self-serve export and deletion in-product; requests received directly by OpenWhispr that identify the Customer's organisation are forwarded to the Customer without undue delay."},"dpias":{"title":"9. DPIAs and prior consultation","body":"OpenWhispr provides reasonable assistance with data protection impact assessments and prior consultation with supervisory authorities, to the extent the required information is available to OpenWhispr."},"audit":{"title":"10. Audit","body":"OpenWhispr makes available on request its current compliance documentation (SOC 2 Type 2 report for the period 1 May 2026 to 31 July 2026, available under NDA; GDPR and HIPAA attestations issued via its compliance platform, 2026; ISO/IEC 27001:2022 certification, certificate KCISM202607012) via its trust center, subject to confidentiality. Where Data Protection Law grants the Customer a mandatory audit right that cannot be satisfied by such reports, OpenWhispr allows an audit by the Customer or an independent auditor not more than once in any 12-month period, and additionally following a confirmed Personal Data breach affecting the Customer's Personal Data, in each case on at least 30 days' notice, subject to confidentiality and OpenWhispr's security policies. The Customer bears its own costs of any such audit. For Restricted Transfers, this Section applies without prejudice to the audit rights in the SCCs."},"customerObligations":{"title":"11. Customer obligations and warranties","body":"The Customer warrants that its instructions and its provision of Personal Data to OpenWhispr comply with Data Protection Law, including having a lawful basis, providing required notices to data subjects, and — where end users may dictate or process special categories of personal data (for example health information) — ensuring an appropriate legal basis and any required agreements are in place (see Section 14 for HIPAA)."},"transfers":{"title":"12. International transfers","intro":"This Section governs transfers of Personal Data originating from the EEA, the UK or Switzerland to a country outside those areas, including any Restricted Transfer. OpenWhispr and its Sub-processors process Personal Data primarily in the United States (per-Sub-processor locations: Annex 3). The parties rely on the first applicable mechanism below for each transfer:","mechanisms":["Adequacy. Where the destination benefits from an adequacy decision, the transfer proceeds on that basis.","Standard Contractual Clauses. For any Restricted Transfer from the EEA, the parties enter into the SCCs, Module Two (Controller → Processor) with the Customer as data exporter and OpenWhispr as data importer, which are incorporated into this DPA by reference and completed as follows: Clause 7 (docking) not included; Clause 9 Option 2 (general authorisation, 30 days' notice); Clause 11 optional redress body not included; Clause 13 / Annex I.C competent supervisory authority: the Irish Data Protection Commission; Clause 17 governing law: the law of Ireland; Clause 18 forum: the courts of Ireland; Annexes 1–3 of this DPA populate SCC Annexes I–III. Onward transfers to Sub-processors are governed by Clauses 8.8 and 9(b) of the Module Two SCCs (same-obligations flow-down).","UK and Swiss transfers. For UK Personal Data the parties enter into the UK Addendum to the EU SCCs, completed using the details in the Annexes; for Swiss Personal Data the SCCs apply as modified by the FADP."],"futureFramework":"Should OpenWhispr self-certify under the EU–US Data Privacy Framework in the future, transfers may additionally rely on that certification.","supplementaryMeasures":"Where a transfer impact assessment identifies that supplementary measures are required, the parties implement them, and OpenWhispr supports the Customer's assessment. OpenWhispr agrees to the obligations in the SCCs, including notifying the Customer of any inability to comply."},"governmentRequests":{"title":"13. Government access requests","body":"If OpenWhispr receives a legally binding request from a public authority for the Customer's Personal Data, it will, to the extent legally permitted: notify the Customer promptly; review the legality of the request and challenge it where there are reasonable grounds; and disclose only the minimum amount of data legally required. OpenWhispr keeps records of such requests for the Customer's review."},"hipaa":{"title":"14. HIPAA","body":"This DPA is not a Business Associate Agreement. Where the Customer is a Covered Entity or Business Associate under HIPAA, end users must not process protected health information through the cloud Services unless a separate Business Associate Agreement is in place between the parties — OpenWhispr offers a BAA to Customers on the Business plan and above, on request via support@openwhispr.com. Where executed, that BAA governs PHI in place of this DPA to the extent of any conflict."},"usStatePrivacy":{"title":"15. US state privacy laws","body":"To the extent the CCPA/CPRA or another US state privacy law applies, OpenWhispr acts as a \"service provider\"/\"processor\": it does not sell or share Personal Data, does not retain, use or disclose it other than to provide the Services or as permitted by law, and certifies that it understands and will comply with these restrictions."},"liability":{"title":"16. Liability and term","body":"Liability for data protection matters is as set out in the Agreement, subject to any caps therein. This DPA continues for as long as OpenWhispr processes Personal Data for the Customer."},"signatures":{"title":"Signatures (optional — for customers requiring an executed copy)","nameRole":"Name & role","signature":"Signature","date":"Date","customer":"For the Customer (Controller)","processor":"For Gizmo Labs Inc. (Processor)","executive":"Gabriel Stein, CEO"},"annex1":{"title":"Annex 1 — Details of processing","parties":"A. Parties (SCC Annex I.A). Data exporter: the Customer (controller) — name, address and contact person as set out in the Agreement or Order Form. Data importer: Gizmo Labs Inc. (processor), 221 W 9th Street, Wilmington, Delaware 19801, USA, contact: support@openwhispr.com; activities: provision of the Services described below.","descriptionTitle":"B. Description of processing (SCC Annex I.B).","field":"Field","detail":"Detail","rows":[{"field":"Subject-matter","detail":"Provision of the OpenWhispr voice-dictation and AI note-taking services (desktop applications and API)"},{"field":"Duration","detail":"Term of the Agreement, plus the post-termination deletion/return period under Section 4(e)"},{"field":"Nature & purpose","detail":"Real-time speech-to-text transcription; AI-assisted text generation, formatting and reasoning; note storage, search and sharing; account management; billing; transactional email"},{"field":"Categories of Personal Data","detail":"Account and identity data (name, email, authentication identifiers); voice audio (processed in real time in the cloud path, not persisted by OpenWhispr); transcribed text and notes (may contain any personal data the end user dictates); conversation history with AI features; usage and device metadata; billing data (handled by the payment Sub-processor)"},{"field":"Special category data","detail":"Possible incidentally: dictated content is user-directed free text and may contain health or other special-category data. Safeguards: encryption in transit and at rest, no audio persistence, local-only and BYOK modes, BAA required for PHI (Section 14)"},{"field":"Categories of data subjects","detail":"The Customer's end users (employees, contractors); persons mentioned in dictated content or notes"},{"field":"Frequency","detail":"Continuous, for the duration of the Services"}]},"annex2":{"title":"Annex 2 — Technical and organisational measures","intro":"Summary of OpenWhispr's security programme (current detail: OpenWhispr trust center — <trustLink>trust.openwhispr.com</trustLink>, Security Controls section):","bullets":["Encryption of Personal Data in transit (TLS 1.2+) and at rest.","Cloud voice audio processed in real time and not persisted by OpenWhispr after transcription (transcription/inference Sub-processors handle audio transiently under their agreements); local-only and bring-your-own-key modes route content around OpenWhispr's cloud entirely.","Access control on least-privilege with MFA and logging/monitoring across production systems; session and login monitoring with monthly review.","Vulnerability and patch management; secure development practices.","Resilience and backup with documented recovery objectives (RTO ≤ 4h / RPO ≤ 1h for the primary datastore) and contingency-plan testing on an annual cadence.","Personnel confidentiality obligations and security training.","Sub-processor governance: contractual flow-down (Section 5), public listing, and BAAs executed with the primary content-storing Sub-processors (database; cloud transcription/LLM), with coverage of the remaining HIPAA-path Sub-processors in progress.","Independently examined: SOC 2 Type 2 report issued for the period 1 May 2026 to 31 July 2026, covering Security, Confidentiality and Availability, available under NDA. Compliance attestations: GDPR and HIPAA (2026, via OpenWhispr's compliance platform). Information security management system certified to ISO/IEC 27001:2022 (certificate KCISM202607012, issued by KVQA Certification Services, accredited by the Standards Council of Canada; initial registration 27 July 2026, valid to 27 July 2027)."]},"annex3":{"title":"Annex 3 — Sub-processors","intro":"Authorised Sub-processors as of 2026-07-17 (live list: OpenWhispr trust center, <trustLink>trust.openwhispr.com</trustLink>, Subprocessors section). Legal entity names and processing locations verified 2026-07-17 against each vendor's terms/DPA. \"US (primary)\" means the vendor's terms name the United States as the primary processing location while permitting processing in other countries by their own sub-processors. \"Clause 9(b) flow-down\" means the same-obligations Sub-processor contract under Clauses 8.8 and 9(b) of the Module Two SCCs, where a Restricted Transfer applies.","headers":{"subprocessor":"Sub-processor","service":"Service","dataCategories":"Data categories","country":"Country","safeguard":"Onward-transfer safeguard"},"rows":[{"subprocessor":"Neon, LLC (an affiliate of Databricks, Inc.)","service":"Primary database and authentication store","dataCategories":"Account data, notes, transcribed text, conversation history","country":"US (AWS US region, fixed at project creation)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Vercel Inc.","service":"Application hosting and delivery","dataCategories":"All data in transit through the Services","country":"US (primary)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"OpenAI OpCo, LLC","service":"Cloud speech-to-text, language models, embeddings (training opted out)","dataCategories":"Voice audio (transient), prompts, text","country":"US (primary)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Parasail, Inc.","service":"Cloud AI inference (default reasoning; zero-data-retention inference)","dataCategories":"Prompts, text (in transit)","country":"US (primary)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Groq, Inc.","service":"Cloud AI inference and transcription failover","dataCategories":"Voice audio (transient), prompts, text (in transit)","country":"US (primary)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Embedchain, Inc. d/b/a Mem0","service":"AI memory for agent features","dataCategories":"Conversation content keyed to user ID","country":"US","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Exa Labs, Inc.","service":"Web search for agent features","dataCategories":"Search query text","country":"US (AWS)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Stripe, LLC","service":"Payment processing","dataCategories":"Billing and payment data","country":"US (primary)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Plus Five Five, Inc. (d/b/a Resend)","service":"Transactional email","dataCategories":"Email address, message content (incl. shared-note titles)","country":"US (AWS us-east-1)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Slack Technologies, LLC (Salesforce)","service":"Internal operations and support communications","dataCategories":"Support-related data (policy: no PHI)","country":"US (default residency)","safeguard":"Clause 9(b) flow-down"}],"footnote":"This annex lists the Sub-processors that process Customer Personal Data within the product Services. Google (OAuth sign-in and the optional Calendar integration) is not a Sub-processor: sign-in is provided by Google as an independent controller, and calendar data is cached only on the user's device. OpenWhispr's broader vendor transparency list (including marketing-site and operational vendors such as RB2B, lemlist and GitHub, for which Gizmo Labs acts as controller or which do not receive Customer Personal Data) is published on the OpenWhispr trust center."}},"blogBestNoteTakers":{"metadata":{"title":"Best AI Note Takers 2026 (Free & Private) | OpenWhispr","description":"Compare the best AI note takers of 2026 — Fathom, Otter, Granola, Fireflies, and OpenWhispr, the free, private, open-source option that runs locally.","keywords":["best ai note taker","best ai note takers","best free ai note taker","open source ai note taker","best ai note taker for privacy","ai note taker comparison","private ai note taker","best ai meeting notes"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"The Best AI Note Takers"},"schema":{"title":"The Best AI Note Takers in 2026 (Including Free and Open-Source Options)","description":"Compare the best AI note takers of 2026 — Fathom, Otter, Granola, Fireflies, and OpenWhispr, the free, private, open-source option that runs locally."},"faqs":[{"question":"What is the best AI note taker?","answer":"There is no single best one — it depends on your priority. For automated meeting summaries with the least setup, Fathom and Otter are excellent. For sales teams that need CRM sync and conversation intelligence, Fireflies is strong. For privacy — keeping audio on your own device — OpenWhispr is the standout because it transcribes locally, is open source, and never sends a bot into your call."},{"question":"What is the best free AI note taker?","answer":"Fathom has one of the most generous free cloud tiers for individuals. If your priority is a free tool that also keeps your data private, OpenWhispr is free and open source with unlimited local transcription — no per-month cap, no card, and nothing uploaded."},{"question":"Is there an open-source AI note taker?","answer":"Yes. OpenWhispr is MIT-licensed and fully open source. Its speech-to-text runs on your own device using OpenAI Whisper and NVIDIA Parakeet, so you can audit exactly how your audio is handled. Meetily is another open-source, local-first option worth a look."},{"question":"What is the best AI note taker for privacy?","answer":"The most private option is one that transcribes on your own device and never uploads your audio. OpenWhispr does this by default — no bot joins the call and the recording stays on your machine — which makes it the safest choice for sensitive meetings like HR, legal, or anything involving customer data. Most cloud note-takers, by contrast, upload and store your meetings on their servers."},{"question":"Do AI note takers need a bot to join the meeting?","answer":"Most do — Otter, Fireflies, and Fathom send a bot that appears in your participant list. A few don't: Granola captures your computer's audio without a bot, and OpenWhispr does the same while also keeping transcription entirely on your device."},{"question":"Which AI note taker works offline?","answer":"Cloud tools like Otter, Fireflies, and Fathom need an internet connection because transcription happens on their servers. OpenWhispr can run fully offline: its local models transcribe on your machine with no connection required, which is useful on flights, in secure facilities, or anywhere you simply don't want audio leaving the device."}],"sections":[{"id":"how-to-choose","title":"How to Choose"},{"id":"comparison","title":"Quick Comparison"},{"id":"picks","title":"The Picks"},{"id":"faq","title":"FAQ"}],"hero":{"title":"The Best AI Note Takers in 2026 (Including Free and Open-Source Options)","subtitle":"Automated meeting summaries, or private on-device notes? A fair, honest comparison of the top AI note takers — what each one is genuinely best at, and where it falls short.","date":"July 6, 2026","tag":"Comparison"},"opening":{"lead":"There is no single best AI note taker — the right one depends on whether you value automation, price, or privacy.","body":"For hands-off cloud summaries, Fathom and Otter are excellent. For sales teams, Fireflies. For keeping your audio on your own device, OpenWhispr. Below is an honest breakdown of each, including where they beat the others.","updated":"Last updated: July 6, 2026. OpenWhispr is on this list — we've tried to be fair about where other tools win."},"howToChoose":{"title":"How to Choose an AI Note Taker","intro":"Almost every AI note taker records a call, transcribes it, and summarizes the result. The real differences come down to three questions:","criteria":[{"lead":"Where does your audio go?","body":"Cloud tools upload and store your meeting on their servers. Local tools keep it on your device. For HR, legal, medical, or customer conversations, this is the deciding factor."},{"lead":"Does a bot join the call?","body":"Most send a visible bot into your participant list. A few capture audio without one, which is less intrusive and avoids awkward ‘who is that?’ moments."},{"lead":"What does the free tier actually cover?","body":"‘Free’ ranges from a few meetings a month to genuinely unlimited. Check the caps before you commit."}]},"comparison":{"title":"Quick Comparison","headers":{"tool":"Tool","bestFor":"Best for","runs":"Runs","bot":"Bot?","freeTier":"Free tier"}},"picks":{"title":"The Best AI Note Takers, One by One","keepInMind":"Keep in mind:","openWhisprLink":"See OpenWhispr","compareLink":"OpenWhispr vs {name}","tools":[{"name":"OpenWhispr","bestFor":"Privacy & open source","runs":"Local","bot":"No bot","freeTier":"Unlimited local, free","blurb":"Full disclosure: this is our tool — so here is the honest case. OpenWhispr transcribes on your own device using OpenAI Whisper and NVIDIA Parakeet, is open source (MIT), and captures meeting audio without sending a bot into the call. It doubles as a system-wide dictation tool for typing with your voice in any app.","pros":["Audio never leaves your device; nothing uploaded","Open source and auditable, on macOS, Windows and Linux","Unlimited local transcription, free forever"],"caveat":"It won't auto-join your scheduled calls with a bot — you start recording yourself — and the polished cloud summaries of a dedicated notetaker take a bit more setup.","href":"/meeting-notes-software"},{"name":"Fathom","bestFor":"Best free cloud notetaker","runs":"Cloud","bot":"Bot","freeTier":"Generous free plan","blurb":"Fathom is the easiest way to get automated meeting recaps, with one of the most generous free tiers among cloud notetakers. Its bot joins your call, records it, and produces clean summaries, highlights, and action items with almost no setup.","pros":["Excellent free tier for individuals","Very low-friction summaries and highlights","Solid Zoom, Teams, and Meet coverage"],"caveat":"It is cloud-based: a bot joins the call and your meeting is uploaded to Fathom's servers.","href":"/compare/fathom"},{"name":"Otter.ai","bestFor":"Live collaboration","runs":"Cloud","bot":"Bot","freeTier":"300 min/mo (historically)","blurb":"Otter is a mature meeting-transcription platform with strong live transcription, sharing, and integrations. It's a good fit for teams that want a shared, searchable record of calls and collaborative highlights.","pros":["Real-time live transcription","Team sharing and workspace features","Broad integrations"],"caveat":"Cloud-based with a bot, and the free tier is capped — heavy users hit limits quickly.","href":"/compare/otter"},{"name":"Fireflies.ai","bestFor":"Sales & CRM","runs":"Cloud","bot":"Bot","freeTier":"Free tier with limits","blurb":"Fireflies leans into revenue teams: its bot joins calls, transcribes them, and layers on conversation intelligence with deep Salesforce and HubSpot sync. If your notes need to feed a sales pipeline, it's a strong pick.","pros":["Conversation intelligence and analytics","Deep CRM integrations","Searchable team transcript library"],"caveat":"Cloud-based with a bot; the most useful features sit on paid, per-seat plans.","href":"/compare/fireflies"},{"name":"Granola","bestFor":"Bot-free cloud notepad","runs":"Cloud","bot":"No bot","freeTier":"Free tier","blurb":"Granola takes a nicer approach than most: it captures your computer's audio without a bot in the participant list and pairs it with a clean notepad and calendar sync. It's a favorite among Mac users who want tidy AI meeting notes.","pros":["No bot joins your call","Clean, fast note-taking UX","Calendar sync"],"caveat":"Processing still happens in the cloud, so your meeting audio leaves your device even though no bot appears.","href":"/compare/granola"},{"name":"Notta","bestFor":"Many languages","runs":"Cloud","bot":"Bot","freeTier":"Free tier with limits","blurb":"Notta is a cloud transcription and meeting tool with especially broad language support, plus file uploads and recording. A good option when your meetings span multiple languages.","pros":["Wide language coverage","Transcribe live calls or uploaded files","Simple exports"],"caveat":"Cloud-based; longer recordings and higher accuracy sit behind paid tiers.","href":null},{"name":"tl;dv","bestFor":"Recording & clips","runs":"Cloud","bot":"Bot","freeTier":"Free recording tier","blurb":"tl;dv is built around recording and sharing meeting moments — timestamped clips, highlights, and handoffs — with a usable free tier. Handy when you share meeting snippets with a team.","pros":["Clip and share key moments","Free recording tier","Zoom, Meet, and Teams support"],"caveat":"Cloud-based with a bot; your recordings live on tl;dv's servers.","href":null}]},"faq":{"title":"Frequently Asked Questions"},"cta":{"title":"Want AI notes that never leave your device?","body":"OpenWhispr is the free, open-source AI note taker that records and transcribes on your own machine — no bot, no cloud upload. Free on macOS, Windows, and Linux.","github":"View on GitHub"}},"blogZoomRecording":{"metadata":{"title":"How to Record a Zoom Meeting (Host or Not) | OpenWhispr","description":"Learn how to record a Zoom meeting as a host or participant — locally, in the cloud, or with no bot. Get a private transcript on your own device.","keywords":["how to record a zoom meeting","record zoom meeting","how to record a zoom meeting as a participant","record zoom meeting without permission","how to record a zoom meeting with audio","where do zoom recordings go","record zoom meeting locally","how to get a transcript of a zoom meeting","zoom meeting transcript","is it legal to record a zoom meeting","record zoom meeting on mac","record zoom meeting on windows"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"How to Record a Zoom Meeting"},"schema":{"title":"How to Record a Zoom Meeting (With or Without Being the Host)","description":"Learn how to record a Zoom meeting as a host or participant — locally, in the cloud, or with no bot. Get a private transcript on your own device."},"faqs":[{"question":"Can you record a Zoom meeting if you are not the host?","answer":"Yes, in two ways. The host can grant you recording permission during the call (Participants → hover your name → More → Allow to Record), which unlocks Zoom's built-in recorder for you. If they won't or can't, you can record your own screen and system audio with a separate recorder on your computer — that captures everything you can see and hear, no host permission required. Either way, tell the other participants you are recording."},{"question":"Where do Zoom recordings get saved?","answer":"Local recordings save to your own computer — by default in a Zoom folder under your Documents, as an MP4 video plus a separate M4A audio file. Cloud recordings (paid plans only) save to Zoom's servers, where you get a shareable link and, on eligible plans, an automatic transcript. If privacy matters, local recording keeps the file on your device."},{"question":"Is it legal to record a Zoom meeting?","answer":"It depends on where the participants are. Some places allow recording with just one person's consent, while others require everyone on the call to agree, and rules like the EU's GDPR treat a recorded voice as personal data. This is not legal advice, but the safe habit everywhere is the same: tell people you are recording and get their okay before you start."},{"question":"How do I record a Zoom meeting with the audio?","answer":"Zoom's built-in recorder captures the meeting audio automatically alongside the video. If you record with a separate screen recorder instead, make sure system audio (not just your microphone) is enabled, or you will end up with silent video. A local voice tool like OpenWhispr captures the system audio directly, so you always get the full conversation."},{"question":"Can you record a Zoom meeting without the host knowing?","answer":"When you use Zoom's built-in recorder, Zoom shows everyone a 'Recording' banner, so the host and participants are notified. A separate screen or audio recorder won't trigger that banner — but recording people secretly can be illegal depending on your location and is a trust risk regardless. Always disclose that you're recording; the point of recording a meeting is to have an accurate shared record, not to hide one."},{"question":"How do I get a transcript of a Zoom meeting?","answer":"On paid plans, Zoom can generate an audio transcript for cloud recordings automatically. For a local recording — or any Zoom plan — you can transcribe the saved audio file yourself. OpenWhispr turns your recording into an accurate, searchable transcript on your own device in 100+ languages, then lets AI summarize it into notes and action items."},{"question":"Can I record a Zoom meeting on my phone?","answer":"If you are the host on a paid plan, you can record to the cloud from the Zoom mobile app. As a participant, use your phone's built-in screen recorder (Control Center on iPhone, Screen Record on Android) and enable microphone/internal audio. Then transcribe the file afterward if you want notes."},{"question":"What is the most private way to record a Zoom meeting?","answer":"Record locally so the file never leaves your device, and avoid cloud note-taker bots that upload your meeting to a third party. OpenWhispr records the system audio and transcribes it entirely on your own machine — no bot joins the call, and nothing is uploaded — which is the safest option for sensitive meetings like HR, legal, or anything with customer data."}],"sections":[{"id":"host","title":"Record as the Host"},{"id":"participant","title":"If You're Not the Host"},{"id":"local","title":"Record Locally, No Bot"},{"id":"mobile","title":"Record on Mobile"},{"id":"transcript","title":"Get a Transcript"},{"id":"legal","title":"Consent & Legality"},{"id":"storage","title":"Where It's Stored"},{"id":"troubleshoot","title":"Troubleshooting"},{"id":"faq","title":"FAQ"}],"hero":{"title":"How to Record a Zoom Meeting (With or Without Being the Host)","subtitle":"Record a Zoom call with audio — as the host, as a participant, in the cloud or locally on your own device — then turn it into a searchable transcript. Plus how to stay on the right side of consent.","date":"July 6, 2026","tag":"Guide"},"opening":{"lead":"To record a Zoom meeting, click the Record button in the toolbar if you are the host or have been given permission.","body":"If you are a participant without permission, you have two options: ask the host to enable recording for you, or record your own screen and system audio with a separate tool on your computer — which needs no permission and captures the whole call. Whatever method you use, tell the other participants first.","scope":"This guide covers every path: recording as the host (local vs cloud), recording as a participant, recording privately on your own device without a bot, doing it on mobile, turning the recording into a transcript, the consent rules to know, and what to do when the Record button is greyed out.","updated":"Last updated: July 6, 2026."},"host":{"title":"Record as the Host (Zoom's Built-in Recorder)","intro":"If you started the meeting, recording takes one click. In the meeting toolbar, press Record. On a paid plan you'll be asked to choose between two destinations:","options":[{"title":"Record on this computer","body":"saves an MP4 video and an M4A audio file to your own machine when the meeting ends. Available on every plan, including free."},{"title":"Record to the cloud","body":"uploads the recording to Zoom's servers and gives you a shareable link plus, on eligible plans, an automatic transcript. Paid plans only."}],"outroPrefix":"Zoom shows every participant a “Recording” indicator the moment you start, so the call is always disclosed. When the meeting ends, a local recording is converted and dropped into a","outroSuffix":"folder inside your Documents by default."},"participant":{"title":"If You Are Not the Host","intro":"Participants can't use Zoom's recorder unless the host allows it. You have two routes:","options":[{"title":"Ask for permission.","body":"The host opens the Participants panel, hovers your name, clicks More, and chooses Allow to Record. Zoom's Record button then appears for you."},{"title":"Record your screen and system audio.","body":"A separate recorder on your computer captures whatever is on screen and coming through your speakers — no host permission required. This is the reliable fallback when the host can't or won't enable recording."}],"warningPrefix":"The catch with option two: a plain screen recorder often captures only your microphone, leaving the other voices out. Make sure you enable","systemAudio":"system audio","warningSuffix":", or use a tool built to record the meeting audio directly — covered next."},"local":{"title":"Record Locally, With No Bot and No Cloud","cloudPrefix":"Most cloud note-takers — Otter, Fireflies, Fathom — join your call with a","bot":"bot","cloudSuffix":"and upload the audio to their servers, which puts a stranger's bot in the participant list and your conversation on someone else's cloud. Even bot-free options still send your meeting off-device to process it — a non-starter for sensitive meetings.","approachPrefix":"takes the opposite approach. It captures the meeting's system audio directly on your machine —","noBot":"no bot joins the call","approachSuffix":", and the audio never leaves your device. It works as a participant, on any Zoom plan, on macOS, Windows and Linux. You still disclose that you're recording, but nothing is uploaded and no third party ever sees the conversation.","bullets":["Records the full call audio, not just your microphone.","No bot in the participant list, and no cloud upload.","Transcribes and summarizes on-device, in 100+ languages."],"github":"View on GitHub"},"mobile":{"title":"Record a Zoom Meeting on Your Phone","paragraph1":"If you host on a paid plan, the Zoom mobile app can record straight to the cloud — tap More then Record to the Cloud. As a participant, the simplest route is your phone's own screen recorder: the iPhone Control Center recorder or Android's Screen Record. Turn on microphone/internal audio so the call is captured, not just muted video.","paragraph2":"Phone recordings are just video files, so you can transcribe them the same way as any other recording once the call ends."},"transcript":{"title":"Turn the Recording Into a Transcript and Notes","paragraph1":"A recording is only useful if you can read and search it. On paid plans, Zoom can auto-generate a transcript for cloud recordings. For a local recording — or on the free plan — you transcribe the saved file yourself.","paragraph2Prefix":"Drop the MP4 or M4A into","paragraph2Suffix":"and it produces an accurate, timestamped transcript on your own device, then lets AI summarize the call into decisions, action items and follow-ups. Because it runs locally, even a confidential recording never gets uploaded to transcribe it."},"legal":{"title":"Consent and Legality","paragraph1":"Whether you can legally record depends on where the participants are. Some places allow recording with just one person's consent; others require everyone on the call to agree, and rules like the EU's GDPR treat a recorded voice as personal data. This isn't legal advice, but the safe habit everywhere is the same: say you're recording and get a clear okay before you start.","paragraph2":"Zoom's built-in recorder discloses automatically with its on-screen banner. A separate screen or local recorder won't show that banner, so the responsibility to disclose is on you. Recording people secretly is a legal and trust risk — and unnecessary, since the whole point is a shared record everyone can rely on."},"storage":{"title":"Where Your Recording Ends Up","localLabel":"Local recordings","localBody":"stay on your computer — an MP4 plus an M4A audio file, in a Zoom folder under Documents by default. Nothing leaves your machine unless you share it.","cloudLabel":"Cloud recordings","cloudBody":"live on Zoom's servers with a shareable link, subject to your plan's storage limits and your organization's retention settings. And","botLabel":"bot-based note-takers","botBody":"upload the whole meeting to a third party, where some providers may use it to train their models. For HR, legal, medical, or anything with customer data, recording locally and transcribing on-device keeps the conversation entirely under your control."},"troubleshoot":{"title":"Troubleshooting: Record Button Missing or Greyed Out","items":[{"title":"You're not the host and lack permission.","body":"Ask the host to allow recording, or fall back to a local screen/audio recorder."},{"title":"Recording is disabled at the account level.","body":"On managed or company accounts, an admin may have turned recording off in the Zoom web settings — you'll need them to enable it."},{"title":"Cloud recording is missing.","body":"Cloud recording requires a paid plan; the free plan supports local recording only."},{"title":"The recording has no sound.","body":"Your screen recorder is capturing only the mic. Enable system audio, or record with a tool built to capture the call audio."}]},"faq":{"title":"Frequently Asked Questions"},"closing":{"title":"Record and transcribe your meetings privately","body":"OpenWhispr captures your Zoom call audio and turns it into notes on your own device — no bot in the call, nothing uploaded. Free on macOS, Windows and Linux.","github":"View on GitHub"}},"blogStreamingStt":{"metadata":{"title":"Streaming Speech-to-Text on CPU: Local & Live | OpenWhispr","description":"How OpenWhispr runs NVIDIA Nemotron streaming ASR on your CPU — cache-aware streaming over a local WebSocket, live partial results, no GPU, no cloud.","keywords":["streaming speech to text","local streaming asr","nemotron asr","cache aware streaming asr","sherpa onnx","real time transcription local","live dictation preview","streaming asr electron","on device speech recognition"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"Streaming Speech-to-Text on Your CPU"},"schema":{"title":"Streaming Speech-to-Text on Your CPU: Shipping NVIDIA Nemotron in a Desktop App","description":"How OpenWhispr runs NVIDIA Nemotron streaming ASR on CPU — cache-aware streaming over a local WebSocket, live partials, no GPU, no cloud."},"faqs":[{"question":"What is streaming speech-to-text?","answer":"Streaming (or 'online') speech-to-text transcribes audio while it's still being spoken, emitting partial results within a fixed latency budget — typically 80 milliseconds to about a second. Batch ('offline') models like Whisper instead wait for the recording to finish and process it all at once. Streaming is what makes live captions, voice agents, and see-it-as-you-speak dictation previews possible."},{"question":"Can I run streaming speech-to-text locally without a GPU?","answer":"Yes. NVIDIA's Nemotron streaming models are 600M parameters, ship as INT8-quantized ONNX files around 650MB, and run in real time on a modern CPU via the sherpa-onnx runtime — a native binary with no Python or PyTorch. OpenWhispr ships this setup on macOS, Windows, and Linux."},{"question":"What is cache-aware streaming ASR?","answer":"It's an architecture (NVIDIA's cache-aware FastConformer) where the model keeps its internal encoder state between audio chunks. Each new chunk is processed once, using cached context from everything said so far, instead of re-encoding overlapping windows of audio. That's what makes low-latency streaming cheap enough to run on a laptop CPU."},{"question":"Why does the live preview text change after it appears?","answer":"Streaming models revise their hypothesis as more audio arrives — a word boundary or a homophone can only be resolved by later context, so partial results are replaced in place. This is by design. When you stop, OpenWhispr flushes the model's tail and commits the settled streamed text as the transcript; if the flush is interrupted or truncated, it automatically re-decodes the full recording instead, so an unstable stream never becomes your pasted text."},{"question":"Is streaming transcription less accurate than batch?","answer":"Slightly, and honestly so: a streaming model can't see future audio. Nemotron's English streaming model averages 6.93% word error rate at 1.12-second chunks on the Open ASR Leaderboard datasets — remarkably close to batch models, but still behind the 5.91% of NVIDIA's best English batch model. Choosing a streaming model means accepting that small gap in exchange for live text and an instant transcript at stop; if maximum accuracy matters more, pick an offline Parakeet model instead."},{"question":"Which languages does streaming support in OpenWhispr?","answer":"Two Nemotron models are available: English-only, and Nemotron 3.5 with 15 transcription-ready languages (English, Spanish, French, Italian, Portuguese, Dutch, German, Turkish, Russian, Arabic, Hindi, Japanese, Korean, Vietnamese, Ukrainian) with automatic language detection. Offline models like Parakeet and Whisper cover more languages but preview via buffered chunks instead of a live stream."}],"sections":[{"id":"the-problem","title":"The Problem: Chunked Previews"},{"id":"cache-aware","title":"What Cache-Aware Streaming Is"},{"id":"architecture","title":"The Architecture"},{"id":"gotchas","title":"What Broke Along the Way"},{"id":"single-pass","title":"Committing the Stream"},{"id":"footprint","title":"Performance & Footprint"},{"id":"limitations","title":"Honest Limitations"},{"id":"model-choice","title":"Choosing a Model"},{"id":"faq","title":"FAQ"}],"hero":{"title":"Streaming Speech-to-Text on Your CPU: Shipping NVIDIA Nemotron in a Desktop App","subtitle":"How OpenWhispr shows live text as you speak — a cache-aware streaming model over a local WebSocket, no GPU and no cloud — and the engineering it took to make that reliable.","date":"July 18, 2026","tag":"Technical"},"opening":{"lead":"Streaming speech-to-text transcribes audio while you're still talking, instead of waiting for the recording to end. As of version 1.7.6, OpenWhispr does this entirely on-device: NVIDIA's Nemotron streaming models (600M parameters, INT8-quantized) run on your CPU through a local sherpa-onnx WebSocket server, the dictation preview updates live as partial results arrive, and the streamed text is committed as your transcript the moment you stop.","body":"No GPU, no Python environment, no audio leaving your machine.","guidePrefix":"This is the engineering write-up: why our old live preview felt laggy, what cache-aware streaming actually is in plain terms, how the pipeline fits together inside an Electron app, the version-pinning bug that mis-decoded everything, and how we went from re-decoding every recording to committing the stream itself. If you want the model-versus-model comparison instead, that's in","comparisonLink":"Parakeet vs Whisper vs Nemotron","guideSuffix":".","updated":"Last updated July 18, 2026. Implementation details reference the streaming transcription feature shipped in OpenWhispr 1.7.6; everything described here is inspectable in the open-source repository."},"factCheck":{"title":"Fact-Check Snapshot (Primary Sources)","items":[{"lead":"Nemotron's cache-aware FastConformer processes only new audio while reusing cached encoder context,","body":"with runtime-configurable chunk sizes from 80ms to 1.12s.","sources":[{"href":"https://huggingface.co/nvidia/nemotron-speech-streaming-en-0.6b","label":"Nemotron Streaming EN model card"},{"href":"https://huggingface.co/blog/nvidia/nemotron-speech-asr-scaling-voice-agents","label":"NVIDIA on cache-aware streaming"}]},{"lead":"The English model averages 6.93% WER at 1.12s chunks while streaming;","body":"the multilingual 3.5 model covers 40 language-locales with automatic detection.","sources":[{"href":"https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b","label":"Nemotron 3.5 ASR model card"}]},{"lead":"All inference runs through sherpa-onnx, a native ONNX runtime with no Python dependency,","body":"using its online (streaming) WebSocket server. We pin and bundle version 1.13.4.","sources":[{"href":"https://github.com/k2-fsa/sherpa-onnx","label":"sherpa-onnx on GitHub"}]},{"lead":"The models are 632MB (English) and 650MB (multilingual) on disk as INT8 ONNX,","body":"downloaded once and cached locally.","sources":[{"href":"https://github.com/OpenWhispr/openwhispr","label":"OpenWhispr model registry (open source)"},{"href":"https://openwhispr.com/models","label":"OpenWhispr models page"}]}]},"problem":{"title":"The Problem: Live Preview Built on a Batch Model","paragraphs":[{"body":"OpenWhispr has always had a live transcription preview — a small window that shows what the model heard while you dictate. Before 1.7.6 it worked the only way you can make a batch model feel live: buffer about 1.5 seconds of microphone audio, run a full transcription on that chunk, append the text, repeat."},{"body":"That approach has three problems, and every one of them is structural. First, latency: your words can't appear faster than the buffer fills, so the preview always trails you by a second and a half plus inference time. Second, boundaries: a word that straddles two chunks gets cut in half, and neither chunk can fix it, because chunks are transcribed independently. Third, context: chunk five doesn't know what you said in chunk four — so “their” versus “there” is a coin flip the model re-tosses every 1.5 seconds."},{"body":"You can shrink the chunks to feel faster, but then accuracy collapses — less context per chunk, more boundaries. You can overlap chunks to fix boundaries, but then you're transcribing the same audio twice. What you actually want is a model that was trained to transcribe a stream. That's what Nemotron is."}]},"cacheAware":{"title":"What Cache-Aware Streaming Actually Means","paragraphs":[{"lead":"A cache-aware streaming model keeps its internal state between chunks.","body":"When the next slice of audio arrives, the encoder doesn't re-read everything you said — it processes only the new frames, attending to a cache of the context it already computed. It's the same trick that makes chatbots fast (KV caching), applied to a speech encoder."},{"body":"Two consequences fall out of this design. Compute stops scaling with recording length — each chunk costs the same whether you're ten seconds or ten minutes in, which is what makes CPU-only streaming viable. And context carries forward — the model resolves “their/there” using everything said so far, and revises earlier words in its running hypothesis when later audio disambiguates them. That revision is why streaming text visibly “settles” a beat after you speak: it's not a bug, it's the model changing its mind with better evidence."},{"body":"The latency budget is a knob, not a constant: Nemotron exposes chunk sizes from 80ms (fastest, least context per step) up to 1.12s (highest accuracy — 6.93% average WER on the leaderboard datasets, within about a point of the best batch models). The training recipe is NVIDIA's cache-aware FastConformer with a transducer decoder; their"}],"engineeringPostLink":"engineering post","engineeringPostSuffix":"covers the theory in depth."},"architecture":{"title":"The Architecture Inside OpenWhispr","paragraphs":[{"body":"The pipeline is deliberately boring. An AudioWorklet captures microphone PCM at 16kHz. Frames are converted to the float32 wire format and pushed over a WebSocket to a sherpa-onnx online server — a native binary OpenWhispr spawns locally and talks to over 127.0.0.1, one persistent connection per recording. The server runs the Nemotron model with INT8 weights and answers with JSON partial results, which the preview window renders by replacing its text in place rather than appending."},{"body":"Why a separate server process instead of in-process inference? Isolation. Native inference runtimes can crash — a bad allocation in an ONNX kernel shouldn't take the whole desktop app down with it. A sidecar process means a crashed runtime is a reconnect, not a lost dictation. It also keeps native code out of Electron's main process entirely: the app talks protocol, the sidecar does math."},{"body":"The same pattern runs our"}],"diarizationLink":"local speaker diarization","diarizationSuffix":"— one sherpa-onnx toolchain, multiple jobs, all on-device."},"gotchas":{"title":"What Broke Along the Way","items":[{"label":"The runtime version mattered more than we expected.","body":"Nemotron models require sherpa-onnx 1.13.4 or newer for accurate decoding — on the older runtime we'd bundled for Parakeet, they didn't fail loudly, they transcribed badly. The fix was mundane (upgrade and re-pin the bundled binaries on every platform), but the lesson generalizes: with model runtimes, “it runs” and “it's correct” are separate tests, and only the second one counts."},{"label":"Streams fail, so the old path stayed.","body":"A port can be taken, a model file can be missing, a server can die mid-recording. If the persistent stream can't start, the preview falls back to the buffered-chunk path automatically — slower, but never blank. Offline models (Parakeet, Whisper) keep the chunked preview by design; each model in our registry declares its runtime as online or offline, and the app picks the right preview path per model."},{"label":"Partials needed different rendering.","body":"The chunked path appends text; a streaming hypothesis revises itself, so the UI replaces the whole preview per partial. Append a revising hypothesis and you get stuttering duplicates — the one-line version of a real UX bug we fixed early."}]},"singlePass":{"title":"Committing the Stream: From Two Passes to One","paragraphs":[{"body":"Our first version treated streaming as a preview, not the product: when you stopped dictating, the app re-decoded the complete recording from scratch and pasted that. Safe — the final pass sees full context — but it meant every dictation paid for two decodes, and the text you'd already watched settle on screen made you wait for a second opinion that almost always agreed with it."},{"body":"So in the same release we made the stream the source of truth. The dictation now runs over the persistent stream whether or not the preview window is open, and when you stop, the app flushes the model's tail and commits the streamed text directly — no second decode. End-of-dictation latency drops to one tail flush, and per-dictation CPU is roughly halved."},{"body":"The safety net didn't go away — it became the fallback. The stop flush is truncation-aware: it extends its deadline while results are still arriving, and if anything is off — a dropped connection, a truncated result — the app discards the stream and runs the classic record-then-transcribe pass over the full recording instead. You only ever get the fast path when it's also the correct one."}]},"footprint":{"title":"Performance & Footprint","items":[{"label":"Models:","body":"Nemotron Speech Streaming EN 0.6B (632MB) and Nemotron 3.5 ASR Streaming 0.6B (650MB, 15 transcription-ready languages, auto-detect) — INT8 ONNX, downloaded once, cached locally."},{"label":"Runtime:","body":"sherpa-onnx 1.13.4, a self-contained native binary. No Python, no PyTorch, no CUDA required."},{"label":"Hardware:","body":"real-time streaming on a modern laptop CPU — Apple Silicon and recent x86 both comfortably keep up with speech."},{"label":"Feel:","body":"partial text lands well under a second behind your voice; at stop, one tail flush commits the transcript — no second decode."}],"privacyPrefix":"Nothing about this path touches the internet. The app talks to the server over 127.0.0.1 (on Windows, a scoped firewall rule additionally closes the port to the network), the model files live in your local cache, and the audio never leaves the machine — the same privacy posture as the rest of","privacyLink":"our local transcription stack","privacySuffix":"."},"limitations":{"title":"Honest Limitations","items":[{"label":"Streaming trades a little accuracy for immediacy.","body":"~6.9% streaming versus 5.91% for NVIDIA's best English batch model on the same benchmarks — and since the committed transcript now comes from the stream, that's the trade you're accepting. If you only care about the end result, an offline Parakeet model is still the accuracy pick."},{"label":"Language coverage is narrower than Whisper.","body":"15 transcription-ready languages in the multilingual model versus Whisper's 99. Streaming in Mandarin or Thai isn't here yet."},{"label":"Partial text moves.","body":"The hypothesis revises itself as context arrives. We think watching it settle is a feature; if you find it distracting, the preview can be turned off and dictation works exactly as before."},{"label":"It's another 650MB on disk","body":"if you already have an offline model — the price of a second engine tuned for a different job."}]},"modelChoice":{"title":"Choosing a Model in OpenWhispr","wantHeader":"You want","pickHeader":"Pick","rows":[{"want":"Live English preview while dictating","pick":"Nemotron Speech Streaming EN 0.6B"},{"want":"Live preview in Spanish, Japanese, Arabic, Hindi…","pick":"Nemotron 3.5 ASR Streaming 0.6B"},{"want":"Maximum English accuracy, don't care about live text","pick":"Parakeet Unified EN 0.6B"},{"want":"A language outside the NVIDIA set","pick":"Whisper (turbo or large)"}],"catalogPrefix":"All of these are a dropdown in Settings — the full catalog with sizes and trade-offs is on the","catalogLink":"models page","catalogSuffix":". OpenWhispr is free, open source, and available for macOS, Windows, and Linux.","githubCta":"Star on GitHub"},"faq":{"title":"Frequently Asked Questions"}},"blogParakeetNemotron":{"metadata":{"title":"Parakeet vs Whisper vs Nemotron: Best Local STT 2026 | OpenWhispr","description":"Honest comparison of the top local speech-to-text models: accuracy (WER), speed, languages, and streaming. Real numbers from primary sources.","keywords":["parakeet vs whisper","nemotron asr","best local speech to text","whisper alternative","nvidia parakeet","local speech to text models","open source speech to text","best speech to text model 2026","nemotron vs whisper","offline speech recognition"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"Parakeet vs Whisper vs Nemotron"},"schema":{"title":"Parakeet vs Whisper vs Nemotron: The Best Local Speech-to-Text in 2026","description":"Honest comparison of the top local speech-to-text models: accuracy (WER), speed, languages, and streaming — with numbers from primary sources."},"faqs":[{"question":"What is the best local speech-to-text model in 2026?","answer":"For English and major European languages, NVIDIA Parakeet TDT 0.6B v3 offers the best overall balance: it averages 6.34% word error rate on the Open ASR Leaderboard benchmarks — ahead of Whisper large-v3's roughly 7.4% — at a quarter of the size and with dramatically higher throughput on CPU. If you need live text while you speak, NVIDIA Nemotron's streaming models are the best choice. If you need a language the NVIDIA models don't cover, use Whisper."},{"question":"Is Parakeet more accurate than Whisper?","answer":"On the Open ASR Leaderboard's English benchmarks, yes — Parakeet TDT 0.6B v3 averages 6.34% WER versus about 7.4% for Whisper large-v3, and the English-only Parakeet Unified model reaches 5.91%. The honest caveat: a ~1-point WER difference is hard to notice in everyday dictation. Parakeet's clearer wins are speed, efficiency, and the fact that it doesn't hallucinate text during silence, a known Whisper failure mode."},{"question":"Do I need a GPU to run these models locally?","answer":"No. All three run on a modern CPU. Parakeet and Nemotron use INT8-quantized ONNX models (roughly 630–680MB on disk) designed for CPU inference. Whisper runs on CPU via whisper.cpp, though its large 3GB model benefits from GPU acceleration — OpenWhispr supports Metal on Apple Silicon, CUDA on NVIDIA, and Vulkan on AMD/Intel GPUs, with automatic CPU fallback."},{"question":"What languages does Nemotron ASR support?","answer":"NVIDIA's Nemotron 3.5 ASR model covers 40 language-locales in a single 600M-parameter model with automatic language detection. In OpenWhispr, the 15 transcription-ready languages are enabled: English, Spanish, French, Italian, Portuguese, Dutch, German, Turkish, Russian, Arabic, Hindi, Japanese, Korean, Vietnamese, and Ukrainian. The English-only Nemotron streaming model is also available."},{"question":"Is Whisper obsolete now?","answer":"No. Whisper still covers 99 languages — far more than Parakeet's 25 or Nemotron's 40 locales — and it remains the safest choice for Asian languages outside Japanese, Korean, and Vietnamese, for low-resource languages, and for heavily mixed-language audio. It's also the most battle-tested model with the largest ecosystem. What has changed is that Whisper no longer holds the accuracy or speed crown for English."},{"question":"Which of these models does OpenWhispr support?","answer":"All of them. OpenWhispr ships six Whisper sizes via whisper.cpp, both Parakeet models and both Nemotron streaming models via sherpa-onnx, and lets you switch between them in Settings. Everything runs on your device — audio never leaves your machine — and the app is free and open source."}],"sections":[{"id":"contenders","title":"The Three Contenders"},{"id":"accuracy","title":"Accuracy"},{"id":"speed","title":"Speed & Efficiency"},{"id":"languages","title":"Language Coverage"},{"id":"streaming","title":"Streaming: The Real Differentiator"},{"id":"hardware","title":"Hardware Requirements"},{"id":"which-to-choose","title":"Which One Should You Use?"},{"id":"try-them","title":"Try All Three"},{"id":"faq","title":"FAQ"}],"hero":{"title":"Parakeet vs Whisper vs Nemotron: The Best Local Speech-to-Text in 2026","subtitle":"Three open models can now turn speech into text entirely on your own computer. Here's how they actually compare on accuracy, speed, languages, and streaming — with honest numbers from primary sources.","date":"July 18, 2026","tag":"Comparison"},"opening":{"lead":"For English and major European languages, NVIDIA Parakeet TDT 0.6B v3 is the best local speech-to-text model in 2026: it beats Whisper large-v3 on measured accuracy at a quarter of the size, and runs dramatically faster on a plain CPU. If you want text that appears live while you speak, NVIDIA's new Nemotron streaming models are the first local models built for exactly that. And if you speak one of the many languages the NVIDIA models don't cover, OpenAI's Whisper — with 99 languages — is still the one to use.","paragraph2":"The surprise of 2026 is that Whisper no longer holds the open-model accuracy crown it wore for three years. But “which model is best” has become a genuinely three-way question, because the models are good at different things: Parakeet optimizes accuracy-per-watt, Nemotron optimizes latency, and Whisper optimizes breadth. This article compares them with numbers from the models' own cards and the public leaderboard — no vibes, and no pretending the differences are bigger than they are.","disclosurePrefix":"Last updated July 18, 2026. We build","disclosureSuffix":", an open-source dictation app that ships all three model families, so we benchmark and support each of them in production. That's also our bias — we don't sell any one of these models, and the honest trade-offs below are the same ones we explain to our own users."},"factCheck":{"title":"Fact-Check Snapshot (Primary Sources)","items":[{"lead":"Parakeet TDT 0.6B v3 averages 6.34% word error rate across the eight Open ASR Leaderboard benchmarks, with a throughput of 3,332× real time.","body":"600M parameters, 25 languages, CC-BY-4.0 license.","sources":[{"href":"https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3","label":"Parakeet TDT v3 model card"},{"href":"https://huggingface.co/spaces/hf-audio/open_asr_leaderboard","label":"Open ASR Leaderboard"}]},{"lead":"Whisper large-v3 averages roughly 7.4% on the same leaderboard and supports 99 languages","body":"with 1.55B parameters under an MIT license.","sources":[{"href":"https://github.com/openai/whisper","label":"OpenAI Whisper repository"},{"href":"https://huggingface.co/openai/whisper-large-v3","label":"Whisper large-v3 model card"}]},{"lead":"Nemotron Speech Streaming EN 0.6B averages 6.93% WER while streaming","body":"(1.12-second chunks), reaches 2.32% on LibriSpeech test-clean, and was trained on 530,000 hours of audio.","sources":[{"href":"https://huggingface.co/nvidia/nemotron-speech-streaming-en-0.6b","label":"Nemotron Streaming EN model card"}]},{"lead":"Nemotron 3.5 ASR covers 40 language-locales in one 600M streaming model","body":"with automatic language detection, released June 4, 2026 under the OpenMDW license.","sources":[{"href":"https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b","label":"Nemotron 3.5 ASR model card"}]},{"lead":"All disk sizes and in-app behavior referenced below are verifiable","body":"in OpenWhispr's open-source model registry.","sources":[{"href":"https://github.com/OpenWhispr/openwhispr","label":"OpenWhispr on GitHub"},{"href":"https://openwhispr.com/models","label":"OpenWhispr models page"}]}]},"contenders":{"title":"The Three Contenders","intro":"All three are free, open-weight models you can download and run offline. They differ in architecture, and the architecture explains almost every practical difference you'll notice.","items":[{"title":"OpenAI Whisper (2022)","body":"The model that made open speech recognition mainstream. An encoder-decoder transformer trained on 680,000 hours of audio, available in six sizes from 75MB (tiny) to 3GB (large-v3), covering 99 languages. It generates text one token at a time, like a language model — which is also why it's the slowest of the three. MIT licensed."},{"title":"NVIDIA Parakeet (2024–2025)","body":"A 600M-parameter transducer (TDT) model built for batch transcription. Instead of generating tokens autoregressively, it emits text in a single pass over the audio — much faster, and it doesn't invent text during silence. Two variants matter: the multilingual v3 (25 languages, 680MB as INT8 ONNX) and the English-only Unified model (631MB), which currently posts state-of-the-art English accuracy. CC-BY-4.0 licensed."},{"title":"NVIDIA Nemotron ASR (2026)","body":"The newest family, built for streaming: a cache-aware FastConformer transducer that transcribes audio as it arrives instead of waiting for the recording to end. The English model shipped in January 2026; the 40-locale multilingual 3.5 model followed in June. Both are 600M parameters (632–650MB as INT8 ONNX) and openly licensed (OpenMDW)."}]},"accuracy":{"title":"Accuracy: Parakeet Wins, But the Gap Is Smaller Than It Looks","definitionPrefix":"Speech models are measured by","definitionTerm":"word error rate (WER)","definitionSuffix":" — the percentage of words the model gets wrong. Lower is better. The standard reference is the Hugging Face Open ASR Leaderboard, which averages WER across eight diverse English datasets (meetings, earnings calls, TED talks, audiobooks, and more), so a single easy dataset can't flatter a model.","analysis":"Read the chart honestly: every model here is production-grade, and the whole spread is about 1.5 points. In day-to-day dictation, a 6.3% and a 7.4% model feel similar on clean audio — the differences show up on hard audio (accents, noise, jargon). What makes the NVIDIA numbers notable is what you pay for them: Parakeet reaches its accuracy at 600M parameters versus Whisper's 1.55B, and Nemotron posts its 6.93% while streaming, without seeing the future audio at all.","calloutTitle":"Whisper's known weakness: hallucination on silence","calloutBody":"Because Whisper's decoder works like a language model, it can generate fluent, entirely invented sentences during silence or noise — a failure mode documented across the ecosystem. Transducer models like Parakeet and Nemotron are structurally resistant to this: no audio evidence, no tokens. For dictation, where recordings start and end with silence, that matters more than a benchmark point."},"speed":{"title":"Speed & Efficiency: Not Even Close","paragraph1":"On the leaderboard's hardware, Parakeet TDT v3 transcribes at 3,332× real time — an hour of audio in about a second. Whisper large-v3 posts well under a tenth of that. The reason is architectural, and it's worth understanding in plain terms: Whisper writes its transcript one token at a time, each step waiting on the last, while a transducer reads the audio once and emits text as it goes. No loop, no per-word round trips.","paragraph2Prefix":"On your laptop the absolute numbers are smaller but the ratio holds: a dictated paragraph that takes Whisper large several seconds to process lands near-instantly with Parakeet — on CPU, with no GPU involved. Independent benchmarks also report the NVIDIA transducers using roughly","energyLink":"8–10× less energy per hour of audio","paragraph2Suffix":"than Whisper at comparable quality — relevant if you dictate all day on battery.","paragraph3Prefix":"Whisper's answer to this is turbo (1.6GB), a distilled large-v3 that trades a little accuracy for roughly 8× Whisper-large speed. It's the right Whisper for dictation — but it narrows the gap rather than closing it. For the full Whisper size-by-size breakdown, see our","guideLink":"Whisper model sizes guide","paragraph3Suffix":"."},"languages":{"title":"Language Coverage: Whisper's Moat","headers":{"model":"Model","languages":"Languages","notes":"Notes"},"rows":[{"model":"Whisper large-v3","languages":"99","notes":"Broadest coverage of any open model, including low-resource languages"},{"model":"Nemotron 3.5 ASR","languages":"40 locales","notes":"Auto language detection; 15 transcription-ready languages including Japanese, Korean, Arabic, and Hindi"},{"model":"Parakeet TDT v3","languages":"25","notes":"European languages plus English and Russian; auto detection"},{"model":"Parakeet Unified / Nemotron EN","languages":"1","notes":"English-only, in exchange for the best accuracy in their class"}],"rule":"The rule of thumb is simple. English or a major European language: any of the three, so pick on speed. Japanese, Korean, Vietnamese, Arabic, Hindi: Nemotron 3.5 now covers them with streaming — a real first for local models. Mandarin, Thai, Indonesian, Swahili, or heavy code-switching between languages: Whisper is still the only serious option, and that's unlikely to change soon."},"streaming":{"title":"Streaming: The Real Differentiator","lead":"Streaming is where Nemotron isn't just incrementally better — it's doing something the other two structurally can't.","paragraph1":"Whisper processes audio in 30-second windows and re-encodes everything it sees, so “live” Whisper is really the same batch model run over and over on overlapping audio — expensive, laggy, and jittery at the seams. Nemotron's cache-aware architecture was trained for the streaming case: it processes only the newest chunk of audio, carries its internal state forward, and emits partial text within a configurable latency budget of 80 milliseconds to 1.12 seconds.","paragraph2Prefix":"The efficiency difference is dramatic at scale — NVIDIA reports","sourceLink":"17× more concurrent streams","paragraph2Middle":"than its previous-generation streaming model on the same GPU — but the part you actually feel on a laptop is simpler: text appears as you speak, from a model running on your own CPU. We wrote a full engineering breakdown of what it took to ship that in a desktop app in","deepDiveLink":"our streaming ASR deep dive","paragraph2Suffix":".","paragraph3":"If you don't care about live text — you dictate, stop, and want the final transcript — streaming buys you nothing, and Parakeet's batch models are the better pick. Streaming exists for live captions, voice agents, and the see-it-as-you-say-it dictation preview."},"hardware":{"title":"Hardware: All Three Run on a Normal Laptop","paragraph1":"None of these models needs a GPU. The NVIDIA models ship as INT8-quantized ONNX files (631–680MB) that run comfortably on any recent CPU — Apple Silicon or x86 — via the sherpa-onnx runtime, a native binary with no Python and no PyTorch. Whisper runs everywhere through whisper.cpp; its small and turbo sizes are fine on CPU, while large-v3 (3GB, ~10GB RAM) genuinely benefits from acceleration.","paragraph2":"If you do have a GPU, use it for Whisper: Metal is built in on Apple Silicon, and OpenWhispr offers one-click CUDA (NVIDIA) and Vulkan (AMD and Intel) runtimes with automatic CPU fallback if the GPU runtime fails. The NVIDIA transducers are fast enough on CPU that GPU acceleration is beside the point for dictation."},"decision":{"title":"Which One Should You Use?","headers":{"situation":"Your situation","use":"Use this"},"rows":[{"situation":"English dictation, want the best accuracy","use":"Parakeet Unified EN 0.6B"},{"situation":"European languages, fast batch transcription","use":"Parakeet TDT 0.6B v3"},{"situation":"Live text as you speak (English)","use":"Nemotron Speech Streaming EN"},{"situation":"Live text in Spanish, Japanese, Korean, Arabic, Hindi…","use":"Nemotron 3.5 ASR Streaming"},{"situation":"A language the NVIDIA models don't cover","use":"Whisper large-v3 or turbo"},{"situation":"Old or low-RAM hardware, quick notes","use":"Whisper tiny or base"}],"privacyPrefix":"And if privacy is the reason you're going local in the first place, all three qualify equally: the audio is processed on your machine and never uploaded. We've written more about that trade-off in","privacyLink":"local vs cloud transcription","privacySuffix":"."},"cta":{"title":"Try All Three in Two Minutes","paragraphPrefix":"The most honest way to choose is to dictate the same paragraph into each model. OpenWhispr ships every model in this article — six Whisper sizes, both Parakeets, both Nemotrons — behind a single settings dropdown, free and open source, on macOS, Windows, and Linux. Download a model, press the hotkey, talk. Our","modelsLink":"models page","paragraphSuffix":"lists everything with sizes and trade-offs.","github":"Star on GitHub"},"faqSection":{"title":"Frequently Asked Questions"}},"blogHipaaNoteTakers":{"metadata":{"title":"HIPAA-Compliant AI Note Takers & Dictation (2026) | OpenWhispr","description":"Which AI note takers sign a BAA, and which quietly exclude the AI features from it. A plain comparison for 2026, plus where OpenWhispr stands on SOC 2 Type II and ISO 27001.","keywords":["hipaa compliant ai note taker","hipaa compliant note taking app","hipaa compliant ai scribe","hipaa compliant transcription","hipaa compliant dictation software","hipaa compliant speech to text","is otter ai hipaa compliant","hipaa compliant ai note taker for therapists","best ai meeting note taker soc 2 hipaa compliant","hipaa compliant meeting transcription"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"HIPAA-Compliant AI Note Takers"},"schema":{"title":"HIPAA-Compliant AI Note Takers and Dictation Software in 2026","description":"Which AI note takers sign a Business Associate Agreement, which exclude their AI features from it, what to ask a vendor, and how local transcription changes the analysis."},"hero":{"title":"HIPAA-Compliant AI Note Takers: What a BAA Actually Covers","subtitle":"Most tools that say “HIPAA compliant” will sign a BAA, then leave the AI features out of it. Here's how to check before patient data reaches a vendor, and where we stand on SOC 2 Type II, ISO 27001, and HIPAA.","date":"August 11, 2026","tag":"Compliance"},"sections":[{"id":"where-we-stand","title":"Where we stand: SOC 2, ISO 27001, HIPAA"},{"id":"what-hipaa-requires","title":"What HIPAA actually requires"},{"id":"ai-exclusion","title":"The trap: when the BAA excludes the AI"},{"id":"comparison","title":"Which tools sign a BAA"},{"id":"checklist","title":"Five questions to ask a vendor"},{"id":"local-processing","title":"Why local processing changes the math"},{"id":"frameworks","title":"SOC 2 and ISO 27001 are not HIPAA"},{"id":"faq","title":"Frequently asked questions"}],"opening":{"lead":"There is no such thing as a HIPAA-certified AI note taker.","body":"No agency certifies software as HIPAA compliant. When a vendor's site says “HIPAA certified,” that line came from their marketing team, not an auditor. The question that actually matters is narrower and much easier to check: will they sign a Business Associate Agreement, and does it cover the feature you want to point at patient data?","thesis":"That second half is where teams get caught. Several large platforms will sign a BAA for their core product and leave the AI features out of it. Which are, of course, the features you came for.","updated":"Last updated August 2026. Vendor terms move quickly, so confirm current coverage with the vendor before you send them anything."},"whereWeStand":{"title":"Where we stand: SOC 2, ISO 27001, HIPAA","disclosure":"We build OpenWhispr, and it is one of the tools in the table below. That is a conflict of interest, so here is our position up front — read the rest of the page with it in mind.","body":"We are a small team. We went through SOC 2 Type II, ISO 27001, and a HIPAA readiness assessment earlier than a company our size usually would, because “trust us” is not an answer when someone asks where their patient audio goes. Concretely, this is what exists today:","items":[{"lead":"SOC 2 Type II and ISO 27001.","body":"Both assessed by an independent firm, with 66 controls monitored continuously and 13 subprocessors listed publicly. Current status is in our Trust Center."},{"lead":"HIPAA, with a BAA.","body":"Available on request from the Business plan up. We are not “HIPAA certified,” because no such certificate exists, and you should be skeptical of anyone who claims otherwise."},{"lead":"A third-party penetration test.","body":"Run by an external firm. We share the report with customers on request."},{"lead":"A way to report problems.","body":"security@openwhispr.com, acknowledged within 48 hours. Publishing an address only helps if someone answers it."},{"lead":"Local mode, which matters more than all of the above.","body":"Pick a local Whisper model and your audio is transcribed on your own machine and never sent to us. There is nothing for a BAA to cover, because we never receive the recording."}],"limitsLabel":"What this does not mean.","limitsBody":"An audit is a floor, not a guarantee. It says a company has controls and follows them; it says nothing about whether a particular feature fits your workflow. OpenWhispr dictates into any EHR field through a global hotkey, so Epic, Cerner, athenahealth, and web charts all work with no integration. What it does not do is write back through an API: there is no FHIR or HL7 integration, no awareness of which patient's chart is open, and no coded or structured data. It is not a clinically validated ambient scribe either. If you need notes filed against a specific encounter with coding suggestions, a dedicated scribe product will fit better than we do.","trustLine":"The SOC 2 report, ISO 27001 certificate, subprocessor list, and monitored controls are published in our <trustLink>Trust Center</trustLink>, and how we handle data is written out on the <securityLink>security page</securityLink>."},"whatHipaaRequires":{"title":"What HIPAA actually requires of an AI note taker","intro":"If a vendor creates, receives, stores, or transmits protected health information (PHI) for you, HIPAA calls them a business associate. That triggers a short and specific list:","items":[{"lead":"A signed BAA.","body":"This is the contract that makes the vendor legally accountable for the PHI you hand over. Without one, sending them PHI is itself a violation, however good their encryption is."},{"lead":"Security Rule safeguards.","body":"Access controls, audit logging, encryption in transit and at rest, and a written incident response process."},{"lead":"Breach notification.","body":"They have to tell you about a breach of unsecured PHI without unreasonable delay, so you can meet your own deadlines."},{"lead":"Minimum necessary use.","body":"They can only use PHI the way the BAA allows. That rules out secondary uses, and training AI models on your patient data is the one to watch."},{"lead":"Subcontractor flow-down.","body":"If your audio gets routed to a third-party model provider, that provider is a subcontractor and needs its own BAA. Ask who is in the chain."}],"mythLabel":"The certification myth.","mythBody":"HIPAA has no certifying body, so “HIPAA certified” is not a claim you can verify. The honest phrasings are “HIPAA compliant,” “we support HIPAA compliance,” or “we sign BAAs.” A vendor advertising a certificate is either shortening “we passed a third-party readiness assessment” or has misunderstood the rule. Either way, ask for the BAA, not the badge."},"aiExclusion":{"title":"The trap: a BAA that excludes the AI","intro":"The common failure in 2026 is not a vendor refusing to sign a BAA. It is a vendor signing one that quietly does not cover the AI. Two of the biggest platforms work this way:","zoomLabel":"Zoom.","zoomBody":"Zoom offers a BAA on Zoom for Healthcare and eligible Business and Enterprise plans, covering Meetings, Webinars, Phone, Team Chat, and Rooms. But turning on HIPAA mode disables a set of AI Companion features, and only some of what remains, such as smart recording and meeting summary, sits inside the BAA. The compliant setup is the less capable one.","zoomSource":"Zoom's HIPAA documentation","notionLabel":"Notion.","notionBody":"Notion will sign a BAA, but only for Enterprise customers above a minimum seat count, and Notion AI is excluded from that coverage on every plan. You can store PHI in a Notion database. You cannot point Notion AI at it.","notionSource":"Notion's HIPAA help page","takeaway":"The pattern repeats because of how these products get built. Compliance reviews the platform, the AI ships later on a different legal footing, and the carve-out never comes up in the sales call. So ask for the BAA's scope in writing, feature by feature, rather than a yes or no about the company."},"comparison":{"title":"Which AI note takers will sign a BAA","intro":"Status as of August 2026, from each vendor's published documentation. The last column is the one that matters most and the one pricing pages answer least often.","headers":{"tool":"Tool","baa":"Signs a BAA?","plan":"Required plan","aiCovered":"AI features covered?"},"vendors":[{"name":"OpenWhispr","baa":"Yes, on request","plan":"Business plan and above","aiCovered":"Yes, or no PHI leaves the device","note":"A BAA is available on request from the Business plan up, a lower bar than the Enterprise-only gate Otter and Fathom apply. In local mode the question does not arise: transcription runs on your machine with open-source Whisper models and the audio is never sent to us. Gizmo Labs, which builds OpenWhispr, holds SOC 2 Type II and ISO 27001 and runs HIPAA and GDPR programs.","href":null},{"name":"Otter.ai","baa":"Yes","plan":"Enterprise only","aiCovered":"Yes, under the BAA","note":"Otter announced HIPAA compliance in July 2025 after an independent assessment and signs a BAA covering the use and disclosure of PHI. Only Enterprise qualifies. Basic, Pro, and Business customers cannot get one, so the compliant tier is a real jump in price.","href":"/compare/otter"},{"name":"Fathom","baa":"Yes","plan":"Enterprise only","aiCovered":"Yes, under the BAA","note":"HIPAA compliance backed by a BAA at the Enterprise level. Fathom's model providers — Anthropic, OpenAI, and Google — are contractually barred from training on Fathom user data, which is the subcontractor question answered properly.","href":"/compare/fathom"},{"name":"Fireflies.ai","baa":"Yes, on request","plan":"Healthcare tier","aiCovered":"Yes, under the BAA","note":"Fireflies launched a healthcare offering in September 2025 with a HIPAA-compliant tier. The BAA is not automatic: default plans are not covered and you have to ask. Fireflies says it has signed BAAs with its own vendors barring ePHI training and storage.","href":"/compare/fireflies"},{"name":"Zoom AI Companion","baa":"Partial","plan":"Healthcare / eligible plans","aiCovered":"Only some features","note":"The BAA covers Zoom's core products, but HIPAA mode switches off a range of AI Companion features and only a subset stays in scope. Check feature by feature before relying on it for clinical documentation.","href":null},{"name":"Granola","baa":"No","plan":"—","aiCovered":"No","note":"Granola does not currently offer HIPAA compliance or a BAA. Good product, but not an option for PHI until that changes.","href":"/compare/granola"},{"name":"Notion AI","baa":"Partial","plan":"Enterprise, 100+ seats","aiCovered":"No, AI is excluded","note":"Notion Enterprise signs a BAA covering pages, databases, wikis, and file uploads, but Notion AI sits outside it on every plan. Fine for storing documentation, unusable as an AI scribe for PHI.","href":null}],"compareLink":"Compare OpenWhispr and {name}","disclaimer":"Compiled from vendor documentation in August 2026. Plan names, tiers, and BAA scope change often, so treat this as a starting point for your own diligence rather than a substitute for it. Confirm current terms with the vendor and your compliance counsel."},"checklist":{"title":"Five questions to ask before you record PHI","intro":"Most vendor security pages answer the easy questions and skip the ones that decide whether you can use the product at all. Put these five in writing:","questions":[{"lead":"Will you sign a BAA, and on which plan?","body":"If the answer is “Enterprise only,” get the price before you get attached. The compliant tier is often several times the number you were first quoted."},{"lead":"Does the BAA name the AI features?","body":"Not “is the product covered” but whether transcription, summarization, and any AI chat or agent features are each in scope. This is exactly where Zoom and Notion diverge from their own marketing."},{"lead":"Which subprocessors receive PHI?","body":"If audio goes to OpenAI, Anthropic, Google, or a hosted speech service, each is a subcontractor that needs its own BAA and a no-training commitment. Ask for the list, not a reassurance."},{"lead":"Is our data used for training or product improvement?","body":"“We don't sell your data” is a different and much weaker promise. You want a contractual ban on training and on humans reviewing your recordings."},{"lead":"How long is audio kept, and can we set it to zero?","body":"Real-time processing with immediate deletion is not the same as ninety-day retention with a delete button. Retention you cannot configure is retention you will have to defend in an audit."}]},"localProcessing":{"title":"Why local processing changes the math","intro":"Every question above exists because PHI is leaving your organization and landing on someone else's servers. When the audio never leaves the device, most of the analysis stops applying:","points":["No disclosure to a business associate, because the vendor never receives the PHI.","No subprocessor chain to audit, because no third-party model provider is in the path.","No retention policy to negotiate, because there is no server-side copy.","No exposure to a vendor changing its policy later, because the model runs on hardware you control.","Works with no network access, which matters in facilities where clinical machines are deliberately offline."],"tradeoff":"The trade-off is real and worth stating. Local models need a reasonably modern machine, the largest and most accurate ones are slower on older hardware, and you give up the convenience of a meeting bot that joins calls on your behalf. For a lot of clinical dictation that is a good trade. For high-volume ambient scribing across a large practice, it may not be.","closing":"This is the design OpenWhispr is built around, and it is the reason we think the local option matters more than any certificate we hold."},"frameworks":{"title":"SOC 2 and ISO 27001 are not HIPAA","intro":"These three get bundled together in procurement conversations, and they answer genuinely different questions. Knowing which is which saves time on both sides of a security review:","items":[{"lead":"SOC 2 is a report, not a certificate.","body":"An independent auditor produces it against the Trust Services Criteria. Type I looks at how controls are designed on one date. Type II tests whether they actually worked over a period, usually three to twelve months, and it is the one worth asking for. Anyone calling themselves “SOC 2 certified” is describing a document that does not work that way."},{"lead":"ISO 27001 really is a certification.","body":"An accredited body certifies an organization's information security management system against the standard, on a three-year cycle with surveillance audits. Ask for the certificate number, who issued it, and above all the scope statement, which tells you what is actually covered."},{"lead":"HIPAA is a law, with no certificate at all.","body":"Compliance is an ongoing obligation you demonstrate through safeguards, policies, risk analysis, and signed BAAs. SOC 2 and ISO 27001 are good evidence that a vendor runs a serious security program. Neither one replaces a BAA."}]},"faq":{"title":"Frequently asked questions","legalNote":"This is general information about how vendors are set up, not legal advice. Your own HIPAA obligations depend on your organization, your risk analysis, and your counsel."},"faqs":[{"question":"Is Otter.ai HIPAA compliant?","answer":"Yes, but only on the Enterprise plan. Otter announced HIPAA compliance in July 2025 after an independent assessment and will sign a BAA covering the use and disclosure of PHI. Basic, Pro, and Business customers cannot get a BAA, so Otter is not HIPAA compliant on those tiers. Contact Otter's sales team to start the process."},{"question":"Is Fathom HIPAA compliant?","answer":"Yes, at the Enterprise level, backed by a BAA. Fathom's AI providers — Anthropic, OpenAI, and Google — are contractually barred from training on Fathom user data. Lower tiers are not covered, so confirm your plan qualifies before recording anything clinical."},{"question":"Is Notion HIPAA compliant?","answer":"Partly. Notion signs a BAA for Enterprise customers above a minimum seat count, covering pages, databases, wikis, and file uploads. Notion AI is excluded from that coverage on every plan, so you cannot use Notion's AI features on protected health information even as an Enterprise customer with a signed BAA."},{"question":"Is Zoom AI Companion HIPAA compliant?","answer":"Only in part. Zoom signs a BAA for Zoom for Healthcare and eligible Business and Enterprise plans, but enabling HIPAA mode disables several AI Companion features. Some, such as smart recording and meeting summary, are in scope; others are not. Confirm the specific features you need are covered before using it for clinical documentation."},{"question":"Is Fireflies HIPAA compliant?","answer":"Yes, on its healthcare tier. Fireflies launched a HIPAA-compliant offering in September 2025, but you have to request the BAA explicitly, as default plans are not covered. Fireflies also reports signing BAAs with its own subprocessors barring them from training on or storing ePHI."},{"question":"Is Granola HIPAA compliant?","answer":"No. Granola does not currently offer HIPAA compliance or a Business Associate Agreement, so it should not be used to record or process protected health information."},{"question":"What is the best HIPAA-compliant AI note taker for therapists?","answer":"It depends on whether you need a meeting bot. Solo and small practices are often best served by on-device dictation, where session audio never leaves the laptop and no BAA is needed for that step. If you need a bot that joins telehealth calls, Otter Enterprise, Fathom Enterprise, and the Fireflies healthcare tier all sign BAAs. Whichever you pick, check the retention setting and the subprocessor list, since therapy notes are among the most sensitive records HIPAA covers."},{"question":"Do I need a BAA if transcription runs locally on my computer?","answer":"Generally no, for that step. If audio is transcribed entirely on-device and never sent to the vendor, the vendor is not creating, receiving, storing, or transmitting PHI for you, so it is not acting as a business associate in that flow. You still need your own safeguards: device encryption, access controls, and a policy for where the resulting notes live. Confirm the analysis with your compliance counsel."},{"question":"Does SOC 2 or ISO 27001 mean a tool is HIPAA compliant?","answer":"No. SOC 2 is an auditor's report on security controls and ISO 27001 certifies an information security management system. Both are meaningful evidence of a serious security program, but neither creates the contractual relationship HIPAA requires. Only a signed BAA does that."},{"question":"What is the difference between HIPAA compliant and HIPAA certified?","answer":"“HIPAA certified” is not a real status, because no government agency or accreditation body certifies software against HIPAA. Vendors using the phrase usually mean they completed a third-party readiness assessment. What you can verify is whether they sign a BAA, what that BAA covers, and which safeguards they have documented."},{"question":"Is there a free HIPAA-compliant AI note taker?","answer":"OpenWhispr's local mode is free and open source, and because the audio never leaves your device there is no disclosure to cover. If you want our cloud instead, a BAA is available from the Business plan up. Most cloud-based note takers tie their BAA to Enterprise pricing, so a free cloud tier is almost never usable with PHI. Check the plan requirement before assuming the free option works."},{"question":"Can AI note takers train on patient data?","answer":"Not under a properly drafted BAA. The agreement should explicitly bar the vendor and every subprocessor from using PHI for model training or product improvement. Ask for that language in writing. A privacy-policy line about not selling data is a much weaker promise than a contractual ban on training."}],"cta":{"title":"Dictation and meeting notes that stay on your machine","body":"OpenWhispr runs open-source Whisper models locally, so your audio never has to leave your device. A BAA is available from the Business plan up if you do need our cloud. Free, open source, and available for macOS, Windows, and Linux.","github":"Star on GitHub"}},"useCaseLawyers":{"breadcrumb":{"home":"Home","useCases":"Use Cases","lawyers":"For Lawyers"},"metadata":{"title":"Free Dictation Software For Lawyers & Attorneys - OpenWhispr","description":"Dictation software for lawyers. OpenWhispr offers Cloud transcription or local processing for sensitive and offline work.","keywords":["legal dictation software","lawyer dictation","attorney dictation software","legal transcription","dictation for law firms","offline legal dictation"]},"schema":{"name":"OpenWhispr","applicationCategory":"Legal dictation software","operatingSystem":"macOS, Windows, Linux, iOS","description":"Open-source dictation software for lawyers. Choose Cloud transcription without a local model download, or local processing for sensitive and offline work.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Dictation software for lawyers & attorneys","title":"Speak the letter. It lands in Word. No typist ever hears the file.","body":"Dictate a letter, a file note or a memo and the clean text drops in wherever your cursor already is: Word, Clio, an email. Choose Cloud for fast transcription or local for highly sensitive and offline work.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Man speaking beside an open laptop on a sunny terrace"}},"trustBar":{"label":"Dictate into any app, including:","names":["Clio","iManage","NetDocuments","MyCase","Smokeball","Microsoft Word"]},"pain":{"label":"What you're choosing between","title":"Saying the letter takes two minutes. Getting it typed is the rest of the afternoon.","subtitle":"Every route from your voice to the file should suit the matter, not keep you typing.","cards":[{"title":"A stranger typing up privileged notes.","body":"With a local model and no Cloud fallback, the dictation is transcribed on the machine you spoke into. No outsourced transcription service handles it."},{"title":"Sensitive matters need a local option.","body":"Choose local models when the matter is highly sensitive or you will work without signal. Your device processes the recording."},{"title":"An hour of typing is an hour you can't bill.","body":"Typing up your own dictation is the slowest way to get a letter out, and you do it after the thinking is already done. Dictate it while the matter is fresh and you review and send instead of type."},{"title":"A tool that only fits one desk.","body":"OpenWhispr runs on macOS, Windows, and Linux. Start with the Free plan, then upgrade when the workflow calls for it."}]},"features":{"label":"From your voice to the file","title":"From your voice to a first draft.","subtitle":"Get the letter out while the matter is fresh. Choose the transcription path that fits the work.","items":[{"image":{"label":"Woman speaking while using a laptop at a desk"},"title":"Cloud transcription without a local model download","body":"Speak a letter, a memo or a file note into your mic. Cloud gives you fast transcription without a local model download, then immediately discards the transcription audio."},{"image":{"label":"Man reviewing a printed document beside his laptop"},"title":"Letters and memos, drafted for you","body":"Tell it what you want and use local AI models for device-only drafting. You read every word before it leaves your desk."},{"image":{"label":"Man speaking and pointing toward a laptop at a table"},"title":"Dictate into Word without an add-in","body":"One hotkey, and the clean punctuated text pastes itself where you're already typing: Word, an email, Clio, your document system. There's nothing for IT to install into any of them."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Privileged audio sent to us","value":"None","footnote":"with a local transcription model the recording never leaves your computer"},{"label":"Cost per lawyer","value":"$$0","footnote":"unlimited local dictation in the Free plan"},{"label":"Processing options","value":"2","footnote":"Cloud transcription or local models"}]},"howItWorks":{"label":"Download to first letter","title":"Three steps from download to draft.","subtitle":"Choose Cloud transcription or download a local model.","image":{"label":"Man standing beside a laptop in a glass-walled office"},"steps":[{"title":"Download it free.","body":"Mac, Windows or Linux. Setup offers OpenWhispr Cloud or local models. Pick local for privileged work and it downloads them once."},{"title":"Dictate the letter.","body":"Speak it the way you'd say it to your assistant, and it comes back as text."},{"title":"Read it, paste it, delete the audio.","body":"Edit the draft, drop it into Word or your document system, and delete the recording. Every one of those is yours to control."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"Why firms keep it","title":"Ten reasons your IT or security team can check for themselves.","subtitle":"Choosing dictation software for lawyers is a confidentiality decision before it's anything else. Most of these are facts about the code, so IT or security can check them instead of taking them from us.","items":[{"title":"The recording stays on your machine","body":"Choose a local transcription model for device-only processing. An optional Cloud fallback can send audio if you turn it on."},{"title":"Free for the whole firm","body":"Unlimited local dictation at $0 for every desk, including your assistant and paralegals."},{"title":"Nothing goes to a typing pool","body":"With local transcription and no Cloud fallback, no outsourced transcription service handles the recording."},{"title":"Works with the wifi off","body":"A courthouse hallway, a deposition, a client's office, a flight. Local models need no signal, so nothing stalls mid-sentence."},{"title":"Dictate into Word or any case system","body":"One system-wide hotkey drops clean text where your cursor already is: Word, Clio, iManage, an email. On a Mac it uses the accessibility permission you grant at setup."},{"title":"Letters and memos, not just transcripts","body":"Ask for a case memo and a local AI model can draft it on your machine. Local AI models are included on every plan."},{"title":"You control storage and deletion","body":"Recordings are files on your machine. Audio is deleted after 30 days out of the box, and Settings moves that or switches it off."},{"title":"Nothing to take on faith","body":"Your IT or security team can read every line of the app on GitHub before a partner installs it."},{"title":"Party names and citations, spelled your way","body":"Put party names, case names, statutes, and citations in the custom dictionary to guide transcription and cleanup."},{"title":"100+ languages","body":"Dictate in 100+ languages."}]},"faq":{"label":"Before you install it","title":"Eight questions law firms ask before they install anything.","subtitle":"Straight answers on confidentiality, cost, accuracy, and what IT will say. More questions? Ask in Discord.","items":[{"question":"Does the audio ever leave my device?","answer":"With local transcription, audio is transcribed on your device. Turn off the optional Cloud fallback when every attempt must stay local. Cloud sends audio only to transcribe it, then immediately discards it."},{"question":"Is this consistent with my duty of confidentiality?","answer":"Confidentiality is your professional duty and no tool discharges it for you. Local processing is the deliberate choice when a highly sensitive matter calls for device-only work. Cloud transcription discards audio immediately after transcription. This is not legal advice, and the rules of your bar or law society still apply."},{"question":"Does dictation software for lawyers get case names and citations right?","answer":"Add party names, case names, Latin, statutes, and citations to the custom dictionary. OpenWhispr passes that list to transcription and cleanup to guide the text."},{"question":"Will IT let me install it on the firm laptop?","answer":"It is a desktop app, not an add-in for Word or your document system. On a Mac, grant accessibility permission for auto-paste. Without it, copy the text from your clipboard instead."},{"question":"Will it work in a courthouse or on a flight with no signal?","answer":"Yes, once the models are on your machine. Local models run entirely offline after the first download. Dictate in the courthouse hallway, after a deposition, or on the flight home, then file the note when you are back in range."},{"question":"Does it draft the letter, or just transcribe?","answer":"Both. Dictation turns speech into text, then an AI model can shape a client letter or case memo. Transcription and AI cleanup are separate settings, so choose a local AI model when the draft must stay on-device."},{"question":"Do I need a separate legal dictation tool?","answer":"OpenWhispr runs on macOS, Windows, and Linux. Try the Free plan with a local model, then move to Cloud transcription when it fits your workflow."},{"question":"Is it actually free?","answer":"Free includes unlimited local dictation and local AI models, plus 2,000 Cloud words a week. Pro adds unlimited Cloud transcription, 20 hours of meeting recordings, sync, mobile, and API/MCP access. Business adds unlimited meeting recordings, Agent mode, and chat over data."}]},"reviews":{"label":"What people tell us","title":"Built by people who wouldn't hand their audio to a stranger either.","subtitle":"From our Discord, GitHub and inboxes. None of them is a lawyer, so read them for the tool, not the practice.","score":{"value":"0","label":"Audio stored on our servers with local transcription","body":"With a local transcription model the audio never leaves your device, so there's nothing on our side to lose, breach or hand over."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free to start.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for lawyers","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Lawyers who want local dictation and 2,000 Cloud words a week"}]},{"parts":[{"text":"Lawyers who want unlimited Cloud transcription plus mobile, sync, and API/MCP access"}]},{"parts":[{"text":"Firms documenting at volume with Agent mode and chat over data"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited local dictation plus 2,000 Cloud words a week"}]},{"parts":[{"text":"Unlimited Cloud transcription plus mobile, sync, and API/MCP access"}]},{"parts":[{"text":"Unlimited meeting recordings plus Agent mode and chat over data"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","Enterprise adds SSO, audit logs, retention controls, and dedicated support."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Why we built it","title":"A client tells you things they'd tell nobody else.","paragraphs":["OpenWhispr started with a simple objection: confidential dictation should have a local option.","We build in public. The app is open source, so your firm can inspect how it works before anyone installs it.","Community support is available in Discord. Paid plans add email or priority support."],"image":{"label":"One of OpenWhispr's founders"}},"finalCta":{"tagline":"$$0 a seat · open source · no add-in","title":"Speak the next letter while the matter is fresh.","body":"Download OpenWhispr, choose Cloud transcription or local processing, and dictate the next letter.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free on macOS, Windows, and Linux. Paid plans add Cloud transcription, mobile, and workflow tools."}}},"useCaseLegalDocuments":{"breadcrumb":{"home":"Home","useCases":"Use Cases","legalDocuments":"Legal Documents"},"metadata":{"title":"Free AI Dictation For Legal Documents - OpenWhispr","description":"AI dictation for legal documents: talk a clause, a brief section or a motion through and OpenWhispr's agent drafts it. Free forever, and it runs locally.","keywords":["ai dictation for legal documents","dictate a contract","brief and motion dictation software","discovery response drafting tool","deposition summary dictation","paralegal and legal ops dictation"]},"schema":{"name":"OpenWhispr","applicationCategory":"Legal document drafting software","operatingSystem":"macOS, Windows, Linux","description":"OpenWhispr is an open-source dictation app for the people who produce legal documents: paralegals, legal assistants and lawyers. They dictate contracts, briefs, motions, discovery responses and deposition summaries, and the AI agent redrafts them into structured text, either in place in the document or in the app's assistant panel.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"AI dictation software for legal documents","title":"Stop typing first drafts. Talk the section through.","body":"Talk a contract clause, a motion section or a discovery response through, then tell the AI agent what to make of it: headings, numbered points, a summary. Pick a local model and it's drafted on your own machine, free and unlimited on every plan.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Woman speaking beside a laptop at an office desk"}},"trustBar":{"label":"Dictate:","names":["Contract clauses","Motion sections","Discovery responses","Deposition summaries","Case memos","Engagement letters"]},"pain":{"label":"Why dictating a contract stalls","title":"Talking it through was never the slow part. Shaping it is.","subtitle":"Speak an engagement letter or a case chronology and you'll get every word back in one unbroken block. Cutting that block into clauses, headings and numbered points is what eats the afternoon, and plain dictation hands that part straight back to you.","cards":[{"title":"One block of speech, not a clause in it.","body":"Dictation cleanup hands back what you said, tidied. It will break the speech into paragraphs and number a list you clearly dictated. What it won't do is impose a heading or restructure a deposition summary into sections. That job belongs to the AI agent, and you have to ask it for it."},{"title":"The motion is the sensitive part, not the audio.","body":"A transcript is words. A motion is the argument you're about to run and the position your client will take. Choose local processing when you want the composing to run on your machine, or cloud tools when they fit your workflow."},{"title":"Paralegals and legal ops draft it too. Per-seat pricing adds up.","body":"An associate dictates the contract, a paralegal checks the cites, an assistant cleans it up, legal ops files it. Charge per seat and the cost follows the draft across four desks. OpenWhispr charges nothing for a seat on Free, so each of them can use it."},{"title":"Party names and statutes come back mangled.","body":"A general-purpose model has never seen your matter. Put the case names, the parties, the Latin and the citations your practice uses in the custom dictionary. A dictation vocabulary only fixes what was heard; this one also reaches the model that drafts, so it sees those names before it writes a word. It moves the odds rather than guaranteeing them, so read what carries legal weight before the document goes out."}]},"features":{"label":"From microphone to document","title":"You dictate a section. You get back something worth editing.","subtitle":"A transcript isn't a document. Dictation, cleanup and the AI agent each hand you something different.","items":[{"image":{"label":"Woman speaking while working on a laptop"},"title":"Say it once, read it back punctuated","body":"Talk an interrogatory response through the way you'd say it out loud, and it comes back transcribed and punctuated with the filler and the false starts taken out. Address the agent instead and it works from the same dictation, handling the disfluencies itself."},{"image":{"label":"Woman reading a printed document beside her laptop"},"title":"Headings and numbered points, not a wall of speech","body":"Cleanup breaks your speech into paragraphs and numbers a list you clearly dictated, and stops there. Ask the AI agent instead, on its own hotkey or by name mid-dictation, and it composes what you asked for: the headings, the numbered points and the lists you talked through."},{"image":{"label":"Man speaking while reviewing a printed document beside a laptop"},"title":"Nothing to install into Word or your DMS","body":"One global hotkey copies the text and pastes it at your cursor, so the clause lands in Word, a matter record, an email or a web form the same way. Auto-paste is on by default, so you edit it where it lands."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Audio kept on your own machine by default","value":"30 days","footnote":"Audio Retention ships on at 30 days. Settings takes it from 1 day to 90, or switches it off"},{"label":"Languages you can dictate a matter in","value":"100+","footnote":"a German framework agreement comes back as German text"},{"label":"Cost to add the paralegal and legal ops","value":"$$0/month","footnote":"no card details, and no per-seat licence to buy"}]},"howItWorks":{"label":"How a contract gets drafted","title":"Talk through the motion, then edit what comes back.","subtitle":"Three steps from talking the argument through to editing it in the document.","image":{"label":"Man speaking beside a laptop in a book-lined office"},"steps":[{"title":"Install it and choose where it runs.","body":"Setup puts a local model and OpenWhispr Cloud in front of you, with Cloud badged Recommended and local warning you it takes a few extra steps. Going local means pulling down a speech model of 75 MB to 3 GB and a separate local LLM of 0.25 GB to 19.8 GB first, depending which two you pick."},{"title":"Talk the clause through, ums and all.","body":"Say the deposition summary the way you'd explain it to a colleague, ums and all. Cleanup hands the same words back with the filler and the false starts stripped out."},{"title":"Ask the agent for headings and numbered points.","body":"Highlight the rough passage, press the agent hotkey and tell it what you want, and it rewrites that passage in place in the document. With nothing highlighted the draft arrives in the app's assistant panel instead, ready to copy across. Either way it's plain markdown text rather than Word styles, so you shape it in the document itself."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"Ten claims you can check","title":"What OpenWhispr does to a legal document, and where it stops.","subtitle":"What it does, what it won't do, and where the document ends up. No asterisks.","items":[{"title":"The setting that would upload a contract arrives switched off","body":"Pick a local model at setup and the audio, the transcript and the composed section all stay on the machine you spoke into. Two deliberate acts send content to us: switching on paid cloud sync, which arrives off and stores your notes, transcripts, agent conversations and your custom dictionary, party names included; and sharing a note or moving it into a team space."},{"title":"Cleanup keeps your wording and won't add a sentence you didn't say","body":"Cleanup is told to keep your voice, your wording and your intent, and never to add a sentence you didn't say. It still fixes grammar and punctuation, splits run-on sentences, drops false starts, and writes spoken numbers and dates out in standard form. That's an instruction to a model rather than a guarantee, so read an indemnity back before it goes out."},{"title":"The agent turns talk into a numbered section","body":"Say its name mid-dictation or press its hotkey, tell it what you want, and it composes that from what you talked through: headings, numbered points, a summary of the argument. Highlight a passage first and it rewrites that passage in place; with nothing highlighted the draft lands in the assistant panel. Pick local at setup and it composes on the LLM you downloaded, on your own machine. Styling stays a Word job."},{"title":"One hotkey, and no add-in for IT to approve","body":"The text goes to the clipboard and pastes itself where your cursor is. On macOS, without Accessibility permission it can copy for you but not paste."},{"title":"Know where the ceiling is before you hit it","body":"For selection edits, select 6,000 characters or fewer."},{"title":"Dictate it into a note now, shape it later","body":"Save the case chronology as a note while it's fresh, and a note action turns it into clean markdown when you come back to it."},{"title":"Recordings are files, so deleting one really deletes it","body":"Recordings are audio files on your machine you can open or delete yourself; notes and transcripts sit in a local database the app deletes rows from. Audio Retention ships on at 30 days, and Settings takes it from 1 day to 90 or switches it off. Transcripts are the other way round, kept until you set a window of your own, so set one alongside the audio. OpenWhispr doesn't encrypt that local data itself, so the folder wants the care you'd give any other matter folder."},{"title":"Send procurement the repo instead of a policy PDF","body":"OpenWhispr is open source. When IT asks what happens to a contract you dictated, they can inspect the code rather than take our word for it."},{"title":"The paralegal and the legal ops lead get the same hotkey","body":"Legal assistants and legal ops draft these documents alongside lawyers. Free is a download rather than a licence, so the person who assembles the interrogatory response has the same hotkey as the person who signs it."},{"title":"Dictate the deal in the language it was done in","body":"Whisper-class models cover 100+ languages, so the deal you did in Spanish comes back transcribed in Spanish."}]},"faq":{"label":"Before the first document","title":"What actually comes back when you dictate a contract.","subtitle":"Anything we've missed, bring it to our Discord, where the people who build OpenWhispr answer.","items":[{"question":"Which legal documents can AI dictation draft: a contract clause, a motion section, a discovery response?","answer":"All three, plus case memos, deposition summaries and file notes. Worth knowing which part does what: plain dictation hands back your own words, tidied up. The AI agent, on its own hotkey or called by name mid-sentence, drafts them into a structured section when you ask it for one. Highlight a passage first and it rewrites that passage in place; with nothing highlighted the draft arrives in the assistant panel to copy across. What you get is markdown text, so ## Heading arrives in Word as those literal characters and not a Word style. Pleading captions and court-ready tables aren't its job either, so the formatting your jurisdiction wants still happens in Word."},{"question":"Who else ends up with a copy of an unsigned contract?","answer":"Nobody, as long as you picked a local model at setup and left the cloud alone. Whisper-class transcription runs on your machine, and the composing runs on the LLM you downloaded, served by llama.cpp at 127.0.0.1, so the clause never travels. Two things send content to us, and both are deliberate acts: cloud sync, a paid setting that stays off until you switch it on, which stores your notes, your dictation transcripts, your agent conversations and your custom dictionary, party names included, in our cloud database; and sharing a note or putting it in a team space, which uploads that note on any signed-in account. One more thing if you sign in: the assistant panel can run a web search, which sends the query the model composed, not your document, to us. Stay signed out and it is never offered."},{"question":"Can it write a whole brief in one go?","answer":"For a brief, work section by section: facts, then argument, then relief."},{"question":"Will party names and citations survive a deposition summary?","answer":"Load the custom dictionary first. It holds the words your matters actually turn on: party names, case names, statutes, the Latin, the citations your practice uses. It's a hint the model sees rather than a find-and-replace, so it moves the odds a long way in your favour without promising a clean sweep. It reaches local Whisper transcription, the cleanup pass, the AI agent and translation, across 100+ languages."},{"question":"Can a paralegal take over the contract after I've dictated it?","answer":"Yes. The hotkey copies the text and pastes it wherever the cursor is, so the words land in the file the two of you already share: Word, the matter record, an email. Auto-paste is on by default. On macOS it needs Accessibility permission, or it falls back to leaving the text on your clipboard."},{"question":"What does it cost to run a matter's documents through it?","answer":"Nothing on the local path, where dictation and drafting are unlimited on every plan, Free included, and neither one is metered. The cloud side has two separate meters. On Free you get 2,000 words a week of cloud speech-to-text, and 5 hours a month of cloud meeting recording. Pro at $8 a month makes cloud speech-to-text unlimited and takes meeting recordings to 20 hours a month. Business at $16 per user makes meeting recordings unlimited and adds team features."}]},"reviews":{"label":"Straight from our users","title":"Rambling in, editable text out.","subtitle":"Nobody was asked for these. One of them stumbles mid-sentence and still gets a clean paragraph back, which is what dictating a clause actually sounds like.","score":{"value":"0","label":"Client documents sitting on our servers","body":"In local mode, signed out, with cloud sync off and nothing shared, nothing reaches us. Turn cloud sync on, a paid setting that arrives switched off, and your notes, dictation transcripts, agent conversations and custom dictionary are stored in our cloud database."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free to get started.","Upgrade when you want more."],"caption":"What each plan covers for legal document work","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Paralegals and legal teams drafting contracts, briefs and motions, who don't mind downloading a speech model and an LLM first"}]},{"parts":[{"text":"Legal assistants who'd rather dictate through OpenWhispr Cloud than manage model downloads, at the desk and on an iPhone"}]},{"parts":[{"text":"Legal ops teams handling document work across a lot of matters, and recording calls all month"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Free plan: $0."}]},{"parts":[{"text":"The cloud speech-to-text meter comes off, and device sync and the iPhone app come with it"}]},{"parts":[{"text":"Cloud meeting recordings go unlimited and team features unlock, at $16 per user per month"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud and does not require local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Read the code yourself","title":"A contract shouldn't need a round trip to get written.","paragraphs":["Turning a transcript into clauses and numbered points takes an AI model. OpenWhispr's AI agent can do that work with a model you downloaded yourself, served by the llama.cpp binary that ships inside the app, on 127.0.0.1.","You don't have to take our word for that. OpenWhispr is open source, so your risk team can inspect the code rather than read a policy about it.","Break something at 11pm the night before a filing deadline and you're talking to the engineers who wrote it, in Discord."],"image":{"label":"The people who build it"}},"finalCta":{"tagline":"$$0 a seat · No card details","title":"Say the clause out loud. Get a numbered section back.","body":"Download it free, pick a local model at setup, then hand the AI agent your rough dictation. Highlight the passage and it rewrites the clause or the case memo in place, as plain markdown text. No card details, no seat to buy.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Available on macOS, Windows, and Linux. Get community support in Discord."}}},"useCaseVeterinary":{"breadcrumb":{"home":"Home","useCases":"Use Cases","veterinary":"For Vets"},"metadata":{"title":"Free Offline Veterinary Dictation Software - OpenWhispr","description":"OpenWhispr is free, open-source veterinary dictation software that turns spoken findings into SOAP notes on your device, even on a farm call with no WiFi.","keywords":["veterinary dictation software","vet dictation app","veterinary transcription","SOAP note dictation","veterinary practice software","offline vet dictation"]},"schema":{"name":"OpenWhispr","applicationCategory":"Veterinary dictation software","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source veterinary dictation software. Dictated findings are transcribed on the vet's own device and never leave it. Free for unlimited local dictation, and works fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Veterinary dictation software for vets & vet techs","title":"Say it once. Get the SOAP note.","body":"OpenWhispr turns that dictation into a SOAP note, a discharge or a client email on your own computer. No per-DVM subscription, and it still works in a barn with no signal.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Veterinarian speaking beside a dog and laptop in a clinic"}},"trustBar":{"label":"Dictate into:","names":["ezyVet","Cornerstone","Avimark","Covetrus Pulse","Shepherd","Microsoft Word"]},"pain":{"label":"The documentation squeeze","title":"Documentation shouldn't cost your evenings and $99 a vet a month.","subtitle":"You want the note written for you. You don't want a subscription per DVM, a recorder in the exam room, or a tool that dies the moment you leave WiFi.","cards":[{"title":"Charting follows you home.","body":"Turn dictated text into a structured note with a custom action."},{"title":"A per-DVM subscription is another monthly line item.","body":"The Free plan includes unlimited local AI models and 100+ language support at $0."},{"title":"No signal, no scribe.","body":"Barn calls, farm visits, a back room with one bar. A cloud scribe can stall exactly where you are, leaving you to rebuild the note from memory that night. OpenWhispr runs on your device, fully offline."},{"title":"A recorder in the euthanasia room.","body":"Some conversations shouldn't be uploaded to a stranger's server to become a note. You dictate the findings afterwards, and with a local model everything is processed on your machine. Nothing sits on the table while you talk a client through it."}]},"features":{"label":"What you get","title":"Records closed before you lock up. No per-DVM bill.","subtitle":"An AI scribe listens to the room so you never have to say it. OpenWhispr waits for you to dictate, then a local model drafts the note from a prompt you saved yourself, on your own laptop.","items":[{"image":{"label":"Veterinarian speaking beside a dog on an exam table and a laptop"},"title":"Dictated at the treatment table, transcribed on your laptop","body":"Signalment, history, exam findings, assessment and plan, straight into the mic. Pick a local model and it's transcribed on your laptop, never uploaded, and you decide what's kept and when it goes."},{"image":{"label":"Veterinarian handing paperwork to a dog owner at a clinic doorway"},"title":"One dictation, two documents","body":"Save your own SOAP, discharge and client-email prompts as one-click actions, then run whichever one the case needs. The same dictation becomes the record for the file and the instructions the client reads in the car park. You check both before either is filed."},{"image":{"label":"Veterinarian using a laptop beside a dog in a rural clinic"},"title":"Types into the PIMS field you're already in","body":"Press a hotkey and dictate into the Cornerstone note field, an ezyVet consult tab, a discharge template or a referral email to the specialist. Clean, punctuated text, processed locally."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Client conversations uploaded","value":"None","footnote":"in local mode, not the estimate conversation, not the goodbye"},{"label":"Cost per DVM, per tech","value":"$$0","footnote":"unlimited local dictation, including the relief vet who works Saturdays"},{"label":"Setup takes","value":"No fixed time","footnote":"Install the app, then choose a local model or OpenWhispr Cloud."}]},"howItWorks":{"label":"How it works","title":"Dictate on the walk back to treatment.","subtitle":"This is one app on the practice laptop, with no PIMS migration or IT project.","image":{"label":"Veterinarian carrying a tablet past a dog in a treatment room"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model during setup."},{"title":"Dictate the case.","body":"Signalment, presenting complaint, exam findings, assessment and plan. In the room, in the treatment area, or in the truck between farms. It's transcribed on your device, and one click turns it into the draft."},{"title":"Review, paste, close the record.","body":"Fix what it got wrong, drop it into the patient's record in ezyVet or Cornerstone, and delete the audio. All of it happens on your laptop."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons to choose OpenWhispr","subtitle":"In a three-vet practice this is a budget decision the owner signs off on, not an IT project. Here are the facts, and the practice manager can hand the repo link to whoever does your IT.","items":[{"title":"Choose local when dictation must stay on your device","body":"Dictation, and note drafts with a local model, are processed on your machine. The open-source code shows there's no upload step."},{"title":"Every DVM and every tech, $0","body":"Unlimited local dictation, forever. Scribes charge per DVM and leave the techs writing records for free anyway."},{"title":"Works in the barn and in the truck","body":"The local model runs with the laptop in airplane mode. Dictate in the truck, file the note when you're back in range."},{"title":"Dictate into the PIMS you already run","body":"A system-wide hotkey drops clean text wherever your cursor is: ezyVet, Cornerstone, Avimark, a discharge template, a referral email. No integration to build, and no reason to change PIMS."},{"title":"One dictation, the SOAP note and the discharge","body":"Save the prompt once. A local model, or your own AI key, turns the same dictation into the record for the file and the instructions the client takes home."},{"title":"No recorder in the room","body":"You dictate afterwards. Nothing sits on the table during the estimate, and nothing is running when you put a hand on a client's shoulder."},{"title":"Choose how long local records stay on the practice laptop","body":"When you enable local audio or transcript retention, files stay on this device and are deleted after the retention period."},{"title":"Read the code before the owner signs anything","body":"Every line is on GitHub. The practice manager can send the link to whoever does your IT and have an answer the same week."},{"title":"Knows cephalexin from Cerenia","body":"The custom dictionary learns the drug names, breeds and procedures you correct, and you add the abbreviations your clinic already writes: OHE, TPR, BAR, QAR."},{"title":"100+ languages","body":"Take the history from a client who would rather explain it in Spanish, transcribed on your laptop."}]},"faq":{"label":"Before you download","title":"The questions vets ask first.","subtitle":"Straight answers on cost, terminology and what happens when you leave WiFi. If yours isn't here, ask us in Discord.","items":[{"question":"Do you record the appointment, like a scribe does?","answer":"No. Nothing joins the consult and nothing sits on the table. You dictate afterwards, at the treatment table or in the truck, and in local mode that dictation is transcribed on your own laptop with no upload step. The code is public, so you can check that instead of taking our word for it."},{"question":"Do vets even need this? HIPAA doesn't cover animals.","answer":"Right. HIPAA covers human patients, not pets. But the record is still confidential, and the audio carries the client's name, their address and the hardest conversation of your day. Local-first keeps all of it off a vendor's server."},{"question":"Does veterinary dictation software get drug names and breeds right?","answer":"Yes, once you teach it yours. It runs Whisper- and Parakeet-class models trained on a wide range of accents and languages, and the custom dictionary learns the terms you correct: Apoquel, Cerenia, brachycephalic, Bichon Frise, OHE, plus whatever abbreviations your clinic already writes."},{"question":"Will it work on a farm call with no signal?","answer":"Yes. Local models run entirely offline. Dictate in the barn or the truck, and file the note once you're back in range. Cloud scribes can't do this, because they stall the moment you leave WiFi."},{"question":"Can it write the discharge as well as the SOAP note?","answer":"Yes, from the same dictation. Run a local model (Llama, Qwen, Gemma) or connect your own AI key, then save your own SOAP, discharge and client-letter prompts as one-click actions and run whichever one the case needs. A local model keeps it all on your machine; your own key means the model and data path are ones you already trust."},{"question":"Is it actually free?","answer":"Yes. The Free plan includes unlimited local AI models, 100+ languages, and desktop availability on macOS, Windows and Linux at $0. Pro is $8/month and adds unlimited OpenWhispr Cloud transcription, sync across devices, and the iPhone and iPad app."}]},"reviews":{"label":"What users say","title":"Built by people who wouldn't put a recorder in the room either.","subtitle":"Three OpenWhispr users describe how they use it.","score":{"value":"0","label":"Consults on a server","body":"You dictate afterwards, so there is no appointment recording to begin with. In local mode the dictation is never uploaded either."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for veterinarians","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Vets and techs who want free, private, offline dictation and are fine with a quick local setup"}]},{"parts":[{"text":"Vets who want easy, reliable cloud transcription"}]},{"parts":[{"text":"Practices and teams documenting at volume"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Free offline dictation, with consult audio staying on your device"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models or keys"}]},{"parts":[{"text":"Unlimited recordings plus advanced workflow features for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","Enterprise includes team features and admin controls, plus dedicated support."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"The goodbye appointment doesn't belong on a vendor's server.","paragraphs":["OpenWhispr started with a simple objection: the three conversations that take the most out of you, the diagnosis, the estimate and the goodbye, shouldn't have to pass through a stranger's server to become a note.","We build in public. The code is on GitHub. When you choose local processing, dictation stays on your device. A practice can verify that in the code.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Last appointment at six. Records closed at six.","body":"Download OpenWhispr and dictate the next case on the walk back to treatment. Choose local models when dictation needs to stay on your laptop and work offline.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Unlimited local AI models are free on macOS, Windows and Linux."}}},"useCaseJournalists":{"breadcrumb":{"home":"Home","useCases":"Use Cases","journalists":"For Journalists"},"metadata":{"title":"Free AI Speech To Text Tool For Journalists - OpenWhispr","description":"Free, open-source AI speech to text for journalists. In local mode the interview is transcribed on your laptop, never uploaded. No per-minute bill.","keywords":["ai speech to text tool for journalists","transcription software for journalists","interview transcription software","offline interview transcription","private transcription for reporters","local speech to text","free interview transcription"]},"schema":{"name":"OpenWhispr","applicationCategory":"Journalism transcription software","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source speech to text software for journalists. In local mode interview audio is transcribed on the reporter's own device and never leaves it. Free for unlimited local transcription.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"AI speech to text for journalists & reporters","title":"Drop in the tape. Get the whole transcript.","body":"Get the whole transcript without uploading the tape. In local mode, OpenWhispr transcribes on your machine, free and open source, with no per-minute bill or signal once the model is downloaded.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Person typing on a laptop beside a field recorder in an open vehicle"}},"trustBar":{"label":"Paste your quotes into:","names":["Google Docs","Microsoft Word","WordPress","Arc XP","Slack","Substack"]},"pain":{"label":"What an upload actually costs you","title":"Source protection shouldn't end at the upload button.","subtitle":"You want the tape transcribed on deadline. You don't want it living in someone else's account to get there.","cards":[{"title":"Keep source audio on your device in local mode.","body":"Upload an interview and a vendor holds the recording. Some services route the audio to human transcriptionists. In local mode OpenWhispr transcribes on your device, so there is no server for the tape to sit on."},{"title":"Your tape becomes someone else's record.","body":"Every upload leaves unpublished material in an account you don't administer, on a retention schedule you didn't set. In local mode there is no upload and no account, so there is nothing on our side for anyone to subpoena."},{"title":"A quote you can't check.","body":"Speech to text mishears names and crosstalk, and every model will occasionally write a sentence nobody said. A summary will not show you where that happened. Record the interview in the app and the transcript is stamped line by line. To check a quote before you file, compare it with the original recording you keep."},{"title":"\"Trust us\" is not source protection.","body":"Closed tools ask you to take a privacy policy on faith. OpenWhispr's code is public on GitHub under the MIT license. Your security desk, or any developer you trust, can read exactly what it does and what it doesn't."}]},"features":{"label":"What you get","title":"Filed faster. The tape stays yours.","subtitle":"Transcription tools usually trade source protection for speed, or bill you by the minute. This does neither. In local mode the interview is transcribed on your machine, at no cost per minute.","items":[{"image":{"label":"Two people talking across a table with an open laptop"},"title":"Interview files transcribed on-device","body":"Drop in an MP3, WAV, M4A or half a dozen other formats, one file or a whole batch. In local mode it's transcribed on your machine, with no length limit and no per-minute bill."},{"image":{"label":"Person reviewing printed pages beside a laptop and phone"},"title":"The whole transcript, never a summary","body":"You get the full transcript, not a digest. Record the interview in the app and every line is stamped on your device, ready to export as .txt, .md, .json or .srt. On an upload, speaker detection stamps each turn locally; per-line stamps need your own Groq or Mistral key, or OpenAI set to whisper-1."},{"image":{"label":"Two people talking at a desk with a field recorder and notebook"},"title":"Live transcript from the room","body":"Put your laptop on the table and it transcribes an in-person interview as you talk, on any platform. On a remote interview it captures your computer's own audio, so no bot joins and nothing shows in the participant list. That capture needs Windows, macOS 14.2 or later, or a Linux desktop whose audio portal allows it. Recording is its own setting, so point that one at a local model too."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Source audio sent to us","value":"None","footnote":"in local mode, your audio is transcribed on your own device"},{"label":"Cost per hour of tape","value":"$$0","footnote":"unlimited local transcription, every reporter, forever"},{"label":"Setup takes","value":"5 min","footnote":"from download to your first transcript, plus the one-time model download"}]},"howItWorks":{"label":"How it works","title":"Transcribed before you write the lede.","subtitle":"No account or credit card required. Choose a local model in Settings, then drop in the tape or record in the app.","image":{"label":"Person at a desk with a laptop, notebook, and audio recorder"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Skip the account and setup hands you a local model. Either way, point uploads and recording at one in Settings."},{"title":"Drop in the tape.","body":"MP3, WAV, M4A, FLAC and more, one file or a batch. In local mode it's transcribed on your device, however long the interview ran."},{"title":"Check it, then file.","body":"Read the transcript in the app and check any line against the audio. A recorded interview exports to .txt, .md, .json or .srt with every line stamped."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"On the record","title":"10 reasons to choose OpenWhispr","subtitle":"Picking an AI speech to text tool is a source-protection decision before it's anything else. Here are the facts, and the ones about your audio you can check in the code yourself.","items":[{"title":"Local mode keeps source audio on your device","body":"Transcription runs on your machine, with no upload step."},{"title":"No bot joins a remote interview","body":"It records your computer's own audio, on Windows, macOS 14.2 or later, or a Linux desktop that allows it. Nothing appears in the participant list, and on a local model no vendor gets a copy of the call."},{"title":"Offline, once the model is downloaded","body":"Courthouse steps, a jail visit, a rural county meeting. Once the model is downloaded, no signal required."},{"title":"No import limits, no per-minute bill","body":"Local transcription is unlimited at $0, on every plan."},{"title":"Your security desk can read the code","body":"MIT licensed, every line on GitHub. A newsroom security review can read exactly what happens to the audio, before you put a source's tape through it."},{"title":"Recorded interviews, stamped line by line","body":"Export to .txt, .md, .json or .srt with every line stamped, so you can point standards or the lawyers at the exact moment."},{"title":"No length limit on your own machine","body":"A ninety-minute interview costs the same as a nine-minute one, and takes as long as your laptop needs."},{"title":"Who said what, on every plan","body":"Two voices in one interview, labelled. Speaker detection runs on your device, so it works on a local model with nothing uploaded."},{"title":"Nothing rewrites your quotes","body":"An uploaded interview is never rewritten. The note is the transcript as the model heard it. In dictation the untouched text sits beside the cleaned-up version, and you can switch cleanup off entirely."},{"title":"100+ languages","body":"Interview a source in their own language, not their second one. In local mode, transcription stays on your machine."}]},"faq":{"label":"Before you download","title":"The questions reporters ask first.","subtitle":"Straight answers, including where it falls short. Anything missing, ask in Discord.","items":[{"question":"Does my source's audio ever leave my device?","answer":"Not in local mode. Transcription runs on your machine and the open-source code proves there's no upload step. Cloud transcription is the default, so switch to a local model in Settings before your first interview. Dictation, recordings and uploaded files each carry their own transcription setting, so change the one you use for tape. If you sign in, point the AI cleanup model at a local one too, or switch cleanup off: that setting writes the note title and it defaults to the cloud."},{"question":"Can it transcribe a file I already recorded?","answer":"Yes. Drag in an MP3, WAV, M4A, FLAC, AAC, OGG, Opus or WebM file, pick it from a file dialog, or queue a batch. On a local model there's no length limit and no per-minute charge. Local video files aren't supported, but you can paste a YouTube or video link and it pulls the audio, up to six hours or 500 MB."},{"question":"Is it accurate enough to quote from?","answer":"It runs Whisper-class and NVIDIA Parakeet models, and no speech to text is perfect on accents, names or crosstalk. That's why you get the whole transcript, never a summary: record the interview in the app and it's stamped line by line, so you can check the quote against the tape before you file. Treat it as a first pass you verify, not a transcript you publish unread."},{"question":"Can it tell my source's voice from mine?","answer":"Yes. On a local model the speaker detection runs on your device too, with nothing uploaded. It separates the voices in an in-person interview and labels who said what, on every plan; the transcript is live as you talk and the labels land when you stop. A recording expects two speakers and goes up to eight, enough for a panel. On an uploaded file, switch labels on before you transcribe and it works out the count for you."},{"question":"If we get a subpoena, what do you have?","answer":"In local mode, nothing. There is no upload and no account, so there is no tape and no transcript on our side to hand anyone. What exists is on your laptop, under your own retention and source-protection rules. That is also why offline works: once the model is downloaded, a courthouse or a jail visit needs no connection."},{"question":"Is it actually free?","answer":"Yes. Unlimited local transcription, speaker labels, 100+ languages, all three desktop platforms, $0 forever. Pro ($8/month) adds unlimited OpenWhispr Cloud speech-to-text and mobile dictation. Business ($16/user/month) adds unlimited cloud meeting recordings and priority support."}]},"reviews":{"label":"What users say","title":"Different jobs. The same instinct about their own audio.","subtitle":"From our Discord, GitHub and inboxes. None of them is a reporter, so read them for the tool, not the beat.","score":{"value":"0","label":"Interviews on a vendor's server","body":"In local mode the interview never leaves your device. There is no copy on our side, because nothing is sent to us."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for journalists","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Reporters who want free, offline transcription and don't mind setting up a local model"}]},{"parts":[{"text":"Reporters who'd rather not manage a local model and are fine with cloud transcription"}]},{"parts":[{"text":"Desks that transcribe at volume, or need unlimited cloud meeting recordings"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"The tape never leaves your laptop, at $0"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models or keys"}]},{"parts":[{"text":"Unlimited cloud meeting recordings and priority support for a desk"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"Some conversations only happen off the record. We built for those.","paragraphs":["OpenWhispr started with a simple objection: a tape of a source who asked for confidentiality shouldn't end up on a stranger's server just so you can read back what they said.","We build in public. The code is on GitHub. When we say the audio stays local, that isn't just a promise in a privacy policy. Anyone on your desk can inspect the code and verify what happens to the audio.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Your tape stays between you, your source and your laptop.","body":"Download OpenWhispr, point it at a local model, and your next interview is transcribed on your machine and nowhere else.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, we'll fix it with you in Discord. Free forever on Mac, Windows & Linux."}}},"useCaseContentCreators":{"breadcrumb":{"home":"Home","useCases":"Use Cases","contentCreators":"For Content Creators"},"metadata":{"title":"Free Dictation Software For Content Creators - OpenWhispr","description":"OpenWhispr is free, open-source dictation software for content creators. Paste your own video link or an audio file, get the transcript, talk out the rest.","keywords":["dictation software for content creators","dictation for youtubers","youtube video transcript software","transcribe your own youtube video","voice to text for creators","speech to text for video scripts"]},"schema":{"name":"OpenWhispr","applicationCategory":"Dictation software for content creators","operatingSystem":"macOS, Windows, Linux, iOS","description":"Local-first, open-source transcription and dictation software for content creators. Paste the link to a video you posted yourself, or drop in an audio file, and get the transcript in a note you can search and quote. Then talk out the title, the description and the captions and the clean text lands wherever the cursor already is. With a local model chosen the audio is transcribed on the creator's own device. Free for unlimited local transcription and dictation on macOS, Windows and Linux, and it works fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Dictation software for content creators","title":"Paste your own video link. Get the transcript.","body":"Get a transcript of your own YouTube video or audio file on your laptop for $0. Then use the same hotkey to dictate the title and description.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"A person speaks to a camera in a home studio with a ring light and laptop nearby"}},"trustBar":{"label":"Dictate into any app, including:","names":["YouTube Studio","CapCut","Notion","Patreon","Buffer","Discord"]},"pain":{"label":"What you're choosing between","title":"Everything you said is in the video. None of it is in text.","subtitle":"A transcript is what turns one upload into chapters, a newsletter and a thread. Getting one means paying per minute, or sitting through your own video with a keyboard.","cards":[{"title":"Your back catalogue is locked inside the videos.","body":"Turning last year's uploads into a newsletter means sitting through them again. Paste the link to one of your own videos and OpenWhispr pulls the audio down and transcribes it. Paste a list of your links, up to 50 at a time, and it works through the queue."},{"title":"The interview is on your disk and nowhere else.","body":"The raw recording, the voice memo, the podcast cut that never went up. Drop the audio file in and it comes back as a transcript in a note you can search. mp3, wav, m4a, flac, aac, opus, ogg and webm, straight off the disk."},{"title":"The description box, still blank at midnight.","body":"The export is done and the title, the description and the pinned comment are not. Press a hotkey in the tab you already have open and say them. Clean, punctuated text lands where the cursor is."},{"title":"Another subscription before the channel pays for itself.","body":"Editor, thumbnails, scheduler, transcripts, newsletter. Five tools billing one channel that does not cover them yet. Unlimited local transcription and dictation on OpenWhispr is $0, on every machine you own, and the free plan is not a trial."}]},"features":{"label":"From the upload to the words around it","title":"The transcript is there before you sit down to write.","subtitle":"Pick local models and both the transcription and the clean-up run on your own computer, unlimited, at no cost, in 100+ languages.","items":[{"image":{"label":"A person works at a laptop in a home editing setup with audio equipment nearby"},"title":"A transcript out of a video you already posted","body":"Paste the link to a video you posted yourself, or a list of them one per line, and OpenWhispr downloads the audio and transcribes it into a note you can search and quote. Up to 50 links at a time, 500 MB and six hours a video. Playlists and live streams are out."},{"image":{"label":"A person speaks beside a camera and laptop at a kitchen table"},"title":"The audio file that never got posted","body":"Interviews, voice memos, the podcast cut, the raw take. Drop in mp3, wav, m4a, flac, aac, opus, ogg or webm and it comes back as a transcript. On a local model there is no per-minute bill, and a file already on your disk has no size limit at all."},{"image":{"label":"A person speaks while using a laptop beside handwritten notes"},"title":"One hotkey, every tab you publish from","body":"Talk into YouTube Studio, your script doc, Notion, an email to a sponsor. The clean text pastes itself where you were already typing, with no plugin in any of them. On a Mac you grant accessibility once, and that is the part that lets it paste at your cursor."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Sent to our servers","value":"None","footnote":"with a local transcription model the audio is transcribed on your own computer"},{"label":"Cost per creator","value":"$$0","footnote":"unlimited local transcription and unlimited local dictation, on every machine you own"},{"label":"Setup takes","value":"A few extra steps","footnote":"download, pick a model, paste a link or set a hotkey"}]},"howItWorks":{"label":"Download to first transcript","title":"Three steps from your posted video to searchable text.","subtitle":"Nothing to connect to the platforms you publish on.","image":{"label":"A person speaks into a microphone at a desk with a laptop and camera gear nearby"},"steps":[{"title":"Download it free.","body":"Mac, Windows or Linux, whatever the edit rig runs. Setup offers OpenWhispr Cloud or local models, and picking local downloads them once."},{"title":"Paste the link, or drop the file.","body":"Your own video's URL, a list of them one per line, or an audio file off your disk. OpenWhispr pulls the audio down and transcribes it into a note."},{"title":"Read it, use it.","body":"Search it, quote it, cut it into chapters and a newsletter. Then hit the hotkey in YouTube Studio and talk the title and the description out."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"Why creators keep it","title":"10 reasons content creators choose OpenWhispr","subtitle":"Every one is in the app or in the repo. Check before you download anything.","items":[{"title":"Paste your own video link, get the transcript.","body":"Give OpenWhispr the YouTube URL of a video you posted, a direct audio URL, or an audio file from your disk. It transcribes the audio and can queue up to 50 links; playlists and live streams are not supported."},{"title":"It types into everything you publish from.","body":"Use the system-wide hotkey in YouTube Studio, your script doc, newsletter or sponsor email. On a Mac, one accessibility permission lets OpenWhispr paste at the cursor."},{"title":"Unlimited local transcription costs nothing.","body":"The free plan is not a trial. With a local model there is no per-minute meter on a file, no word cap on dictation, and no card to enter."},{"title":"Your niche words come out spelled right.","body":"The custom dictionary takes your channel name, your recurring guests and the vocabulary only your audience uses. Dictation hands that list to the transcription model as a hint, and to the clean-up model after it."},{"title":"The ums and the false starts come out.","body":"The clean-up model strips filler, stutters and restarts, and is told to keep your voice, your wording, your formality and your intent, with your jargon and names exactly as spoken."},{"title":"100+ languages.","body":"Dictate the description in the language your audience actually comments in. Same hotkey, same model, no separate tool."},{"title":"Location shoots with no signal.","body":"A rented studio, a hotel, a plane, a shoot in the middle of nowhere. Local models need no network at all."},{"title":"Nothing is uploaded when you keep it local.","body":"Pick a local transcription model and the audio is transcribed on your own machine, so none of it reaches us. Point the clean-up at a local model too and the text stays put as well."},{"title":"The code is public.","body":"OpenWhispr is open source on GitHub, where 5,000+ engineers have starred it. Every privacy claim on this page is a file you can open and read."},{"title":"Mac, Windows and Linux, one free download.","body":"Whatever you edit on. The same app, the same free price, and a hotkey you choose yourself."}]},"faq":{"label":"Before you install it","title":"The questions creators ask before they download it","subtitle":"Short answers, and where in the app or the repo you can check them.","items":[{"question":"Can I get a transcript of a video I already published?","answer":"Paste the link to your own video. OpenWhispr accepts a YouTube URL or a direct audio file URL, downloads the audio and transcribes it into a note. Paste a list of your links, one per line and up to 50 at a time, and it works through them as a queue. Playlists and live streams are not supported, and the download caps at 500 MB and six hours a video. On a local model that cap is the only one. On OpenWhispr Cloud the free plan transcribes files up to 25 MB, and past that it is a paid plan."},{"question":"Can I upload the video file itself?","answer":"Not the video file. The upload panel takes audio: mp3, wav, m4a, webm, ogg, oga, flac, aac and opus. For something you already posted, paste its link instead and OpenWhispr pulls the audio down for you. For a cut that is not posted anywhere yet, export the audio out of your editor and drop that in."},{"question":"Is my audio uploaded anywhere?","answer":"Not if you pick a local transcription model: the audio is transcribed on your own computer and nothing is sent. One thing worth knowing, because it is two settings and not one. Transcription and the AI clean-up are chosen separately, and if you are signed in the clean-up still defaults to OpenWhispr Cloud. On that default the transcript text, never the audio, is sent to us to be tidied. Set the clean-up model to a local one too and neither leaves the machine."},{"question":"Is it really free, or is the free plan a trial?","answer":"Free is free, with no card. Unlimited local dictation and unlimited local transcription are $0 on every plan, forever. Pro ($8/month) makes OpenWhispr Cloud transcription unlimited, and Business ($16/month) adds unlimited cloud meeting recordings and agent mode. Nothing on the free plan expires."},{"question":"Will it get my channel name and my niche jargon right?","answer":"Add them to the custom dictionary once. Your channel name, your recurring guests, the in-jokes and the product names your audience uses. Dictation passes that list to the transcription model as a hint, and the clean-up model gets the same list. One exception worth knowing: the local Parakeet model does not take the hint, so use a local Whisper model if the dictionary matters."},{"question":"Does it work on a shoot with no WiFi?","answer":"Yes, once the local models are downloaded. Local transcription and a local AI model both run on your own machine with no network, so a hotel, a plane or a location with no signal is the same as your desk. Cloud transcription is the option that needs a connection."}]},"reviews":{"label":"What people tell us","title":"They tried it on their own machine and stopped typing.","subtitle":"From our Discord, GitHub and inbox, word for word. None of them runs a channel, so read them for the tool.","score":{"value":"0","label":"Plugins to install in YouTube Studio","body":"The hotkey works the same in a browser tab, your edit suite and your newsletter draft. Nothing to connect, nothing to log into, no integration to keep alive."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for content creators","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"One-person channels that want unlimited local transcription and dictation at no cost and will set up a local model"}]},{"parts":[{"text":"Creators who would rather pay $8 than look after a local model"}]},{"parts":[{"text":"Channels where an editor, a writer and a producer all touch the same content"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Transcripts of your own videos, and everything you write around them, for nothing, with the audio staying on your machine"}]},{"parts":[{"text":"Cloud transcription that never runs out, with no models or API keys to look after"}]},{"parts":[{"text":"Unlimited cloud meeting recordings and agent mode for everyone on the channel"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Why we built it","title":"The fastest way to sound like yourself is to say it out loud.","paragraphs":["OpenWhispr started with an objection: talking is how people are fluent and typing is how they get stiff, and the software that turns one into the other kept routing everything through somebody else's server.","We build in public. Every release and every line of the app is on GitHub. So when we say the dictation can stay on your machine, you do not have to take our word for it. You can open the file and read it.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"One of OpenWhispr's founders"}},"finalCta":{"tagline":"$$0 · open source · no plugin to install","title":"Get the transcript, then write around it.","body":"Download OpenWhispr, pick a local model at setup, and paste the link to the video you posted this morning. No card, no per-minute bill, no second subscription.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free forever on Mac, Windows & Linux."}}},"useCasePodcasters":{"breadcrumb":{"home":"Home","useCases":"Use Cases","podcasters":"For Podcasters"},"metadata":{"title":"Best Speech-to-Text Software For Podcasting - OpenWhispr","description":"OpenWhispr is free, open-source speech-to-text for podcasters. Transcribe unlimited episode audio on your own machine and never pay per minute again.","keywords":["speech to text for podcasting","podcast transcription software","free podcast transcription","dictation software for podcasters","podcast show notes software","offline podcast transcription","open source transcription software"]},"schema":{"name":"OpenWhispr","applicationCategory":"Podcast transcription software","operatingSystem":"macOS, Windows, Linux, iOS","description":"Local-first, open-source speech-to-text for podcasters. With a local model chosen, episodes and show notes are transcribed on the creator's own device. Free for unlimited local transcription and dictation, and works fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Speech-to-text software for podcasters","title":"Record the episode. Get the transcript. Pay nothing per minute.","body":"Turn each recorded episode into a transcript, then dictate show notes, descriptions and posts into any app. Choose local transcription for offline work, or Cloud for high accuracy and speed without downloading models.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Person speaking into a studio microphone beside a laptop and handwritten notes"}},"trustBar":{"label":"Write show notes in:","names":["Notion","Buzzsprout","Transistor","Spotify for Creators","Substack","YouTube Studio"]},"pain":{"label":"The problem with per-minute transcription","title":"Transcription shouldn't cost more than your hosting.","subtitle":"A transcript and show notes are part of shipping many episodes. Many tools bill by the minute or cap usage by the month.","cards":[{"title":"Paying by the minute to transcribe your own voice.","body":"A weekly one-hour show, metered per minute, turns into a real monthly bill that never ends. OpenWhispr transcribes on your machine: unlimited hours, $0, and the Free plan is not a trial."},{"title":"Show notes take longer than the episode.","body":"You talked for an hour, then typed for two. Press a hotkey and speak the summary, the guest bio and the links straight into your host or your docs. Clean, punctuated text, at speaking speed."},{"title":"Auto-captions butcher your guests' names.","body":"A transcript that mangles your guest's name or your show's title is a correction job. Add the name to OpenWhispr's dictionary to improve how it appears in dictated notes and AI-drafted summaries."},{"title":"Pulling one quote means scrubbing the whole episode.","body":"Clips, quote graphics and newsletters all start with finding the line. With a transcript on your machine you search text, copy the quote and move on. No scrubbing."}]},"features":{"label":"What you get","title":"Transcripts, show notes, summaries. No meter on any of them.","subtitle":"Make transcripts and show notes on your own machine in 100+ languages, without adding another monthly transcription bill.","items":[{"image":{"label":"Person monitoring audio at a microphone with a laptop and recording interface"},"title":"Unlimited local transcription","body":"Import the episode you already cut, and it is transcribed on your device with local Whisper-class models. No per-minute fees or monthly hour caps for local transcription, with nothing uploaded in local mode."},{"image":{"label":"Person speaking into a desktop microphone while working on a laptop"},"title":"Show notes at speaking speed","body":"Press a hotkey and speak your summary, description or newsletter straight into Notion, Buzzsprout, Substack, wherever your cursor is."},{"image":{"label":"Two people recording with microphones while one reviews notes beside a laptop"},"title":"A dictionary that knows your show","body":"Add your show's title, recurring guests and niche vocabulary to the custom dictionary. It improves how those names and terms are handled in dictation and AI-drafted notes."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Unreleased audio uploaded","value":"None","footnote":"in local mode the episode never leaves your machine"},{"label":"Cost per episode","value":"$$0","footnote":"unlimited local transcription and dictation, every episode, forever"},{"label":"Transcription mode","value":"Local or Cloud","footnote":"Local runs on your device. Cloud prioritizes accuracy and speed with no model download."}]},"howItWorks":{"label":"How it works","title":"A transcript and show notes before you close the editor.","subtitle":"Choose local transcription when you need offline work or want audio to stay on your machine. Choose Cloud for high accuracy and speed without downloading local models.","image":{"label":"Person holding headphones beside a closed laptop and handwritten notes"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model during setup."},{"title":"Record or dictate.","body":"Record a session for a local transcript, or press the hotkey and speak your show notes into any app."},{"title":"Review, paste, publish.","body":"Edit the text, paste it into your host or your docs, and ship the episode. You control all of it."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons podcasters choose OpenWhispr","subtitle":"Transcription cost recurs with every episode. These are the practical details to compare.","items":[{"title":"Unlimited local transcription","body":"No per-minute fees or monthly hour caps for local transcription on the Free plan."},{"title":"Free for your whole show","body":"Unlimited local dictation and transcription at $0: you, your co-host, your editor. No per-seat licence."},{"title":"Dictate straight into your host","body":"A system-wide hotkey drops clean text wherever your cursor is: Buzzsprout, Notion, Substack, YouTube Studio."},{"title":"Summaries drafted, not just transcribed","body":"Connect your own AI key, or run a local model, to turn a transcript into an episode summary or description."},{"title":"A dictionary for your guests' names","body":"Teach it the show title, recurring guests and niche terms once. The custom dictionary improves how it handles them in dictated notes and drafted summaries."},{"title":"Works fully offline","body":"Transcribe on the flight home from the conference. Local models never need a signal."},{"title":"Pull quotes without scrubbing","body":"The transcript is text on your machine. Search it, copy the line, make the clip."},{"title":"Unreleased audio stays on your machine","body":"In local mode nothing is uploaded. Embargoed interviews and unshipped episodes never touch a server."},{"title":"Open source, auditable by anyone","body":"OpenWhispr is open source under the MIT License."},{"title":"100+ languages","body":"Record in the language your audience actually speaks, transcribed locally."}]},"faq":{"label":"Before you download","title":"The questions podcasters ask first.","subtitle":"Straight answers on cost, accuracy and privacy. If yours isn't here, ask us in Discord.","items":[{"question":"What's the best speech-to-text software for podcasting?","answer":"OpenWhispr is a strong option if you want to transcribe recorded episodes locally without per-minute fees. It can turn imported audio into a transcript and lets you dictate show notes wherever your cursor is. Its open-source code makes the workflow inspectable. Trying it costs a download."},{"question":"Is it really free?","answer":"Yes. Local dictation and transcription are included at $0 on macOS, Windows and Linux. The Free plan also includes 2,000 words a week of OpenWhispr Cloud transcription and 5 hours a month of meeting recordings. Pro is $8/month with unlimited Cloud transcription and 20 meeting-recording hours. Business includes unlimited meeting recordings."},{"question":"Will it get my guests' names and niche terms right?","answer":"OpenWhispr supports 100+ languages. Its custom dictionary lets you add your show's title, guest names and niche vocabulary to improve how they appear in dictation and AI-drafted show notes."},{"question":"Does it transcribe my recorded episodes, or only what I say live?","answer":"Both. Import an episode you've already recorded, or record a session in the app, and the transcript is made on your device. Press the hotkey to dictate show notes into any app, and connect your own AI key, or a local model, to draft summaries from the transcript."},{"question":"Does it work offline?","answer":"Yes, once the models are downloaded at setup. After that they run entirely offline. Transcribe an interview on the plane and publish when you land."},{"question":"Do unreleased episodes stay private?","answer":"In local mode, the audio stays on your machine. OpenWhispr is open source under the MIT License. Cloud transcription is optional: choose local for offline work or Cloud for high accuracy and speed without downloading models."}]},"reviews":{"label":"What users say","title":"Built for people who'd rather talk than type.","subtitle":"What people say about OpenWhispr.","score":{"value":"0","label":"Invoices for your own words","body":"Local transcription runs on your machine. There is no meter, no hour cap, and no bill at the end of the month."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for podcasters","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Get unlimited Cloud transcription"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Podcasters who want free, unlimited local transcription and are fine with a quick local setup"}]},{"parts":[{"text":"Podcasters who want easy cloud transcription and mobile dictation"}]},{"parts":[{"text":"Networks and teams producing shows at volume"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited local transcription, with your audio staying on your machine"}]},{"parts":[{"text":"Unlimited cloud speech-to-text without managing local models or keys"}]},{"parts":[{"text":"Unlimited cloud recordings plus advanced workflow features for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"Your words are the show. You shouldn't rent them back.","paragraphs":["OpenWhispr started with a simple objection: turning your own voice into text shouldn't cost by the minute or pass through someone else's server.","We build in public on GitHub. Local transcription is not just a privacy-policy promise. It is a property of the code you can inspect.","And if anything breaks the night before an episode drops, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Stop paying by the minute for your own voice.","body":"Download OpenWhispr, pick a local model at setup, and your next episode gets transcribed, summarized and shipped on your own computer.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free desktop app for macOS, Windows and Linux."}}},"useCaseProductTeams":{"breadcrumb":{"home":"Home","useCases":"Use Cases","productTeams":"For Product Teams"},"metadata":{"title":"Free AI Meeting Notes For Product Teams, No Bot - OpenWhispr","description":"OpenWhispr is free, open-source AI meeting notes for product teams. No bot joins the call, and with a local model the roadmap never leaves your machine.","keywords":["AI meeting notes","meeting notes for product teams","local meeting transcription","private meeting notes","bot-free notetaker","offline meeting notes"]},"schema":{"name":"OpenWhispr","applicationCategory":"Meeting notes software","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source AI meeting notes for product teams. With a local model the call is transcribed on the user's own machine with no bot joining, and written up as decisions, action items and owners. Local meeting notes are unlimited and free, and work fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"AI meeting notes for product teams","title":"Run the meeting. Get the decisions written up.","body":"Get a clean meeting write-up with decisions and next steps. OpenWhispr captures the call on your own computer; pick a local model and the recording never reaches a server, ours or anyone's.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Two people listen to a colleague during a meeting around a table"}},"trustBar":{"label":"Paste the notes into:","names":["Linear","Jira","Notion","Confluence","Slack","Productboard"]},"pain":{"label":"The trade-offs product teams notice","title":"Your roadmap shouldn't live on a notetaker's server.","subtitle":"You want the decisions written down. You don't want a bot in every call, or your strategy sitting on a transcription vendor's cloud.","cards":[{"title":"A bot shows up to every meeting.","body":"Meeting bots can make a sensitive discovery call feel different. OpenWhispr captures your computer's audio, so it does not join the call."},{"title":"Your strategy on a vendor's server.","body":"Roadmap, pricing, competitive analysis, reorg talk: product meetings can cover sensitive work. Choose a local model when you want transcription to run on your device."},{"title":"Someone stops contributing to take notes.","body":"The PM types instead of facilitating, and the decision still gets lost. OpenWhispr writes up the decisions, action items and owners, so everyone stays in the discussion."},{"title":"Per-seat pricing for the whole org.","body":"Choose the mode and plan that fit the team. Local transcription is unlimited on every plan; Free includes 5 hours of meeting recordings a month, Pro adds unlimited Cloud transcription and 20 recording hours, and Business includes unlimited meeting recordings."}]},"features":{"label":"What you get","title":"AI meeting notes for product teams, made on your own laptop.","subtitle":"Pick local models and the roadmap discussion is transcribed and summarized on your machine.","items":[{"image":{"label":"A person gestures while speaking to a colleague across a table"},"title":"Captures the call from your computer, no bot","body":"Listens to your computer's audio on Zoom, Meet, Teams or a Slack huddle, or to the mic for a standup in the room. Transcribed on your device with a local model, and nothing joins the call."},{"image":{"label":"Two colleagues review a laptop together during a group discussion"},"title":"Decisions, action items and owners","body":"Turns your rough notes and the transcript into a structured write-up: what was decided, what happens next, who owns it. You review before it's shared."},{"image":{"label":"A person gestures while working on a laptop at a table"},"title":"Paste it where the work lives","body":"Copy the clean write-up into Linear, Jira, Notion or Slack. It's text, so there's nothing to connect to your tracker."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Roadmap calls sent to us","value":"None","footnote":"pick a local model at setup and there is no upload step in the code"},{"label":"Cost per person on the call","value":"$$0","footnote":"unlimited local meeting notes, the whole org, forever"},{"label":"Setup takes","value":"5 min","footnote":"from download to your first meeting written up"}]},"howItWorks":{"label":"How it works","title":"Leave the call with the decisions already written down.","subtitle":"Download the app, choose a model, then start the call.","image":{"label":"A person holding a tablet speaks in an open office"},"steps":[{"title":"Download for free.","body":"macOS 14.2+, Windows or Linux. Pick a local model during setup."},{"title":"Start the meeting.","body":"Open Zoom, Meet or Teams as usual. OpenWhispr captures your computer's audio and transcribes it on your device."},{"title":"Review, paste, move on.","body":"Hit Generate Notes, edit the write-up and paste it where the work lives. With a local model, transcription runs on your device."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons product teams choose OpenWhispr","subtitle":"Choosing a notetaker is a security decision before it's a note-taking one. Here are the facts, and your security team can verify the privacy ones line by line in the repo.","items":[{"title":"The roadmap call never reaches a server","body":"Pick a local model and transcription runs on your machine. The open-source code shows no upload step on that path."},{"title":"No bot in the call","body":"It captures your computer's audio, so nothing appears in the participant list. Your customer never sees a notetaker."},{"title":"Every seat is free, including the quiet ones","body":"Unlimited local meeting notes at $0, for everyone who ever sits in a call. Only the cloud side is metered."},{"title":"Works with any meeting app","body":"Zoom, Meet, Teams, Slack huddles, Webex, a standup in the room. It captures audio, not a specific app."},{"title":"Decisions and owners, not just transcripts","body":"The write-up is structured with decisions, action items and follow-ups. Where an owner is clear in the transcript, it uses that speaker label. You don't write the prompt."},{"title":"No meeting-app integration to approve","body":"No meeting-app plugin or bot account to approve. It's a desktop app that listens to your computer's audio."},{"title":"The recording is deleted after the meeting","body":"Temporary diarization audio is removed after processing. You control the recordings and transcripts saved in the app."},{"title":"Read the code before you pick the tool","body":"OpenWhispr is open source. Your security team can read the code on GitHub."},{"title":"No signal, same write-up","body":"A standup in a room with bad WiFi, a customer site, a plane. Local models need no signal."},{"title":"Who said what, worked out on your machine","body":"Speaker labels are computed on your own device, free on every plan. The write-up says who owns cutting scope, rather than merely that someone said it."}]},"faq":{"label":"Before you download","title":"The questions product teams ask first.","subtitle":"Straight answers on bots, your data and cost. If yours isn't here, ask us in Discord.","items":[{"question":"Does the meeting audio ever leave my device?","answer":"Not in local mode. Pick a local model during setup and transcription runs on your machine with no upload step. Cloud transcription is optional, and we never train on your data."},{"question":"Does a bot join the call?","answer":"No. OpenWhispr captures your computer's audio, so nothing appears in the participant list. Zoom, Meet, Teams, a Slack huddle or a standup in the room all work the same way because it listens instead of joining. With a local model, transcription runs on your device. Tell people you're recording, same as you would with a bot."},{"question":"Will it get our feature names and acronyms right?","answer":"Add them to the custom dictionary, free on every plan. The write-up is told to use those exact spellings, so codenames and feature names come out right in the notes you paste."},{"question":"Do the AI meeting notes capture decisions and owners, not just a transcript?","answer":"Yes. Jot rough notes during the call, then run the write-up when it ends. AI combines them with the transcript into decisions, next steps and action items. Where an owner is clear, it uses that speaker label. You review before anything is shared."},{"question":"Will it work for an in-person standup with no call?","answer":"Yes. It records from your mic, and local models run entirely offline. Capture the standup, then paste the write-up into Linear or Notion when you're done."},{"question":"Is it actually free?","answer":"Yes. Unlimited local meeting notes, 100+ languages, all three desktop platforms: $0 forever, for every seat. Only the cloud side is metered: cloud meeting recordings run 5 hours a month free, 20 on Pro ($8/month), unlimited on Business ($16/user/month)."}]},"reviews":{"label":"What users say","title":"Built by people who don't want a bot in their meetings either.","subtitle":"From our Discord, GitHub and inboxes, the same instinct that matters when the roadmap's on screen.","score":{"value":"0","label":"Roadmap calls on a vendor's server","body":"In local mode, transcription runs on your device. Cloud transcription is optional, and its audio is discarded after transcription."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for product teams","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Product teams who want private meeting notes and are fine with a quick local setup"}]},{"parts":[{"text":"Product teams who want easy, reliable cloud transcription"}]},{"parts":[{"text":"Product orgs in calls all day"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Free local meeting notes, with the audio staying on your device"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models or keys"}]},{"parts":[{"text":"Unlimited recordings plus advanced workflow features for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for top accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","tooltip":"Cloud meeting recordings. Recording and transcribing meetings with a local model is unlimited on every plan.","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"Some meetings decide the roadmap. We built for those.","paragraphs":["OpenWhispr started with a simple objection: the meeting where you decide the roadmap shouldn't have to pass through a stranger's server to become a write-up.","OpenWhispr is open source. Your security team can inspect how local transcription works before putting it in roadmap meetings.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Keep the call on your laptop.","body":"Download OpenWhispr, pick a local model, and your next meeting gets transcribed, written up and ready to paste on your computer.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, we'll fix it with you in Discord. Free forever on Mac, Windows & Linux."}}},"useCaseStudents":{"breadcrumb":{"home":"Home","useCases":"Use Cases","students":"For Students"},"metadata":{"title":"Free Speech-to-Text Software For Students - OpenWhispr","description":"OpenWhispr is free, open-source speech-to-text software for students. Record the lecture, get the notes written up on your own laptop, at $0 with no card.","keywords":["speech to text software for students","free dictation software for students","lecture transcription app","record lectures and get notes","speech to text for university students","free speech to text for students","student lecture notes app"]},"schema":{"name":"OpenWhispr","applicationCategory":"Speech-to-text software for students","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source speech-to-text software for students. Lectures are recorded and transcribed on the student's own laptop, written up as structured notes with a local AI model, and essays are dictated the same way. Unlimited local transcription is free, and it works fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Speech-to-text software for students","title":"Hit record in the lecture. Get the notes written up.","body":"Turn a lecture into readable notes on your laptop. OpenWhispr records the room or the stream, and one tap on Generate Notes turns the transcript into headings and key points for $0 with a local model.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"A person smiling with a laptop while relaxing on a sofa."}},"trustBar":{"label":"Record lectures from:","names":["Zoom","Microsoft Teams","Panopto","Echo360","Google Meet","YouTube"]},"pain":{"label":"The problem with lecture notes","title":"You cannot listen and write at the same time.","subtitle":"A lecture runs at the speed of speech. Writing notes pulls your attention away from following the argument.","cards":[{"title":"The notes you take are the lecture you missed.","body":"Every sentence you write down is a sentence you stop listening to. Let the recording take the words and keep your attention on the whiteboard instead."},{"title":"A two-hour recording takes two hours to replay.","body":"Audio is hard to revisit when you cannot skim it. OpenWhispr hands back text and structured notes you can scan before replaying anything."},{"title":"Pick the right mode for exam month.","body":"Use local transcription for unlimited on-device transcription at $0, or Cloud for top accuracy and speed without downloading a model. Free includes 2,000 words of Cloud transcription a week and 5 hours of meeting recordings a month; Pro adds unlimited Cloud transcription and 20 recording hours."},{"title":"Your lectures, sitting on a company's server.","body":"Most transcription tools upload the audio to process it. In local mode OpenWhispr keeps the recording on your laptop, and the code is public on GitHub so you can check that yourself."}]},"features":{"label":"What you get","title":"The lecture, written up before you get home.","subtitle":"Recording is the easy half. What makes a lecture useful in week nine is text you can search and notes you can read.","items":[{"image":{"label":"A person listening beside a laptop in a tiered lecture hall."},"title":"Records the room or the stream","body":"In the lecture hall it records through the mic. For a Zoom, Teams or Panopto lecture it captures the audio playing on your laptop. Mac, Windows and Linux."},{"image":{"label":"A person sitting beside a laptop in a campus lounge."},"title":"Notes, not a wall of transcript","body":"The transcript goes through a write-up step that returns headings, key points and action items in markdown. With a local AI model that step runs on your laptop too."},{"image":{"label":"A person speaking while working on a laptop at a library desk."},"title":"Dictation for everything after","body":"Press a hotkey anywhere and talk. Clean text lands in Google Docs, Word or Overleaf, so the essay that follows the lecture gets said out loud instead of typed."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Lecture audio sent to us","value":"None","footnote":"in local mode, recording and transcription run on your laptop"},{"label":"Cost per student","value":"$$0","footnote":"unlimited local transcription and write-ups, no card, no expiry"},{"label":"Setup takes","value":"5 min","footnote":"from download to your first recorded lecture"}]},"howItWorks":{"label":"How it works","title":"Three steps, and you leave with the write-up.","subtitle":"The whole setup is one download and one model choice. There is nothing to sign up for.","image":{"label":"A person wearing an earbud beside a laptop and handwritten notes."},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model during setup and the whole thing runs offline."},{"title":"Hit record when the lecture starts.","body":"It takes the room through your mic, or the stream through your laptop's audio. Then you just listen."},{"title":"Tap Generate Notes.","body":"The transcript comes back as headings, key points and action items. Search those in revision week instead of scrubbing audio."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons students choose OpenWhispr","subtitle":"On a student budget, cost comes first.","items":[{"title":"Free forever, no card details","body":"Unlimited local transcription and note write-ups at $0, with no trial or semester limit."},{"title":"No bot joins the call","body":"OpenWhispr captures audio directly from your system; it does not join meetings as a participant."},{"title":"Your own notes go in the same file","body":"Type a line during the lecture and the write-up folds it in with what was said."},{"title":"Works with no WiFi","body":"Local models run offline. A basement lecture theatre, the train home, a flight: it still records and still transcribes."},{"title":"Recordings stay on your laptop","body":"In local mode the audio is transcribed on your device, and the open-source code shows the local path never uploads it."},{"title":"Searchable in revision week","body":"Search a term in your write-ups instead of scrubbing audio."},{"title":"Use a custom dictionary in the write-up","body":"Add lecturer names, module codes and technical terms to the custom dictionary to give the write-up the spellings you use."},{"title":"100+ languages","body":"Sit a lecture in your second language and read the notes in it, or dictate your essay in your first."},{"title":"Upgrade for unlimited cloud transcription","body":"The free plan includes 2,000 words a week of OpenWhispr Cloud transcription. Pro is $8/month for unlimited Cloud transcription. Students get 40% off for 12 months; email support@openwhispr.com with proof of enrolment for a code."},{"title":"Open source, under the MIT license","body":"Read the code before you install it. No privacy policy to take on faith."}]},"faq":{"label":"Before you download","title":"The questions students ask first.","subtitle":"Straight answers on cost, permission and what your university will think. If yours is not here, ask us in Discord.","items":[{"question":"Is it actually free, or free for a month?","answer":"Actually free. Local transcription, local note write-ups, 100+ languages and all three desktop platforms cost $0 forever, with no card details. The free plan also includes 2,000 words a week of OpenWhispr Cloud transcription and 5 hours a month of cloud meeting recordings. Pro is $8/month if you want unlimited Cloud transcription. Students get 40% off for 12 months; email support@openwhispr.com with proof of enrolment for a code."},{"question":"Am I allowed to record a lecture?","answer":"That is your university's call, not ours, and it varies by institution and by country. Most have a policy on recording teaching, often tied to accessibility. Check yours, and ask the lecturer. OpenWhispr does not record anything until you press the button."},{"question":"Does it work for online lectures as well as in-person ones?","answer":"Yes. In a lecture hall it records through your mic. For a Zoom, Teams, Google Meet or Panopto lecture it captures the audio playing on your laptop, on Windows, Linux and macOS 14.2 or newer. Already have the file, from your phone or a lecture-capture download? Drop it in and it transcribes that too, with no size limit on a local model."},{"question":"Will it get my subject's terminology right?","answer":"It runs Whisper-class models trained on a wide range of accents and 100+ languages. For the words a general model has never seen, add them to the custom dictionary in Settings: lecturer names, module codes and technical terms. Those spellings carry into your dictation and into the write-up of your notes."},{"question":"Is using the AI write-up cheating?","answer":"Transcription is not: the words are the ones your lecturer said, and reading them back is no different from reading your own notes. The write-up step is different, because a model summarises the transcript for you, so check your university's rules on AI before any of it goes into work you hand in."},{"question":"Where does my recording actually go?","answer":"It depends on the mode you pick during setup. OpenWhispr Cloud is the default: audio is sent only to transcribe it, and is not stored afterwards. Choose a local model instead and the answer is nowhere. It is transcribed on your laptop, and the open-source code shows the local path has no upload step. You can switch in Settings any time."}]},"reviews":{"label":"What users say","title":"Built by people who take notes in bad WiFi too.","subtitle":"From our Discord, GitHub and inboxes.","score":{"value":"0","label":"Lecture recordings kept on our servers","body":"In local mode the recording never leaves your laptop. On cloud, audio is sent only to transcribe it and is not stored after that."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for students","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Students recording lectures on their own laptop, with unlimited local transcription and notes"}]},{"parts":[{"text":"Students who want cloud transcription that just works, with no local models to download"}]},{"parts":[{"text":"Research groups and societies recording at volume"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"No meter and no card on the local side, with the recording staying on your laptop"}]},{"parts":[{"text":"Unlimited cloud speech-to-text without managing local models or keys"}]},{"parts":[{"text":"Unlimited cloud meeting recordings plus speaker labels for group work"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"A lecture you can read shouldn't need a subscription.","paragraphs":["OpenWhispr is for students who want to turn lectures into readable notes without a subscription.","We build in public. The code and releases are on GitHub, where 5,000+ engineers have starred the project. When we say the recording stays on your laptop, that's not a promise in a privacy policy. It's a property of the code, and you do not have to take our word for it.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Turn your 9am into notes with one tap.","body":"Download OpenWhispr, hit record when the lecture starts, and read the notes on the way home. On your laptop, at $0, with no card details.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free forever on Mac, Windows & Linux."}}},"useCaseSpanishVeterinary":{"breadcrumb":{"home":"Home","useCases":"Use Cases","spanishVeterinary":"For Spanish-speaking vets"},"metadata":{"title":"Free Spanish Veterinary Dictation Software - OpenWhispr","description":"OpenWhispr turns Spanish veterinary dictation into text for QVET, Provet Cloud, or any clinical record, with free local models available.","keywords":["spanish veterinary dictation software","veterinary dictation software spanish","spanish vet dictation","veterinary transcription spanish","dictado veterinario en español","software de dictado veterinario"]},"schema":{"name":"OpenWhispr","applicationCategory":"Spanish veterinary dictation software","operatingSystem":"macOS, Windows, Linux","description":"Free, open-source dictation software for veterinary teams who document in Spanish. Select Spanish, dictate, review the transcript, and paste it into the clinical record you already use.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Spanish veterinary dictation software","title":"Dictate in Spanish. Get text ready for your PIMS.","body":"Put the cursor in the historia clínica, press the hotkey, and dictate in Spanish. Review the text before you file it.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Veterinarian working at a clinic computer beside a dog"}},"trustBar":{"label":"Use alongside:","names":["QVET","Provet Cloud","GVET","WinVet","Wirevet","Microsoft Word"]},"pain":{"label":"The problem with Spanish clinical documentation","title":"The consult ends. The historia clínica is still blank.","subtitle":"You still owe the anamnesis, exploración, diagnóstico, tratamiento, informe de alta, and sometimes a prescription. Every line still needs your review.","cards":[{"title":"The Spanish note starts after the patient leaves.","body":"The findings are fresh, but the historia clínica is empty. Dictate the note while you still remember the sequence, then review it before filing."},{"title":"Drug names, doses, and accents need a final check.","body":"Select Spanish before you dictate. Store recurring patient names, drug names, and clinic abbreviations in the custom dictionary, then check every dose and clinical term before filing."},{"title":"Your PIMS does not need a new integration.","body":"OpenWhispr does not connect to QVET, Provet Cloud, GVET, WinVet, Wirevet, or Word. It produces text you can paste into a standard editable field, so you keep the workflow you already use."},{"title":"A prescription still needs a veterinarian.","body":"Dictation can type the details. It cannot examine the animal, choose treatment, validate required fields, or sign the prescription. You stay responsible for reviewing the record in the approved system."}]},"features":{"label":"What you get","title":"Pick Spanish. Keep your PIMS.","subtitle":"Select Spanish, press the hotkey, then review what lands in the field. No migration and no new record system.","items":[{"image":{"label":"Veterinarian examining a cat beside an open laptop"},"title":"Spanish is an actual speech setting","body":"Select Spanish during setup or in settings. The shipped app lists Spanish for local Whisper and Parakeet models as well as supported cloud transcription."},{"image":{"label":"Veterinarian using a laptop beside a cat in a carrier"},"title":"From the microphone to a PIMS-ready transcript","body":"Use the system-wide hotkey in a clinical history, discharge, referral, email, or Word document. Automatic paste is a setting and needs the required OS permission. The clipboard option keeps text available for manual paste."},{"image":{"label":"Man using a laptop in a stable tack room"},"title":"A local option when the connection drops","body":"Download a local model once, then Spanish dictation can run on the computer without internet. Local models are unlimited and included on every plan."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"New PIMS required","value":"None","footnote":"paste into the veterinary record system you already use"},{"label":"Cost to start","value":"$$0","footnote":"unlimited local models are included on the Free plan"},{"label":"Setup takes","value":"5 min","footnote":"from download to the first Spanish dictation"}]},"howItWorks":{"label":"How it works","title":"From spoken Spanish to the right clinical field.","subtitle":"OpenWhispr handles the dictation step. You keep the PIMS, the clinical judgment, and the final review.","image":{"label":"Veterinarian using a laptop while comforting a dog on an exam table"},"steps":[{"title":"Download for free.","body":"Install on Mac, Windows, or Linux. Select Spanish as a spoken language."},{"title":"Put the cursor in the record.","body":"Open the relevant field in QVET, Provet Cloud, GVET, WinVet, Wirevet, or Word, then press the hotkey and speak."},{"title":"Review every clinical term.","body":"Check patient names, drug names, doses, accents, and required fields before you save or sign anything."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons Spanish-speaking vets choose OpenWhispr","subtitle":"Spanish is a speech setting in the app. The hotkey, local models, dictionary, and automatic-paste or clipboard workflow stay the same across your desktop.","items":[{"title":"Spanish appears in the language picker","body":"Spanish is a selectable spoken language in the shipped onboarding and settings flow."},{"title":"Two local model families list Spanish","body":"The app's model registry includes Spanish for local Whisper and multilingual Parakeet transcription."},{"title":"No dedicated PIMS integration required","body":"If a PIMS field accepts pasted text, use automatic paste or the clipboard. No dedicated QVET, Provet Cloud, GVET, WinVet, or Wirevet integration is claimed."},{"title":"Use the sections your team already knows","body":"Dictate anamnesis, exploración, diagnóstico, tratamiento, evolución, or informe de alta without changing the record's structure."},{"title":"Keep recurring terms in one list","body":"The custom dictionary stores the patient names, drug names, procedures, and abbreviations you add. You still verify each result."},{"title":"Local Spanish dictation can work offline","body":"After the local model is downloaded, transcription runs on the computer without an internet connection."},{"title":"Unlimited local models cost $0","body":"The Free plan includes unlimited local models, Spanish support, and the custom dictionary."},{"title":"Mac, Windows, and Linux","body":"Use the same Spanish dictation workflow across all three supported desktop platforms."},{"title":"The cleanup instruction is set for Spanish","body":"The instruction asks for Spanish grammar, spelling, and punctuation while keeping naturally spoken English technical terms. Review the result before filing."},{"title":"It does not pretend to prescribe","body":"OpenWhispr types what you dictate. The veterinarian still examines the animal, chooses treatment, checks the required fields, and signs in the approved system."}]},"faq":{"label":"Before you download","title":"The questions Spanish-speaking vets ask first.","subtitle":"Straight answers on Spanish support, terminology, PIMS workflow, offline use, prescriptions, and cost.","items":[{"question":"Does OpenWhispr actually support Spanish dictation?","answer":"Yes. Spanish is a selectable speech language in the shipped app. The code lists it for local Whisper, local multilingual Parakeet, and supported cloud transcription. This page does not claim equal accuracy for every accent or vocabulary."},{"question":"Will it recognize every veterinary term and drug name?","answer":"No blanket accuracy claim. The closed product sources do not document veterinary-specific or Spanish-veterinary tuning for the included local models. Add recurring terms to the custom dictionary, then review patient names, drugs, doses, abbreviations, and diagnoses before filing."},{"question":"Does it work with QVET, Provet Cloud, GVET, WinVet, and Wirevet?","answer":"No dedicated integration is claimed. If the field accepts pasted text, press the system-wide hotkey and dictate. Automatic paste needs the relevant setting and OS permission; the clipboard option supports manual paste."},{"question":"Can Spanish veterinary dictation work without internet?","answer":"Yes, in local mode after you download a model. Local models run on the computer with no internet required. Cloud transcription still needs a connection."},{"question":"Can OpenWhispr create or submit a veterinary prescription?","answer":"No. It can type the details you dictate into the active field, but it does not prescribe, validate a jurisdiction's required fields, submit to a prescription platform, or sign on your behalf. The veterinarian reviews and completes that workflow."},{"question":"Is it actually free?","answer":"Yes. The Free plan is $0 and includes unlimited local models, Spanish support, the custom dictionary, and Mac, Windows, and Linux. OpenWhispr Cloud on Free is limited to 2,000 words a week; Pro is $8 a month and makes cloud transcription unlimited."}]},"reviews":{"label":"What users say","title":"Real users. Their words, unchanged.","subtitle":"These reviews are not presented as veterinary endorsements. They describe the dictation experience the page is built on.","score":{"value":"ES","label":"Selectable speech language","body":"Spanish appears in the shipped language registry for local and cloud transcription."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for Spanish-speaking veterinary teams","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Vets and techs who want unlimited local Spanish dictation and are fine with a quick model download"}]},{"parts":[{"text":"Vets who want unlimited OpenWhispr Cloud transcription in Spanish"}]},{"parts":[{"text":"Veterinary teams that want unlimited recordings and advanced workflows"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Spanish dictation with local models at $0"}]},{"parts":[{"text":"Unlimited cloud transcription without downloading a local model"}]},{"parts":[{"text":"Unlimited recordings plus Agent mode and chat over your data"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" stores terms you add"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" stores terms you add"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" stores terms you add"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Why we built it","title":"Spanish should be a setting, not a workaround.","paragraphs":["OpenWhispr gives veterinary teams a direct path from spoken Spanish to the clinical field already open on screen. The app handles dictation. The veterinarian keeps the terminology review, clinical judgment, and signature.","We build in public. Every release and every line of the app is on GitHub, where 5,000+ engineers have starred and forked the project. So when we say the dictation can stay on your machine, nobody in your practice has to take our word for it. They can read the code.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"One of OpenWhispr's founders"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Put the cursor in the historia clínica. Dictate in Spanish.","body":"Review the text while the case is still fresh, then file it in the system you already use.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, we'll fix it with you in Discord. Free forever on Mac, Windows & Linux."}}},"useCasePhysicians":{"breadcrumb":{"home":"Home","useCases":"Use Cases","physicians":"For Physicians"},"metadata":{"title":"Free Physician Dictation Software for EHR Notes - OpenWhispr","description":"OpenWhispr is free physician dictation software for HPI, exam and assessment text at your EHR cursor, with on-device transcription.","keywords":["physician dictation software","medical dictation for physicians","EHR dictation software","clinical note dictation","voice recognition for physicians","offline physician dictation"]},"schema":{"name":"OpenWhispr","applicationCategory":"Physician dictation software","operatingSystem":"macOS, Windows, Linux, iOS","description":"Open-source physician dictation software with optional on-device transcription. Put HPI, exam and assessment text at the active cursor with unlimited local dictation at no cost.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Physician dictation software for clinical notes","title":"Press the hotkey. Fill the chart field you're in.","body":"Dictate the HPI, exam, assessment and plan into the EHR field you already opened. Choose a local model so transcription runs on your computer. OpenWhispr is free and open source.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Physician dictating at a computer in an exam room"}},"trustBar":{"label":"Physicians chart in:","names":["Epic","Oracle Health","athenaOne","eClinicalWorks","MEDITECH","Microsoft Word"]},"pain":{"label":"The problem with charting by keyboard","title":"The visit ended. The chart did not.","subtitle":"HPI, exam, assessment, plan, inbox. Every empty field is another piece of the visit waiting to be typed.","cards":[{"title":"Open charts stack up between patients.","body":"The encounter is over, but the HPI and assessment are still blank. Dictate while the details are fresh."},{"title":"Every field breaks your train of thought.","body":"Click HPI, type. Click exam, type. Click assessment, type again. OpenWhispr leaves the EHR open and puts the transcript at the active cursor."},{"title":"One drug name becomes five corrections.","body":"Generic dictation trips over medication names, eponyms and local abbreviations. Add the terms you use to OpenWhispr's custom dictionary."},{"title":"A five-minute trial needs an IT ticket.","body":"You want to test dictation on one chart, not provision a practice. Local mode starts with a desktop download and no account."}]},"features":{"label":"What you get","title":"Less keyboard. Fewer open charts.","subtitle":"Physician dictation software should fit the note template already on screen. OpenWhispr turns speech into text at your cursor without making you rebuild the workflow around it.","items":[{"image":{"label":"Physician dictating into a desktop microphone at a workstation"},"title":"HPI and assessment at the active cursor","body":"Open the note, place the cursor and press the hotkey. Your dictation is transcribed and pasted into that field, ready for review."},{"image":{"label":"Physician holding a medication bottle beside a computer"},"title":"Your medication names in the dictionary","body":"Add the drugs, procedures, clinician names and abbreviations that keep coming back wrong. OpenWhispr passes those terms to transcription."},{"image":{"label":"Clinician dictating beside a laptop inside a van"},"title":"Keep dictating when ward Wi-Fi drops","body":"Download a local model once. Then transcribe on the workstation in clinic, at home or on rounds without an internet connection."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Audio uploaded in local mode","value":"None","footnote":"in local mode, transcription runs on the physician's device"},{"label":"Cost per physician","value":"$$0","footnote":"unlimited local dictation on every desktop plan"},{"label":"Setup takes","value":"5 min","footnote":"from download to your first dictated chart field"}]},"howItWorks":{"label":"How it works","title":"Finish the chart you already have open.","subtitle":"Keep the EHR workflow you already know. OpenWhispr adds a hotkey and local transcription, with no account or card required to start.","image":{"label":"Physician working on a laptop at night"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Choose a local model during setup."},{"title":"Put the cursor in the chart.","body":"Open the HPI, exam or assessment field, press the hotkey and dictate."},{"title":"Review and sign.","body":"Read the inserted text, correct what needs correcting, then finish the note in the EHR."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons physicians choose OpenWhispr","subtitle":"Try one workstation, one chart field and one local model. These are the ten things to check.","items":[{"title":"No copy-and-paste detour","body":"Dictate inside the open chart field. Review the inserted text there, then continue the note."},{"title":"One hotkey across desktop apps","body":"Use the same dictation control in the EHR, a portal reply, an email or a document."},{"title":"A dictionary for your clinical vocabulary","body":"Add drug names, procedures, eponyms and abbreviations as recognition hints for supported transcription and cleanup paths."},{"title":"Local mode stays on the workstation","body":"Choose a local model and dictation is transcribed on the device. Cloud fallback is a separate opt-in setting."},{"title":"Works fully offline","body":"Once the local model is downloaded, dictation keeps working without internet in clinic, at home or on a ward."},{"title":"Unlimited local dictation at $0","body":"Every desktop plan includes unlimited use of local speech-to-text models."},{"title":"Mac, Windows and Linux","body":"Use it on the desktop operating system your workstation already runs."},{"title":"Open source, inspectable by IT","body":"Your organization can review the code on GitHub instead of taking the local-mode claim on faith."},{"title":"100+ languages","body":"OpenWhispr supports 100+ languages. Availability varies by local model and cloud provider."},{"title":"Local or cloud dictation in Settings","body":"Choose Local or Cloud for dictation during setup, and change that dictation setting later when your workflow changes."}]},"faq":{"label":"Before you download","title":"What physicians ask before changing dictation software.","subtitle":"Straight answers on EHR fields, terminology, patient data, offline use and cost.","items":[{"question":"What is the best physician dictation software for EHR notes?","answer":"If you want a hotkey that puts the transcript at your active cursor, unlimited local dictation at $0, and code your IT team can inspect, try OpenWhispr on one workstation. It does not replace the physician's review or the EHR's signing workflow."},{"question":"Does OpenWhispr integrate with my EHR?","answer":"OpenWhispr is not an EHR integration. It transcribes your speech and pastes the text at the active cursor, so you can use it in editable chart fields and other desktop apps. Test it with your organization's EHR and virtual-desktop setup before a wider rollout."},{"question":"How does it handle medication names and clinical abbreviations?","answer":"OpenWhispr includes a custom dictionary. Add the medication names, procedures, eponyms and abbreviations you use, and those terms are supplied to transcription. You still review the note before signing."},{"question":"Can I dictate patient information?","answer":"Your organization decides which tools are approved for patient information. In local mode, transcription runs on the device without uploading the audio. Confirm your organization's policy before use."},{"question":"Will it work when the hospital network or Wi-Fi does not?","answer":"Yes, after you download a local model. Local transcription runs on Mac, Windows and Linux without internet. Cloud transcription remains a separate mode you can choose in Settings."},{"question":"Is it actually free?","answer":"Yes. Every desktop plan includes unlimited local dictation, a custom dictionary and 100+ languages at $0. Pro is $8 per month and adds unlimited cloud transcription and the iPhone and iPad app."}]},"reviews":{"label":"What users say","title":"What users notice when voice replaces the keyboard.","subtitle":"Verbatim notes from our Discord, GitHub and inboxes.","score":{"value":"0","label":"Per-physician licences for local dictation","body":"Unlimited local dictation is included at $0 on every desktop plan."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Start free with local dictation.","Upgrade for cloud or team features."],"caption":"Compare OpenWhispr plans for physicians","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Physicians who want free offline dictation at the active cursor and can set up a local model"}]},{"parts":[{"text":"Physicians who want unlimited cloud transcription and mobile dictation"}]},{"parts":[{"text":"Practices and teams documenting at volume"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited on-device dictation with no per-physician fee"}]},{"parts":[{"text":"Unlimited cloud dictation without downloading a local model"}]},{"parts":[{"text":"Unlimited recordings and priority support for a practice"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"The path from microphone to chart should be inspectable.","paragraphs":["Physicians should not have to guess what a dictation tool does between the microphone and the chart. OpenWhispr makes that path visible in public code.","Your organization can inspect the local-mode implementation on GitHub instead of relying only on a privacy page.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"OpenWhispr founder and community"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Finish the chart before you open the next one.","body":"Review the inserted text, sign the note and move on.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, ask us for help in Discord. Free forever on Mac, Windows & Linux."}}},"useCaseDoctors":{"breadcrumb":{"home":"Home","useCases":"Use Cases","doctors":"For Doctors"},"metadata":{"title":"Free Doctor Dictation Software, No Middleman - OpenWhispr","description":"Doctor dictation software with no middleman: pick a local model and OpenWhispr types your letters on your own machine in seconds. Free and open source.","keywords":["doctor dictation software","physician dictation software","dictation app for doctors","voice to text for doctors","referral letter dictation","offline dictation for physicians"]},"schema":{"name":"OpenWhispr","applicationCategory":"Physician dictation software","operatingSystem":"macOS, Windows, Linux, iOS","description":"Local-first, open-source dictation software for doctors. With a local model chosen, referral letters, reports and notes are transcribed on the doctor's own device and never leave it. Free for unlimited local dictation, and works fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Dictation software for doctors & physicians","title":"Speak the referral. Send it before the patient leaves.","body":"Choose a local model at setup and the referral, report or patient letter is typed and drafted on your own computer. No transcription service, no per-seat licence, no one else hearing the audio.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Doctor speaking into a desk microphone at a computer"}},"trustBar":{"label":"Dictate into:","names":["DrChrono","Practice Fusion","Tebra","AdvancedMD","Outlook","Microsoft Word"]},"pain":{"label":"The problem with dictation services","title":"You already dictate. Now choose where it runs.","subtitle":"Choose local processing for sensitive or offline dictation, or Cloud for top accuracy and speed without models to download.","cards":[{"title":"Keep sensitive dictation on your device.","body":"With a local model, OpenWhispr transcribes the recording on your machine."},{"title":"Get the text while the case is still fresh.","body":"OpenWhispr turns your dictation into text as you work, so you can review it before sending."},{"title":"Free local dictation for every doctor.","body":"OpenWhispr includes unlimited local AI models on every plan, free. Choose Cloud when you want top accuracy and speed without models to download."},{"title":"The code is the confidentiality agreement.","body":"Closed tools hand you a privacy policy and ask for faith. OpenWhispr's code is public on GitHub, so anyone in the practice can verify that in local mode, patient audio never leaves the machine."}]},"features":{"label":"What you get","title":"Instant turnaround. Nobody else in the loop.","subtitle":"Doctor dictation software used to mean a slow service or a per-seat licence. Keep the typed letter and drafted report without sending dictation through a service or buying another seat.","items":[{"image":{"label":"Doctor using a desktop computer in an office"},"title":"Typed on your machine, in seconds","body":"Dictate the referral, report or reply into your mic. The words land on your screen seconds after you stop talking, transcribed locally and never uploaded."},{"image":{"label":"Doctor writing in a notebook beside a computer"},"title":"Dictate the detail. Review the text.","body":"Speak the gist, then review the transcription, edit it and use it in your referral, report or patient instructions."},{"image":{"label":"Doctor typing at a clinical workstation"},"title":"One hotkey. Any window, any EHR.","body":"Press the hotkey and speak into whatever is open: the chart, a reply to a colleague, a report in Word. The text lands at your cursor, punctuated, processed locally."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Audio sent to a transcription service","value":"None","footnote":"in local mode, dictation is transcribed on your device and goes nowhere else"},{"label":"Cost per doctor","value":"$$0","footnote":"unlimited local dictation with no per-doctor licence"},{"label":"Local setup","value":"Choose your model","footnote":"install the app and choose local models"}]},"howItWorks":{"label":"How it works","title":"Turnaround time: the length of the dictation.","subtitle":"Old dictation goes out to a service and comes back tomorrow. Local mode runs on your machine, with no account or card required to start.","image":{"label":"Doctor opening a laptop at a desk"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Install the app and choose a local model."},{"title":"Dictate as you always have.","body":"Speak the referral, report or note. It's typed on your device seconds after you stop."},{"title":"Review, send, delete.","body":"Check the draft, paste it into the EHR or an email, and delete the audio. You keep control of all of it."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons doctors choose OpenWhispr","subtitle":"Your dictation carries a patient's history, so check the code and the settings, not just the brochure.","items":[{"title":"No transcription service in the loop","body":"The recording stays on your device instead of going out for overnight transcription."},{"title":"The audio has nowhere to go","body":"Local mode has no upload step. The recording is transcribed on your machine and stays there."},{"title":"Free for every doctor in the practice","body":"Unlimited local dictation at $0, every seat, forever. No per-doctor licence to renew each year."},{"title":"Works fully offline","body":"Home visits, a locum shift on a ward with no signal, a flight to a conference. Once the model is on your machine, dictation never stalls."},{"title":"Types into the chart, the inbox and Word","body":"One hotkey puts the text at your cursor: a referral template, a colleague's inbox, the record itself."},{"title":"Letters and reports, not just transcripts","body":"Run a local AI model, or connect your own key, to turn a dictated summary into a formatted referral, report or patient letter."},{"title":"Your letters are your files","body":"Recordings and drafts sit on your disk. Keep, encrypt or shred them on your schedule, not a vendor's retention policy."},{"title":"Auditable by whoever runs your IT","body":"Your IT team can inspect the public code instead of relying only on a vendor claim."},{"title":"Learns your drug names and eponyms","body":"A custom dictionary picks up the drugs, procedures and abbreviations your specialty uses, so they come out spelled right."},{"title":"100+ languages","body":"Dictate in the language you think in, transcribed locally. Useful when the note is in one language and the referral goes out in another."}]},"faq":{"label":"Before you download","title":"What doctors ask before switching dictation software.","subtitle":"Confidentiality, cost and accuracy, answered straight. Anything else, ask in Discord.","items":[{"question":"What's the best dictation software for doctors?","answer":"OpenWhispr gives doctors a choice: local dictation on their device for offline or sensitive work, or Cloud for top accuracy and speed without models to download. The free plan includes unlimited local models, 100+ languages and a custom dictionary."},{"question":"Who hears my dictation?","answer":"In local mode, no one: it is transcribed on your machine. In Cloud mode, audio is sent only to transcribe it and is never stored. Cloud is the fast, high-accuracy choice when you do not want to download models."},{"question":"Can I use it with patient information under HIPAA?","answer":"In local mode, patient audio is transcribed on your device and never reaches a third party, so there is no vendor to cover with a BAA. For cloud transcription of PHI, OpenWhispr signs a BAA on the Business plan and up, and holds a SOC 2 Type 2 report and ISO 27001 certification. Whether your workflow is compliant still depends on how you use it; no tool can grant that on its own."},{"question":"Will it get the drug names right?","answer":"It runs Whisper-class models that handle a wide range of accents, and the custom dictionary learns the drug names, procedures and abbreviations your specialty uses. Add a term once and it comes out spelled right from then on."},{"question":"Does it work on home visits, or on a ward with no signal?","answer":"Yes. Once the local model is on your machine, it works offline. When you are online, Cloud provides top accuracy and speed without a model download."},{"question":"Is it really free for the whole practice?","answer":"Yes. Unlimited local dictation on Mac, Windows and Linux, in 100+ languages, at $0, for every doctor in the practice. Pro at $8/month adds unlimited cloud transcription and the iPhone app when you want them."}]},"reviews":{"label":"What users say","title":"Nobody here mails their voice to a stranger either.","subtitle":"From our Discord, GitHub and inboxes.","score":{"value":"0","label":"Consults heard by a transcriber","body":"In local mode the audio never leaves your device. No server stores it, and no transcriber ever hears it."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for doctors","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Doctors who want free, private, offline dictation and are comfortable choosing local models during setup"}]},{"parts":[{"text":"Doctors who want easy, reliable cloud transcription and mobile dictation"}]},{"parts":[{"text":"Practices documenting at volume, or that need a signed BAA"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Free offline dictation, with patient audio staying on your device"}]},{"parts":[{"text":"Unlimited cloud transcription with no local models to manage"}]},{"parts":[{"text":"Unlimited recordings plus advanced workflow features for a practice"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"A patient's history is not training data.","paragraphs":["OpenWhispr started with a simple objection: turning speech into text shouldn't mean the audio lands on someone else's server, or in someone's training set. A patient's history least of all.","We build in public on GitHub. When we say local-mode dictation stays on your device, your practice can inspect the code instead of relying only on a privacy policy.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Keep dictating. Just stop paying someone to type it.","body":"Your next referral goes from your voice to the recipient without passing through anyone else's hands. No service, no seat licence, no waiting.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, we'll fix it with you in Discord. Free forever on Mac, Windows & Linux."}}},"useCaseProductivity":{"breadcrumb":{"home":"Home","useCases":"Use Cases","productivity":"For Productivity"},"metadata":{"title":"Free Dictation App For Productivity - OpenWhispr","description":"OpenWhispr is a free, open-source dictation app for productivity. Press one hotkey to put tasks, notes and replies in the app already open.","keywords":["dictation app for productivity","productivity dictation app","voice typing for productivity","speech to text productivity app","dictate tasks and notes","offline dictation app"]},"schema":{"name":"OpenWhispr","applicationCategory":"Productivity dictation app","operatingSystem":"macOS, Windows, Linux, iOS","description":"Local-first, open-source dictation app for productivity. Press a system-wide hotkey and speak tasks, notes and replies into the app already open. With local models selected, transcription and clean-up run on your own computer and work offline. Unlimited local dictation is free on macOS, Windows and Linux.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Dictation app for productivity","title":"Say the task. Get it into the list.","body":"Capture the thought before it disappears, right in the app you already have open; no separate inbox to check later. Use OpenWhispr Cloud for fast, accurate transcription, or local models to keep the audio on your machine.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Woman speaking while working at a laptop beside a notebook and material samples"}},"trustBar":{"label":"Dictate into the app already open, including:","names":["Todoist","Notion","Obsidian","Gmail","Google Docs","Slack"]},"pain":{"label":"The capture problem","title":"The thought is quick. Typing it down is the slow part.","subtitle":"A task arrives while you are already doing something else. The longer capture takes, the more likely the thought stays in your head.","cards":[{"title":"The task arrives in the middle of another task.","body":"You are in Slack when you remember the invoice, or in a doc when you spot tomorrow's follow-up. Put the cursor in your task list, press the hotkey and say it before the thought disappears."},{"title":"The note is longer than the time you have.","body":"A useful idea becomes three thin bullet points because typing the whole thought feels like work. Talk it into Notion or Obsidian instead. OpenWhispr removes filler and false starts, then pastes the cleaned text at the cursor."},{"title":"Five two-minute replies become an hour of avoidance.","body":"The answer is already in your head. Press the same hotkey in Gmail or Slack, say the reply once and review the text before you send it. No separate recorder to empty later."},{"title":"Your system has names only your system understands.","body":"Project codenames, client names, acronyms and recurring places get mangled by generic dictation. Add them to the custom dictionary once so the transcription and clean-up get the spellings you use."}]},"features":{"label":"One hotkey wherever you type","title":"Keep the system you built. Replace the typing.","subtitle":"One system-wide hotkey works wherever the cursor already is. Choose local models and it keeps working without a connection.","items":[{"image":{"label":"Man speaking into a desk microphone beside a monitor and open notebook"},"title":"Your apps stay exactly where they are","body":"Todoist for tasks, Notion for projects, Obsidian for notes, Gmail for replies. OpenWhispr types into all four. It pastes the finished text into the active field, so there is no new inbox to process and no integration to maintain."},{"image":{"label":"Woman speaking beside a laptop and architectural model"},"title":"A false start does not become a messy task","body":"Say 'Thursday, no, Friday' and the clean-up keeps the correction. It removes filler, fixes punctuation and turns spoken dates into written ones while keeping your wording and intent."},{"image":{"label":"Man working on a laptop and notebook aboard a train"},"title":"The capture habit still works offline","body":"Download local speech-to-text and a local clean-up model once. Then tasks, notes and replies can be transcribed and tidied on your computer with no connection and no per-word bill."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"AI processing location","value":"On your device","footnote":"when local transcription and local clean-up models are selected"},{"label":"Cost to capture locally","value":"$$0","footnote":"unlimited local dictation on every desktop you use"},{"label":"Local setup","value":"A few steps","footnote":"download, choose models, set the hotkey and start speaking"}]},"howItWorks":{"label":"Download to first captured task","title":"Three steps from a thought to the right inbox.","subtitle":"Nothing to migrate and no productivity system to rebuild.","image":{"label":"Woman typing on a laptop backstage beside camera equipment"},"steps":[{"title":"Download it free.","body":"Install OpenWhispr on Mac, Windows or Linux. Choose OpenWhispr Cloud for speed, or download local models for offline use."},{"title":"Put the cursor where the thought belongs.","body":"Open the task field, note, email or message you already use. Press your hotkey and say the whole thought, including the context you would normally skip."},{"title":"Review it and move on.","body":"OpenWhispr cleans the wording and pastes it at the cursor. Check the text, save or send it, and get back to the work you were doing."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"Why it earns a place in the stack","title":"10 reasons people use OpenWhispr in the apps they already use","subtitle":"Each one removes friction from capture without asking you to replace the apps that already work.","items":[{"title":"One hotkey across your apps.","body":"Tasks in Todoist, project notes in Notion, linked thoughts in Obsidian, replies in Gmail and Slack. If automatic paste is unavailable, the text stays on the clipboard."},{"title":"No new inbox to empty.","body":"The words land in the tool where they belong. You do not record a voice memo now and remember to move it later."},{"title":"You approve the text before it becomes a task.","body":"The cleaned text lands in the field first. Fix a name, date or label before you save or send it."},{"title":"Dates look like dates.","body":"Spoken dates, times, numbers and currency are converted to standard written forms before the text reaches your task, note or reply."},{"title":"Your project vocabulary follows you.","body":"Add client names, codenames, acronyms and product terms to the custom dictionary. The same list helps across the apps you dictate into."},{"title":"Unlimited local dictation is free.","body":"No trial clock, no word cap and no card on every desktop plan when you run the speech-to-text model on your own machine."},{"title":"Offline means the capture habit does not break.","body":"Local transcription and a local clean-up model keep working on a plane, during an outage or anywhere the connection drops."},{"title":"100+ languages.","body":"Capture the thought in the language you actually think in. The same app and the same hotkey support more than 100 languages."},{"title":"The privacy claim is code you can inspect.","body":"OpenWhispr is open source. You can inspect the local transcription and paste paths yourself."},{"title":"Mac, Windows and Linux.","body":"Use the same free app on the desktop at work and the laptop at home, with a hotkey you choose for each machine."}]},"faq":{"label":"Before you add voice input to your productivity stack","title":"What people ask before they put OpenWhispr in their stack","subtitle":"What it does, what it does not do and what stays on your machine.","items":[{"question":"Does OpenWhispr create and schedule tasks for me?","answer":"OpenWhispr types finished text where your cursor is. It does not organise or schedule tasks. Todoist, Notion and other apps still apply their own date and label rules, and you review the text before saving."},{"question":"Do I need an integration for every productivity app?","answer":"No. The system-wide hotkey pastes text into the active app, so there is no Todoist plugin, Notion connection or Gmail extension to set up. On Mac, OpenWhispr needs accessibility permission to paste automatically. If automatic paste fails, the transcription remains available to paste from the clipboard."},{"question":"Will it clean up a rambling thought?","answer":"Yes, when clean-up is enabled. The clean-up prompt removes filler words, false starts, stutters and accidental repetition. It fixes grammar and punctuation, converts dates and times to written forms and is explicitly told to preserve your wording, formality and intent."},{"question":"Can OpenWhispr process my dictation locally?","answer":"Yes. Choose local transcription and local clean-up so audio and transcript processing stay on your computer. Once text is pasted, the destination app's own storage and privacy rules apply. OpenWhispr Cloud audio is transcribed and discarded, not stored."},{"question":"Is unlimited dictation really free?","answer":"Unlimited local dictation is $0 on every desktop plan and does not expire. The Free plan also includes 2,000 words a week through OpenWhispr Cloud. Pro is $8 per user each month for unlimited cloud transcription, and Business is $16 per user each month for unlimited cloud meeting recordings and agent mode."},{"question":"Will it spell project names and acronyms correctly?","answer":"Add the exact spellings to the custom dictionary. OpenWhispr passes that vocabulary into supported transcription paths and the clean-up step so client names, project codenames and internal acronyms have more context. You can edit the dictionary any time."}]},"reviews":{"label":"What people tell us","title":"Better than typing. Clean after a stumble. Local without WiFi.","subtitle":"Read them for the tool, not for a productivity method.","score":{"value":"0","label":"Productivity apps to replace","body":"OpenWhispr is the voice input across the stack you already use. Your tasks, notes, docs, email and chat stay where they are."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for productivity","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"People who want unlimited local dictation across their personal productivity stack for $0"}]},{"parts":[{"text":"People who want unlimited cloud dictation without managing local models"}]},{"parts":[{"text":"Teams that want agent mode and unlimited cloud meeting recordings alongside dictation"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Capture tasks, notes and replies locally with no word cap, trial clock or card"}]},{"parts":[{"text":"Unlimited OpenWhispr Cloud transcription with no local model to download"}]},{"parts":[{"text":"Unlimited cloud meeting recordings and agent mode for shared workflows"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Why we built it","title":"A productivity system is only useful if the thought gets into it.","paragraphs":["OpenWhispr turns speech into clean text where the thought belongs: your task list, notes or reply field.","OpenWhispr is open source, so you can inspect the local transcription and paste paths yourself. With local models selected, dictation can stay on your machine.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"One of OpenWhispr's founders"}},"finalCta":{"tagline":"$$0 · open source · your apps stay put","title":"Press the hotkey. Get the thought into the app already open.","body":"Download OpenWhispr, choose a model and dictate into your task list, notes or reply field. No card and no migration.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Unlimited local dictation is free on Mac, Windows and Linux."}}},"useCasePathology":{"breadcrumb":{"home":"Home","useCases":"Use Cases","pathology":"For Pathologists"},"metadata":{"title":"Pathology Dictation Software for Reports - OpenWhispr","description":"OpenWhispr is free pathology dictation software for macroscopic descriptions and microscopic findings, with auto-paste in supported report fields.","keywords":["pathology dictation software","pathologist dictation software","pathology report dictation","anatomic pathology dictation","histopathology dictation software","pathology speech recognition","offline pathology dictation"]},"schema":{"name":"OpenWhispr","applicationCategory":"Pathology dictation software","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source pathology dictation software that transcribes spoken report text for automatic paste into supported fields. Free for unlimited local dictation.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Pathology dictation software","title":"Dictate the findings. Get report text at your cursor.","body":"Keep the specimen or slide in view. Dictate the macroscopic description, microscopic findings or final diagnosis. OpenWhispr types into supported fields. You review and sign out.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Laboratory professional reviewing a computer beside a microscope and slide tray"}},"trustBar":{"label":"Reporting environments to test with OpenWhispr:","names":["Epic Beaker","Oracle Health PathNet","Clinisys PowerPath","Clinisys WinPath","Sectra Digital Pathology","Microsoft Word"]},"pain":{"label":"The problem with typing pathology reports","title":"The slide is in focus. The keyboard pulls you away.","subtitle":"The finding is clear. The field is empty. Every trip to the keyboard interrupts the read.","cards":[{"title":"Your eyes leave the slide for the keyboard.","body":"You see the finding, then look away to type it. OpenWhispr lets you speak the report text while the morphology is still in view."},{"title":"The gross description waits on your hands.","body":"Measurements, margins, sections and the block key arrive in sequence at the cut-up bench. Dictate each observation while it is in front of you, then review the text before it enters the final report."},{"title":"Structured fields still need precise prose.","body":"Synoptic fields capture element-response pairs. Free-text microscopic descriptions, comments and addenda can still need your words. OpenWhispr transcribes those words without building or validating the protocol."},{"title":"One unfamiliar term becomes a correction loop.","body":"Specimen labels, eponyms and local abbreviations recur case after case. Add them to the custom dictionary as context hints, then check every occurrence before sign-out."}]},"features":{"label":"What you get","title":"Keep the case in view. Put the words in the report.","subtitle":"Pathology dictation software should fit the reporting screen already open. OpenWhispr turns speech into text at the active cursor. It does not replace your LIS, protocols or judgment.","items":[{"image":{"label":"Person typing beside magnified tissue images on two monitors"},"title":"Report text at the active cursor","body":"Open an editable report field, place the cursor and press the hotkey. OpenWhispr pastes the transcript when the field and session support automatic paste. Otherwise it copies the text for manual paste."},{"image":{"label":"Laboratory professional using a desktop microphone at a workstation"},"title":"Local transcription when you choose it","body":"Download a local speech model and select Local in Settings. With cloud fallback disabled, dictation stays on the workstation and continues without internet after the model is installed."},{"image":{"label":"Laboratory professional holding a specimen tube and smartphone beside a microscope"},"title":"Your recurring terms as context hints","body":"Add specimen names, eponyms, stains and local abbreviations to the custom dictionary. OpenWhispr feeds them to supported speech models as hints. It does not promise specialist terminology accuracy."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Direct LIS integrations included","value":"None","footnote":"OpenWhispr uses active-cursor paste in compatible fields instead"},{"label":"Cost per pathologist","value":"$$0","footnote":"unlimited local dictation on every desktop plan"},{"label":"Setup takes","value":"5 min","footnote":"from download to your first dictated report text"}]},"howItWorks":{"label":"How it works","title":"Findings in the report before you leave the case.","subtitle":"Speech becomes editable text in the report field. You check it against the case and complete sign-out in the reporting system.","image":{"label":"Healthcare worker in blue scrubs opening a laptop in a hospital corridor"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model during setup."},{"title":"Dictate the findings.","body":"Put the cursor in an editable field, press the hotkey and speak the macroscopic description, microscopic findings, ancillary-study results, addendum or diagnosis text."},{"title":"Review, then sign out.","body":"Check the text against the specimen, slides, ancillary studies and required report elements. The final report remains your call."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons OpenWhispr fits pathology reporting","subtitle":"A pathology report carries exact observations and a final professional judgment. OpenWhispr stays in the narrower lane of turning your speech into editable text.","items":[{"title":"Automatic paste in supported fields","body":"Use the same hotkey in a compatible report field, an email or a document. If automatic paste is unavailable, OpenWhispr leaves the text on your clipboard."},{"title":"No direct LIS plug-in required","body":"Compatible editable fields use desktop dictation and automatic paste. OpenWhispr does not claim a direct integration with Epic Beaker, PathNet, PowerPath, WinPath or another LIS. Test your reporting system and virtual-desktop setup."},{"title":"A local transcription path","body":"Choose Local in Settings and keep cloud fallback disabled. The speech model runs on the workstation and the dictation audio is not uploaded for transcription on that path."},{"title":"Works offline after model setup","body":"Once a local model is downloaded, dictation continues without internet at the microscope, gross bench or home workstation."},{"title":"A dictionary for recurring terms","body":"Add specimen names, stains, eponyms and abbreviations as context hints. The pathologist still checks the transcript before sign-out."},{"title":"Unlimited local dictation at $0","body":"Every desktop plan includes unlimited use of local speech-to-text models."},{"title":"Mac, Windows and Linux","body":"Use OpenWhispr on the desktop operating system your reporting workstation already runs."},{"title":"Open source, inspectable by IT","body":"The desktop app code is public on GitHub, so your technical team can inspect the local and cloud transcription paths."},{"title":"100+ languages","body":"OpenWhispr supports 100+ languages. Availability varies by local model and cloud provider."},{"title":"Local or cloud in Settings","body":"Choose the transcription path during setup and change it later in Settings. Cloud audio is sent only to transcribe it and is never stored by OpenWhispr."}]},"faq":{"label":"Before you download","title":"What pathologists ask before changing dictation software.","subtitle":"Straight answers on reporting fields, terminology, diagnostic boundaries, local processing and cost.","items":[{"question":"Does OpenWhispr integrate with my LIS or digital pathology viewer?","answer":"OpenWhispr is not an LIS or viewer integration. It transcribes speech and pastes text at the active cursor, so you can use it in compatible editable fields and other desktop apps. Test it with your reporting system and virtual-desktop setup before wider use."},{"question":"Does it generate a diagnosis or complete a synoptic report?","answer":"No. OpenWhispr turns what you dictate into editable text. It does not inspect slides, interpret findings, choose a diagnosis, populate required protocol elements or sign out a case. The pathologist reviews every word in the reporting system."},{"question":"Will it recognize pathology terminology?","answer":"OpenWhispr includes a custom dictionary. Add recurring specimen names, stains, eponyms and abbreviations, and supported speech models receive them as context hints. This is not a guarantee of specialist terminology accuracy, so review every report before sign-out."},{"question":"Where is the dictation processed?","answer":"Choose Local and transcription runs on the workstation. Choose OpenWhispr Cloud and the audio is sent only for transcription, then never stored by OpenWhispr. Your laboratory decides which mode and workflow are approved for its data."},{"question":"Will it work without internet?","answer":"Yes, after you download a local speech model. Local transcription runs on Mac, Windows and Linux without internet. Cloud transcription remains a separate mode you can choose in Settings."},{"question":"Is it actually free?","answer":"Yes. Every desktop plan includes unlimited local dictation and a custom dictionary at $0. OpenWhispr supports 100+ languages across available models and providers; local-model availability varies. Pro is $8 per month and adds unlimited cloud transcription, more meeting recording hours, mobile and device sync."}]},"reviews":{"label":"What users say","title":"The dictation is fast. The judgment stays yours.","subtitle":"Existing OpenWhispr user feedback. These testimonials are about the tool, not pathology reporting.","score":{"value":"0","label":"Diagnostic decisions made by OpenWhispr","body":"OpenWhispr transcribes what you say. It does not interpret a specimen or slide, choose a diagnosis or sign out a report."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for pathologists","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Pathologists who want free offline dictation at the active report cursor"}]},{"parts":[{"text":"Pathologists who want unlimited OpenWhispr Cloud transcription"}]},{"parts":[{"text":"Laboratories and teams documenting at volume"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited local report dictation at $0"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models"}]},{"parts":[{"text":"Unlimited recordings and priority support for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users and OpenWhispr features"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" supplies names and terms as context hints"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" supplies names and terms as context hints"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" supplies names and terms as context hints"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" runs user-configured actions on transcripts"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic OpenWhispr team use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, SAML, SCIM, audit logs, retention controls, admin controls, or dedicated support for a larger team? These are enterprise security and data-governance options, not clinical or regulatory certification.","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"The report remains yours. OpenWhispr handles the typing.","paragraphs":["OpenWhispr handles the mechanical part: speech becomes editable text at your cursor. The interpretation, structured elements, final diagnosis and sign-out remain with the pathologist and the laboratory's reporting system.","OpenWhispr is open source under the MIT License, and its source is available on GitHub. When we say dictation can run locally, your laboratory can inspect the code.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"OpenWhispr founder portrait"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Finish the report while the case is still on screen.","body":"Put the cursor in the editable field, dictate the findings and review every word before sign-out.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work with your workstation, ask us for help in Discord. Free forever on Mac, Windows & Linux."}}},"useCaseRadiology":{"breadcrumb":{"home":"Home","useCases":"Use Cases","radiology":"For Radiology"},"metadata":{"title":"Best Radiology Dictation Software, Free & Local - OpenWhispr","description":"OpenWhispr is free, open-source radiology dictation software that transcribes locally and pastes text at the active cursor for review.","keywords":["best radiology dictation software","radiology dictation software","radiologist speech recognition software","offline radiology dictation","local radiology transcription","radiology voice to text"]},"schema":{"name":"OpenWhispr","applicationCategory":"Radiology dictation software","operatingSystem":"macOS, Windows, Linux, iOS","description":"Local-first, open-source dictation software for radiologists. Choose local speech-to-text and the audio is transcribed on the workstation, then text is pasted at the active cursor. Cleanup is configured separately; there is no native radiology-system integration.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Best radiology dictation software for imaging specialists","title":"Keep the images in view. Get dictated text at the cursor.","body":"Choose a local speech-to-text model, place the cursor in an editable report field, and speak. Audio is transcribed on the workstation; use local cleanup or turn cleanup off to keep the text there too.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Radiologist dictating beside chest scans at a reading station"}},"trustBar":{"label":"Common reporting tools, not integrations:","names":["PowerScribe One","Epic Radiant","Sectra IDS7","Visage 7","InteleViewer","Microsoft Word"]},"pain":{"label":"Where radiology dictation breaks","title":"The images need your eyes. The report still needs your words.","subtitle":"A radiology dictation tool has to respect the active field, the vocabulary and the final read. Miss one, and the keyboard comes back.","cards":[{"title":"Your eyes leave the study for the keyboard.","body":"Place the cursor once, then hold the hotkey and dictate the finding. OpenWhispr pastes the transcript into the active field so you can stay with the images."},{"title":"The same term comes back wrong every case.","body":"Add names, modalities, anatomy and local abbreviations to the custom dictionary. They are supplied as hints to supported transcription paths."},{"title":"One missing 'no' changes the report.","body":"Speech recognition can change clinical meaning. OpenWhispr supplies text, not clinical quality control, so the radiologist still proofreads before signing."},{"title":"A simple dictation test turns into an integration project.","body":"OpenWhispr is a desktop speech-to-text layer, not a PACS or RIS replacement. Try it in an editable field without changing the reporting stack."}]},"features":{"label":"What you get","title":"Cursor dictation, not another reporting platform.","subtitle":"OpenWhispr handles speech to text. Your worklist, images, templates, measurements and signing workflow stay in the systems that already own them.","items":[{"image":{"label":"Person working at two monitors beside a desktop computer"},"title":"Findings transcribed on the workstation","body":"Choose a local speech-to-text model and audio is processed on the device. Use local cleanup or turn cleanup off if report text must stay there too; after model download, speech-to-text works offline."},{"image":{"label":"Radiologist checking a reference card beside a medical scan"},"title":"A dictionary for the terms you actually dictate","body":"Add recurring anatomy, eponyms, protocol names and abbreviations. OpenWhispr supplies the list as recognition hints, and you still check the result."},{"image":{"label":"Radiologist typing beside monitors displaying medical scans"},"title":"One hotkey at the active report cursor","body":"Open the report field, place the cursor, hold the hotkey and speak. The transcript is pasted there for review, with no claim of a native PACS, RIS or EHR integration."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Dictated audio uploaded in local mode","value":"None","footnote":"with local-to-cloud fallback off, audio stays on the radiologist's workstation"},{"label":"Cost per radiologist","value":"$$0","footnote":"unlimited local dictation for every radiologist"},{"label":"Setup takes","value":"5 min","footnote":"from download to a test dictation at the cursor"}]},"howItWorks":{"label":"How it works","title":"From spoken finding to text in the open report.","subtitle":"Keep the reporting workflow you already know. OpenWhispr adds a system-wide hotkey and a local transcription option.","image":{"label":"Radiologist working at a laptop beside brain scans"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Choose a local model during setup."},{"title":"Put the cursor in the report.","body":"Open an editable findings or impression field, then press the hotkey and dictate."},{"title":"Proofread and sign.","body":"Check terminology, numbers, negation and laterality, then finish the report in your reporting system."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons radiologists choose OpenWhispr","subtitle":"Use it for the narrow job it ships to do: turn speech into text at the active cursor. Keep the clinical and reporting workflow under your control.","items":[{"title":"Local transcription at the reading station","body":"Choose a local model and, with local-to-cloud fallback off, audio is transcribed on the device without an upload step."},{"title":"Text lands at the active cursor","body":"The desktop app pastes the transcript into the focused editable field. If automatic paste is unavailable, it can leave the text on the clipboard."},{"title":"No reporting-stack replacement","body":"Keep the worklist, viewer, report template and signing system your practice already uses."},{"title":"A dictionary for radiology vocabulary","body":"Add anatomy, eponyms, protocol names and abbreviations as hints for supported transcription and cleanup paths."},{"title":"Keeps working off the hospital network","body":"Once a local model is downloaded, dictation does not need the reading-room network or internet."},{"title":"$$0 local dictation per radiologist","body":"Every desktop plan includes unlimited use of local speech-to-text models."},{"title":"Three desktop operating systems","body":"Run the same desktop dictation app on the operating system your workstation uses."},{"title":"Your IT team can inspect the paste path","body":"Your organization can inspect the transcription and paste paths on GitHub."},{"title":"You read the images. OpenWhispr handles speech to text.","body":"It is not a diagnostic image reader and has no native patient, study or radiology-system context. Agent Mode can use optional screen context, but it does not diagnose or replace review."},{"title":"100+ spoken languages","body":"OpenWhispr supports 100+ languages. Availability and accuracy vary by local model and cloud provider."}]},"faq":{"label":"Before you download","title":"What radiologists ask before testing dictation software.","subtitle":"Straight answers on report fields, terminology, integrations, patient data, offline use and cost.","items":[{"question":"What is the best radiology dictation software?","answer":"It depends on the job. If you need a full reporting platform with patient and study context, templates, structured data, quality checks and enterprise interfaces, use a radiology reporting system built for that. If you want free, local-first speech-to-text that pastes at an active cursor, test OpenWhispr in your actual report field."},{"question":"Does OpenWhispr integrate with PACS, RIS or EHR software?","answer":"No native integration. OpenWhispr does not receive patient or study context from PACS, RIS, EHR, DICOM, HL7 or FHIR. It transcribes speech and pastes text at the active cursor. Optional Agent Mode can use screen context, which is not diagnostic integration. Test your reporting software and virtual desktop."},{"question":"Can it navigate templates or build a structured radiology report?","answer":"OpenWhispr ships no built-in radiology structured-reporting integration, report macros, voice-command field navigation or template logic. Generic Agent Mode can reformat text from your prompt, but it does not know the report fields. Use your existing system template and cursor."},{"question":"How does it handle terminology, numbers and laterality?","answer":"The custom dictionary lets you add recurring anatomy, eponyms, protocol names and abbreviations as recognition hints. It is not a radiology lexicon or clinical error checker. Proofread every report for terminology, measurements, negation and laterality before signing."},{"question":"Can I use it offline with patient information?","answer":"After a local model is downloaded, local speech-to-text runs offline and does not upload audio when the separate local-to-cloud fallback is off. Cleanup is configured separately. Your organization still decides which devices and applications are approved for patient information; cloud transcription is a separate global mode."},{"question":"Is local radiology dictation really free?","answer":"Yes. Every desktop plan includes unlimited local dictation, a custom dictionary and 100+ languages at $0. Pro is $8 per month and adds unlimited OpenWhispr Cloud transcription, more meeting hours, mobile and device sync."}]},"reviews":{"label":"What users say","title":"What users say about local dictation.","subtitle":"These quotes are verbatim. None is rewritten as a radiologist endorsement.","score":{"value":"1","label":"Job in the reporting stack","body":"OpenWhispr supplies speech-to-text at the active cursor. Your reporting system keeps the patient, study, template and signature workflow."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Start free with local dictation.","Upgrade for cloud or team features."],"caption":"Compare OpenWhispr plans for radiologists","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Radiologists who want unlimited local cursor dictation and can run a model on the workstation"}]},{"parts":[{"text":"Radiologists who want unlimited OpenWhispr Cloud transcription"}]},{"parts":[{"text":"Teams that want OpenWhispr Cloud, recordings and priority support"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited local speech-to-text at $0, with audio staying on the workstation"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models or keys"}]},{"parts":[{"text":"Unlimited recordings and priority support for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"We built the speech-to-text layer, not a radiology stack.","paragraphs":["Radiologists already work across a worklist, image viewer and reporting system. OpenWhispr does one narrower job: turn speech into text and paste it at the active cursor.","We build in public. Your reading-room or IT team can inspect the local transcription and paste paths on GitHub instead of relying only on a product page.","And if a report field will not accept the paste, you talk to the people who built it, in Discord."],"image":{"label":"OpenWhispr founder and community"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Test one dictated finding at the cursor.","body":"Start in a blank text field, then confirm cursor paste in an approved report field before clinical use.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, ask us for help in Discord. Free forever on Mac, Windows & Linux."}}}},"now":"$undefined","timeZone":"UTC","children":"$L18"}] 19:I[89554,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"ThemeProvider"] 1a:I[50103,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"GitHubStarsProvider"] 1b:I[28473,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"ConsentProvider"] 1c:I[51627,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"default"] 1d:I[31494,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js","/_next/static/immutable/chunks/01pbao-w8op81.js"],"default"] 1e:I[16508,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"default"] 1f:I[95338,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"default"] 20:I[88667,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"AttributionCapture"] 21:I[24901,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"MetaPixel"] 22:I[21307,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"Rb2b"] 23:I[2355,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"Analytics"] 27:I[27201,["/_next/static/immutable/chunks/0r3sg558voqp8.js"],"IconMark"] 18:["$","$L19",null,{"attribute":"class","defaultTheme":"system","enableSystem":true,"disableTransitionOnChange":true,"children":["$","$L1a",null,{"children":[["$","$L1b",null,{"children":[["$","$L1c",null,{}],["$","main",null,{"className":"flex-1","children":["$","$L2",null,{"parallelRouterKey":"children","error":"$undefined","errorStyles":"$undefined","errorScripts"</article></body></html>

1:"$Sreact.fragment" 2:I[39756,["/_next/static/immutable/chunks/0r3sg558voqp8.js"],"default"] 3:I[37457,["/_next/static/immutable/chunks/0r3sg558voqp8.js"],"default"] 4:I[22016,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js","/_next/static/immutable/chunks/1jjsyi7c0q8kj.js"],""] 7:I[97367,["/_next/static/immutable/chunks/0r3sg558voqp8.js"],"OutletBoundary"] 8:"$Sreact.suspense" b:I[97367,["/_next/static/immutable/chunks/0r3sg558voqp8.js"],"ViewportBoundary"] d:I[97367,["/_next/static/immutable/chunks/0r3sg558voqp8.js"],"MetadataBoundary"] f:I[68027,[],"default",1] :HL["/_next/static/immutable/chunks/1hl1q8f21akxa.css","style"] :HL["/_next/static/immutable/chunks/0o4yun1hxa7er.css","style"] :HL["/_next/static/immutable/media/0c89a48fa5027cee-s.p.1ii7211jqpzi_.woff2","font",{"crossOrigin":"","type":"font/woff2"}] a:X 0:{"P":null,"c":["","en","download"],"q":"","i":false,"f":[[["",{"children":[["locale","en","d",["api","providers","unsubscribe"]],{"children":["download",{"children":["__PAGE__",{},"$undefined","$undefined",4608]},"$undefined","$undefined",4608]},"$undefined","$undefined",4624]},"$undefined","$undefined",4624],[["$","$1","c",{"children":[null,["$","$L2",null,{"parallelRouterKey":"children","error":"$undefined","errorStyles":"$undefined","errorScripts":"$undefined","template":["$","$L3",null,{}],"templateStyles":"$undefined","templateScripts":"$undefined","notFound":[["$","html",null,{"lang":"en","suppressHydrationWarning":true,"children":["$","body",null,{"className":"flex min-h-screen flex-col items-center justify-center bg-background px-6 text-center","children":[["$","p",null,{"className":"text-7xl font-bold text-stone-200 select-none","children":"404"}],["$","h1",null,{"className":"mt-4 text-2xl font-bold text-foreground","children":"Page not found"}],["$","p",null,{"className":"mt-2 max-w-md text-muted-foreground","children":"The page you're looking for might have been moved, deleted, or never existed."}],["$","$L4",null,{"href":"/","className":"mt-6 inline-flex items-center rounded-lg bg-blue-600 px-5 py-2.5 text-sm font-medium text-white transition-colors hover:bg-blue-700","children":"Take me home"}]]}]}],[["$","link","0",{"rel":"stylesheet","href":"/_next/static/immutable/chunks/1hl1q8f21akxa.css","precedence":"next","crossOrigin":"$undefined","nonce":"$undefined"}]]],"forbidden":"$undefined","unauthorized":"$undefined"}]]}],{"children":[["$","$1","c",{"children":[[["$","link","0",{"rel":"stylesheet","href":"/_next/static/immutable/chunks/1hl1q8f21akxa.css","precedence":"next","crossOrigin":"$undefined","nonce":"$undefined"}],["$","link","1",{"rel":"stylesheet","href":"/_next/static/immutable/chunks/0o4yun1hxa7er.css","precedence":"next","crossOrigin":"$undefined","nonce":"$undefined"}],["$","script","script-0",{"src":"/_next/static/immutable/chunks/1p2ndgirh2-it.js","async":true,"nonce":"$undefined"}],["$","script","script-1",{"src":"/_next/static/immutable/chunks/3zvxazlv9iyny.js","async":true,"nonce":"$undefined"}],["$","script","script-2",{"src":"/_next/static/immutable/chunks/3sqnre1wqw5c3.js","async":true,"nonce":"$undefined"}],["$","script","script-3",{"src":"/_next/static/immutable/chunks/3q341afyljkz1.js","async":true,"nonce":"$undefined"}],["$","script","script-4",{"src":"/_next/static/immutable/chunks/2mda3sy1qsv_a.js","async":true,"nonce":"$undefined"}],["$","script","script-5",{"src":"/_next/static/immutable/chunks/2w6h9kjzbvct2.js","async":true,"nonce":"$undefined"}]],"$L5"]}],{"children":[["$","$1","c",{"children":[null,["$","$L2",null,{"parallelRouterKey":"children","error":"$undefined","errorStyles":"$undefined","errorScripts":"$undefined","template":["$","$L3",null,{}],"templateStyles":"$undefined","templateScripts":"$undefined","notFound":"$undefined","forbidden":"$undefined","unauthorized":"$undefined"}]]}],{"children":[["$","$1","c",{"children":["$L6",[["$","script","script-0",{"src":"/_next/static/immutable/chunks/1jjsyi7c0q8kj.js","async":true,"nonce":"$undefined"}]],["$","$L7",null,{"children":["$","$8",null,{"name":"Next.MetadataOutlet","children":"$@9"}]}]]}],{},null,false,null]},null,false,"$a"]},null,false,null]},null,false,"$a"],["$","$1","h",{"children":[null,["$","$Lb",null,{"children":"$Lc"}],["$","div",null,{"hidden":true,"children":["$","$Ld",null,{"children":["$","$8",null,{"name":"Next.Metadata","children":"$Le"}]}]}],["$","meta",null,{"name":"next-size-adjust","content":""}]]}],false]],"m":"$undefined","G":["$f",[]],"S":true,"h":null,"r":"$undefined","s":"$undefined","a":"$undefined","l":"$undefined","p":"$undefined","d":"$undefined","b":"ZYkEsdd5Jb_xJZ1hUs2qs"} 11:I[19598,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"GoogleTagManager"] a:C 10:T579,window.dataLayer=window.dataLayer||[];function gtag(){dataLayer.push(arguments);} gtag('consent','default',{ad_storage:'denied',ad_user_data:'denied',ad_personalization:'denied',analytics_storage:'denied',functionality_storage:'granted',personalization_storage:'denied',security_storage:'granted',region:["AT","BE","BG","HR","CY","CZ","DK","EE","FI","FR","DE","GR","HU","IE","IT","LV","LT","LU","MT","NL","PL","PT","RO","SK","SI","ES","SE","IS","LI","NO","GB","CH"],wait_for_update:500}); gtag('consent','default',{ad_storage:'granted',ad_user_data:'granted',ad_personalization:'granted',analytics_storage:'granted',functionality_storage:'granted',personalization_storage:'granted',security_storage:'granted'}); try{var _m=document.cookie.match(/(?:^|; )ow_consent=([^;]*)/);if(_m){var _v=JSON.parse(decodeURIComponent(_m[1]));if(_v&&_v.choices&&typeof _v.choices.marketing==='boolean'&&typeof _v.choices.analytics==='boolean'&&typeof _v.v==='string'&&/^\d{4}-\d{2}-\d{2}$/.test(_v.v)&&_v.v>='2026-06-01'){gtag('consent','update',{analytics_storage:_v.choices.analytics?'granted':'denied',ad_storage:_v.choices.marketing?'granted':'denied',ad_user_data:_v.choices.marketing?'granted':'denied',ad_personalization:_v.choices.marketing?'granted':'denied'});}}}catch(e){} if(navigator.globalPrivacyControl){gtag('consent','update',{ad_storage:'denied',ad_user_data:'denied',ad_personalization:'denied'});}5:["$","html",null,{"lang":"en","dir":"ltr","className":"space_grotesk_e6988195-module__RNs2Mq__variable","suppressHydrationWarning":true,"children":[["$","head",null,{"children":[["$","script",null,{"dangerouslySetInnerHTML":{"__html":"$10"}}],["$","link",null,{"rel":"preconnect","href":"https://ddwl4m2hdecbv.cloudfront.net"}]]}],["$","body",null,{"className":"flex min-h-screen flex-col","children":[["$","$L11",null,{"gtmId":"GTM-NPK2WBC7"}],"$L12"]}]]}] 14:I[59302,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js","/_next/static/immutable/chunks/1jjsyi7c0q8kj.js"],"default"] 6:["$L13",["$","$L14",null,{}]] 13:["$","script",null,{"type":"application/ld+json","dangerouslySetInnerHTML":{"__html":"{\"@context\":\"https://schema.org\",\"@type\":\"BreadcrumbList\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https://openwhispr.com\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Download\",\"item\":\"https://openwhispr.com/download\"}]}"}}] c:[["$","meta","0",{"charSet":"utf-8"}],["$","meta","1",{"name":"viewport","content":"width=device-width, initial-scale=1"}]] 15:I[75696,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"default"] 16:T4cc,This Data Processing Addendum ("DPA") forms part of, and is incorporated by reference into, the OpenWhispr Terms of Service or, where one exists, the master services or enterprise agreement between the customer that has entered into it (the "Customer", "Controller") and Gizmo Labs Inc. (the "Agreement"). It applies whenever Gizmo Labs Inc., a company incorporated in Delaware, with its registered address at 221 W 9th Street, Wilmington, Delaware 19801 ("OpenWhispr", "Processor"), processes Personal Data on behalf of the Customer in the course of providing the OpenWhispr services (the "Services"). No signature is required for this DPA to take effect; signature blocks are provided for customers whose procurement requires an executed copy. Where this DPA conflicts with the Agreement on data protection matters, this DPA prevails; where this DPA conflicts with the SCCs (or the UK Addendum) in respect of a Restricted Transfer, the SCCs (or the UK Addendum) prevail to the extent of the conflict. Notices under this DPA go to the parties' account contacts and, for OpenWhispr, to support@openwhispr.com.17:T413,OpenWhispr processes Personal Data only as a processor on behalf of the Customer. The subject-matter, duration, nature and purpose of processing, the types of Personal Data and categories of data subjects are set out in Annex 1. For clarity: where end users run OpenWhispr in local-only mode or with their own model API keys ("bring your own key"), audio and transcription/inference content are sent directly from the user's device to the user's chosen provider or processed on-device; those content flows are directed by the user, and the third-party providers involved act on the user's instructions, not as OpenWhispr Sub-processors. Account data, cloud sync (where enabled), usage and billing data remain processed by OpenWhispr and its Sub-processors under this DPA in all modes. OpenWhispr may also process data that has been aggregated and de-identified so that it is no longer Personal Data (for example, service performance metrics) to operate, secure and improve the Services, and will not attempt to re-identify it.12:["$","$L15",null,{"formats":"$undefined","locale":"en","messages":{"common":{"nav":{"features":"Features","pricing":"Pricing","openMenu":"Open menu","closeMenu":"Close menu","useCases":"Use Cases"},"features":{"notepad":{"label":"AI Meeting Notes","description":"Private, bot-free meeting notes"},"aiChat":{"label":"AI Chat","description":"Chat that knows your meetings"},"audioUpload":{"label":"Audio to Text","description":"Transcribe any audio file locally"},"dictation":{"label":"Dictation Software","description":"Voice to text in any app"}},"footer":{"tagline":"Open source voice-to-text that respects your privacy and puts you in control.","columns":{"product":"Product","resources":"Resources","legal":"Legal","followUs":"Follow Us","company":"Company"},"links":{"aiNotepad":"AI Meeting Notes","aiChat":"AI Chat","audioUpload":"Audio to Text","features":"Features","pricing":"Pricing","faq":"FAQ","models":"Models","documentation":"Documentation","trustCenter":"Trust Center","comparisons":"Comparisons","useCases":"Use Cases","blog":"Blog","changelog":"Changelog","privacy":"Privacy Policy","cookies":"Cookie Policy","terms":"Terms of Service","security":"Security","contact":"Contact","download":"Download","dpa":"DPA"},"copyright":"© 2026 Gizmo Labs Inc.","madeWith":"Made with {heart} for the open source community","ariaLabel":"Footer","legalAriaLabel":"Legal","socialLabel":"Social media","logoAlt":"OpenWhispr logo","postalAddress":"221 W 9th Street, Wilmington, Delaware 19801, USA"},"actions":{"github":"GitHub","download":"Download","starUsOnGitHub":"Star us on GitHub","noAccountOfflineOpenSource":"No account required · Works offline · Open source forever","close":"Close"},"localeSwitcher":{"label":"Language","select":"Select language"},"useCases":{"medical":{"label":"Medical Dictation","description":"Private dictation for clinicians"},"mac":{"label":"Mac Dictation","description":"AI voice-to-text for macOS"},"windows":{"label":"Windows","description":"AI voice-to-text on Windows"},"legal":{"label":"Legal Dictation","description":"Private dictation for lawyers"}},"downloadButton":{"loading":"Loading...","downloadOpenWhispr":"Download OpenWhispr","downloadForPlatform":"Download for {platform}","allPlatforms":"Available on macOS, Windows, Linux and iOS","platforms":{"macos":"macOS","windows":"Windows","linux":"Linux","unknown":"your OS","macosAppleSilicon":"macOS (Apple Silicon)","macosIntel":"macOS (Intel)","ios":"iOS"},"getStartedOnIPhone":"Get started on iPhone"},"theme":{"toggle":"Toggle theme","light":"Light mode","dark":"Dark mode"},"siteMetadata":{"title":"OpenWhispr | Open Source AI Voice Dictation","description":"Turn your voice into text instantly. Fully private, open source, and powered by OpenAI Whisper and NVIDIA Parakeet.","imageAlt":"OpenWhispr — private, open source AI voice dictation and meeting notes"},"adminMetadata":{"title":"Admin — OpenWhispr","description":"Internal administration dashboard for OpenWhispr."},"workspaceSsoMetadata":{"title":"Workspace SSO — OpenWhispr","description":"Configure single sign-on for your OpenWhispr workspace."}},"featureNotepad":{"metadata":{"title":"AI Meeting Notes Software (Free Download) - OpenWhispr","description":"OpenWhispr is AI meeting notes software that transcribes calls locally. Get private transcripts, summaries and action items with no bot joining.","keywords":["AI meeting notes","AI notepad for meetings","meeting notes app","meeting transcription","AI meeting assistant","open source meeting notes","private meeting notes","granola alternative","granola open source alternative","meeting notes without bots","meeting summary AI","automated meeting notes","meeting note taker","AI note taking app","meeting minutes AI","meeting notes privacy","meeting recording app","AI meeting summary","meeting action items","meeting notes organizer"]},"breadcrumb":{"home":"Home","notepad":"Notepad"},"schema":{"description":"AI notepad for meetings. Transcribe, take notes, and let AI enhance them. Open source, private by design, available on macOS, Windows, Linux and iOS.","offerDescription":"Free and open source. Optional Pro plan available.","featureList":["AI-powered meeting note enhancement","Real-time meeting transcription","No meeting bots required","Custom AI actions and templates","Searchable meeting history","Folder organization","Cross-platform support","Zero data retention","Open source"]},"hero":{"title":"AI Meeting Notes, Without the Bots","subtitle":"The notepad that cleans up after your meetings. Take rough notes and OpenWhispr listens in the background, turning them into clear, organized summaries — decisions, action items, and follow-ups.","availability":"Available on macOS, Windows, Linux and iOS"},"howItWorks":{"eyebrow":"How it works","title":"Three steps. Zero setup.","subtitle":"No plugins, no integrations, no meeting bots. Just open the app and start your meeting."},"steps":[{"number":"01","title":"Start your meeting","description":"Open any meeting on Zoom, Google Meet, Teams, or any other app. OpenWhispr captures your computer's audio directly — no bots join your call, and nobody knows it's running."},{"number":"02","title":"Take notes your way","description":"Write quick bullets, messy thoughts, whatever works for you. OpenWhispr transcribes the full conversation in the background while you stay focused on the meeting."},{"number":"03","title":"Get clean, structured notes","description":"When you're done, AI turns your rough notes and the full transcript into clear meeting minutes — decisions, action items, and key takeaways, ready to share."}],"enhancedNotes":{"eyebrow":"AI-enhanced notes","title":"You write three bullets. AI does the rest.","paragraph1":"Jot down whatever comes to mind during your meeting. When it's over, OpenWhispr combines your notes with the full transcript and turns everything into structured meeting minutes — decisions, action items, and next steps.","paragraph2":"Switch between your raw notes, the full transcript, and the AI-enhanced version anytime. Nothing gets overwritten — your original notes are always there."},"platforms":{"eyebrow":"No bots, no plugins","title":"Works with every meeting app.","subtitle":"OpenWhispr captures your computer's audio directly — no bots join your call, no plugins to install, no integrations to configure.","footnote":"Plus any other app that plays audio on your computer."},"aiActions":{"eyebrow":"AI actions","title":"Turn your notes into emails, summaries, and action items.","paragraph1":"After any meeting, run AI actions on your notes — summarize the discussion, draft a follow-up email, extract all action items, or format everything using your team's template.","paragraph2":"Create your own custom actions for the things you do after every meeting. Save time on the repetitive work so you can focus on what matters."},"organization":{"eyebrow":"Stay organized","title":"Folders for every project, team, and client.","paragraph1":"Create folders to keep your meeting notes organized however you work. Group notes by project, by team, by client — whatever makes sense for you.","paragraph2":"Every meeting is automatically saved and easy to find. No more digging through a single long list of notes."},"search":{"eyebrow":"Search","title":"Find anything, from any meeting.","subtitle":"Search across all your transcripts, notes, and summaries. Instantly find what was discussed, who said what, and what was decided — even months ago."},"privacy":{"eyebrow":"Private by design","title":"Your meetings are your business.","subtitle":"Your meeting recordings and notes are yours. We don't store your audio, we don't keep copies, and we never use your data to train models.","points":["Zero data retention — audio is processed and immediately discarded","We never train on your data or use it to improve our models","Your notes are stored locally on your machine","Open source — anyone can verify the code on GitHub"],"cardTitle":"Zero data retention","cardBody":"Your audio is processed in real time and immediately discarded — no recordings stored, no copies made, ever. Your notes live on your device, and our code is open source so you can verify every claim yourself.","cardCtaPrefix":"Don't take our word for it.","cardCtaLink":"Read the source code","cardCtaSuffix":"."},"comparison":{"eyebrow":"Why OpenWhispr","title":"A better way to handle meeting notes.","subtitle":"Same capabilities as other tools, but your data stays yours.","headerFeature":"Feature","headerOthers":"Others","rows":{"zeroRetention":"Zero data retention","neverTrains":"Never trains on your data","noBots":"No meeting bots","openSource":"Open source","freeTier":"Free tier","crossPlatform":"Cross-platform"},"values":{"freeTierOpenwhispr":"5 hrs/month free","freeTierOthers":"Limited","crossPlatformOpenwhispr":"macOS, Windows, Linux, iOS","crossPlatformOthers":"macOS, Windows"}},"faq":{"title":"Common questions","subtitle":"What people usually want to know about the AI Notepad."},"faqs":[{"question":"How does OpenWhispr Notepad work?","answer":"OpenWhispr listens to your computer's audio during meetings — no bots join your call. You take notes as usual, and when the meeting ends, AI combines your notes with the full transcript to create structured meeting minutes with decisions, action items, and key takeaways."},{"question":"Does a bot join my meeting?","answer":"No. OpenWhispr captures audio directly from your computer's speakers and microphone. No bot joins your Zoom, Google Meet, Teams, or any other call. Other participants never know you're recording."},{"question":"Is it free?","answer":"Yes. OpenWhispr is free and open source. You can run everything on your own computer with no account required. A Pro plan is available if you want cloud-powered AI for faster processing."},{"question":"Is my data private?","answer":"Yes. OpenWhispr is private by design. Your audio is processed in real time and immediately discarded — nothing is stored, no copies are made. We never use your data to train models. Your notes are stored locally on your device, and our code is fully open source."},{"question":"Which meeting apps are supported?","answer":"All of them. OpenWhispr works with Zoom, Google Meet, Microsoft Teams, Slack, Discord, Webex, FaceTime, and any other app that plays audio on your computer. No plugins or integrations needed."},{"question":"Does OpenWhispr store my recordings?","answer":"No. Your audio is processed in real time and immediately discarded. We never store recordings, and we never use your data to train AI models. Your notes stay on your device."},{"question":"What operating systems are supported?","answer":"OpenWhispr is available on macOS, Windows, Linux and iOS."},{"question":"What are AI actions?","answer":"AI actions let you run tasks on your notes with one click — like summarizing the meeting, drafting a follow-up email, extracting action items, or formatting notes using your team's template. You can also create your own custom actions."}],"graphics":{"notesTransform":{"title":"Sprint Planning","yourNotesLabel":"Your notes","note1":"- ship voice pipeline by march","note2":"- talk to design team re: new UI","note3":"- budget Q: ask finance","enhancedLabel":"Enhanced","decisionsTitle":"Decisions","decisionsText":"Voice pipeline is top priority — ship by end of March","actionItemsTitle":"Action Items","actionItemsText":"Gabriel: follow up with finance on Q2 budget by Friday","nextStepsTitle":"Next Steps","nextStepsText":"Design review for new editor — Tuesday 2pm"},"aiActions":{"title":"AI Actions","actions":[{"name":"Summarize meeting","desc":"Key points and decisions"},{"name":"Draft follow-up email","desc":"Email based on your notes"},{"name":"Sync to template","desc":"Format using team template"},{"name":"List action items","desc":"Extract all to-dos"},{"name":"Create your own...","desc":"Custom AI action"}]},"folders":{"foldersLabel":"Folders","folders":[{"name":"Product","count":8},{"name":"Design","count":3},{"name":"Clients","count":5},{"name":"1:1s","count":12}],"recentLabel":"Recent","recent":[{"title":"Sprint Planning","time":"10:30"},{"title":"Product Sync","time":"9:15"},{"title":"Design Review","time":"Yest."}],"detailTitle":"Sprint Planning","detailTime":"Today, 10:30 AM","decisionsTitle":"Decisions","decisionsText":"Voice pipeline ships by end of March. UI redesign deferred until pipeline lands.","actionItemsTitle":"Action Items","actionItemsText":"Sarah: frontend migration by end of sprint. Alex: API refactor + benchmarks."},"search":{"query":"budget discussion","results":[{"title":"Q2 Budget Planning","date":"Mar 8","snippet":"Approved $50k for new hires. Follow up with finance..."},{"title":"Sprint Planning","date":"Mar 11","snippet":"...need headcount approval by Friday for Q2 budget"},{"title":"Team Standup","date":"Mar 5","snippet":"Budget review pushed to next week, waiting on final..."}]}}},"featureAiChat":{"meta":{"title":"AI Meeting Notes & Chat — Private, Local | OpenWhispr","description":"Local AI chat for your meeting notes — ask questions, get summaries, pull action items. Runs on-device. Open source and free.","keywords":["ai meeting assistant","best ai meeting assistant","ai assistant for meetings","chat with your meeting notes","ask your meetings","meeting Q&A","voice AI assistant","private AI assistant","open source AI assistant"]},"breadcrumbs":{"home":"Home","aiChat":"AI Chat"},"structuredData":{"description":"AI chat that knows your meetings. Ask questions, get summaries, and extract action items by voice or text. Open source, private by design.","offerDescription":"Free and open source. Optional Pro plan available."},"hero":{"title":"AI chat that already knows your meetings.","subtitle":"Ask questions about your work by voice or text. Get instant answers, summaries, and action items pulled directly from your meeting notes.","platforms":"Available on macOS, Windows, Linux and iOS"},"graphics":{"label":"AI Chat"},"chatExchange":{"question":"What were the key decisions from this week's product meetings?","intro":"Based on 3 meetings this week:","points":["Voice pipeline ships by end of March — top priority","UI redesign deferred until pipeline is done","Q2 budget approved — need headcount request by Friday"],"followUpPlaceholder":"Ask a follow-up question..."},"askAnything":{"eyebrow":"Ask anything","title":"Stop digging through notes. Just ask.","paragraph1":"AI Chat has full context from your meetings — transcripts, notes, and summaries. Ask about decisions, action items, or what someone said, and get the answer instantly.","paragraph2":"Speak your question by holding a hotkey, or type it out. Either way, the answer comes from your actual meetings, not generic AI."},"exampleQuestions":{"title":"Your meetings, answerable.","subtitle":"Ask anything you'd normally dig through notes to find.","items":["What were the key decisions from last week?","List all action items assigned to Sarah","What did we decide about the API timeline?","Draft a follow-up email for today's meeting","Summarize the budget discussions from Q2","Who raised concerns about the migration?"]},"lifecycle":{"eyebrow":"Your whole workday","title":"Helpful before, during, and after every meeting.","subtitle":"AI Chat isn't just for after meetings. Use it to prepare, to ask smarter questions during the call, and to follow up when it's over.","stages":[{"label":"Before","title":"Get prepared","desc":"Ask about past meetings to catch up before your next call"},{"label":"During","title":"Ask better questions","desc":"Look up context in real time without leaving the meeting"},{"label":"After","title":"Get things done","desc":"Draft emails, summarize decisions, plan next steps"}]},"contentGeneration":{"eyebrow":"Get things done","title":"Turn conversations into action.","subtitle":"AI Chat helps you write documents, follow-up emails, and briefs — all based on what was actually discussed in your meetings.","items":["Draft a follow-up email","Summarize the meeting","Write a Slack update","Create action item list","Draft a project brief","Write meeting minutes","Create a proposal","Draft a status update"]},"folderContext":{"eyebrow":"Folder context","title":"Ask questions across a whole project.","paragraph1":"Scope your questions to a specific folder — like a project, team, or client. AI Chat searches across all meetings in that folder to find patterns, recurring topics, and trends.","paragraph2":"Perfect for preparing quarterly reviews, gathering feedback themes, or catching up on a project you've been away from.","scope":"in Product folder","question":"What feature requests have come up the most?","intro":"Across 12 meetings in the Product folder:","items":["Real-time collaboration (mentioned 8 times)","Mobile app support (mentioned 6 times)","Custom templates (mentioned 5 times)"]},"privacy":{"eyebrow":"Private by design","title":"Your conversations stay yours.","subtitle":"AI Chat works with your notes stored on your device. Your questions and answers stay private — you choose whether to use local or cloud AI models.","bullets":["Chat history stored locally on your device","Local AI models available for complete offline privacy","Cloud models available as an opt-in for faster responses","Open source — verify exactly what the code does"],"cardTitle":"Your data, your choice","cardBody":"Run AI Chat with local models for complete privacy — nothing leaves your device. Or choose cloud models when you want faster, more capable responses. Either way, your meeting data is never stored on external servers.","sourcePrefix":"Don't take our word for it.","sourceLink":"Read the source code","sourceSuffix":"."},"comparison":{"eyebrow":"Why OpenWhispr","title":"AI chat that respects your privacy.","subtitle":"Context-aware AI that works with your meetings, not against your privacy.","headerFeature":"Feature","headerOthers":"Others","rows":[{"feature":"Voice-powered Q&A","openwhispr":true,"others":false},{"feature":"Context from your meetings","openwhispr":true,"others":true},{"feature":"Private by design","openwhispr":true,"others":false},{"feature":"Open source","openwhispr":true,"others":false},{"feature":"Works offline","openwhispr":true,"others":false},{"feature":"Free tier","openwhispr":"5 hrs/month free","others":"Limited"},{"feature":"Cross-platform","openwhispr":"macOS, Windows, Linux, iOS","others":"macOS, Windows"}]},"faq":{"title":"Common questions","subtitle":"What people usually want to know about AI Chat."},"faqs":[{"question":"What is AI Chat?","answer":"AI Chat is a built-in assistant that knows the context of your meetings. Ask it questions about what was discussed, who said what, or what needs to happen next — by voice or text. It pulls answers directly from your meeting notes and transcripts."},{"question":"How does it know about my meetings?","answer":"AI Chat has access to your meeting notes and transcripts stored on your device. When you ask a question, it searches through your notes to give you accurate, context-aware answers."},{"question":"Can I talk to it by voice?","answer":"Yes. Hold your hotkey, speak your question naturally, and AI Chat will transcribe and respond. You can also type if you prefer."},{"question":"Is my data private?","answer":"Yes. Your notes and chat history stay on your device. You choose whether to use local AI models (fully private, works offline) or cloud models (faster responses). Even with cloud models, your meeting data is never stored externally."},{"question":"What kind of questions can I ask?","answer":"Anything about your meetings — summarize discussions, list action items, find who said what, look up decisions, draft follow-up emails, or search across all your meeting history. You can also scope questions to a specific folder or project."},{"question":"Is this included in the free plan?","answer":"Yes. AI Chat works with local models at no cost. Cloud-powered models are available on the Pro plan for faster, more capable responses."}]},"featureAudioUpload":{"metadata":{"title":"Transcribe Audio to Text — Free & Private | OpenWhispr","description":"Transcribe any audio file to text on your own device. OpenWhispr converts MP3, WAV, M4A and more into accurate transcripts — private, offline, and free.","ogTitle":"Audio Upload & Transcription — Private by Design | OpenWhispr","ogDescription":"Upload any audio or video file and get accurate transcripts in seconds. Private by design, open source.","keywords":["audio transcription","audio to text","transcribe audio file","local audio transcription","private audio transcription","offline transcription","MP3 to text","WAV transcription","meeting recording transcription","voice memo transcription","AI audio transcription","open source transcription","whisper transcription","audio file to notes","transcribe recording locally"]},"breadcrumbs":{"home":"Home","audioUpload":"Audio Upload"},"schema":{"name":"OpenWhispr Audio Upload","description":"Upload any audio file and get accurate transcripts. Private by design, works offline, open source. Supports 99+ languages.","offerDescription":"Free and open source. Optional Pro plan available."},"hero":{"title":"Transcribe Audio to Text on Your Device","subtitle":"Turn any recording into notes. Drop an audio file and get an accurate transcript in seconds, then use AI to summarize, extract action items, and organize everything into notes.","availability":"Available on macOS, Windows, Linux and iOS"},"howItWorks":{"eyebrow":"How it works","title":"Drop, transcribe, done.","subtitle":"No accounts, no sign-ups, no waiting. Just drop your file and get a transcript."},"steps":[{"number":"01","title":"Drop your file","description":"Drag and drop any audio file — meeting recordings, voice memos, lectures, interviews. Supports MP3, WAV, M4A, and many more formats."},{"number":"02","title":"Get your transcript","description":"OpenWhispr transcribes your audio using AI. Processing happens on your device by default — your recording never gets uploaded anywhere."},{"number":"03","title":"Turn it into notes","description":"Your transcript becomes a searchable note. Run AI actions to summarize, extract action items, or format it with a template."}],"transcriptGraphic":{"filename":"meeting-recording.mp3","transcriptLabel":"Transcript","enhancedLabel":"Enhanced","quote1":"“I think we should go ahead with the new design system. Sarah, can you lead the frontend migration?”","quote2":"“Sure, I can start the component audit this sprint. Alex, are you good to handle the API side?”","quote3":"“Yeah, I’ll scope the refactor and have an RFC by Wednesday.”","decisionsHeading":"Decisions","decisionsBody":"Moving forward with the new design system","actionItemsHeading":"Action Items","actionItem1":"Sarah: Component audit this sprint","actionItem2":"Alex: API refactor RFC by Wednesday"},"beyondTranscription":{"eyebrow":"Beyond transcription","title":"From raw recording to structured notes.","paragraph1":"Don’t just get a wall of text. OpenWhispr turns your transcript into organized notes — decisions, action items, key points, and next steps, all pulled from what was actually said.","paragraph2":"Run AI actions on any transcript — summarize it, draft a follow-up email, or format it using your team’s template."},"formats":{"title":"Every audio format you need","subtitle":"Upload recordings from any device or app. OpenWhispr handles all major audio formats out of the box.","footnote":"Plus any format supported by FFmpeg.","list":[{"name":"MP3","popular":true},{"name":"WAV","popular":true},{"name":"M4A","popular":true},{"name":"WebM","popular":false},{"name":"OGG","popular":false},{"name":"FLAC","popular":false},{"name":"AAC","popular":false}]},"useCasesSection":{"eyebrow":"Use cases","title":"Transcribe anything.","subtitle":"Meeting recordings, voice memos, interviews, lectures — if it’s audio, OpenWhispr can turn it into searchable, organized notes."},"useCases":[{"title":"Meeting recordings","description":"Upload recordings from Zoom, Teams, or any platform and get structured meeting minutes with decisions and action items."},{"title":"Voice memos","description":"Turn quick voice notes into searchable text. Capture ideas on the go and organize them later."},{"title":"Interviews & research","description":"Transcribe interviews, user research sessions, or focus groups. Extract key quotes and themes."},{"title":"Lectures & talks","description":"Convert class recordings or conference talks into study notes with key takeaways."}],"languageGraphic":{"label":"Supported languages","total":"99+ total","more":"+ 87 more","languages":["English","Spanish","French","German","Portuguese","Japanese","Chinese","Korean","Italian","Dutch","Russian","Arabic"]},"multiLanguage":{"eyebrow":"Multi-language","title":"99+ languages. Automatic detection.","paragraph1":"Transcribe audio in dozens of languages without any setup. OpenWhispr automatically detects the language and produces accurate transcripts.","paragraph2":"Whether your meetings are in English, Spanish, Japanese, or Arabic — it just works."},"privacy":{"eyebrow":"Private by design","title":"Your recordings stay yours.","intro":"Most transcription services upload your audio to the cloud. OpenWhispr processes your recordings on your device by default — sensitive audio never has to leave your computer.","checklist":["Audio files processed on your device by default","Cloud processing available as an opt-in for speed","Transcripts stored locally — you own your data","Open source — anyone can verify the code on GitHub"],"cardTitle":"Perfect for sensitive recordings","cardBody":"Legal depositions, medical notes, financial calls, HR interviews — when your recordings contain sensitive information, local processing means they never touch an external server.","cardFootnotePrefix":"Don’t take our word for it. ","cardFootnoteLink":"Read the source code","cardFootnoteSuffix":"."},"comparison":{"eyebrow":"Why OpenWhispr","title":"A better way to transcribe audio.","subtitle":"Accurate transcription that keeps your recordings private.","featureHeader":"Feature","openwhisprHeader":"OpenWhispr","othersHeader":"Cloud services","rows":[{"feature":"Private by design","openwhispr":true,"others":false},{"feature":"Open source","openwhispr":true,"others":false},{"feature":"Works offline","openwhispr":true,"others":false},{"feature":"No file size limit","openwhispr":true,"others":false},{"feature":"Free tier","openwhispr":"5 hrs/month free","others":"Limited minutes"},{"feature":"AI note enhancement","openwhispr":true,"others":"Add-on"},{"feature":"Multi-language","openwhispr":"99+ languages","others":"Varies"}]},"faqSection":{"title":"Common questions","subtitle":"What people usually want to know about audio upload."},"faqs":[{"question":"What audio formats can I upload?","answer":"All the common ones — MP3, WAV, M4A, WebM, OGG, FLAC, and AAC. If you recorded it on your phone, laptop, or any other device, it will work."},{"question":"How accurate is the transcription?","answer":"Very accurate. OpenWhispr uses state-of-the-art AI models for transcription. It handles different accents, background noise, and multiple speakers well. You can also choose between local models (private, works offline) or cloud models (faster, even more accurate)."},{"question":"Is there a file size or duration limit?","answer":"No artificial limits. You can upload recordings of any length. Longer files take more time to process, but there's no cap on what you can transcribe."},{"question":"What can I do after transcribing?","answer":"Your transcript becomes a note in OpenWhispr. From there you can run AI actions on it — summarize it, extract key points, format it with a template, or draft a follow-up email. It's not just a transcript, it's a starting point."},{"question":"Is my audio kept private?","answer":"Yes. OpenWhispr is private by design. With local processing, your audio never leaves your device. If you opt for cloud processing for faster results, your audio is processed in real time and immediately discarded — nothing is stored."},{"question":"What languages are supported?","answer":"OpenWhispr supports transcription in dozens of languages including English, Spanish, French, German, Portuguese, Japanese, Chinese, and many more. It can also detect the language automatically."},{"question":"Is it free?","answer":"Yes. OpenWhispr is free and open source. Local transcription costs nothing and works offline. A Pro plan is available if you want cloud-powered AI for faster, more accurate results."}]},"compareHandy":{"metadata":{"title":"OpenWhispr vs Handy: Open Source Dictation Compared","description":"Compare OpenWhispr and Handy, two free open source dictation apps. See how they differ on AI models, local privacy, and features. Pick your tool.","keywords":["OpenWhispr vs Handy","Handy speech to text","open source dictation","free voice dictation app","Whisper dictation app","nvidia parakeet","local speech to text","privacy voice dictation","cross-platform dictation","Handy alternative","Tauri speech to text","AI dictation comparison"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","handy":"vs Handy"},"hero":{"title":"OpenWhispr vs Handy","subtitle":"Two free, open source dictation apps that keep your voice data private. Here's how they compare on features, flexibility, and workflow.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","handyIntro":"is a lightweight, open source speech-to-text app built with Tauri and Rust. It focuses on doing one thing extremely well: press a key, speak, and get your words pasted into any text field. With support for both Whisper and Parakeet V3 models, Silero voice activity detection, and GPU acceleration, it delivers fast local transcription with a minimal footprint. It has earned over 14,000 stars on GitHub and has an active community.","openWhisprIntro":"shares the same open source, privacy-first DNA but extends the concept further. Beyond local transcription, it offers Agent Mode with custom AI system prompts for post-processing your dictation, optional cloud transcription for faster results, and a polished system-wide dictation experience. It works on macOS, Windows, Linux and iOS with both free and optional Pro tiers.","chooseHandy":"Choose Handy if: You want a minimal, Rust-powered tool with fast local-only transcription and a small system footprint.","chooseOpenWhispr":"Choose OpenWhispr if: You want AI post-processing, cloud transcription options, and a more feature-rich dictation workflow."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"handy":{"status":"full","details":"MIT licensed, fully open source"}},{"feature":"Privacy & Local Processing","openWhispr":{"status":"full","details":"100% local option, bring your own API keys"},"handy":{"status":"full","details":"Fully offline, no cloud transmission"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"handy":{"status":"full","details":"macOS, Windows, Linux"}},{"feature":"AI Model Variety","openWhispr":{"status":"full","details":"OpenAI Whisper and NVIDIA Parakeet models"},"handy":{"status":"full","details":"Whisper models + Parakeet V3"}},{"feature":"Agent Mode / AI Post-Processing","openWhispr":{"status":"full","details":"Custom system prompts for formatting, translation, etc."},"handy":{"status":"none","details":"Not available"}},{"feature":"Cloud Transcription Option","openWhispr":{"status":"full","details":"OpenWhispr Cloud or BYOK for fast results"},"handy":{"status":"none","details":"Local only, no cloud option"}},{"feature":"Voice Activity Detection","openWhispr":{"status":"partial","details":"Basic silence detection"},"handy":{"status":"full","details":"Silero VAD for intelligent filtering"}},{"feature":"GPU Acceleration","openWhispr":{"status":"partial","details":"Depends on model and platform"},"handy":{"status":"full","details":"GPU support for Whisper (Intel, AMD, NVIDIA)"}},{"feature":"Real-Time Dictation UX","openWhispr":{"status":"full","details":"System-wide dictation with overlay"},"handy":{"status":"full","details":"Push-to-talk with auto-paste"}},{"feature":"Tech Stack","openWhispr":{"status":"partial","details":"Electron + React + TypeScript"},"handy":{"status":"full","details":"Tauri (Rust) + React, smaller footprint"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","agentMode":{"title":"Agent Mode & AI Post-Processing:","body":"Go beyond raw transcription. Use custom system prompts to reformat, translate, summarize, or clean up your dictation automatically."},"cloud":{"title":"Cloud Transcription Option:","body":"Use OpenWhispr Cloud or bring your own API keys for lightning-fast results when local processing isn't fast enough."},"overlay":{"title":"Polished Dictation Overlay:","body":"A system-wide dictation experience with a visual overlay that works seamlessly across all your apps."},"pricing":{"title":"Free + Optional Pro:","body":"Free forever with all core features. Optional Pro ($8/user/mo) for unlimited cloud transcription, more meeting hours, and the mobile app."}},"whyHandy":{"title":"Why Handy?","architecture":{"title":"Lightweight Rust/Tauri Architecture:","body":"Built with Tauri and Rust instead of Electron, resulting in a smaller binary, less memory usage, and faster startup times."},"parakeet":{"title":"Parakeet V3 Model Support:","body":"Offers NVIDIA's Parakeet V3 as an alternative to Whisper, with automatic language detection and CPU-only operation at ~5x real-time speed."},"vad":{"title":"Advanced Voice Activity Detection:","body":"Uses Silero VAD for intelligent silence filtering, so only your actual speech gets transcribed."},"community":{"title":"Large Community & MIT License:","body":"Over 14,000 GitHub stars and an active community. MIT licensed for maximum flexibility."}},"pricing":{"title":"The Real Cost Comparison","ow":{"freeLabel":"Forever Free","free1":"All core features included","free2":"Local processing","free3":"Agent mode & system prompts","free4":"Community support","proLabel":"Pro (Optional)","pro1":"Unlimited cloud transcription","pro2":"20h meeting recordings/month","pro3":"Mobile app","pro4":"Email support"},"handy":{"freeLabel":"Free & Open Source","free1":"All features included","free2":"Local-only processing","free3":"Multiple model options","free4":"Community support","noPaidTier":"No Paid Tier","sponsorsLabel":"GitHub Sponsors optional","paid1":"No premium features gated","paid2":"Everything is free","paid3":"Sponsor to support development"},"callout":{"lead":"Both apps are free:","body":"This is a rare comparison where cost isn't a differentiator. Both tools are genuinely free and open source. The question is which feature set and architecture fits your workflow better."}},"featureShowdown":{"title":"Feature Showdown","bothHaveTitle":"Features Both Have:","bothHave":["Open source and free forever","Local Whisper model processing","Privacy-first, no cloud required","macOS, Windows, and Linux","Global keyboard shortcut activation","Works in any application","Push-to-talk recording mode","Multiple Whisper model sizes"],"openWhisprExclusivesTitle":"OpenWhispr Exclusives:","openWhisprExclusives":["Agent mode with custom AI system prompts","Cloud transcription (OpenWhispr Cloud or BYOK)","Visual dictation overlay across apps","Optional Pro tier with unlimited cloud and mobile app"],"handyExclusivesTitle":"Handy Exclusives:","handyExclusives":["Parakeet V3 model (CPU-only, auto language detection)","Tauri/Rust architecture (smaller footprint)","Silero voice activity detection","Broad GPU acceleration (Intel, AMD, NVIDIA)"]},"evidence":{"title":"Evidence-Based Notes","body":"This comparison is based on vendor documentation, pricing pages, and public product docs rather than anecdotal testimonials."},"migration":{"title":"Switching from Handy to OpenWhispr","intro":"Already using Handy and curious about AI-powered post-processing? Here's how to try OpenWhispr:","step1":{"title":"Download OpenWhispr","body":"Download for free. Optional Pro ($8/user/mo) for unlimited cloud transcription, more meeting hours, and the mobile app."},"step2":{"title":"Set Up Your Hotkey","body":"Configure the same push-to-talk shortcut you use in Handy for a familiar experience."},"step3":{"title":"Try Agent Mode","body":"Set up a system prompt to automatically format, clean up, or translate your dictation — the feature Handy doesn't have."},"callout":{"lead":"Keep both!","body":"Since both apps are free and open source, you don't have to choose. Many users run both and use each for different workflows."}},"bottomLine":{"title":"The Bottom Line","intro":"This is a genuinely close comparison. Both OpenWhispr and Handy are free, open source, privacy-focused dictation tools that work across macOS, Windows, and Linux. Neither locks you into a subscription or sends your voice to the cloud without consent.","teamHandy":{"title":"Team Handy","body":"You want the leanest possible tool with a Rust-powered backend, Parakeet V3 support, and excellent voice activity detection. You prefer local-only processing and want the smallest possible resource footprint on your system."},"teamOpenWhispr":{"title":"Team OpenWhispr","body":"You want dictation that goes beyond transcription. Agent mode, cloud processing options, a visual dictation overlay, and a growing feature set make OpenWhispr ideal for writers, developers, and professionals who want AI-enhanced voice input."},"closing":"Two great open source projects with the same values but different strengths. The best part? Both are free, so you can try them both."},"finalCta":{"title":"Ready to Try AI-Enhanced Dictation?","body":"Get the same local transcription you love from Handy, plus Agent Mode, cloud options, and a polished system-wide overlay. Free forever.","starButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever"},"viewAllComparisons":"View all dictation app comparisons"},"compareMacwhisper":{"metadata":{"title":"MacWhisper Alternative: Free Open-Source Dictation","description":"MacWhisper alternative for real-time dictation. OpenWhispr is free, open-source, and works on Mac, Windows, Linux & iOS — try it free today.","keywords":["MacWhisper alternative","OpenWhispr vs MacWhisper","cross-platform whisper app","open source dictation","real-time dictation app","whisper transcription","nvidia parakeet","free dictation software","MacWhisper Windows alternative","MacWhisper Linux alternative"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","macwhisper":"vs MacWhisper"},"hero":{"title":"OpenWhispr vs MacWhisper","subtitle":"Two AI-powered speech tools, two different philosophies. Real-time dictation meets file transcription.","githubButton":"View on GitHub"},"quickTake":{"title":"The Quick Take","paragraph1":"by Jordi Bruin is a highly-regarded macOS app built for transcribing audio and video files. Drag in a podcast, meeting recording, or video file and get a polished transcript with speaker labels, timestamps, and multiple export formats. It's excellent at what it does — post-recording transcription — and has earned a 4.5-star rating from over 2,000 users.","paragraph2":"solves a different problem: real-time voice dictation that works system-wide in any application. Press a hotkey, speak, and your words appear wherever your cursor is. It's open source, cross-platform (Mac, Windows, Linux, iOS), and free forever — powered by OpenAI Whisper and NVIDIA Parakeet.","chooseMacWhisper":"Choose MacWhisper if: You need to transcribe existing audio/video files with speaker labels, timestamps, and subtitle export.","chooseOpenWhispr":"Choose OpenWhispr if: You want real-time dictation in any app, cross-platform support, or open-source freedom."},"featureTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Primary Use Case","openWhispr":{"status":"full","details":"Real-time dictation into any app"},"macWhisper":{"status":"full","details":"File transcription and audio processing"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"macWhisper":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"macWhisper":{"status":"partial","details":"macOS (iOS/iPad added recently)"}},{"feature":"Pricing","openWhispr":{"status":"full","details":"Free forever, optional $8/mo Pro"},"macWhisper":{"status":"partial","details":"Free basic tier with optional paid tiers (pricing varies by channel)"}},{"feature":"Privacy & Local Processing","openWhispr":{"status":"full","details":"100% local, BYOK for cloud APIs"},"macWhisper":{"status":"full","details":"Local Whisper processing on-device"}},{"feature":"AI Model Options","openWhispr":{"status":"full","details":"OpenAI Whisper and NVIDIA Parakeet models"},"macWhisper":{"status":"full","details":"Multiple models including WhisperKit"}},{"feature":"System-Wide Dictation","openWhispr":{"status":"full","details":"Push-to-talk hotkey, works everywhere"},"macWhisper":{"status":"partial","details":"Has dictation mode, but primarily file-focused"}},{"feature":"Batch File Processing","openWhispr":{"status":"none","details":"Focused on live dictation"},"macWhisper":{"status":"full","details":"Drag-and-drop batch transcription"}},{"feature":"Export Formats","openWhispr":{"status":"partial","details":"Clipboard paste, text history"},"macWhisper":{"status":"full","details":"SRT, VTT, HTML, text, subtitles"}},{"feature":"Speaker Identification","openWhispr":{"status":"none","details":"Not available"},"macWhisper":{"status":"full","details":"Auto speaker recognition (v12+)"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","realTimeTitle":"Real-Time Dictation:","realTimeText":"Speak and see your words appear instantly in any app. Not just transcribing files after the fact.","crossPlatformTitle":"Cross-Platform:","crossPlatformText":"Works on Mac, Windows, Linux and iOS. MacWhisper is Apple-only.","openSourceTitle":"Open Source Freedom:","openSourceText":"Inspect the code, modify features, contribute improvements. No black boxes.","freeTitle":"Forever Free:","freeText":"All core features at no cost. Pro is optional at $8/month for convenience."},"whyMacWhisper":{"title":"Why MacWhisper?","fileTranscriptionTitle":"File Transcription:","fileTranscriptionText":"Purpose-built for transcribing audio and video files. Drag, drop, and get a polished transcript.","speakerTitle":"Speaker Recognition:","speakerText":"Automatic speaker identification and labeling with color-coded transcripts. Great for meetings and interviews.","exportTitle":"Rich Export Options:","exportText":"Export to SRT, VTT, HTML, text, and subtitle formats. Built-in video player with synced playback.","appleTitle":"Apple Ecosystem:","appleText":"Native macOS design with recent iPhone and iPad support. WhisperKit models optimized for Apple Silicon."},"pricing":{"title":"The Real Cost Comparison","openWhispr":{"free":{"label":"Forever Free","features":["Real-time dictation in any app","Whisper and Parakeet models included","AI cleanup via system prompts","Cross-platform support"]},"pro":{"label":"Pro","features":["Unlimited cloud transcription","20h meeting recordings/month","Mobile app","Email support"]}},"macWhisper":{"free":{"price":"Free","label":"Basic Version","features":["Basic Whisper transcription","Smaller models only","Local processing"]},"pro":{"price":"Pro","label":"One-Time Purchase","features":["All models including WhisperKit","Speaker recognition","Advanced export formats","Voice Activity Detection"]}},"calloutLead":"Different tools, different jobs:","calloutText":"MacWhisper is a great investment for transcription workflows. OpenWhispr is free for real-time dictation — and you might want both."},"featureShowdown":{"title":"Feature Showdown","sharedTitle":"Features Both Share:","shared":["Powered by OpenAI Whisper and NVIDIA Parakeet","100% local on-device processing","Privacy-first architecture","Multiple language support (99+)","Free tier available","High-accuracy transcription"],"openWhisprExclusivesTitle":"OpenWhispr Exclusives:","openWhisprExclusives":["Real-time dictation with push-to-talk hotkey","Works system-wide in any text field","Cross-platform (Mac, Windows, Linux, iOS)","Open source and forkable","AI cleanup with custom system prompts","BYOK for OpenAI/Groq cloud APIs"],"macWhisperExclusivesTitle":"MacWhisper Exclusives:","macWhisperExclusives":["Batch audio/video file transcription","Automatic speaker identification","Subtitle export (SRT, VTT)","Built-in video player with synced playback","Auto-translation of transcripts","WhisperKit + Parakeet model support"]},"evidence":{"title":"Evidence-Based Notes","text":"This comparison is based on vendor documentation, pricing pages, and public product docs rather than anecdotal testimonials."},"migration":{"title":"Adding Real-Time Dictation to Your Workflow","intro":"Already use MacWhisper for file transcription? OpenWhispr is the perfect complement for live dictation:","step1Title":"Download OpenWhispr","step1Text":"Download for free. Optional Pro ($8/user/mo) for unlimited cloud transcription, more meeting hours, and the mobile app.","step2Title":"Set Your Hotkey","step2Text":"Configure push-to-talk (hold or toggle) and pick your preferred Whisper model for speed vs accuracy.","step3Title":"Dictate Anywhere","step3Text":"Use OpenWhispr for live dictation in emails, docs, and chat — and MacWhisper for file transcription. Best of both worlds.","calloutLead":"Works on any platform:","calloutText":"If you also use Windows or Linux, OpenWhispr brings the same dictation experience to every machine you own."},"bottomLine":{"title":"The Bottom Line","paragraph":"OpenWhispr and MacWhisper are genuinely different tools that happen to use similar AI speech engines. MacWhisper excels at post-recording transcription — turning audio files into text with speaker labels, timestamps, and subtitle exports. OpenWhispr excels at real-time dictation — speaking your thoughts directly into any application as you work.","teamMacWhisperTitle":"Team MacWhisper","teamMacWhisperText":"You regularly transcribe podcasts, meetings, or video files. You need speaker labels, subtitle formats, and batch processing. You're in the Apple ecosystem.","teamOpenWhisprTitle":"Team OpenWhispr","teamOpenWhisprText":"You want to dictate in real time into any app. You value open source, need cross-platform support, or want a free tool you can customize and extend.","closing":"Many users run both — MacWhisper for transcription, OpenWhispr for dictation. They complement each other perfectly."},"cta":{"title":"Ready for Real-Time Dictation?","subtitle":"Get the free, open-source voice dictation tool that works in every app on every platform. Powered by OpenAI Whisper and NVIDIA Parakeet.","githubButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever"},"compareHubLink":"← Compare OpenWhispr with other tools"},"compareHub":{"metadata":{"title":"Compare OpenWhispr — See How We Stack Up | OpenWhispr","description":"Honest, detailed dictation app comparisons. See how OpenWhispr compares to SuperWhisper, WisprFlow, Dragon NaturallySpeaking, Otter.ai, and more.","keywords":["dictation app comparison","voice to text comparison","openwhispr vs","best dictation app","dictation software comparison","speech to text alternatives"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons"},"hero":{"title":"Compare OpenWhispr","subtitle":"Side-by-side comparisons with every major dictation app. Features, pricing, privacy, and trade-offs.","description":"We believe in transparency. Every comparison covers features, pricing, privacy, platform support, and real trade-offs — including where the competitor genuinely excels."},"card":{"versus":"vs {name}","readFullComparison":"Read full comparison →"},"competitors":{"granola":{"name":"Granola","description":"Cloud-based AI meeting notepad with bot-free capture and calendar sync.","differentiator":"Local-first open source alternative with on-device speaker diarization"},"meetily":{"name":"Meetily","description":"Open source, local-first AI meeting assistant using Whisper.","differentiator":"Same local-first values + calendar, diarization, and chat with notes shipped today"},"otter":{"name":"Otter.ai","description":"Cloud-based meeting transcription and note-taking service.","differentiator":"Private local dictation vs cloud meeting transcription"},"macwhisper":{"name":"MacWhisper","description":"macOS app for transcribing audio files using Whisper models.","differentiator":"Cross-platform real-time dictation vs Mac-only file transcription"},"superwhisper":{"name":"SuperWhisper","description":"Polished proprietary dictation app across macOS, Windows, and iOS.","differentiator":"Open source local control vs proprietary managed experience"},"wisprflow":{"name":"WisprFlow","description":"Proprietary AI dictation tool with context-aware auto-editing.","differentiator":"Open source freedom vs proprietary convenience"},"willowvoice":{"name":"WillowVoice","description":"Proprietary cloud-based voice writing app for desktop and mobile.","differentiator":"Local open-source dictation vs cloud-managed voice writing"},"dragon":{"name":"Dragon Professional","description":"Nuance Dragon Professional Speech Recognition (formerly NaturallySpeaking), now owned by Microsoft.","differentiator":"Free open-source AI vs $699 Windows-only legacy dictation"},"windowsSpeech":{"name":"Windows Speech Recognition","description":"Built-in Windows voice typing with basic dictation features.","differentiator":"Powerful AI dictation vs basic built-in tool"},"appleDictation":{"name":"Apple Dictation","description":"Built-in macOS and iOS dictation with limited customization.","differentiator":"Customizable AI dictation vs limited built-in feature"},"talon":{"name":"Talon Voice","description":"Complex voice control system designed for hands-free computing.","differentiator":"Simple dictation vs complex voice programming"},"handy":{"name":"Handy","description":"Lightweight Tauri-based speech-to-text app with Parakeet and Whisper support.","differentiator":"Full-featured dictation vs lightweight local transcription"},"voiceink":{"name":"VoiceInk","description":"Native macOS dictation app with screen-aware Smart Modes and Power Mode.","differentiator":"Cross-platform freedom vs polished native macOS experience"},"spokenly":{"name":"Spokenly","description":"macOS and iOS dictation app with Agent Mode and free local Whisper models.","differentiator":"Cross-platform open source vs polished Apple-only experience"},"typeless":{"name":"Typeless","description":"AI voice keyboard with smart editing, filler removal, and tone adaptation.","differentiator":"Unlimited offline dictation vs AI-polished cloud voice typing"},"fathom":{"name":"Fathom","description":"Cloud AI meeting notetaker with a generous free tier and automatic summaries.","differentiator":"Private on-device capture vs a cloud bot that records your meetings"},"fireflies":{"name":"Fireflies.ai","description":"Cloud AI meeting assistant whose bot joins calls with CRM sync and analytics.","differentiator":"Local, no-bot meeting notes vs a cloud bot that uploads your calls"}},"whyCompare":{"title":"Why Compare Dictation Apps?","paragraph1":"Choosing a voice-to-text tool is a personal decision that depends on your platform, privacy requirements, budget, and workflow. Some dictation apps excel on macOS but leave Windows and Linux users behind. Others require cloud processing that sends your voice data to remote servers. A few lock you into expensive subscriptions for features that should be free.","paragraph2":"OpenWhispr is open source, cross-platform, and privacy-first — but that does not mean it is the best choice for everyone. These comparison pages give you the information you need to decide for yourself. We cover the strengths of each competitor honestly, because the best tool is the one that fits your needs.","paragraph3":"Whether you are looking for a free alternative to Dragon NaturallySpeaking, a cross-platform replacement for SuperWhisper, or simply want to understand how AI-powered dictation compares to built-in speech recognition, you will find a detailed breakdown above."}},"compareAppleDictation":{"metadata":{"title":"Apple Dictation Alternative — OpenWhispr for Mac","description":"Apple Dictation alternative with local AI, offline privacy, and open-source code. OpenWhispr adds agent mode and cross-platform support. Free.","keywords":["Apple Dictation alternative","better Mac dictation","upgrade Mac dictation","Mac dictation app","better than Apple Dictation","Whisper dictation Mac","nvidia parakeet","AI dictation macOS","macOS dictation alternative","voice to text Mac"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","appleDictation":"vs Apple Dictation"},"hero":{"title":"OpenWhispr vs Apple Dictation","subtitle":"Apple Dictation is built in and convenient. OpenWhispr gives you AI-powered accuracy, agent mode, and total control.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","appleIntro":"is free, built into every Mac, and has improved significantly in recent macOS versions. It now supports on-device processing for about 29 languages, auto-punctuation in 19 languages, and works in any text field with no timeout. For quick, casual dictation it's genuinely solid.","openWhisprIntro":"is built on OpenAI Whisper and NVIDIA Parakeet -- state-of-the-art speech recognition used by researchers and developers worldwide. It supports 99+ languages, offers multiple model sizes for the accuracy/speed tradeoff you need, and adds agent mode with custom system prompts that can clean up, reformat, or transform your dictated text with AI. Plus it's open source, cross-platform, and free forever.","stickWithApple":"Stick with Apple Dictation if: You only need basic dictation for short notes and messages.","upgradeToOpenWhispr":"Upgrade to OpenWhispr if: You want better accuracy, AI text cleanup, custom prompts, or cross-platform support."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"AI Model Quality","openWhispr":{"status":"full","details":"Whisper and Parakeet models, state-of-the-art accuracy"},"appleDictation":{"status":"partial","details":"Apple's speech model, good but less accurate on technical content"}},{"feature":"Agent Mode / AI Cleanup","openWhispr":{"status":"full","details":"Custom system prompts reformat, summarize, or clean up text"},"appleDictation":{"status":"none","details":"No AI post-processing or text transformation"}},{"feature":"Privacy & Local Processing","openWhispr":{"status":"full","details":"100% local with Whisper and Parakeet, optional BYOK cloud APIs"},"appleDictation":{"status":"full","details":"On-device and server-assisted modes depending on language and Apple feature settings"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"appleDictation":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"appleDictation":{"status":"partial","details":"macOS and iOS only"}},{"feature":"Language Support","openWhispr":{"status":"full","details":"99+ languages via Whisper and Parakeet models"},"appleDictation":{"status":"partial","details":"Language and auto-punctuation coverage varies by language and region"}},{"feature":"Customization","openWhispr":{"status":"full","details":"Hotkeys, system prompts, model selection, BYOK, forkable"},"appleDictation":{"status":"partial","details":"Keyboard shortcut and language selection only"}},{"feature":"Auto-Punctuation","openWhispr":{"status":"full","details":"Whisper and Parakeet natively punctuate in all languages"},"appleDictation":{"status":"partial","details":"Supported language coverage varies by region and Apple feature availability"}}],"whyOpenWhispr":{"title":"Why Upgrade to OpenWhispr?","agentMode":{"title":"Agent Mode & System Prompts:","body":"AI post-processing cleans up filler words, reformats text, fixes grammar, or transforms dictation into bullet points, emails, or code comments."},"models":{"title":"State-of-the-Art AI Models:","body":"Choose from Whisper and Parakeet models. These consistently outperforms built-in dictation on technical vocabulary, accents, and noisy environments."},"crossPlatform":{"title":"Cross-Platform:","body":"Works on macOS, Windows, Linux and iOS. Your workflow isn't locked to Apple hardware."},"openSource":{"title":"Open Source & Hackable:","body":"Inspect the code, add features, integrate with your workflow. No black boxes, no vendor lock-in."}},"whereAppleShines":{"title":"Where Apple Dictation Shines","zeroSetup":{"title":"Zero Setup Required:","body":"Already on your Mac. Press the microphone key or Fn twice and start talking. No downloads or configuration."},"osIntegration":{"title":"Deep OS Integration:","body":"Works natively in every text field. Seamless with iOS, iPadOS, and iCloud. No clipboard paste needed."},"free":{"title":"Completely Free:","body":"Included with macOS at no additional cost. No accounts, no subscriptions, no upsells."},"voiceCommands":{"title":"Voice Commands:","body":"Built-in commands for punctuation, capitalization, new lines, and emoji insertion by name."}},"pricing":{"title":"The Real Cost Comparison","ow":{"freeLabel":"Forever Free","free1":"Whisper and Parakeet models included","free2":"Agent mode & system prompts","free3":"99+ languages","free4":"macOS, Windows, Linux","proLabel":"Pro","pro1":"Everything in Free","pro2":"Unlimited cloud transcription","pro3":"20h meeting recordings/month","pro4":"Mobile app","pro5":"Email support"},"apple":{"builtInLabel":"Built into macOS","pro1":"Basic dictation","pro2":"Auto-punctuation (19 languages)","pro3":"Voice commands for formatting","con1":"No agent mode or AI cleanup","con2":"No model selection","con3":"Apple hardware only"},"callout":{"lead":"Both are free.","body":"The difference is what you get: Apple Dictation gives you basic speech-to-text. OpenWhispr gives you a full AI dictation toolkit with Whisper and Parakeet accuracy, agent mode, and the freedom to customize everything."}},"featureShowdown":{"title":"Feature Showdown","bothHaveTitle":"Features Both Have:","bothHave":["On-device speech recognition","Works in any text field","Auto-punctuation","No time limit on dictation","Free to use","Multi-language support"],"openWhisprExclusivesTitle":"OpenWhispr Exclusives:","openWhisprExclusives":["Agent mode with custom system prompts","Multiple AI model sizes (Whisper and Parakeet)","99+ languages with full punctuation","Cross-platform (macOS, Windows, Linux, iOS)","BYOK cloud API support","Open source and forkable"],"appleDictationExclusivesTitle":"Apple Dictation Exclusives:","appleDictationExclusives":["Built into macOS (zero setup)","Native text field integration (no clipboard)","Voice commands (capitalization, emoji)","iOS and iPadOS support","Ambiguous word highlighting"]},"evidence":{"title":"Evidence-Based Notes","body":"This comparison is based on vendor documentation, pricing pages, and public product docs rather than anecdotal testimonials."},"migration":{"title":"Upgrading from Apple Dictation to OpenWhispr","intro":"Ready for more accurate, AI-powered dictation? Here's how to get started:","step1":{"title":"Download OpenWhispr","body":"Download for free. Optional Pro ($8/user/mo) for unlimited cloud transcription, more meeting hours, and the mobile app."},"step2":{"title":"Set Your Hotkey","body":"Configure a push-to-talk hotkey. Many users pick the same Fn key they used for Apple Dictation."},"step3":{"title":"Pick an AI Model","body":"Choose from OpenAI Whisper or NVIDIA Parakeet. Start with \"base\" for a balance of speed and accuracy, or go straight to \"large\" for maximum quality."},"step4":{"title":"Try Agent Mode","body":"Set up a system prompt to clean up your dictation automatically. This is the feature Apple Dictation simply doesn't have."},"callout":{"lead":"You can use both side by side.","body":"Keep Apple Dictation for quick messages and use OpenWhispr when you need accuracy, formatting, or AI cleanup."}},"bottomLine":{"title":"The Bottom Line","paragraph1":"Apple Dictation is a capable built-in tool that's gotten significantly better in recent macOS versions. For basic dictation -- quick messages, short notes -- it does the job well.","paragraph2":"But if you need more from your voice input -- better accuracy on technical content, AI-powered text cleanup, custom formatting, or the ability to work on Windows and Linux -- OpenWhispr is the upgrade you've been looking for.","keepApple":{"title":"Keep Apple Dictation","body":"You only dictate short texts and emails on your Mac. You don't need AI cleanup or custom formatting. The built-in solution works fine for your needs."},"upgradeOpenWhispr":{"title":"Upgrade to OpenWhispr","body":"You want Whisper and Parakeet-level accuracy, agent mode with custom prompts, cross-platform support, or the transparency of open source. Everything Apple Dictation does, plus everything it doesn't."},"closing":"Apple Dictation is a good starting point. OpenWhispr is where you go when you outgrow it."},"finalCta":{"title":"Ready to Upgrade Your Mac Dictation?","body":"Get AI-powered dictation with Whisper and Parakeet accuracy, agent mode, and full customization. Free forever, open source, works everywhere.","starButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever","viewAllComparisons":"View all comparisons"}},"compareDragon":{"metadata":{"title":"Dragon Dictation Alternative — Free & Open Source","description":"OpenWhispr is a free, open source Dragon alternative with local, private dictation on Mac, Windows, Linux and iOS. No $699 license needed.","keywords":["Dragon NaturallySpeaking alternative","Dragon Professional alternative","Dragon Professional Speech Recognition alternative","Nuance Dragon alternative","free Dragon alternative","open source alternative to Dragon","Dragon dictation replacement","Dragon NaturallySpeaking replacement","Dragon Home discontinued alternative","Dragon Professional v16 alternative","Dragon Professional Anywhere alternative","voice dictation software free","Whisper dictation","nvidia parakeet","open source dictation","speech recognition software free","Dragon for Mac alternative","Microsoft Dragon alternative"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","dragon":"vs Dragon Professional"},"hero":{"title":"OpenWhispr vs Dragon Professional Speech Recognition","subtitle":"The free, open-source alternative to Nuance Dragon. Modern AI dictation that runs on every platform — not just Windows.","note":"Formerly Dragon NaturallySpeaking — now Dragon Professional v16, owned by Microsoft via Nuance.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","dragonIntro":"(originally Dragon NaturallySpeaking, now Dragon Professional Speech Recognition) defined dictation for over 25 years. Originally developed by Dragon Systems, it was acquired by Nuance, which Microsoft then purchased in 2022 for $19.7 billion. Since the acquisition, Microsoft has prioritized healthcare AI — launching Microsoft Dragon Copilot for clinicians — while the general-purpose desktop product has stagnated. Dragon Home was discontinued in February 2023 with no replacement. Dragon Professional v16 remains available at ~$699 but hasn't had a major update since its 2023 release.","openWhisprIntro":"is the modern, open-source alternative to Dragon Professional. Built on OpenAI Whisper and NVIDIA Parakeet, it's completely free, runs on macOS, Windows, Linux and iOS, and supports 99+ languages. For the millions of Dragon Home users left without an upgrade path — and anyone who doesn't want to pay $699 for Windows-only dictation — OpenWhispr delivers state-of-the-art accuracy at zero cost.","legacy":"Dragon's legacy: It pioneered consumer dictation and earned a reputation for industry-leading accuracy over 25+ years. That legacy deserves respect.","reality":"The reality today: Dragon Home is dead, Dragon for Mac was killed in 2018, Professional costs $699, and Microsoft's focus is healthcare AI — not desktop dictation. OpenWhispr is free, cross-platform, and actively developed."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature","headerDragon":"Dragon Professional"},"comparisonData":[{"feature":"Pricing","openWhispr":{"status":"full","details":"Free forever, optional $8/mo Pro"},"dragon":{"status":"none","details":"Dragon Professional v16: ~$699 one-time; Anywhere: ~$660/year"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"dragon":{"status":"partial","details":"Windows only (Mac discontinued Oct 2018)"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"dragon":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Consumer Version","openWhispr":{"status":"full","details":"Free for everyone, no restrictions"},"dragon":{"status":"none","details":"Dragon Home discontinued Feb 2023, no replacement"}},{"feature":"AI Technology","openWhispr":{"status":"full","details":"OpenAI Whisper and NVIDIA Parakeet models"},"dragon":{"status":"full","details":"Nuance Deep Learning engine"}},{"feature":"Privacy & Local Processing","openWhispr":{"status":"full","details":"100% local, your data never leaves your device"},"dragon":{"status":"partial","details":"Desktop v16: local; Anywhere: cloud-based"}},{"feature":"Accuracy","openWhispr":{"status":"full","details":"State-of-the-art Whisper and Parakeet models, 99+ languages"},"dragon":{"status":"full","details":"Up to 99% accuracy, Deep Learning engine"}},{"feature":"Voice Commands","openWhispr":{"status":"partial","details":"AI cleanup via system prompts, basic commands"},"dragon":{"status":"full","details":"Advanced custom voice commands and macros"}},{"feature":"Active Development","openWhispr":{"status":"full","details":"Active open source community, frequent releases"},"dragon":{"status":"partial","details":"Desktop v16 stagnant since 2023; cloud products updated"}},{"feature":"Multilingual Support","openWhispr":{"status":"full","details":"99+ languages with auto-detection"},"dragon":{"status":"partial","details":"Limited language packs, sold separately"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","free":{"title":"Free Forever:","body":"No $699 license fee. No subscription required. Bring your own API keys for cloud models on the free plan. Optional Pro ($8/mo) for unlimited cloud transcription and meeting recordings."},"crossPlatform":{"title":"Cross-Platform:","body":"Works on macOS, Windows, Linux and iOS. Dragon desktop is Windows-only — Dragon for Mac was discontinued in October 2018 and doesn't run on modern macOS."},"modernAi":{"title":"Modern AI:","body":"Powered by OpenAI Whisper and NVIDIA Parakeet with 99+ languages. Dragon's desktop engine hasn't had a major update since v16 launched in 2023."},"openSource":{"title":"Open Source:","body":"Inspect the code, modify features, contribute improvements. No vendor lock-in, no end-of-life surprises like Dragon Home users experienced."}},"dragonStrengths":{"title":"Dragon's Strengths","macros":{"title":"Voice Commands & Macros:","body":"Deep system-level voice commands, custom macros, and auto-text rules refined over 25+ years. Unmatched for power users who rely on complex voice workflows."},"enterprise":{"title":"Enterprise & Industry Solutions:","body":"Dragon Legal, Dragon Medical One, and Dragon Professional Anywhere offer specialized vocabularies and centralized management for organizations."},"accuracy":{"title":"Proven Accuracy:","body":"Decades of refinement with up to 99% out-of-box accuracy. Voice profiles adapt to your speech patterns over time for even better results."},"windows":{"title":"Deep Windows Integration:","body":"Application-specific voice commands, direct text insertion, and Auto Transcribe Folder Agent for batch processing audio files."}},"pricing":{"title":"The Real Cost Comparison","ow":{"freeLabel":"Forever Free","free1":"Unlimited local dictation","free2":"Whisper and Parakeet models","free3":"macOS, Windows, Linux","free4":"99+ languages","free5":"Bring your own API keys","proLabel":"Pro (optional)","pro1":"Unlimited cloud transcription","pro2":"20h meeting recordings/month","pro3":"Email support"},"dragonPro":{"heading":"Dragon Professional v16","oneTimeLabel":"One-time license","pro1":"Advanced voice commands","pro2":"Custom vocabularies","pro3":"Nuance Deep Learning engine","con1":"Windows only","con2":"No major updates since 2023"},"dragonHome":{"label":"Dragon Home (Discontinued)","con1":"Discontinued Feb 2023","con2":"No upgrade path to v16","con3":"Windows only"},"dragonAnywhere":{"subLabel":"Professional Anywhere (subscription)","pro1":"Cloud-based, auto-updates","pro2":"No voice profile training needed","pro3":"Mobile app included","con1":"Requires internet","con2":"Audio sent to cloud"},"dragonMobile":{"label":"Dragon Anywhere Mobile","pro1":"iOS and Android","con1":"Mobile only","con2":"Cloud processing required"},"callout":{"lead":"The math:","body":"Dragon Professional v16 costs $699 for a Windows-only license with no guarantee of future updates. Dragon Professional Anywhere costs $660 per year — that's $82/mo. OpenWhispr is free, runs on every platform, and is powered by cutting-edge AI. Even Pro at $8/mo is 10x cheaper than Dragon's subscription."}},"landscape":{"title":"The Dragon Product Landscape in 2026","intro":"Since Microsoft acquired Nuance, the Dragon product line has fragmented. Here's what exists today and what was discontinued:","available":{"title":"Still Available","item1":"— $699 desktop license, Windows only, last updated 2023","item2":"— ~$660/yr cloud subscription, receives regular updates","item3":"— Specialized legal dictation","item4":"— Cloud clinical documentation for healthcare","item5":"— AI clinical assistant (launched March 2025)"},"discontinued":{"title":"Discontinued","item1":"— Discontinued Feb 2023, no v16 replacement","item2":"— Discontinued Oct 2018, last version v6.0","item3":"— Brand retired in 2015, now \"Dragon Professional\"","item4":"— No longer sold since Feb 2023","item5":"— Replaced by Dragon Medical One (cloud)"}},"featureShowdown":{"title":"Feature Showdown","bothHaveTitle":"Features Both Have:","openWhisprExclusivesTitle":"OpenWhispr Exclusives:","dragonExclusivesTitle":"Dragon Exclusives:","bothHave":["High-accuracy speech recognition","Local / offline processing option","Works in any application","Continuous dictation","Push-to-talk hotkey activation","Custom vocabulary support"],"openWhisprExclusives":["Cross-platform (macOS, Windows, Linux, iOS)","Completely free and open source","99+ languages with automatic detection","AI agent mode for text cleanup and reformatting","Bring your own API keys (OpenAI, Anthropic, etc.)","Meeting recording with automatic detection","Active community development"],"dragonExclusives":["Advanced voice command macros and auto-text","Specialized legal and medical vocabularies","Enterprise management console (Nuance Management Center)","Auto Transcribe Folder Agent for batch processing","Adaptive voice profiles that learn over time","25+ years of refinement for English dictation"]},"evidence":{"title":"Evidence-Based Notes","item1":"Dragon Professional v16 pricing (~$699) and Dragon Professional Anywhere (~$660/yr) based on Nuance's official pricing and authorized reseller listings as of March 2026.","item2":"Dragon Home discontinuation confirmed by Nuance's February 2023 product advisory notice. Dragon for Mac discontinued October 2018, last version 6.0.","item3":"Microsoft acquired Nuance Communications in March 2022 for approximately $19.7 billion. Microsoft Dragon Copilot for healthcare launched March 2025.","item4":"Dragon Professional v16 system requirements: Windows 10 (1909+) or Windows 11, Intel Core i3+ or AMD Ryzen 3+, 4 GB RAM minimum. No macOS or Linux support.","item5":"This comparison is based on vendor documentation, official release notes, and public product information — not anecdotal testimonials."},"migration":{"title":"Switching from Dragon to OpenWhispr","intro":"Whether Dragon Home was discontinued on you, Dragon for Mac left you behind, or you're done paying $699 for a stagnating Windows-only product — here's how to switch:","step1":{"title":"Download OpenWhispr","body":"Free for macOS, Windows, and Linux. Bring your own API keys for cloud models at no extra cost. Optional Pro ($8/mo) for unlimited cloud transcription and meeting recordings."},"step2":{"title":"Choose Your AI Model","body":"Pick from OpenAI Whisper or NVIDIA Parakeet models. Start with \"base\" or \"small\" for a good balance of speed and accuracy — models download automatically."},"step3":{"title":"Set Up Your Hotkey","body":"Configure a push-to-talk hotkey just like Dragon's mic toggle. Hold to record, release to transcribe — works system-wide in any application."},"step4":{"title":"Enable AI Agent Mode (Optional)","body":"Use agent mode with system prompts to automatically clean up, reformat, and improve your transcriptions — similar to Dragon's formatting rules but powered by modern LLMs."},"callout":{"lead":"Former Dragon Home user?","body":"Your v15 license has no upgrade path — Nuance discontinued Dragon Home in February 2023 and Windows 10 reached end of support in October 2025. OpenWhispr is future-proof, open source, and works on Windows 11, macOS, and Linux."}},"bottomLine":{"title":"The Bottom Line","intro":"Dragon NaturallySpeaking earned its legendary status. For decades, it was simply the best dictation software available. But Microsoft's acquisition of Nuance shifted the focus to healthcare AI, leaving general-purpose desktop dictation in maintenance mode.","dragonIf":{"title":"Dragon Still Makes Sense If...","body":"You're a Windows power user who depends on advanced voice macros, need specialized legal or medical vocabularies, or your enterprise already has Dragon infrastructure. Dragon Professional Anywhere's cloud model also suits organizations that want centralized management."},"openWhisprIf":{"title":"OpenWhispr Is the Better Choice If...","body":"You want free, modern dictation that works on any platform. If you're a former Dragon Home or Dragon for Mac user, need Linux support, value open source, or simply refuse to pay $699 for Windows-only speech-to-text that hasn't been meaningfully updated since 2023."},"closing":"Dragon opened the door. Whisper, Parakeet, and open source kicked it wide open. The future of dictation is free, cross-platform, and community-driven."},"finalCta":{"title":"Ready for Modern Dictation?","body":"Join thousands of former Dragon users who've switched to OpenWhispr. Same great accuracy, zero cost, every platform.","starButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever","viewAllComparisons":"← View all comparisons"}},"compareGranola":{"metadata":{"title":"Granola Alternative — Local, Open Source AI Meeting Notes","description":"Compare OpenWhispr vs Granola. Get local AI meeting notes, on-device transcription, and speaker diarization — free and open source. Try it today.","keywords":["Granola alternative","Granola ai alternative","private meeting notes","local ai meeting notes","open source granola alternative","Granola vs OpenWhispr","ai meeting notepad privacy","on-device meeting transcription","local speaker diarization","chat with meeting notes","bot-free meeting transcription","granola for linux"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","granola":"vs Granola"},"hero":{"title":"OpenWhispr: The Private, Open-Source Granola Alternative","subtitle":"Looking for a Granola alternative that keeps your meetings private? OpenWhispr gives you the same AI meeting-notes workflow, but transcription runs locally on your device and the code is fully open source.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","granolaIntro":"is a polished AI notepad for back-to-back meetings. It captures your system audio without a bot, transcribes in the cloud, enhances your notes with AI, and lets you chat across meetings. It syncs with Google Calendar and Outlook, and integrates with Slack, Notion, HubSpot, Attio, and Zapier.","openWhisprIntro":"is the private, open-source Granola alternative. It gives you the same meeting workflow, but the pipeline runs on your device. Audio is transcribed locally with Whisper or Parakeet, speakers are diarized locally, and your notes are stored locally by default. It also doubles as a system-wide dictation tool.","chooseGranola":"Choose Granola if: You are deep in a cloud-first workflow, want the most polished commercial experience, and do not mind your meeting audio leaving your machine.","chooseOpenWhispr":"Choose OpenWhispr if: You want everything Granola does, but with local transcription, on-device speaker diarization, and the freedom of open source."},"privacy":{"title":"Privacy, Side by Side","openWhisprPoints":["Audio captured and transcribed on your device","Speaker diarization runs locally","Notes stored locally by default","Optional cloud sync is opt-in and encrypted","Zero data retention on free tier cloud transcription","Open source — you can verify every claim"],"granolaPoints":["Audio uploaded for cloud transcription","AI processing happens on Granola's servers","Transcripts stored server-side with auto-deletion policies","Closed source — you cannot audit the pipeline","Data controls depend on plan tier and admin settings"]},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"Local-first: audio and transcription run on your device by default"},"granola":{"status":"none","details":"Cloud-based transcription and AI processing on Granola's servers"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub — audit every line"},"granola":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Audio Never Leaves Your Device","openWhispr":{"status":"full","details":"100% on-device transcription with Whisper and Parakeet models"},"granola":{"status":"none","details":"Audio is uploaded and processed in the cloud"}},{"feature":"Local Speaker Diarization","openWhispr":{"status":"full","details":"Who-said-what recognition runs entirely on your machine"},"granola":{"status":"none","details":"No on-device speaker diarization — processing is cloud-side"}},{"feature":"Bot-Free Meeting Capture","openWhispr":{"status":"full","details":"Captures system audio directly — no bot joins your call"},"granola":{"status":"full","details":"Captures system audio directly — no bot joins your call"}},{"feature":"AI Meeting Notes & Summaries","openWhispr":{"status":"full","details":"Auto-generated notes, action items, and custom templates"},"granola":{"status":"full","details":"Auto-generated notes with customizable templates"}},{"feature":"Chat With Your Notes & Data","openWhispr":{"status":"full","details":"Ask questions across all your notes and transcripts"},"granola":{"status":"full","details":"AI chat within and across meetings"}},{"feature":"Calendar Integration","openWhispr":{"status":"full","details":"Google Calendar sync for upcoming meetings"},"granola":{"status":"full","details":"Google Calendar and Microsoft Outlook sync"}},{"feature":"General Voice Dictation","openWhispr":{"status":"full","details":"Push-to-talk dictation in any app or text field"},"granola":{"status":"none","details":"Not a dictation tool — focused on meeting capture"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"granola":{"status":"partial","details":"macOS, Windows, iPhone — no Linux or Android"}},{"feature":"Works Offline","openWhispr":{"status":"full","details":"Full transcription and notes without internet using local models"},"granola":{"status":"none","details":"Requires internet connection — cloud processing"}},{"feature":"Free Tier","openWhispr":{"status":"full","details":"Unlimited local transcription, 5h meeting recordings/month"},"granola":{"status":"partial","details":"Limited meeting history on Basic tier"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","localAudio":{"title":"Your Audio Never Leaves Your Device:","body":"Transcription runs on-device with Whisper or Parakeet. No cloud upload, no third-party servers, no data broker relationships. Your meetings are yours."},"diarization":{"title":"Local Speaker Diarization:","body":"Know who said what — computed entirely on your machine. Granola does not offer on-device speaker diarization."},"chat":{"title":"Chat With Your Data:","body":"Ask questions across every meeting, note, and transcript. Agent mode can pull from your notes, calendar, and the web to draft follow-ups."},"openSource":{"title":"Open Source:","body":"Inspect, fork, and audit every line on GitHub. Verify exactly where your data goes — no blind trust required."},"platforms":{"title":"Linux + Windows + macOS:","body":"First-class native apps on all three desktop platforms. Granola ships macOS and Windows only."},"free":{"title":"Free for Individuals:","body":"Unlimited local transcription forever. Granola's Business tier is $14/user/month."}},"whyGranola":{"title":"Why Granola?","ux":{"title":"Very Polished UX:","body":"Granola has invested heavily in visual design and meeting template ergonomics. The product feels great to use."},"calendar":{"title":"Outlook + Google Calendar:","body":"Supports both Google and Microsoft calendar out of the box. OpenWhispr currently supports Google Calendar."},"integrations":{"title":"Deep Cloud Integrations:","body":"Attio, HubSpot, Affinity, Notion, Slack, and Zapier on paid tiers. If your team is CRM-heavy, this matters."},"admin":{"title":"Team Admin Controls:","body":"SSO, centralized billing, org-wide retention policies, and admin sharing controls on Enterprise."}},"pricing":{"title":"The Real Cost Comparison","ow":{"freeLabel":"Free — Forever","free":["Unlimited local transcription","Local speaker diarization","5h meeting recordings/month","100+ languages","Works offline","macOS, Windows, Linux"],"proLabel":"Pro","pro":["Everything in Free","20h meeting recordings/month","Unlimited cloud transcription","Mobile companion app"],"businessLabel":"Business","business":["Unlimited meeting recordings","Agent mode","Chat over your data"]},"granola":{"basicLabel":"Basic","basic":["✓ Limited meeting history","✓ AI chat within meetings","✓ Shared folders","✗ No unlimited notes"],"businessLabel":"Business","business":["Unlimited meeting notes","Advanced integrations","Centralized billing","MCP access"],"enterpriseLabel":"Enterprise","enterprise":["SSO","Priority support","Org-wide retention controls","Admin controls"]},"callout":{"lead":"The cost of cloud vs. local:","body":"Granola's Business tier is $168 per user per year. OpenWhispr Pro is $96/year for the same meeting workflow — plus your data never leaves your machine."}},"featureShowdown":{"title":"Feature Showdown","bothDoWellTitle":"What Both Do Well:","bothDoWell":["Bot-free meeting capture","AI-enhanced meeting notes","Customizable note templates","Chat with your meeting data","Calendar sync for upcoming meetings","Action item extraction"],"openWhisprExclusivesTitle":"OpenWhispr Exclusives:","openWhisprExclusives":["100% local transcription (audio never uploaded)","Local speaker diarization (who-said-what on-device)","Fully open source on GitHub","Works offline end-to-end","Native Linux support","Push-to-talk voice dictation in any app","BYOK for OpenAI, Anthropic, Groq, Ollama"],"granolaExclusivesTitle":"Granola Exclusives:","granolaExclusives":["Native Microsoft Outlook calendar support","Deep CRM integrations (Attio, HubSpot, Affinity)","Zapier workflows on paid tiers","Mature team admin and SSO controls","iPhone companion app (today)"]},"evidence":{"title":"Evidence-Based Notes","bodyPrefix":"This comparison is based on Granola's public pricing page, help center, and product marketing as of 2026. Vendor docs, plan inclusions, and platform support can change — refer to","bodySuffix":"for the latest."},"migration":{"title":"Switching from Granola to OpenWhispr","intro":"If you love Granola's workflow but want your meeting audio to stay on your machine, the switch takes about five minutes:","step1":{"title":"Download OpenWhispr","body":"Free download for macOS, Windows, and Linux. No account required to start."},"step2":{"title":"Pick Your Transcription Model","body":"Choose a local Whisper or NVIDIA Parakeet model that fits your hardware — or bring your own OpenAI/Groq key for cloud transcription you control."},"step3":{"title":"Connect Google Calendar","body":"Optional — sync upcoming meetings so notes attach to the right event automatically."},"step4":{"title":"Enable Local Speaker Diarization","body":"Turn on speaker recognition so your meeting notes say who spoke when — all computed on your device."},"step5":{"title":"Chat With Your Notes","body":"Ask questions across every meeting you have captured. Draft follow-ups with agent mode, using your notes, calendar, and the web as context."}},"bottomLine":{"title":"The Bottom Line","intro":"Granola and OpenWhispr compete on the same meeting workflow: capture the audio from your computer, let AI enhance your notes, and make the whole transcript chattable. The real difference is where that work happens.","teamGranola":{"title":"Team Granola","body":"Your team is already on HubSpot/Attio/Notion and wants the most polished commercial experience. Cloud processing is not a dealbreaker."},"teamOpenWhispr":{"title":"Team OpenWhispr","body":"You want Granola's workflow without Granola's cloud. You value local processing, local speaker diarization, open source, Linux support, and the freedom to bring your own keys."},"closing":"The best meeting notepad is the one that does not make you trade privacy for intelligence. OpenWhispr gives you both."},"finalCta":{"title":"Ready for Private, Local AI Meeting Notes?","body":"Stop uploading your meetings to the cloud. OpenWhispr gives you AI notes, speaker diarization, calendar sync, and chat with your data — all running on your machine.","starButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever","viewAllComparisons":"View all comparisons"}},"compareMeetily":{"metadata":{"title":"Meetily Alternative: Local Open Source AI Notes","description":"Meetily alternative: OpenWhispr is open source and local-first with calendar sync, speaker diarization, and chat with your notes today. Try it free.","keywords":["Meetily alternative","Meetily vs OpenWhispr","open source meeting assistant","local meeting transcription","local speaker diarization","open source meeting notes","calendar integration meeting notes","chat with meeting notes","private meeting assistant","on-device meeting transcription","ai meeting notepad open source","bot-free meeting notes"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","meetily":"vs Meetily"},"hero":{"title":"OpenWhispr vs Meetily","subtitle":"Two open source, local-first AI meeting assistants. See how OpenWhispr adds calendar sync, local speaker diarization, and chat with your notes — today, not on a roadmap.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","meetilyIntro":"is an open source, local-first AI meeting assistant. It records, transcribes, and summarizes meetings using local Whisper models, with an optional local LLM or BYOK for summaries. Your data never leaves your device. It is MIT licensed, actively developed, and free for individual use.","openWhisprIntro":"shares Meetily's local-first, open source values — and ships the features Meetily has on its roadmap. Calendar integration, local speaker diarization, and chat with your notes are all available today. OpenWhispr is also a full voice-dictation app: push-to-talk into any text field, anywhere on your desktop.","chooseMeetily":"Choose Meetily if: You want a minimal, MIT-licensed meeting recorder and are happy to wait for calendar sync, diarization, and chat features.","chooseOpenWhispr":"Choose OpenWhispr if: You want the same local-first guarantees plus the full feature set — calendar sync, local speaker diarization, chat with your data, and general voice dictation — already shipped."},"sharedValues":{"title":"Where We Agree","body":"OpenWhispr and Meetily share the same foundational commitments — and we think that is a good thing. The local-first, open source meeting-notes category is better with both of us pushing it forward.","items":["Audio never leaves your device","Open source — auditable pipeline","Local Whisper-based transcription","BYOK for cloud LLMs, if you want them","Free community edition","Bot-free system audio capture"]},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"Local-first: audio and transcription run on your device by default"},"meetily":{"status":"full","details":"Local-first: all processing happens on device"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"meetily":{"status":"full","details":"Open source under MIT license"}},{"feature":"Audio Never Leaves Your Device","openWhispr":{"status":"full","details":"On-device transcription with Whisper and Parakeet"},"meetily":{"status":"full","details":"On-device transcription with Whisper models"}},{"feature":"Local Speaker Diarization","openWhispr":{"status":"full","details":"Who-said-what recognition runs entirely on your machine today"},"meetily":{"status":"none","details":"Listed on the roadmap — not available today"}},{"feature":"Chat With Your Notes & Data","openWhispr":{"status":"full","details":"Ask questions across all your notes and transcripts today"},"meetily":{"status":"none","details":"Listed on the roadmap — not available today"}},{"feature":"Calendar Integration","openWhispr":{"status":"full","details":"Google Calendar sync — meetings auto-attach to notes"},"meetily":{"status":"none","details":"Listed on the roadmap — not available today"}},{"feature":"AI Meeting Notes & Summaries","openWhispr":{"status":"full","details":"Auto-generated notes, action items, and custom templates"},"meetily":{"status":"full","details":"AI summaries with six built-in templates"}},{"feature":"Bot-Free Meeting Capture","openWhispr":{"status":"full","details":"Captures system audio directly — no bot joins your call"},"meetily":{"status":"full","details":"Captures system audio directly — no bot joins your call"}},{"feature":"General Voice Dictation","openWhispr":{"status":"full","details":"Push-to-talk dictation in any app or text field"},"meetily":{"status":"none","details":"Meeting-only — no general dictation"}},{"feature":"Mobile App","openWhispr":{"status":"partial","details":"iOS companion app"},"meetily":{"status":"none","details":"No mobile app — listed as coming soon"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"Native macOS, Windows, Linux, iOS"},"meetily":{"status":"partial","details":"Native macOS and Windows; Linux via Docker"}},{"feature":"Works Offline","openWhispr":{"status":"full","details":"Full transcription and notes without internet"},"meetily":{"status":"full","details":"Full transcription without internet using local Whisper"}},{"feature":"Free Tier","openWhispr":{"status":"full","details":"Unlimited local transcription, 5h meeting recordings/month"},"meetily":{"status":"full","details":"Free community edition with all core features"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","diarization":{"title":"Local Speaker Diarization Today:","body":"Who-said-what recognition runs entirely on your device. Meetily lists advanced speaker diarization on its roadmap — OpenWhispr ships it now."},"calendar":{"title":"Calendar Integration Today:","body":"Google Calendar sync attaches meeting notes to the right event automatically. Meetily's calendar integration is on the roadmap."},"chat":{"title":"Chat With Your Notes Today:","body":"Ask questions across every meeting, note, and transcript. Draft follow-ups with agent mode. Meetily's chat is on the roadmap."},"dictation":{"title":"General Voice Dictation:","body":"Push-to-talk into any app — emails, docs, code, terminal. Meetily is meeting-only."},"linux":{"title":"Native Linux App:","body":"Built for Linux, not run through Docker. A proper first-class desktop app on macOS, Windows, and Linux."},"mobile":{"title":"Mobile Companion:","body":"iOS companion app for voice capture on the go."}},"whyMeetily":{"title":"Why Meetily?","mit":{"title":"MIT License:","body":"Permissively licensed — fork, rebrand, or embed Meetily in your own tools with minimal friction."},"scope":{"title":"Minimal Scope:","body":"If you only need recording and summaries, Meetily's focused scope matches that exactly — no dictation, no calendar, no chat."},"docker":{"title":"Docker Deployment:","body":"Self-host via Docker on any Linux distro or server if that matches your workflow."},"community":{"title":"Community Momentum:","body":"Active GitHub project with strong open source community interest in local-first meeting tooling."}},"pricing":{"title":"The Real Cost Comparison","ow":{"freeLabel":"Free — Forever","free1":"Unlimited local transcription","free2":"Local speaker diarization","free3":"Google Calendar sync","free4":"5h meeting recordings/month","free5":"Voice dictation into any app","free6":"100+ languages","proLabel":"Pro","pro1":"20h meeting recordings/month","pro2":"Unlimited cloud transcription","pro3":"Mobile companion app","businessLabel":"Business","business1":"Unlimited meeting recordings","business2":"Agent mode","business3":"Chat over your data"},"meetily":{"communityLabel":"Community Edition","community1":"Real-time transcription","community2":"AI summaries","community3":"6 templates","community4":"Local Whisper models","community5":"No diarization","community6":"No calendar sync","community7":"No chat with notes","proLabel":"Pro (annual billing only — no monthly option)","pro1":"Enhanced accuracy","pro2":"Advanced exports","pro3":"Auto-detect features","pro4":"No monthly billing available","enterprisePrice":"Custom","enterpriseLabel":"Enterprise","enterprise1":"Dedicated support","enterprise2":"Compliance frameworks"},"callout":{"lead":"Similar pricing, different feature depth:","body":"Both tools offer a free tier with local transcription. OpenWhispr's free tier includes speaker diarization, calendar sync, and chat-with-notes — features that are roadmap items on Meetily."}},"featureShowdown":{"title":"Feature Showdown","openWhisprExclusivesTitle":"OpenWhispr Exclusives:","openWhisprExclusives":["Local speaker diarization (who-said-what on-device)","Google Calendar sync","Chat with your notes across all meetings","Agent mode (drafts, follow-ups, web + notes + calendar)","Push-to-talk voice dictation in any app","NVIDIA Parakeet models in addition to Whisper","Native Linux desktop app (not Docker)","iOS companion"],"meetilyExclusivesTitle":"Meetily Exclusives:","meetilyExclusives":["MIT license (permissive vs OpenWhispr's license)","Docker-based Linux deployment","Narrower scope if you only need recording + summaries"]},"evidence":{"title":"Evidence-Based Notes","bodyBefore":"Meetily feature availability is based on their public roadmap, GitHub project, and","bodyAfter":"as of 2026. Calendar integration, advanced speaker diarization, chat with meeting notes, and mobile apps are listed as planned/coming-soon by Meetily at the time of writing. Check their site for the latest."},"migration":{"title":"Switching from Meetily to OpenWhispr","intro":"Already using Meetily and want the features still on its roadmap? OpenWhispr keeps the same local-first values you chose Meetily for:","step1":{"title":"Download OpenWhispr","body":"Native installers for macOS, Windows, and Linux. Free to start, no account required for local use."},"step2":{"title":"Pick a Local Model","body":"Same Whisper model family you know from Meetily, plus NVIDIA Parakeet as an option. All runs on-device."},"step3":{"title":"Enable Speaker Diarization","body":"Turn on local diarization so your transcripts attribute lines to the right speaker — no cloud required."},"step4":{"title":"Connect Google Calendar","body":"Optional — auto-attach notes to the right meeting and let agent mode read your upcoming schedule."},"step5":{"title":"Chat With Your Meetings","body":"Ask questions across every meeting you have recorded. Draft follow-up emails with one command."}},"bottomLine":{"title":"The Bottom Line","intro":"Meetily and OpenWhispr belong to the same category: open source, local-first AI meeting assistants that keep your audio on your device. The difference is feature depth and scope.","teamMeetily":{"title":"Team Meetily","body":"You want a small, focused, MIT-licensed meeting recorder. You are happy to wait on diarization, chat, and calendar features."},"teamOpenWhispr":{"title":"Team OpenWhispr","body":"You want the same local-first, open source guarantees plus local speaker diarization, Google Calendar sync, chat with your notes, and general voice dictation — shipped today."},"closing":"Two projects, same values. OpenWhispr just ships more of the meeting-assistant stack today."},"finalCta":{"title":"Ready for the Full Local-First Meeting Stack?","body":"OpenWhispr is open source, local-first, and fully featured. Meeting notes, speaker diarization, calendar sync, and chat with your data — all running on your machine.","starButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever","viewAllComparisons":"View all comparisons"}},"compareOtter":{"metadata":{"title":"Otter.ai Alternative — Private, Local & Open Source","description":"The free, open-source Otter.ai alternative. OpenWhispr runs voice dictation 100% on-device — offline, private, no cloud, no usage caps. Try free.","keywords":["Otter.ai alternative","Otter alternative free","private dictation app","Otter.ai vs OpenWhispr","local voice to text","private transcription alternative","open source dictation","offline dictation app","nvidia parakeet"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","otter":"vs Otter.ai"},"hero":{"title":"The Private, Open-Source Otter.ai Alternative","subtitle":"OpenWhispr is an Otter.ai alternative for people who want voice-to-text that runs 100% on their own device. Your audio never leaves your machine, it works offline, and the whole app is open source.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","otterIntro":"is a powerful cloud-based meeting transcription platform. It joins your Zoom, Teams, and Google Meet calls, transcribes conversations in real time, generates summaries, extracts action items, and integrates with tools like Salesforce and HubSpot. It is excellent at what it does — but it is fundamentally a meeting tool, not a dictation tool.","openWhisprIntro":"takes a completely different approach. It is a privacy-first dictation tool that runs entirely on your computer using OpenAI Whisper and NVIDIA Parakeet. It also includes an on-device AI Notepad for capturing meeting notes without a bot joining the call.","chooseOtter":"Choose Otter.ai if: You need automated meeting transcription with team collaboration and CRM integrations.","chooseOpenWhispr":"Choose OpenWhispr if: You want private, local voice dictation that works in any app without sending audio to the cloud."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"100% local processing, your data never leaves your machine"},"otter":{"status":"none","details":"Cloud-based processing and storage per Otter policies"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"otter":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"otter":{"status":"full","details":"Web, macOS, Windows, iOS, Android"}},{"feature":"Works Offline","openWhispr":{"status":"full","details":"Full functionality without internet"},"otter":{"status":"none","details":"Requires internet connection for all features"}},{"feature":"General Dictation","openWhispr":{"status":"full","details":"Built for dictation in any text field"},"otter":{"status":"partial","details":"Primarily meeting-focused, limited dictation use"}},{"feature":"Meeting Transcription","openWhispr":{"status":"partial","details":"Meeting recording included (5h free, 20h Pro)"},"otter":{"status":"full","details":"Core feature with Zoom/Teams/Meet integration"}},{"feature":"AI Cleanup & Reformatting","openWhispr":{"status":"full","details":"Custom system prompts for any formatting task"},"otter":{"status":"partial","details":"Auto-generated summaries and action items"}},{"feature":"Free Tier Limits","openWhispr":{"status":"full","details":"Unlimited, all core features included"},"otter":{"status":"partial","details":"300 minutes/month, 3 lifetime file imports"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","privacy":{"title":"True Privacy — Local Processing:","body":"Your audio never leaves your device. No cloud uploads, no third-party servers, nothing stored off-device. Otter.ai processes everything in the cloud and stores recordings in Otter's cloud."},"dictation":{"title":"Built for Dictation:","body":"OpenWhispr is designed for typing with your voice — drafting emails, writing code comments, filling forms. Otter is designed for transcribing meetings."},"noLimits":{"title":"No Usage Limits:","body":"Dictate as much as you want, forever. Otter's free tier caps you at 300 minutes per month, and even Pro limits you to 1,200 minutes."},"openSource":{"title":"Open Source & Hackable:","body":"Inspect every line of code. Fork it, extend it, audit it. Know exactly what happens with your audio data."}},"whyOtter":{"title":"Why Otter.ai?","meetings":{"title":"Meeting Transcription:","body":"Otter automatically joins Zoom, Teams, and Meet calls, transcribes in real time, and identifies speakers. This is its core strength."},"collaboration":{"title":"Team Collaboration:","body":"Share transcripts, highlight key moments, assign action items to teammates. Built for teams who need shared meeting notes."},"summaries":{"title":"AI Summaries & Action Items:","body":"Automatically generates meeting summaries, extracts decisions, and creates action items you can assign and track."},"integrations":{"title":"Deep Integrations:","body":"Connects with 50+ tools including Salesforce, HubSpot, Slack, Notion, Jira, and Asana. Powerful for sales and project management workflows."}},"pricing":{"title":"The Real Cost Comparison","ow":{"freeLabel":"Forever Free","free1":"Unlimited dictation","free2":"Whisper and Parakeet models included","free3":"100% local processing","free4":"AI cleanup via system prompts","free5":"macOS, Windows, Linux","proLabel":"Pro (Optional)","pro1":"Everything in Free","pro2":"Unlimited cloud transcription","pro3":"20h meeting recordings/month","pro4":"Mobile app","pro5":"Email support"},"otter":{"basicLabel":"Basic","basic1":"300 minutes/month","basic2":"3 file imports (lifetime)","basic3":"Meeting integrations","basic4":"Limited storage","proLabel":"Pro ($8.33/mo billed annually)","pro1":"1,200 minutes/month","pro2":"10 file imports/month","pro3":"Advanced search & export","pro4":"90-min max meeting length","businessLabel":"Business ($13.33/mo billed annually)","business1":"Unlimited recordings","business2":"4-hour meeting limit","business3":"Usage analytics","business4":"Priority support"},"callout":{"lead":"Different tools, different costs:","body":"Otter.ai is priced for teams who need meeting transcription. OpenWhispr gives you unlimited private dictation for free — and even the Pro tier costs less than half of Otter's cheapest paid plan."}},"featureShowdown":{"title":"Feature Showdown","bothHaveTitle":"Features Both Have:","bothHave":["Voice-to-text transcription","Multi-language support","Punctuation and formatting","AI-powered text processing","Free tier available"],"openWhisprExclusivesTitle":"OpenWhispr Exclusives:","openWhisprExclusives":["100% local/offline processing","Open source (inspect and fork the code)","Works in any text field via clipboard paste","Custom AI system prompts (agent mode)","Unlimited free usage, no minute caps","Linux support","BYOK for OpenAI and Groq APIs"],"otterExclusivesTitle":"Otter.ai Exclusives:","otterExclusives":["Automated meeting bot (joins calls for you)","Real-time team collaboration on transcripts","CRM integrations (Salesforce, HubSpot)","Auto-generated meeting summaries","Action item extraction and assignment","iOS and Android mobile apps","Enterprise features (SSO, HIPAA, admin controls)"]},"evidence":{"title":"Evidence-Based Notes","body":"This comparison is based on vendor documentation, pricing pages, and public product docs rather than anecdotal testimonials."},"migration":{"title":"Switching from Otter.ai to OpenWhispr for Dictation","intro":"Using Otter.ai for dictation and want something purpose-built? Here is how to make the switch:","step1":{"title":"Download OpenWhispr","body":"Download for free. Optional Pro ($8/user/mo) for unlimited cloud transcription, more meeting hours, and the mobile app. Works on macOS, Windows, Linux and iOS."},"step2":{"title":"Pick Your AI Model","body":"Choose from OpenAI Whisper or NVIDIA Parakeet models based on your hardware and needs. Start with \"base\" for a good balance."},"step3":{"title":"Set Up Your Hotkey","body":"Configure push-to-talk (hold to record or toggle mode). Then just speak — your text appears wherever your cursor is."},"step4":{"title":"Customize AI Cleanup","body":"Set up custom system prompts to format your dictation automatically — fix grammar, format as email, translate, or anything else you need."},"callout":{"lead":"Keep Otter for meetings:","body":"OpenWhispr and Otter.ai solve different problems. Many users keep Otter for meeting transcription while using OpenWhispr for everyday dictation — the best of both worlds."}},"bottomLine":{"title":"The Bottom Line","intro":"Otter.ai and OpenWhispr are built for different jobs. Comparing them head-to-head on every metric misses the point — the right choice depends entirely on what you need.","teamOtter":{"title":"Team Otter.ai","body":"You are in meetings all day, need transcripts shared across your team, and want AI-generated summaries piped into Salesforce or Slack. Otter is purpose-built for this and does it exceptionally well."},"teamOpenWhispr":{"title":"Team OpenWhispr","body":"You want to dictate text privately, without cloud processing, usage limits, or subscription costs. You value open source, work across multiple platforms, and want a tool that does one thing perfectly: turn your voice into text."},"closing":"The best tool is the one built for your actual workflow. For private dictation, OpenWhispr is the clear choice. For meeting transcription, Otter.ai earns its reputation."},"finalCta":{"title":"Ready for Private, Open Source Dictation?","body":"Stop sending your voice to the cloud. OpenWhispr gives you fast, accurate dictation that runs entirely on your machine — free and open source.","starButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever","viewAllComparisons":"View all comparisons"}},"compareSpokenly":{"metadata":{"title":"OpenWhispr vs Spokenly — Detailed Comparison | OpenWhispr","description":"Compare OpenWhispr and Spokenly side by side. Open-source cross-platform dictation vs a polished macOS and iOS voice-to-text app with Agent Mode.","keywords":["spokenly alternative","spokenly vs openwhispr","mac dictation app","whisper dictation","voice to text mac","offline dictation"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","spokenly":"vs Spokenly"},"hero":{"title":"OpenWhispr vs Spokenly","subtitle":"Spokenly is a popular macOS and iOS dictation app with Agent Mode. Here is how it compares to the open-source, cross-platform alternative.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","spokenlyIntro":"is a polished voice-to-text app for macOS and iOS with 100,000+ active users. Its free local models (Whisper and Parakeet) have no usage limits, and its Agent Mode lets you control your Mac with voice commands. Pro costs $9.99/month for managed cloud models.","openWhisprIntro":"offers a different value proposition. It is completely open source, works on macOS, Windows, Linux and iOS, and gives you full control over your data and workflow. Local processing is free forever with optional $8/month pro for cloud features and mobile access.","chooseSpokenly":"Choose Spokenly if: You are on Mac/iOS and want Agent Mode, smart prompts, and a polished native experience.","chooseOpenWhispr":"Choose OpenWhispr if: You need Windows or Linux support, value open source, or want to customize your tools."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"100% local option, bring your own API keys"},"spokenly":{"status":"full","details":"Local models available, cloud audio processed and deleted"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"spokenly":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"spokenly":{"status":"partial","details":"macOS and iOS only"}},{"feature":"Pricing Model","openWhispr":{"status":"full","details":"Free forever, optional $8/mo pro"},"spokenly":{"status":"full","details":"Local models free and unlimited, Pro $9.99/mo for cloud"}},{"feature":"AI Model Options","openWhispr":{"status":"full","details":"OpenAI Whisper and NVIDIA Parakeet models"},"spokenly":{"status":"full","details":"Whisper, Parakeet, plus cloud models via BYOK or Pro"}},{"feature":"Voice Commands","openWhispr":{"status":"partial","details":"Basic punctuation commands"},"spokenly":{"status":"full","details":"Agent Mode with Mac automation, app launching, web search"}},{"feature":"Offline Support","openWhispr":{"status":"full","details":"Complete offline functionality"},"spokenly":{"status":"full","details":"Local-only mode blocks all network requests"}},{"feature":"Customization","openWhispr":{"status":"full","details":"Fully customizable, fork-friendly"},"spokenly":{"status":"partial","details":"Smart prompts and settings, not extensible"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","crossPlatform":{"title":"Cross-Platform:","body":"Works on macOS, Windows, Linux and iOS. Spokenly is limited to macOS and iOS."},"openSource":{"title":"Open Source Freedom:","body":"Inspect the code, modify features, contribute improvements. No black boxes."},"pricing":{"title":"Lower Pro Cost:","body":"Pro at $8/month vs Spokenly Pro at $9.99/month, with meeting recordings and mobile included."},"hackable":{"title":"Hackable:","body":"Add features, create plugins, integrate with your workflow however you want."}},"whySpokenly":{"title":"Why Spokenly?","agentMode":{"title":"Agent Mode:","body":"Control your Mac with voice commands — search the web, launch apps, and run shortcuts hands-free."},"nativeApple":{"title":"Native Apple Experience:","body":"Purpose-built for macOS and iOS with Apple Silicon optimization and a 4.9-star App Store rating."},"cloudModels":{"title":"Broad Cloud Model Support:","body":"Bring your own keys for OpenAI, Claude, Gemini, Deepgram, and Groq — or let Pro handle it."},"mcp":{"title":"MCP Integration:","body":"Built-in MCP server for AI coding agents, letting you dictate directly into development workflows."}},"pricing":{"title":"The Real Cost Comparison","ow":{"freeLabel":"Forever Free","free1":"All local features included","free2":"Local Whisper and Parakeet models","free3":"No usage limitations","free4":"macOS, Windows, and Linux","proLabel":"Pro","pro1":"Unlimited cloud transcription","pro2":"20h meeting recordings/month","pro3":"Mobile app","pro4":"Email support"},"spokenly":{"freeLabel":"Local Models","free1":"Whisper and Parakeet, unlimited","free2":"No sign-up required","free3":"macOS and iOS only","proLabel":"Pro","pro1":"Managed cloud models included","pro2":"Priority support","pro3":"Covers macOS and iPhone"},"callout":{"lead":"Both offer free local models:","body":"Spokenly and OpenWhispr both let you run Whisper and Parakeet locally at no cost. The key difference is platform support and open-source availability."}},"featureShowdown":{"title":"Feature Showdown","bothHaveTitle":"Features Both Have:","bothHave":["Free local Whisper and Parakeet models","Complete offline functionality","Hotkey activation","Works in any application","100+ language support","Real-time transcription"],"openWhisprExclusivesTitle":"OpenWhispr Exclusives:","openWhisprExclusives":["Windows and Linux support","Fully open source and forkable","Meeting recording and transcription","Community-driven development"],"spokenlyExclusivesTitle":"Spokenly Exclusives:","spokenlyExclusives":["Agent Mode for macOS automation","BYOK support for multiple cloud providers","Built-in MCP server for AI coding agents","Smart prompts for grammar and formatting"]},"evidence":{"title":"Evidence-Based Notes","body":"This comparison is based on Spokenly's official website, App Store listing, and public documentation. Spokenly reports 100,000+ active users and a 4.9/5 App Store rating. Pricing and features are accurate as of March 2026."},"migration":{"title":"Switching from Spokenly to OpenWhispr","intro":"Already using Spokenly on Mac? Here is how to add OpenWhispr to your workflow:","step1":{"title":"Download OpenWhispr","body":"Get the free version from GitHub or go Pro for cloud features."},"step2":{"title":"Match Your Settings","body":"Choose the same Whisper or Parakeet model you use in Spokenly and set your preferred hotkey."},"step3":{"title":"Expand to Other Platforms","body":"Take your voice dictation workflow to Windows or Linux machines too."},"callout":{"lead":"Missing Agent Mode?","body":"OpenWhispr focuses on dictation accuracy and cross-platform reach. If you rely heavily on voice-driven Mac automation, you can run both tools side by side."}},"bottomLine":{"title":"The Bottom Line","intro":"Spokenly and OpenWhispr share a lot of common ground — both offer free local Whisper and Parakeet models, strong privacy, and system-wide dictation. The choice comes down to platform needs and philosophy:","teamSpokenly":{"title":"Team Spokenly","body":"You are all-in on Apple, want Agent Mode for Mac automation, and appreciate a polished native experience with broad cloud model support via BYOK."},"teamOpenWhispr":{"title":"Team OpenWhispr","body":"You need Windows or Linux, believe in open source, want meeting recordings, or prefer the freedom to customize your tools completely."},"closing":"Both are excellent dictation tools. Pick the one that fits your platforms and workflow."},"finalCta":{"title":"Ready to Try Open Source Dictation?","body":"Get the same great local models as Spokenly, plus cross-platform support and the freedom of open source.","starButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever","viewAllComparisons":"View all comparisons"}},"compareSuperwhisper":{"metadata":{"title":"SuperWhisper Alternative: Open-Source & Free","description":"SuperWhisper alternative that's free, open-source, and runs locally on Mac, Windows, Linux and iOS. No subscriptions — just powerful on-device dictation.","keywords":["superwhisper alternative","superwhisper vs openwhispr","mac dictation app","whisper dictation","nvidia parakeet"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","superwhisper":"vs SuperWhisper"},"hero":{"title":"OpenWhispr vs SuperWhisper","subtitle":"SuperWhisper now supports macOS, Windows, and iOS. Here is a factual comparison of the open-source option vs the polished proprietary experience.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","superWhisperIntro":"is a polished dictation product that started Mac-first and now ships on macOS, Windows, and iOS. It offers free monthly credits plus paid plans for heavier usage and additional cloud features.","openWhisprIntro":"offers a different value proposition. It's completely free, open source, and works across all platforms - not just Mac. You get powerful OpenAI Whisper and NVIDIA Parakeet models, with the freedom to modify, extend, and truly own your tools. Plus, it's backed by a community that values privacy and transparency.","chooseSuperWhisper":"Choose SuperWhisper if: You want a polished proprietary app with built-in cross-device sync.","chooseOpenWhispr":"Choose OpenWhispr if: You value open source, need cross-platform support, or want to customize."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"100% local option, bring your own API keys"},"superWhisper":{"status":"full","details":"Local processing, optional cloud features"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"superWhisper":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"superWhisper":{"status":"partial","details":"macOS, Windows, iOS (no Linux)"}},{"feature":"Pricing Model","openWhispr":{"status":"full","details":"Free forever, optional $8/mo pro"},"superWhisper":{"status":"partial","details":"Free plan plus paid credits/subscriptions (see official pricing)"}},{"feature":"AI Model Options","openWhispr":{"status":"full","details":"OpenAI Whisper and NVIDIA Parakeet models"},"superWhisper":{"status":"full","details":"Multiple Whisper models available"}},{"feature":"System Integration","openWhispr":{"status":"full","details":"Works in any application"},"superWhisper":{"status":"full","details":"Deep macOS integration"}},{"feature":"Customization","openWhispr":{"status":"full","details":"Fully customizable, fork-friendly"},"superWhisper":{"status":"partial","details":"Settings and preferences only"}},{"feature":"Voice Commands","openWhispr":{"status":"partial","details":"Basic punctuation commands"},"superWhisper":{"status":"full","details":"Advanced commands and shortcuts"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","openSource":{"title":"Open Source Freedom:","body":"Inspect the code, modify features, contribute improvements. No black boxes."},"crossPlatform":{"title":"Cross-Platform:","body":"Works on Mac, Windows, Linux and iOS. Take your workflow anywhere."},"foreverFree":{"title":"Forever Free:","body":"No purchase necessary. Pro version is optional and only $8/month."},"hackable":{"title":"Hackable:","body":"Add features, create plugins, integrate with your workflow however you want."}},"whySuperWhisper":{"title":"Why SuperWhisper?","nativeFeel":{"title":"Native macOS Feel:","body":"Designed specifically for Mac with beautiful UI and smooth animations."},"polish":{"title":"Polish & Design:","body":"Every detail crafted for the best user experience. It's genuinely delightful to use."},"advancedFeatures":{"title":"Advanced Features:","body":"Voice commands, custom modes, app-specific settings out of the box."},"support":{"title":"Professional Support:","body":"Dedicated support team and regular updates from a focused development team."}},"pricing":{"title":"The Real Cost Comparison","ow":{"freeLabel":"Forever Free","free1":"All features included","free2":"Local processing","free3":"No limitations","free4":"Community support","proLabel":"Pro","pro1":"Unlimited cloud transcription","pro2":"20h meeting recordings/month","pro3":"Mobile app","pro4":"Email support"},"super":{"freeAmount":"Free","freeLabel":"Free monthly credits","free1":"All pro features trial","free2":"Small AI models","free3":"Basic functionality","paidAmount":"Paid plans","paidLabel":"Paid plans","paid1":"Unlimited use","paid2":"Large AI models","paid3":"All features","paid4":"macOS + Windows + iOS"},"callout":{"lead":"Money isn't everything:","body":"While SuperWhisper offers great value with its pricing tiers, OpenWhispr gives you something money can't buy - complete control and transparency."}},"featureShowdown":{"title":"Feature Showdown","bothHaveTitle":"Features Both Have:","bothHave":["Multiple AI model options (Whisper and Parakeet)","Local processing for privacy","Hotkey activation","Works in any application","High accuracy transcription","Multiple language support"],"openWhisprExclusivesTitle":"OpenWhispr Exclusives:","openWhisprExclusives":["Cross-platform (Mac, Windows, Linux, iOS)","Open source and forkable","Free forever option","Community-driven development"],"superWhisperExclusivesTitle":"SuperWhisper Exclusives:","superWhisperExclusives":["Native macOS UI/UX","Advanced voice commands","App-specific modes","Menubar integration"]},"evidence":{"title":"Evidence-Based Notes","body":"This comparison is based on vendor documentation, pricing pages, and public product docs rather than anecdotal testimonials."},"migration":{"title":"Switching from SuperWhisper to OpenWhispr","intro":"Love SuperWhisper but curious about open source? Here's how to make the switch:","step1":{"title":"Download OpenWhispr","body":"Get the free version from GitHub or go Pro for convenience."},"step2":{"title":"Match Your Settings","body":"Set your preferred Whisper model and hotkey to match your SuperWhisper setup."},"step3":{"title":"Enjoy the Freedom","body":"Keep using it on Mac, or take your workflow to Windows or Linux machines too."},"callout":{"lead":"Missing voice commands?","body":"The OpenWhispr community is actively working on advanced features. Join us on GitHub to help shape the future!"}},"bottomLine":{"title":"The Bottom Line","intro":"Both OpenWhispr and SuperWhisper are excellent tools that respect your privacy and deliver powerful AI transcription. The choice comes down to your values and needs:","teamSuperWhisper":{"title":"Team SuperWhisper","body":"You're a Mac user who values polish, native integration, and having everything work perfectly out of the box. $29 for a tool you'll use daily is a bargain."},"teamOpenWhispr":{"title":"Team OpenWhispr","body":"You believe in open source, want cross-platform support, or love the idea of customizing your tools. Free forever with the option to support development."},"closing":"The best tool is the one that fits your workflow and philosophy. Both are winners in their own right."},"finalCta":{"title":"Ready to Join the Open Source Movement?","body":"Experience the power of community-driven development. Get the same great features as SuperWhisper, plus the freedom to make it your own.","starButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever","viewAllComparisons":"View all comparisons"}},"compareTalon":{"metadata":{"title":"OpenWhispr vs Talon Voice: Simple AI Dictation","description":"OpenWhispr is the simple Talon Voice alternative for dictation. AI cleanup, instant setup, fully open source and free. No scripting needed.","keywords":["Talon Voice alternative","Talon vs OpenWhispr dictation","simple voice dictation","voice typing without Talon","Talon Voice alternative for dictation","easy voice to text","voice dictation no scripting","OpenWhispr vs Talon","nvidia parakeet"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","talon":"vs Talon Voice"},"hero":{"title":"OpenWhispr vs Talon Voice","subtitle":"Two different tools for two different needs. Simple AI dictation vs full hands-free computer control.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","talonIntro":"is a powerful hands-free input system built for people who need or want to control their entire computer by voice. It supports eye tracking, noise recognition, and a Python scripting layer that lets you customize every aspect of voice interaction. It excels at hands-free coding and is invaluable for accessibility. But it requires significant setup — installing community script files, learning a phonetic alphabet, and writing custom commands.","openWhisprIntro":"is focused on one thing: turning speech into text as accurately and effortlessly as possible. Download it, press a hotkey, speak naturally, and your words appear — cleaned up and formatted by AI. No scripting, no phonetic alphabet, no learning curve. It's built for anyone who wants fast, private, AI-powered dictation.","chooseTalon":"Choose Talon if: You need full hands-free computer control, voice coding, or have accessibility needs that go beyond dictation.","chooseOpenWhispr":"Choose OpenWhispr if: You want simple, accurate voice dictation with AI cleanup and zero setup."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Primary Purpose","openWhispr":{"status":"full","details":"Voice dictation and text input"},"talon":{"status":"full","details":"Full hands-free computer control"}},{"feature":"Setup Complexity","openWhispr":{"status":"full","details":"Download, install, start dictating"},"talon":{"status":"partial","details":"Requires scripting and community files"}},{"feature":"Learning Curve","openWhispr":{"status":"full","details":"Minimal — speak naturally"},"talon":{"status":"partial","details":"Steep — phonetic alphabet, commands, scripting"}},{"feature":"AI Text Cleanup","openWhispr":{"status":"full","details":"AI formatting via system prompts"},"talon":{"status":"none","details":"Raw transcription only"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"talon":{"status":"partial","details":"Community scripts open, core is proprietary"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"talon":{"status":"full","details":"macOS, Windows, Linux (X11)"}},{"feature":"General Dictation","openWhispr":{"status":"full","details":"Optimized for natural speech dictation"},"talon":{"status":"partial","details":"Possible but not the primary focus"}},{"feature":"Hands-Free Coding","openWhispr":{"status":"partial","details":"Dictation in any text field"},"talon":{"status":"full","details":"Purpose-built for voice coding"}},{"feature":"Eye Tracking","openWhispr":{"status":"none","details":"Not supported"},"talon":{"status":"full","details":"Tobii eye tracker integration"}},{"feature":"Pricing","openWhispr":{"status":"full","details":"Free forever, optional $8/mo Pro"},"talon":{"status":"full","details":"Free core app, with optional paid Patreon tiers for early features and support"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","instantSetup":{"title":"Instant Setup:","body":"Download, install, press your hotkey, and start dictating. No scripting, no configuration files, no phonetic alphabet to memorize."},"aiCleanup":{"title":"AI Text Cleanup:","body":"System prompts automatically format, punctuate, and clean up your speech. Talon gives you raw transcription."},"naturalSpeech":{"title":"Natural Speech:","body":"Speak naturally and get polished text. No need to learn special commands like \"word cap hello\" or \"phrase hello world\"."},"freeForever":{"title":"Free Forever:","body":"Fully functional at no cost. Pro is optional at $8/mo. Talon's beta features require $25/mo on Patreon."}},"whyTalon":{"title":"Why Talon Voice?","fullControl":{"title":"Full Computer Control:","body":"Navigate your entire OS by voice — window management, app switching, file browsing, and more. OpenWhispr focuses only on text input."},"eyeTracking":{"title":"Eye Tracking:","body":"Integrates with Tobii eye trackers for mouse control. Essential for users who cannot use a mouse at all."},"voiceCoding":{"title":"Voice Coding:","body":"Purpose-built commands for writing code — language-specific snippets, IDE integration, and terminal control via voice."},"accessibility":{"title":"Accessibility:","body":"Genuinely life-changing for people with RSI, motor disabilities, or other conditions that limit hand use. Talon serves a critical need."}},"pricing":{"title":"The Real Cost Comparison","ow":{"freeLabel":"Forever Free","free1":"All features included","free2":"Local Whisper and Parakeet processing","free3":"AI text cleanup","free4":"99+ languages","proLabel":"Pro","pro1":"Unlimited cloud transcription","pro2":"20h meeting recordings/month","pro3":"Mobile app","pro4":"Email support"},"talon":{"freeAmount":"Free","freeLabel":"Public Release","free1":"Core voice control","free2":"Conformer speech engine","free3":"Community scripts","betaLabel":"Beta (Patreon)","beta1":"Early access to new features","beta2":"Priority support","beta3":"Beta channel access","beta4":"VIP Slack channel"},"callout":{"lead":"Different tools, different costs:","body":"Talon's Patreon supports one developer working full-time on accessibility software. OpenWhispr's free tier gives you full-featured dictation at zero cost. Both are fair models."}},"featureShowdown":{"title":"Feature Showdown","bothHaveTitle":"Features Both Have:","bothHave":["Cross-platform (macOS, Windows, Linux)","Local speech processing","Privacy-first design","Customizable via open community","Works offline","Active community support"],"openWhisprExclusivesTitle":"OpenWhispr Exclusives:","openWhisprExclusives":["AI text cleanup and formatting","Instant setup — no scripting needed","Natural speech dictation (speak normally)","BYOK cloud API support","Fully open source (core included)"],"talonExclusivesTitle":"Talon Voice Exclusives:","talonExclusives":["Eye tracking mouse control","Full OS control by voice","Voice coding with language-specific commands","Noise recognition (pop/hiss clicks)","Python scripting for custom commands"]},"evidence":{"title":"Evidence-Based Notes","body":"This comparison is based on vendor documentation, pricing pages, and public product docs rather than anecdotal testimonials."},"migration":{"title":"Coming from Talon? Here's What to Expect","intro":"If you use Talon for dictation and want something simpler, or want to add a dedicated dictation tool alongside Talon:","step1":{"title":"Download OpenWhispr","body":"Download for free. Optional Pro ($8/user/mo) for unlimited cloud transcription, more meeting hours, and the mobile app."},"step2":{"title":"Set Your Hotkey","body":"Pick a push-to-talk key that doesn't conflict with your Talon commands. Both can run side by side."},"step3":{"title":"Dictate Naturally","body":"No phonetic alphabet, no \"phrase\" prefix. Just speak and let the AI handle formatting and punctuation."},"callout":{"lead":"They work great together:","body":"Use Talon for hands-free coding and OS control. Use OpenWhispr when you want to dictate long-form text, emails, or notes with AI cleanup."}},"bottomLine":{"title":"The Bottom Line","intro":"Talon Voice and OpenWhispr solve fundamentally different problems. Talon is a comprehensive hands-free input system — it replaces your keyboard and mouse entirely. OpenWhispr is a focused dictation tool — it turns your speech into polished text. Comparing them directly is like comparing a Swiss Army knife to a chef's knife: both cut, but they're designed for different jobs.","teamTalon":{"title":"Team Talon","body":"You need hands-free computer control, voice coding, or eye tracking. You're willing to invest time learning the system because the payoff — complete voice-driven computing — is worth it. Talon is genuinely life-changing for accessibility."},"teamOpenWhispr":{"title":"Team OpenWhispr","body":"You want accurate, AI-enhanced dictation that works instantly. You speak, it writes — clean, formatted, and ready to use. No scripting, no learning curve. Free, open source, and private."},"closing":"Many users run both — Talon for hands-free coding and computer control, OpenWhispr for fast natural dictation. They complement each other well."},"finalCta":{"title":"Ready for Effortless Dictation?","body":"Skip the setup and start dictating in minutes. AI-powered, private, open source, and free.","starButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever"},"viewAllComparisons":"View all comparisons"},"compareTypeless":{"metadata":{"title":"Typeless Alternative — OpenWhispr vs Typeless","description":"Typeless alternative with true offline processing. OpenWhispr is open-source, runs locally with no word limits, and supports Linux. Try it free.","keywords":["typeless alternative","typeless vs openwhispr","ai voice dictation","voice typing app","dictation software comparison","smart voice dictation"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","typeless":"vs Typeless"},"hero":{"title":"OpenWhispr vs Typeless","subtitle":"Typeless is an AI voice keyboard with smart editing across desktop and mobile. Here is how it compares to the open-source, privacy-first alternative.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","typelessIntro":"is an AI-powered voice keyboard that transforms natural speech into polished text. It automatically removes filler words, handles self-corrections, and adapts tone based on which app you are using. It works on macOS, Windows, iOS, and Android with a free tier of 4,000 words per week.","openWhisprIntro":"takes a different approach. It is fully open source, runs entirely offline with local Whisper and Parakeet models, and gives you complete control over your data. It supports Linux in addition to macOS and Windows, with no weekly word limits on local processing.","chooseTypeless":"Choose Typeless if: You want AI-powered smart editing, mobile voice typing, and automatic filler word removal.","chooseOpenWhispr":"Choose OpenWhispr if: You need offline support, Linux compatibility, open source, or unlimited local dictation."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"100% local option, bring your own API keys"},"typeless":{"status":"partial","details":"Zero cloud data retention, but processing is not fully local"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"typeless":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"typeless":{"status":"full","details":"macOS, Windows, iOS, Android (no Linux)"}},{"feature":"Pricing Model","openWhispr":{"status":"full","details":"Free forever, optional $8/mo pro"},"typeless":{"status":"partial","details":"Free 4,000 words/week, Pro $12/mo (annual) or $30/mo"}},{"feature":"Smart Editing","openWhispr":{"status":"partial","details":"AI post-processing with cloud models"},"typeless":{"status":"full","details":"Auto filler removal, self-correction, tone adaptation"}},{"feature":"Offline Support","openWhispr":{"status":"full","details":"Complete offline functionality"},"typeless":{"status":"none","details":"Requires internet for transcription"}},{"feature":"Customization","openWhispr":{"status":"full","details":"Fully customizable, fork-friendly"},"typeless":{"status":"partial","details":"Personal dictionary, tone preferences per app"}},{"feature":"Mobile Support","openWhispr":{"status":"partial","details":"iOS and Android with Pro plan"},"typeless":{"status":"full","details":"Native iOS and Android apps included free"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","privacy":{"title":"True Offline Privacy:","body":"Process everything locally. Your voice never leaves your device. Typeless requires an internet connection."},"openSource":{"title":"Open Source Freedom:","body":"Inspect, modify, and extend the code. No vendor lock-in, ever."},"noLimits":{"title":"No Word Limits:","body":"Unlimited local dictation on the free plan. Typeless caps free usage at 4,000 words per week."},"linux":{"title":"Linux Support:","body":"The only option here that runs on Linux. Typeless covers Mac, Windows, and mobile only."}},"whyTypeless":{"title":"Why Typeless?","smartEditing":{"title":"Smart Editing:","body":"Automatically removes filler words, handles self-corrections, and formats lists and steps without manual cleanup."},"tone":{"title":"Context-Aware Tone:","body":"Adapts writing style based on the active app — professional in email, casual in messaging."},"mobile":{"title":"Native Mobile Apps:","body":"Full iOS and Android keyboard apps included free, replacing your default voice typing."},"voiceEditing":{"title":"Voice Editing Commands:","body":"Select text and speak to make it shorter, change tone, generate replies, or translate on the fly."}},"pricing":{"title":"The Real Cost Comparison","ow":{"freeLabel":"Forever Free","free1":"Unlimited local dictation","free2":"Local Whisper and Parakeet models","free3":"Full offline support","free4":"macOS, Windows, and Linux","proLabel":"Pro","pro1":"Unlimited cloud transcription","pro2":"20h meeting recordings/month","pro3":"Mobile app","pro4":"Email support"},"typeless":{"freeLabel":"4,000 words/week","free1":"AI auto-editing included","free2":"100+ language support","free3":"Personal dictionary","proLabel":"Pro (billed annually)","pro1":"Unlimited words per week","pro2":"Priority feature requests","pro3":"Early access to new features","pro4":"$$30/mo if billed monthly"},"callout":{"lead":"The key trade-off:","body":"Typeless offers polished AI editing but requires internet and caps free usage. OpenWhispr gives unlimited offline dictation with full source code access."}},"bestFor":{"title":"Best For Different Users","openWhisprTitle":"OpenWhispr is Best For:","openWhisprItems":["Privacy-conscious users handling sensitive data offline","Linux users with no other smart dictation option","Heavy users who need unlimited dictation without weekly caps","Developers who want to inspect and extend their tools","Teams needing meeting recording and transcription"],"typelessTitle":"Typeless is Best For:","typelessItems":["Writers who want auto-cleaned, polished output","Mobile-first users on iOS and Android","Professionals who switch between formal and casual contexts","Multilingual users who mix languages mid-sentence","Users who prefer a managed service with no setup"]},"evidence":{"title":"Evidence-Based Notes","body":"This comparison is based on Typeless's official website, pricing page, App Store and Play Store listings, and public product documentation. Pricing and features are accurate as of March 2026."},"migration":{"title":"Switching from Typeless to OpenWhispr","intro":"Using Typeless and want to try the open-source approach? Here is how:","step1":{"title":"Download OpenWhispr","body":"Get the free version from GitHub or go Pro for cloud and mobile features."},"step2":{"title":"Choose Your Model","body":"Pick a local Whisper or Parakeet model. No internet required, no word limits."},"step3":{"title":"Dictate Without Limits","body":"No weekly word caps, no subscription needed for core features."},"callout":{"lead":"Missing smart editing?","body":"OpenWhispr supports AI post-processing with cloud models on Pro. The output is clean and punctuated, though Typeless's filler removal and tone adaptation are more advanced."}},"bottomLine":{"title":"The Bottom Line","intro":"Typeless and OpenWhispr solve the same problem — turning speech into text — but with fundamentally different approaches. Typeless bets on AI-powered smart editing and broad mobile reach. OpenWhispr bets on privacy, openness, and unlimited offline access.","teamTypeless":{"title":"Team Typeless","body":"You want polished, ready-to-send text from your voice. Filler removal, tone matching, and native mobile keyboard support make dictation feel effortless."},"teamOpenWhispr":{"title":"Team OpenWhispr","body":"You want full control. Offline processing, no word limits, Linux support, and the ability to inspect every line of code that handles your voice data."},"closing":"Different philosophies, both effective. Choose the one that matches how you work."},"finalCta":{"title":"Ready to Own Your Voice Data?","body":"Unlimited offline dictation, no weekly word caps, and the transparency of open source. Try OpenWhispr free.","starButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever","viewAllComparisons":"View all comparisons"}},"compareVoiceink":{"metadata":{"title":"VoiceInk Alternative — Free, Cross-Platform Dictation","description":"Compare two open-source dictation apps. OpenWhispr is free forever and runs on Mac, Windows, Linux and iOS — unlike VoiceInk's macOS-only approach.","keywords":["OpenWhispr vs VoiceInk","VoiceInk alternative","voice dictation comparison","open source dictation","whisper voice to text","nvidia parakeet","macOS dictation app","cross-platform voice dictation","AI transcription comparison","local speech to text","privacy voice dictation","best dictation app 2025","VoiceInk review"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","voiceink":"vs VoiceInk"},"hero":{"title":"OpenWhispr vs VoiceInk","subtitle":"Two open-source, privacy-first dictation apps built on Whisper. Let's compare the cross-platform contender with the native Mac powerhouse.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","voiceInkIntro":"is a native macOS dictation app built in Swift with impressive context-aware AI features. It uses whisper.cpp for local transcription, offers Power Mode for per-app configurations, and includes a personal dictionary for custom terminology. It's open source under GPLv3 with a strong community of 3,700+ GitHub stars. At $39.99 as a paid license, and it can be strong value for Mac users who want a polished, offline-first voice typing experience.","openWhisprIntro":"takes a different approach. It's completely free, works across macOS, Windows, Linux and iOS, and offers both local and cloud transcription options. With Agent mode for AI post-processing and custom system prompts, it gives you flexible control over how your voice becomes text. The optional Pro tier ($8/user/mo) adds unlimited cloud transcription, more meeting hours, and the mobile app, but the core app is fully functional at no cost.","chooseVoiceInk":"Choose VoiceInk if: You're on a Mac with Apple Silicon and want screen-aware AI modes, a personal dictionary, and per-app settings out of the box.","chooseOpenWhispr":"Choose OpenWhispr if: You need cross-platform support, want a free-forever option, or prefer BYOK cloud flexibility alongside local processing."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"100% local option, bring your own API keys"},"voiceInk":{"status":"full","details":"100% offline processing, no cloud required"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"voiceInk":{"status":"full","details":"Open source under GPLv3 license"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"voiceInk":{"status":"partial","details":"macOS only (Apple Silicon required)"}},{"feature":"Pricing Model","openWhispr":{"status":"full","details":"Free forever, optional $8/mo pro"},"voiceInk":{"status":"partial","details":"Paid license (check official pricing)"}},{"feature":"AI Model Options","openWhispr":{"status":"full","details":"OpenAI Whisper and NVIDIA Parakeet models"},"voiceInk":{"status":"full","details":"Whisper + Parakeet models via whisper.cpp"}},{"feature":"System Integration","openWhispr":{"status":"full","details":"Works in any application"},"voiceInk":{"status":"full","details":"System-wide with Power Mode per-app settings"}},{"feature":"AI Post-Processing","openWhispr":{"status":"full","details":"Agent mode with custom system prompts"},"voiceInk":{"status":"full","details":"Smart Modes with screen-aware context"}},{"feature":"Customization","openWhispr":{"status":"full","details":"Fully customizable, fork-friendly"},"voiceInk":{"status":"full","details":"Personal dictionary, text replacements, Power Mode"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","crossPlatform":{"title":"True Cross-Platform:","body":"Works on Mac, Windows, Linux and iOS. VoiceInk requires Apple Silicon Macs only — no Intel Macs, no Windows, no Linux."},"freeCore":{"title":"Free Forever Core:","body":"All essential features at $0. No $39.99 upfront cost — get started immediately and only pay if you want Pro convenience."},"flexibility":{"title":"Cloud + Local Flexibility:","body":"Choose local processing for privacy or cloud transcription for speed. Bring your own API keys for full control."},"hackable":{"title":"Hackable & Extensible:","body":"Built with Electron and React — a familiar stack for web developers. Fork it, extend it, make it yours."}},"whyVoiceInk":{"title":"Why VoiceInk?","native":{"title":"Native macOS Performance:","body":"Built in Swift with Apple Silicon optimization. Leverages the Neural Engine for fast, efficient on-device transcription."},"smartModes":{"title":"Screen-Aware Smart Modes:","body":"Context-aware AI that reads your screen content to deliver more relevant transcriptions and formatting."},"powerMode":{"title":"Power Mode:","body":"Automatically applies different settings per app and URL. Your dictation adapts to Slack, email, code editors, and more."},"dictionary":{"title":"Personal Dictionary:","body":"Train the system on your custom terminology, names, acronyms, and jargon. Plus text replacements for shortcuts."}},"pricing":{"title":"The Real Cost Comparison","ow":{"freeLabel":"Forever Free","free1":"All core features included","free2":"Local + cloud transcription","free3":"Agent mode with system prompts","free4":"macOS, Windows, Linux","proLabel":"Pro","pro1":"Unlimited cloud transcription","pro2":"20h meeting recordings/month","pro3":"Mobile app","pro4":"Email support"},"voiceInk":{"freeAmount":"Free","freeLabel":"Trial","free1":"Try before you buy","free2":"Build from source (GPLv3)","free3":"Core transcription features","paidLabel":"One-Time Purchase","paid1":"Lifetime access to all features","paid2":"Future updates included","paid3":"Priority support","paid4":"Student discounts available"},"callout":{"lead":"Two honest models:","body":"VoiceInk's $39.99 paid plans can be strong value if you're committed to macOS. OpenWhispr is free forever with an optional Pro subscription — ideal if you want to try before committing or need cross-platform support."}},"featureShowdown":{"title":"Feature Showdown","bothShareTitle":"Features Both Share:","bothShare":["Whisper-based local transcription","100% offline privacy option","Global hotkey activation","Works system-wide in any app","Open source code available","AI-enhanced text formatting","Multiple language support","Active community development"],"openWhisprExclusivesTitle":"OpenWhispr Exclusives:","openWhisprExclusives":["Cross-platform (Mac, Windows, Linux, iOS)","Works on Intel and Apple Silicon Macs","Cloud transcription with BYOK API keys","Agent mode with custom system prompts","Free forever with no feature gating"],"voiceInkExclusivesTitle":"VoiceInk Exclusives:","voiceInkExclusives":["Native Swift app with Apple Silicon optimization","Screen-aware Smart Modes for context","Power Mode with per-app/URL settings","Personal dictionary and text replacements","Parakeet model support alongside Whisper"]},"evidence":{"title":"Evidence-Based Notes","body":"This comparison is based on vendor documentation, pricing pages, and public product docs rather than anecdotal testimonials."},"migration":{"title":"Switching from VoiceInk to OpenWhispr","intro":"Love VoiceInk but need cross-platform support or want to avoid the upfront cost? Here's how to make the switch:","step1":{"title":"Download OpenWhispr","body":"Download for free. Optional Pro ($8/user/mo) for unlimited cloud transcription, more meeting hours, and the mobile app. Works on any platform."},"step2":{"title":"Set Up Your Hotkey & Model","body":"Match your VoiceInk keyboard shortcut and choose a Whisper model size. Start with \"small\" for a good balance of speed and accuracy."},"step3":{"title":"Configure Agent Mode","body":"Recreate your VoiceInk Smart Mode behavior using OpenWhispr's Agent mode with custom system prompts for formatting and context."},"step4":{"title":"Go Multi-Platform","body":"Install on your Windows or Linux machines too. Same workflow, same quality — everywhere you work."},"callout":{"lead":"Missing Power Mode?","body":"OpenWhispr's Agent mode with custom system prompts gives you similar AI post-processing flexibility. The community is also building more context-aware features — join us on GitHub to help shape the roadmap!"}},"bottomLine":{"title":"The Bottom Line","intro":"This is a rare comparison where both tools are open source and privacy-focused. The choice really comes down to your platform and workflow preferences:","teamVoiceInk":{"title":"Team VoiceInk","body":"You're all-in on macOS with Apple Silicon, you want native Swift performance, and features like Power Mode and screen-aware Smart Modes fit perfectly into your workflow. the paid license can be worthwhile for what you get."},"teamOpenWhispr":{"title":"Team OpenWhispr","body":"You work across platforms, want a free dictation tool with no upfront cost, or need the flexibility of cloud transcription with your own API keys. Agent mode gives you powerful AI post-processing at zero cost."},"closing":"Both apps champion open source and privacy. The best choice depends on whether you live in the Apple ecosystem or need the freedom to work everywhere."},"finalCta":{"title":"Ready to Try the Cross-Platform Alternative?","body":"Get the same powerful Whisper and Parakeet transcription as VoiceInk — on every platform, completely free. No Apple Silicon required, no upfront purchase, no compromises on privacy.","starButton":"Star us on GitHub","footnote":"No credit card required • Works offline • Open source forever"},"viewAllComparisons":"← View all dictation app comparisons"},"compareWillowvoice":{"metadata":{"title":"Willow Voice Alternative: Open-Source, Local & Free","description":"Compare OpenWhispr and Willow Voice side by side. Open-source local dictation vs proprietary cloud-based voice writing across desktop and mobile.","keywords":["willow voice alternative","willow voice vs openwhispr","cloud dictation app","voice writing software","nvidia parakeet"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","willowVoice":"vs Willow Voice"},"hero":{"title":"OpenWhispr vs Willow Voice","subtitle":"Comparing two powerful voice-to-text solutions. One open and free, one proprietary. Let's help you choose.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","willowVoiceIntro":"is a proprietary dictation tool focused on fast cloud-based voice writing, command workflows, and rewrite/tone presets. It supports desktop and mobile use cases with account-based syncing and managed updates.","openWhisprIntro":"takes a different approach. It's completely open source, works across all your applications (not just browsers), and gives you full control over your data. You can run it entirely offline or use your own API keys. Plus, it's free forever if you're comfortable with open source.","chooseWillowVoice":"Choose Willow Voice if: You want a managed cloud workflow with built-in rewrite/tone tools and account syncing.","chooseOpenWhispr":"Choose OpenWhispr if: You value privacy, want complete control, or prefer open source software."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"100% local processing option, self-hosted API keys"},"willowVoice":{"status":"partial","details":"Cloud-first processing with account-level controls"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"willowVoice":{"status":"none","details":"Proprietary software"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"Works in any app system-wide"},"willowVoice":{"status":"full","details":"macOS, Windows, and iPhone app support"}},{"feature":"Pricing Model","openWhispr":{"status":"full","details":"Free forever with paid pro options"},"willowVoice":{"status":"partial","details":"Free tier includes 10,000 words with 2,000 words/week refill; paid plans from $15/mo"}},{"feature":"Voice Commands","openWhispr":{"status":"partial","details":"Basic commands for punctuation"},"willowVoice":{"status":"full","details":"Advanced dictation commands, presets, and editing flows"}},{"feature":"Accuracy","openWhispr":{"status":"full","details":"State-of-the-art Whisper and Parakeet models"},"willowVoice":{"status":"full","details":"Cloud AI dictation with rewrite/tone options"}},{"feature":"Customization","openWhispr":{"status":"full","details":"Fully customizable, fork-friendly"},"willowVoice":{"status":"partial","details":"Configurable presets and commands (not open-source extensible)"}},{"feature":"Offline Support","openWhispr":{"status":"full","details":"Complete offline functionality"},"willowVoice":{"status":"partial","details":"Internet required for core dictation pipeline"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","privacy":{"title":"True Privacy:","body":"Process everything locally. Your voice never leaves your device unless you choose to use cloud APIs."},"crossPlatform":{"title":"True Cross-Platform:","body":"Available on macOS, Windows, Linux and iOS. Willow Voice supports macOS and Windows, plus an iPhone app."},"openSource":{"title":"Open Source Freedom:","body":"Modify it, extend it, or contribute back. No vendor lock-in, ever."},"cost":{"title":"Cost Effective:","body":"Free forever with local processing, or use your own API keys for cloud features."}},"whyWillowVoice":{"title":"Why Willow Voice?","voiceCommands":{"title":"Voice Commands:","body":"Built-in command workflows for editing and formatting dictated text quickly."},"autoEditing":{"title":"AI Auto-Editing:","body":"Rewrite, tone, and writing presets designed for cloud AI post-processing."},"managedService":{"title":"Managed Service:","body":"Managed cloud product with account sync and automatic updates."},"commandMode":{"title":"Command Mode:","body":"Preset-driven dictation and editing flows for common writing tasks."}},"pricing":{"title":"Pricing Breakdown","ow":{"freeLabel":"Forever Free","freeDetails":"Local processing, unlimited use","proLabel":"Pro","proDetails":"Unlimited cloud, mobile app, support"},"willow":{"freeLabel":"Free","freeWords":"10,000 words + 2,000/week refill","freeDetails":"Starter usage with account required","proLabel":"Pro / Team plans","proDetails":"More words, advanced features, and team options"}},"useCases":{"title":"Best For Different Users","openWhisprTitle":"OpenWhispr is Best For:","willowVoiceTitle":"Willow Voice is Best For:","openWhisprList":["Developers who want to customize their tools","Privacy-conscious users handling sensitive data","Teams needing system-wide dictation support","Budget-conscious individuals and startups","Users who work across multiple applications"],"willowVoiceList":["Users who prefer managed cloud dictation workflows","Writers using rewrite/tone presets","Teams that want account sync and centralized management","People who do not want self-hosted or open-source setup","Users comfortable with cloud processing requirements"]},"migration":{"title":"Switching from Willow Voice to OpenWhispr","intro":"Making the switch is easier than you think. Here's how:","step1":{"title":"Download OpenWhispr","body":"Get the free version from GitHub or go Pro for instant setup."},"step2":{"title":"Configure Your Preferences","body":"Set your hotkey, choose local or cloud processing, and select your model."},"callout":{"lead":"Pro tip:","body":"You can use both tools side-by-side during your transition. OpenWhispr won't interfere with Willow Voice's desktop/mobile workflow."}},"finalCta":{"title":"Ready to Take Control of Your Voice Data?","body":"Join thousands who've switched to OpenWhispr for true privacy, system-wide support, and the freedom of open source.","starButton":"Explore on GitHub","footnote":"No credit card required • Works offline • Open source forever","viewAllComparisons":"View all comparisons"}},"compareWindowsSpeech":{"metadata":{"title":"Windows Voice Typing Alternative — OpenWhispr","description":"Switch from Windows Voice Typing to OpenWhispr: open-source AI dictation with Whisper models, 100% offline privacy, and free cross-platform use.","keywords":["Windows voice typing alternative","better Windows dictation","Windows speech recognition upgrade","Windows speech recognition alternative","Win+H alternative","Whisper voice typing Windows","nvidia parakeet","offline dictation Windows","AI dictation Windows"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","windowsSpeech":"vs Windows Speech"},"hero":{"title":"OpenWhispr vs Windows Voice Typing","subtitle":"Windows has built-in dictation, but is it enough? Compare the free built-in tool with state-of-the-art AI voice recognition.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","windowsIntro":"(Win+H) is a convenient built-in dictation tool that ships free with every copy of Windows 11. It handles basic dictation well, supports 45+ languages, and offers voice commands for editing. However, it relies on an internet connection for best results, uses Microsoft's proprietary speech engine, and offers minimal customization. The newer \"Fluid Dictation\" feature on Copilot+ PCs adds grammar correction, but only in English.","openWhisprIntro":"brings OpenAI Whisper and NVIDIA Parakeet models to your desktop -- the same technology that powers some of the most accurate speech recognition in the world. It runs 100% locally, works on Windows, Mac, and Linux, and gives you full control over model size, hotkeys, and AI-powered text cleanup via agent mode. It's free, open source, and never sends your voice data anywhere.","stickWith":"Stick with Windows Voice Typing if: You only need quick, occasional dictation and don't want to install anything.","chooseOpenWhispr":"Choose OpenWhispr if: You want significantly better accuracy, offline privacy, AI text cleanup, or work across multiple platforms."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Speech Recognition Engine","openWhispr":{"status":"full","details":"OpenAI Whisper and NVIDIA Parakeet models"},"windowsSpeech":{"status":"partial","details":"Microsoft proprietary engine"}},{"feature":"Offline Processing","openWhispr":{"status":"full","details":"100% local, no internet required"},"windowsSpeech":{"status":"partial","details":"Basic offline; best accuracy requires internet"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"windowsSpeech":{"status":"none","details":"Closed source, proprietary"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"macOS, Windows, Linux, iOS"},"windowsSpeech":{"status":"partial","details":"Windows only"}},{"feature":"AI Model Selection","openWhispr":{"status":"full","details":"Choose from tiny, base, small, medium, large"},"windowsSpeech":{"status":"none","details":"No model choice, single engine"}},{"feature":"Agent Mode / AI Cleanup","openWhispr":{"status":"full","details":"System prompts for formatting & cleanup"},"windowsSpeech":{"status":"none","details":"No AI post-processing"}},{"feature":"Language Support","openWhispr":{"status":"full","details":"99+ languages via Whisper and Parakeet"},"windowsSpeech":{"status":"partial","details":"45+ languages (see Microsoft language list)"}},{"feature":"Customization","openWhispr":{"status":"full","details":"Hotkeys, models, system prompts, BYOK"},"windowsSpeech":{"status":"partial","details":"Auto-punctuation, mic selection"}}],"whyOpenWhispr":{"title":"Why Upgrade to OpenWhispr?","accuracy":{"title":"Whisper and Parakeet Accuracy:","body":"OpenAI Whisper and NVIDIA Parakeet deliver state-of-the-art accuracy across accents, background noise, and technical vocabulary. Choose from tiny (fast) to large (maximum accuracy)."},"offline":{"title":"True Offline Privacy:","body":"Every word is processed on your machine. No cloud, no data collection, no internet required -- ever."},"agentMode":{"title":"Agent Mode Cleanup:","body":"AI-powered system prompts automatically clean up grammar, remove filler words, and format text -- beyond basic dictation."},"crossPlatform":{"title":"Cross-Platform:","body":"Works on Windows, macOS, and Linux. One tool, every OS. Take your workflow anywhere."}},"whyWindows":{"title":"What Windows Voice Typing Does Well","zeroSetup":{"title":"Zero Setup:","body":"Already installed on every Windows 11 PC. Press Win+H and start talking -- no downloads, no configuration."},"voiceCommands":{"title":"Voice Commands:","body":"Built-in commands for \"delete that,\" \"select that,\" \"undo that,\" and punctuation by name. Handy for quick edits."},"autoPunctuation":{"title":"Auto-Punctuation:","body":"Automatic punctuation insertion without speaking punctuation names. Simple toggle in settings."},"fluidDictation":{"title":"Fluid Dictation (Copilot+ PCs):","body":"On newer Copilot+ hardware, grammar correction and filler word removal runs on-device. English only."}},"pricing":{"title":"Both Free -- But Not Equal","ow":{"freeLabel":"Free & Open Source","free1":"Whisper and Parakeet models included","free2":"Agent mode with system prompts","free3":"100% offline, no internet needed","free4":"macOS, Windows, Linux","free5":"99+ languages","free6":"BYOK cloud API support","proLabel":"Pro (Optional)","pro1":"Unlimited cloud transcription","pro2":"20h meeting recordings/month","pro3":"Mobile app","pro4":"Email support"},"windows":{"builtInLabel":"Built into Windows 11","pro1":"No installation required","pro2":"Basic dictation & voice commands","pro3":"Auto-punctuation","con1":"Best accuracy requires internet","con2":"No model selection","con3":"No AI cleanup or formatting","con4":"Windows only"},"copilot":{"title":"Copilot+ PC Required","subtitle":"For Fluid Dictation","pro1":"On-device grammar correction","pro2":"Filler word removal","con1":"English only","con2":"Requires specific hardware"},"callout":{"lead":"Both are free, but \"free\" isn't the whole story.","body":"Windows Voice Typing is convenient but limited. OpenWhispr gives you the accuracy of Whisper and Parakeet AI, full offline privacy, agent mode for smart text cleanup, and the freedom of open source -- all for $0."}},"featureShowdown":{"title":"Feature Showdown","bothHaveTitle":"Features Both Have:","bothHave":["Free to use","Works in any text field","Multiple language support","Automatic punctuation","Push-to-talk activation","Real-time transcription"],"openWhisprExclusivesTitle":"OpenWhispr Exclusives:","openWhisprExclusives":["State-of-the-art Whisper and Parakeet AI models","Agent mode: AI-powered text cleanup & formatting","Choose your model size (speed vs accuracy)","100% offline -- no internet, no exceptions","Cross-platform (macOS, Windows, Linux, iOS)","Open source -- inspect, modify, contribute","99+ languages with high accuracy","BYOK: bring your own cloud API keys"],"windowsExclusivesTitle":"Windows Voice Typing Exclusives:","windowsExclusives":["Pre-installed, zero setup required","Built-in voice editing commands","Fluid Dictation on Copilot+ PCs","Integrated with Windows accessibility"]},"evidence":{"title":"Evidence-Based Notes","body":"This comparison is based on vendor documentation, pricing pages, and public product docs rather than anecdotal testimonials."},"migration":{"title":"Upgrading from Windows Voice Typing","intro":"Ready for more accurate, private, AI-powered dictation? Here's how to get started:","step1":{"title":"Download OpenWhispr","body":"Download for free. Optional Pro ($8/user/mo) for unlimited cloud transcription, more meeting hours, and the mobile app."},"step2":{"title":"Pick Your AI Model","body":"Start with \"base\" for a balance of speed and accuracy. Upgrade to \"large\" when you need maximum precision."},"step3":{"title":"Set Your Hotkey","body":"Configure push-to-talk with your preferred key. Works just like Win+H, but with your own shortcut."},"step4":{"title":"Try Agent Mode","body":"Set up system prompts to automatically clean up grammar, remove filler words, or format text for specific use cases."},"callout":{"lead":"You can keep both.","body":"OpenWhispr and Windows Voice Typing work side by side. Try OpenWhispr for your next writing session -- we think you'll notice the difference in accuracy immediately."}},"bottomLine":{"title":"The Bottom Line","intro":"Windows Voice Typing is a solid built-in tool for casual dictation. It's convenient, free, and good enough for quick notes. But if you care about accuracy, privacy, or power, it falls short.","windowsFineFor":{"title":"Windows Voice Typing Is Fine For","body":"Quick messages, casual notes, and situations where convenience matters more than accuracy. No installation, no configuration -- just press Win+H."},"openWhisprBuiltFor":{"title":"OpenWhispr Is Built For","body":"Writers, developers, professionals, and anyone who demands the best accuracy, complete privacy, AI-powered text cleanup, and the freedom of open source. Works everywhere -- not just Windows."},"closing":"Windows gave you the basics. OpenWhispr gives you the upgrade."},"finalCta":{"title":"Ready to Upgrade Your Windows Dictation?","body":"Get the accuracy of Whisper and Parakeet, the privacy of local processing, and the power of agent mode -- all free, all open source.","starButton":"Star us on GitHub","badge1":"No credit card required","badge2":"Works 100% offline","badge3":"Open source forever","viewAllComparisons":"View all comparisons"}},"compareWisprflow":{"metadata":{"title":"Wispr Flow Alternative: Free, Local & Open Source","description":"OpenWhispr is the free, open-source Wispr Flow alternative that runs entirely on your device. Private, offline dictation — no subscription needed.","keywords":["wisprflow alternative","wisprflow vs openwhispr","AI dictation","voice to text","nvidia parakeet"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","wisprflow":"vs WisprFlow"},"hero":{"title":"OpenWhispr: The Open-Source Wispr Flow Alternative","subtitle":"Wispr Flow vs OpenWhispr, compared. If you want a private, open-source voice-to-text tool that runs entirely on your own device, here is how the two stack up.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","wisprFlowIntro":"is an AI-powered dictation tool that works in any application, from Slack to email to AI chat interfaces. It focuses on fast cloud dictation with advanced voice commands and auto-editing.","openWhisprIntro":"takes a different approach. It's completely open source, works across all your applications (not just browsers), and gives you full control over your data. You can run it entirely offline or use your own API keys. Plus, it's free forever if you're comfortable with open source.","chooseWisprFlow":"Choose Wispr Flow if: You want AI-powered auto-editing, context-aware writing, and advanced voice commands.","chooseOpenWhispr":"Choose OpenWhispr if: You value privacy, want complete control, or prefer open source software."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"100% local processing option, self-hosted API keys"},"wisprFlow":{"status":"partial","details":"Cloud-based with privacy controls"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"Fully open source on GitHub"},"wisprFlow":{"status":"none","details":"Proprietary software"}},{"feature":"Platform Support","openWhispr":{"status":"full","details":"Works in any app system-wide"},"wisprFlow":{"status":"full","details":"Works everywhere - any text field in any app"}},{"feature":"Pricing Model","openWhispr":{"status":"full","details":"Free forever with paid pro options"},"wisprFlow":{"status":"partial","details":"Free tier (2,000 words/week desktop), paid Pro plans available"}},{"feature":"Voice Commands","openWhispr":{"status":"partial","details":"Basic commands for punctuation"},"wisprFlow":{"status":"full","details":"Advanced voice commands and shortcuts"}},{"feature":"Accuracy","openWhispr":{"status":"full","details":"State-of-the-art Whisper and Parakeet models"},"wisprFlow":{"status":"full","details":"High accuracy with AI-powered auto-editing"}},{"feature":"Customization","openWhispr":{"status":"full","details":"Fully customizable, fork-friendly"},"wisprFlow":{"status":"partial","details":"Limited to provided settings"}},{"feature":"Offline Support","openWhispr":{"status":"full","details":"Complete offline functionality"},"wisprFlow":{"status":"partial","details":"Some features require internet"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","privacy":{"title":"True Privacy:","body":"Process everything locally. Your voice never leaves your device unless you choose to use cloud APIs."},"crossPlatform":{"title":"True Cross-Platform:","body":"Available on macOS, Windows, Linux and iOS. WisprFlow only supports Mac and Windows."},"openSource":{"title":"Open Source Freedom:","body":"Modify it, extend it, or contribute back. No vendor lock-in, ever."},"cost":{"title":"Cost Effective:","body":"Free forever with local processing, or use your own API keys for cloud features."}},"whyWisprFlow":{"title":"Why WisprFlow?","voiceCommands":{"title":"Voice Commands:","body":"Advanced voice control system with custom commands and macros for complex workflows."},"autoEditing":{"title":"AI Auto-Editing:","body":"Context-aware AI that knows if you're writing an email, text, or document and adapts accordingly."},"managedService":{"title":"Managed Service:","body":"No setup required. Updates and improvements handled automatically."},"commandMode":{"title":"Command Mode:","body":"Voice editing commands like \"make this formal\" or \"turn into bullet points\"."}},"pricing":{"title":"Pricing Breakdown","ow":{"freeLabel":"Forever Free","freeDetails":"Local processing, unlimited use","proLabel":"Pro","proDetails":"Unlimited cloud, mobile app, support"},"wisprFlow":{"freeLabel":"Free","freeSub":"Limited free tier","freeDetails":"Basic features, limited usage","proLabel":"Pro","proSub":"Subscription","proDetails":"Unlimited words, all features"}},"useCases":{"title":"Best For Different Users","openWhisprTitle":"OpenWhispr is Best For:","openWhisprList":["Developers who want to customize their tools","Privacy-conscious users handling sensitive data","Teams needing system-wide dictation support","Budget-conscious individuals and startups","Users who work across multiple applications"],"wisprFlowTitle":"WisprFlow is Best For:","wisprFlowList":["Power users needing advanced voice commands","Writers who need AI-powered auto-editing","Users wanting pre-built workflow automation","Those preferring managed services over DIY","Users who need multilingual cloud dictation"]},"migration":{"title":"Switching from WisprFlow to OpenWhispr","intro":"Making the switch is easier than you think. Here's how:","step1":{"title":"Download OpenWhispr","body":"Get the free version from GitHub or go Pro for instant setup."},"step2":{"title":"Configure Your Preferences","body":"Set your hotkey, choose local or cloud processing, and select your model."},"callout":{"lead":"Pro tip:","body":"You can use both tools side-by-side during your transition. OpenWhispr won't interfere with WisprFlow's browser extension."}},"finalCta":{"title":"Ready to Take Control of Your Voice Data?","body":"Join thousands who've switched to OpenWhispr for true privacy, system-wide support, and the freedom of open source.","starButton":"Explore on GitHub","footnote":"No credit card required • Works offline • Open source forever","viewAllComparisons":"View all comparisons"}},"useCasesHub":{"metadata":{"title":"Dictation Software for Every Workflow | OpenWhispr","description":"Explore dictation software use cases: medical, legal, developers, writers, Mac, Windows & Linux. Private, local, open-source. Find your workflow.","keywords":["voice dictation use cases","speech to text for developers","dictation for writers","voice typing workflows","openwhispr use cases"]},"breadcrumb":{"home":"Home","useCases":"Use Cases"},"hero":{"title":"Use Cases","subtitle":"Voice dictation is not one-size-fits-all. Explore how OpenWhispr fits into your specific workflow — from writing code to drafting legal briefs.","description":"Every use case page covers setup tips, real-world examples, and the features that matter most for your profession or platform."},"learnMore":"Learn more →","useCases":[{"title":"For Developers","href":"/use-cases/developers","description":"Dictate code comments, documentation, LLM prompts, and commit messages without leaving your IDE.","icon":"💻","accent":"from-blue-500 to-blue-600"},{"title":"For Product Teams","href":"/use-cases/product-teams","description":"Local AI meeting notes for product teams. No bot joins the call, and with a local model the roadmap never leaves your machine.","icon":"🗺️","accent":"from-violet-500 to-violet-600"},{"title":"For Linux","href":"/use-cases/linux","description":"The best AI dictation app for Linux. System-wide voice typing that works in any X11 or Wayland application.","icon":"🐧","accent":"from-orange-500 to-orange-600"},{"title":"For Writers","href":"/use-cases/writers","description":"Talk your draft out and get your own sentences back. Books, articles and newsletters, transcribed on your machine.","icon":"✍️","accent":"from-purple-500 to-purple-600"},{"title":"For Students","href":"/use-cases/students","description":"Free speech-to-text for students. Record a lecture, get the notes written up, and dictate the essay after, all on your own laptop at $0.","icon":"🎓","accent":"from-green-500 to-green-600"},{"title":"For Journalists","href":"/use-cases/journalists","description":"Private AI speech to text for journalists. Pick a local model and interview audio is transcribed on your own device, never uploaded.","icon":"📰","accent":"from-fuchsia-500 to-fuchsia-600"},{"title":"For Clinicians","href":"/use-cases/medical","description":"Private on-device dictation for clinicians. Dictate clinical notes and SOAP notes into any EHR without sending patient audio to the cloud.","icon":"🩺","accent":"from-blue-500 to-blue-600"},{"title":"For Radiology","href":"/use-cases/radiology","description":"Put dictated findings at the active report cursor with local speech-to-text built for radiologists who keep the images in view.","icon":"🩻","accent":"from-violet-500 to-indigo-600"},{"title":"For Pathologists","href":"/use-cases/pathology","description":"Dictate macroscopic descriptions and microscopic findings into a compatible report field, then review every word before sign-out.","icon":"🔬","accent":"from-fuchsia-500 to-fuchsia-600"},{"title":"For Doctors","href":"/use-cases/doctors","description":"Referral letters, reports and notes typed on your own machine the moment you stop talking. Free for every doctor.","icon":"⚕️","accent":"from-emerald-500 to-emerald-600"},{"title":"For Physicians","href":"/use-cases/physicians","description":"Put HPI, exam and assessment text at the EHR cursor with one hotkey. Free on-device dictation for physicians.","icon":"🩻","accent":"from-cyan-500 to-cyan-600"},{"title":"For Vets","href":"/use-cases/veterinary","description":"Private on-device dictation for vets and vet techs. Dictate SOAP notes, discharges and client emails into any PIMS, offline on a farm call.","icon":"🐾","accent":"from-emerald-500 to-emerald-600"},{"title":"Legal Dictation","href":"/use-cases/lawyers","description":"Private dictation for lawyers and legal teams. Draft briefs, motions, memos, and intake notes with local transcription for privileged work.","icon":"⚖️","accent":"from-indigo-500 to-indigo-600"},{"title":"Legal Documents","href":"/use-cases/legal-documents","description":"Dictate a contract clause, a discovery response or a brief section and a local AI model drafts it into headings and numbered points on your machine.","icon":"📄","accent":"from-rose-500 to-rose-600"},{"title":"For Mac","href":"/use-cases/mac","description":"Modern AI dictation for macOS. A Dragon Dictate replacement with Whisper and Parakeet accuracy, Apple Silicon performance, and offline privacy.","icon":"⌘","accent":"from-zinc-500 to-zinc-700"},{"title":"For Windows","href":"/use-cases/windows","description":"Modern AI dictation for Windows. A powerful upgrade over built-in speech recognition with Whisper and Parakeet-level accuracy.","icon":"🪟","accent":"from-sky-500 to-sky-600"},{"title":"For Email","href":"/use-cases/email","description":"Draft emails by speaking. Compose messages in Gmail, Outlook, or any email client at the speed of thought.","icon":"📧","accent":"from-red-500 to-red-600"},{"title":"For Productivity","href":"/use-cases/productivity","description":"Capture tasks, notes and replies by voice in the productivity app already open.","icon":"⚡","accent":"from-emerald-500 to-emerald-600"},{"title":"For Accessibility","href":"/use-cases/accessibility","description":"Hands-free typing for users with RSI, mobility impairments, or anyone who prefers voice input.","icon":"♿","accent":"from-teal-500 to-teal-600"},{"title":"Offline Use","href":"/use-cases/offline","description":"Transcribe without an internet connection. Run OpenAI Whisper and NVIDIA Parakeet models entirely on-device for air-gapped workflows.","icon":"✈️","accent":"from-amber-500 to-amber-600"},{"title":"For Content Creators","href":"/use-cases/content-creators","description":"Paste your own video link or an audio file and get the transcript, then talk the title and the description out instead of typing them.","icon":"🎬","accent":"from-fuchsia-500 to-fuchsia-600"},{"title":"For Translators","href":"/use-cases/translators","description":"Dictate translations at speech speed in 99+ languages. Works in any CAT tool with fully local, private processing.","icon":"🌐","accent":"from-cyan-500 to-cyan-600"},{"title":"For Spanish-speaking vets","href":"/use-cases/spanish-veterinary","description":"Spanish dictation for veterinary teams. Speak the clinical note, review the transcript, then paste it into the PIMS you already use.","icon":"🐾","accent":"from-emerald-500 to-emerald-600"},{"title":"For Podcasters","href":"/use-cases/podcasters","description":"Free, unlimited episode transcription on your own machine. Turn recordings into transcripts and speak your show notes into any app.","icon":"🎙️","accent":"from-rose-500 to-rose-600"}],"oneTool":{"title":"One Tool, Every Workflow","paragraph1":"OpenWhispr is a system-wide dictation tool — it works in any text field, in any application, on macOS, Windows, and Linux. That flexibility means it adapts to your workflow rather than forcing you to adapt to it.","paragraph2":"Developers use it to dictate documentation and LLM prompts without switching context. Writers draft entire chapters at the speed of speech. Healthcare professionals keep patient notes private with fully local processing. And anyone with repetitive strain injuries can give their hands a break without sacrificing productivity.","paragraph3":"Whether you need AI-powered dictation for a specific platform like Linux or Windows, a particular profession, or a specialized requirement like offline transcription, the guides above walk you through exactly how OpenWhispr fits in."},"sections":{"workflow":{"eyebrow":"Explore by workflow","title":"Made for the way you work"},"platform":{"eyebrow":"More ways to use OpenWhispr","title":"Choose a platform or capability"}}},"useCaseAccessibility":{"metadata":{"title":"Dictation Software for Accessibility | OpenWhispr","description":"Free, open-source dictation software for RSI, carpal tunnel, and motor disabilities. Runs locally, works offline, and types by voice in any app.","keywords":["voice dictation accessibility","hands free typing","RSI dictation software","speech to text disability","accessible voice input","motor disability typing","nvidia parakeet","openai whisper"]},"breadcrumb":{"home":"Home","useCases":"Use Cases","accessibility":"Accessibility"},"hero":{"badge":"Hands-Free Typing","title":"Voice Dictation for Accessibility","subtitle":"Accurate, reliable hands-free text input. Built with OpenAI Whisper and NVIDIA Parakeet, open source, and free — because tools that help people communicate shouldn't have a paywall.","viewOnGitHub":"View on GitHub"},"problem":{"title":"When Typing Isn't an Option","paragraph1":"For people living with RSI, carpal tunnel syndrome, arthritis, or motor disabilities, typing can range from painful to impossible. The need to communicate through text doesn't go away — work emails, messages, documents, and forms still need to be written.","paragraph2":"Most built-in dictation tools are mediocre at best. They misrecognize words frequently, require constant corrections (which means more typing), and often need an internet connection. Commercial alternatives like Dragon are expensive and limited to specific platforms.","paragraph3":"OpenWhispr provides a different path: highly accurate voice-to-text powered by OpenAI Whisper and NVIDIA Parakeet, running locally on your computer, working in every application, on every major operating system, completely free."},"howItHelps":{"title":"How OpenWhispr Helps","accuracy":{"title":"High-Accuracy Speech Models","body":"Fewer misrecognitions means fewer corrections. OpenAI Whisper and NVIDIA Parakeet models deliver exceptional accuracy across accents, speaking speeds, and vocabulary — reducing the frustration of fixing transcription errors."},"pushToTalk":{"title":"Push-to-Talk or Toggle Mode","body":"Hold a key to dictate (push-to-talk), or press once to start and once to stop (toggle mode). Toggle mode is especially helpful when holding a key is difficult or uncomfortable."},"everyApp":{"title":"Works in Every Application","body":"OpenWhispr pastes text via clipboard, so it works anywhere you can type: email, documents, messaging apps, web forms, code editors, social media. No special integrations needed."},"noInternet":{"title":"No Internet Required","body":"All processing happens on your computer. No internet connection needed, no cloud dependency. It works reliably every time you need it, regardless of your network situation."}},"scenarios":{"title":"How People Use OpenWhispr","rsi":{"title":"Managing RSI While Working Full-Time","body":"Repetitive strain injuries don't mean you stop working — but they mean every keystroke costs something. OpenWhispr lets you compose emails, write documents, and respond to messages entirely by voice. Toggle mode means you don't need to hold anything down. The high accuracy of Whisper and Parakeet means you spend less time correcting and more time resting your hands. Whisper and Parakeet both run locally, so you always have reliable access."},"motor":{"title":"Staying Productive with Motor Disabilities","body":"When fine motor control is limited, a keyboard becomes a barrier instead of a tool. Voice dictation removes that barrier entirely. OpenWhispr works on macOS, Windows, Linux and iOS, so it fits whatever setup you already have. And because it's free and open source, there's no financial barrier either."},"prevention":{"title":"Preventing Injury Before It Starts","body":"You don't have to wait for pain to start using voice dictation. If you type for hours daily — writing, coding, emailing — alternating between keyboard and voice input can significantly reduce strain. OpenWhispr's push-to-talk hotkey makes it easy to switch between the two throughout the day."}},"whyChoose":{"title":"Why OpenWhispr for Accessibility","free":{"title":"Free forever.","body":"Assistive technology shouldn't be gated behind a subscription. OpenWhispr's core features are free, always. Optional Pro ($8/user/mo) adds unlimited cloud transcription, more meeting hours, and the mobile app."},"crossPlatform":{"title":"Cross-platform.","body":"Works on macOS, Windows, Linux and iOS. Use whatever computer and operating system works best for you — OpenWhispr runs on all of them."},"private":{"title":"Completely private.","body":"Your voice data never leaves your device. No cloud processing, no accounts, no data collection. Your conversations and documents stay yours."},"openSource":{"title":"Open source and customizable.","bodyPrefix":"The source code is on","githubLink":"GitHub","bodySuffix":". If you need a specific feature or modification, the community can help — or you can build it yourself."},"stats":{"languages":"99+ languages","aiCleanup":"AI cleanup via agent mode","history":"Transcription history"}},"internalLinks":{"intro":"Learn more about OpenWhispr:","offlineLink":"offline voice-to-text","offlineSuffix":"(no internet required),","windowsLink":"Windows dictation","windowsSuffix":", and","emailLink":"email dictation","emailSuffix":". Or","compareLink":"compare with other tools","compareSuffix":"."},"cta":{"title":"Try OpenWhispr","description":"Download OpenWhispr and set up hands-free dictation in minutes. No account, no payment, no internet connection required.","footer":"Free forever · macOS, Windows, Linux · Works offline"}},"useCaseDevelopers":{"breadcrumb":{"home":"Home","useCases":"Use Cases","developers":"For Developers"},"metadata":{"title":"Free Local Dictation Software For Developers - OpenWhispr","description":"OpenWhispr is free, MIT-licensed dictation for developers. Talk the prompt through and it lands at your cursor, transcribed on your own machine.","keywords":["best dictation software for developers","dictation software for developers","speech to text for developers","voice dictation for programmers","open source dictation app","local speech to text","dictate prompts to AI coding tools"]},"schema":{"name":"OpenWhispr","applicationCategory":"Dictation software for developers","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source dictation software for developers. Prompts, pull request descriptions and commit messages are transcribed on the developer's own machine. Free for unlimited local dictation.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Dictation software for developers & engineers","title":"Talk the change through. Get the prompt typed.","body":"One hotkey, and the clean text lands where your cursor is: Cursor, a terminal, a GitHub review box. Pick a local model at setup and the audio never leaves your machine.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Person seated on the floor, reaching toward a laptop on a low table."}},"trustBar":{"label":"Dictate into any app, including:","names":["Cursor","VS Code","Claude Code","GitHub","Linear","Slack"]},"pain":{"label":"What you're putting up with","title":"So much of your typing isn't code.","subtitle":"Prompts, pull request descriptions, commit messages, review comments, standup updates. None of it is code. All of it is typing.","cards":[{"title":"The prompt takes longer than the run.","body":"An agent needs a paragraph of context to do the right thing, so you type the paragraph, and it finishes in under a minute. Say it instead and the paragraph lands in the composer."},{"title":"You already wired Whisper to a hotkey once.","body":"Then the OS update broke the paste, the venv rotted, or you never got the clipboard back. OpenWhispr is that script, finished and maintained."},{"title":"A privacy claim you can't check.","body":"Closed-source dictation asks you to believe a sentence on a marketing page. Ours is a repo you can read, and a local path you can test by pulling the cable out and carrying on."},{"title":"Built-in dictation has never heard of kubectl.","body":"macOS and Windows dictation guess at library names, CLI flags and service names. Put your stack in the custom dictionary to help the local Whisper model recognize those spellings, on the free plan."}]},"features":{"label":"What the hotkey actually does","title":"A paragraph of context. None of it typed.","subtitle":"Transcription runs on your machine. So can the cleanup. Neither costs anything.","items":[{"image":{"label":"Person speaking beside an open laptop at a shared desk."},"title":"Whisper runs on your machine, not ours","body":"Pick a local model at setup and Whisper or Parakeet transcribes the audio on your own machine, with no signal needed once the model is downloaded."},{"image":{"label":"Person gesturing while using a keyboard at a dimly lit desk."},"title":"One hotkey, any window","body":"The clean text pastes where your cursor already is: Cursor, VS Code, a terminal, a GitHub review box, Slack. Nothing to install per editor, and your clipboard is put back the way you left it."},{"image":{"label":"Person speaking at a workstation in an office at dusk."},"title":"The system prompt is yours to edit","body":"The prompt that cleans up your dictation ships as text you can open, rewrite, test against sample input and save."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Audio uploaded in local mode","value":"None","footnote":"the recording is transcribed on your own machine"},{"label":"Cost per developer","value":"$$0","footnote":"unlimited local dictation, every machine you own, forever"},{"label":"Setup takes","value":"A few steps","footnote":"download, pick a model, bind a hotkey, talk"}]},"howItWorks":{"label":"What the install involves","title":"Talking to your agent starts here.","subtitle":"No account, no card, and no telemetry unless you switch it on.","image":{"label":"Person speaking at a home desk with a keyboard, laptop and monitor."},"steps":[{"title":"Download it free.","body":"Mac, Windows or Linux. Setup offers OpenWhispr Cloud or local models. Pick local and it pulls them down once."},{"title":"Hit the hotkey and talk.","body":"Describe the change you want the way you would say it to a colleague at the next desk."},{"title":"Read it, then send it.","body":"The clean text is already at your cursor. Edit it, hit enter, and the recording is yours to keep or delete."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"What you can check yourself","title":"Ten reasons, and the repo is the citation for most of them.","subtitle":"Choosing dictation software as a developer comes down to what you'll take on trust. These are the parts you don't have to.","items":[{"title":"The audio is transcribed on your own disk","body":"A local Whisper or Parakeet model does the listening, on your machine. Drop the network and it keeps working, which is the shortest audit anyone has run on us."},{"title":"Free on every machine you own","body":"Unlimited local dictation at $0, forever, on the laptop, the desktop and the one in the office. There's no per-seat license to renew."},{"title":"MIT licensed, so fork it","body":"The desktop app is on GitHub under the MIT license. Read the transcription path before you install it, and change it if you want it to behave differently."},{"title":"It types into a terminal","body":"It pastes where your cursor is, including in a terminal. There's no editor plugin to install or maintain."},{"title":"Your identifiers, spelled your way","body":"Put library names, service names and the CLI tools you live in into the custom dictionary for sharper transcripts."},{"title":"The cleanup prompt is editable","body":"Tell it to leave code blocks alone, keep your shorthand, or stop it capitalising your service names. It's your prompt, and it ships as text."},{"title":"It runs with the wifi off","body":"A plane, a train, a customer's locked-down network. Local models need no signal once they're on disk."},{"title":"A CLI, and reading your notes is free","body":"npm install -g @openwhispr/cli, then openwhispr --local notes list reads them off your own machine with no login and no API key. Sync them to the cloud on a paid plan and Claude, ChatGPT and Cursor reach them over MCP."},{"title":"Telemetry is off unless you turn it on","body":"Analytics ship disabled out of the box. Recordings are files on your machine, deleted after 30 days by default, and Settings moves that or switches it off."},{"title":"100+ languages","body":"Standups, reviews and issue threads in the language your team actually speaks, transcribed on your own machine."}]},"faq":{"label":"Before you install it","title":"Six questions developers ask first.","subtitle":"Straight answers on what leaves the machine, what it types into, and what it won't do. If yours isn't here, ask us in Discord.","items":[{"question":"Does anything I dictate leave my machine?","answer":"Not the audio, once you pick a local transcription model at setup. Two more things worth knowing while you're in Settings. If you sign in, the AI cleanup that tidies your dictation runs through OpenWhispr Cloud until you point it at a local model too, so the text of what you said goes up even though the audio doesn't; signed out, it only leaves your machine if you've pointed cleanup at your own API key. And there's a cloud fallback for when a local model fails, which ships switched off and stays off until you turn it on."},{"question":"Does it work in a terminal, in Cursor and in VS Code?","answer":"Yes, and in anything else with a text field, because it pastes at your cursor rather than plugging into an editor. There's no extension to install per app. On a Mac it asks for the accessibility permission at setup, which is what lets it paste; without it the text goes to your clipboard and you paste it yourself. When it does paste for you, your previous clipboard contents are put back afterwards. On Linux, X11 works out of the box and a Wayland session needs ydotool installed before auto-paste works."},{"question":"Can I write code by voice with it?","answer":"Not the way Talon Voice does, and we'd rather say so. Talon is purpose-built for voice coding with language-specific commands and hands-free control of the whole machine. This is dictation in any text field, which in practice means the prose around the code: the prompt, the pull request description, the commit message, the review comment, the Slack reply."},{"question":"Does dictation software for developers get library names and CLI flags right?","answer":"That's what the custom dictionary is for, and it's free. Put your stack in it and the spellings are handed both to the local Whisper model doing the listening and to the AI model that tidies the transcript afterwards, so a service name doesn't get corrected back out after it was heard correctly. One exception worth knowing: the Parakeet model doesn't take the dictionary at the speech-to-text step, only the cleanup does."},{"question":"Why not just wire up whisper.cpp myself?","answer":"You can, and the parts that take the weekend aren't the transcription. They're the global hotkey on three platforms, the paste that works in a terminal and a browser, restoring the clipboard afterwards, model downloads and the cleanup pass that turns speech into something you'd send. All of that is written and MIT licensed on GitHub, so forking it is faster than starting."},{"question":"Is it actually free, and how does it compare to Wispr Flow?","answer":"Unlimited local dictation is $0 forever on Mac, Windows and Linux, and so is the custom dictionary. Our Pro is $8 a month, which makes OpenWhispr Cloud transcription unlimited and adds API and MCP access and the iPhone and iPad app."}]},"reviews":{"label":"What people tell us","title":"The people who found it before you did.","subtitle":"What people tell us.","score":{"value":"0","label":"Recordings kept on our servers","body":"With a local transcription model the audio never leaves your machine. Cloud audio is sent only to be transcribed, then never stored."},"items":[{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for developers","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Developers who want free, private, offline dictation and are fine with a local setup"}]},{"parts":[{"text":"Developers who want unlimited cloud transcription and MCP in their AI tools"}]},{"parts":[{"text":"Teams documenting and reviewing at volume"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited local dictation, with the audio staying on your own machine"}]},{"parts":[{"text":"Unlimited cloud transcription plus API and MCP access, without managing local models"}]},{"parts":[{"text":"Unlimited recordings plus Agent mode and chat over your own notes"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"For individuals and teams"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with team features and admin, SSO, SAML and SCIM, audit logs, retention controls, and dedicated support."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Why we built it","title":"We wanted this for our own terminal first.","paragraphs":["OpenWhispr started as the thing we kept half-building for ourselves: a hotkey that turns a spoken paragraph into text, without shipping the paragraph to anybody.","We build in public. The desktop app is MIT licensed and on GitHub. When we say dictation can stay on your machine, you can check the code yourself.","And if it breaks the night before a release, you talk to the people who built it, in Discord."],"image":{"label":"One of OpenWhispr's founders"}},"finalCta":{"tagline":"$$0 · open source · no extension to install","title":"Say the next prompt. Don't type it.","body":"Download OpenWhispr, pick a local model at setup, and the next paragraph you owe an agent gets spoken, cleaned and dropped at your cursor. No card, no seat to buy.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free forever on Mac, Windows & Linux."}}},"useCaseEmail":{"metadata":{"title":"Voice Dictation for Email | Draft Emails Faster | OpenWhispr","description":"Draft emails 3x faster by speaking. AI agent mode auto-formats as professional email. Works with Gmail, Outlook, Superhuman, and any email client.","keywords":["voice dictation for email","dictate emails","speech to text email","voice typing gmail","dictation for outlook","email voice to text","nvidia parakeet","openai whisper"]},"breadcrumb":{"home":"Home","useCases":"Use Cases","forEmail":"For Email"},"hero":{"badge":"Email Dictation","title":"Draft Emails 3x Faster by Speaking","subtitle":"Stop typing emails one key at a time. Speak naturally and let AI format your words into polished, professional messages — in any email client.","viewOnGitHub":"View on GitHub"},"problem":{"title":"Email Is Eating Your Day","paragraph1":"The average professional spends 2.5 hours per day on email. A significant chunk of that time is spent composing replies — carefully typing, editing tone, and formatting. You know what you want to say. The bottleneck is getting it from your brain to the screen.","paragraph2":"Voice dictation changes the equation. You speak at roughly 150 words per minute versus 40 words per minute typing. But raw dictation is messy — that's where OpenWhispr's agent mode comes in. It takes your natural speech and transforms it into clean, well-formatted email text."},"howItHelps":{"title":"How OpenWhispr Transforms Email","agentMode":{"title":"Agent Mode: \"Format as Professional Email\"","body":"Set a system prompt like \"format as a professional email\" and speak freely. OpenWhispr's AI cleans up filler words, adds proper salutations, fixes grammar, and structures your thoughts into a polished message."},"anyClient":{"title":"Works in Any Email Client","body":"Gmail, Outlook, Apple Mail, Superhuman, Fastmail, Proton Mail — OpenWhispr pastes via clipboard, so it works everywhere you compose email. No plugins or extensions needed."},"pushToTalk":{"title":"Push-to-Talk for Quick Replies","body":"Hold your hotkey, say \"Sounds good, let's go with option two. I'll send the details by Friday.\" Release, and it's pasted. Reply to emails in seconds instead of minutes."},"toneControl":{"title":"Tone and Grammar Control","body":"Create custom system prompts for different contexts: casual for teammates, formal for clients, concise for executives. Switch between them instantly."}},"scenarios":{"title":"Real-World Email Scenarios","inboxBlitz":{"title":"Morning Inbox Blitz","body":"You open your inbox to 30 unread emails. Instead of typing each reply, you work through them with voice: hold hotkey, speak your response, release. Agent mode formats each one properly. What used to take 45 minutes now takes 15. You haven't touched the keyboard once."},"proposal":{"title":"Long-Form Client Proposal","body":"You need to write a detailed project proposal email. Rather than staring at a blank compose window, you talk through your thinking naturally. OpenWhispr captures it all, and agent mode restructures your stream of consciousness into organized paragraphs with clear next steps."},"multilingual":{"title":"Multilingual Correspondence","body":"You work with international clients and write emails in multiple languages. OpenAI Whisper and NVIDIA Parakeet support 99+ languages with strong accuracy. Dictate in Spanish to one client, switch to English for the next, then German for a partner — without changing any settings."}},"why":{"title":"Why OpenWhispr for Email","private":{"title":"Private by Default","body":"Your email content stays on your machine. No cloud processing means sensitive business correspondence stays confidential."},"faster":{"title":"Faster Than Typing","body":"Speak at 150 WPM vs typing at 40 WPM. Agent mode eliminates the editing step. Net result: emails drafted 3x faster."},"anyClient":{"title":"Any Client, Any Platform","body":"Works on macOS, Windows, and Linux. Pastes into any email client — web or desktop. No integrations to configure."},"proTip":{"lead":"Pro tip:","body":"Create multiple system prompts — one for casual internal emails, one for formal client emails, and one for concise executive summaries. Switch between them with a click."}},"internalLinks":{"intro":"See more ways to use OpenWhispr:","windows":"best dictation for Windows","sep1":",","accessibility":"accessibility","sep2":", and","offline":"offline voice-to-text","sep3":". Or","compare":"compare OpenWhispr","outro":"with other dictation tools."},"bottomCta":{"title":"Start Dictating Emails Today","description":"Reclaim hours of your week. Download OpenWhispr, set up a system prompt for email, and watch your inbox zero time shrink dramatically.","footer":"Free forever · Works in any email client · No account required"}},"useCaseLinux":{"metadata":{"title":"Linux Dictation Software — Free & Open Source | OpenWhispr","description":"One of the few actively maintained dictation apps for Linux. Open source, 100% local processing via Whisper — works on Ubuntu, Arch, Fedora and more.","keywords":["dictation app for linux","voice to text linux","linux speech to text","best linux dictation","linux voice recognition","open source dictation linux","speech recognition ubuntu","nvidia parakeet","openai whisper"]},"breadcrumb":{"home":"Home","useCases":"Use Cases","forLinux":"For Linux"},"hero":{"badge":"For Linux Users","title":"Linux Finally Gets Great Dictation","subtitle":"Most dictation apps ignore Linux entirely. OpenWhispr is open source, runs on every distro, and processes everything locally — no cloud dependency, no workarounds, no compromises.","viewOnGitHub":"View on GitHub"},"problem":{"title":"The Linux Dictation Problem","paragraph1":"If you've searched for \"best dictation app for Linux,\" you already know the frustration. Dragon NaturallySpeaking? Windows only. Apple Dictation? macOS only. Google's built-in speech-to-text? Chrome OS. Almost every polished dictation tool treats Linux as an afterthought — or ignores it completely.","paragraph2":"The options that do exist are usually browser-based (requiring you to send audio to Google or Microsoft servers), abandoned open source projects with outdated speech models, or hacks involving multiple tools stitched together with shell scripts.","callout":"OpenWhispr changes this. It's a real, maintained, cross-platform dictation app that treats Linux as a first-class citizen — with the same features Mac and Windows users get."},"different":{"title":"What Makes OpenWhispr Different on Linux","native":{"title":"Native Linux Support","body":"Not a Wine wrapper. Not a browser app. OpenWhispr runs natively on Linux as an Electron app with full system integration. Works on Ubuntu, Fedora, Arch, Debian, and every other major distro."},"openSource":{"title":"Fully Open Source","body":"The entire codebase is on GitHub. Build from source, audit the code, submit patches, package it for your distro. This is how Linux software should work — transparent, community-driven, and yours to control."},"local":{"title":"100% Local — No Cloud Needed","body":"All transcription runs on your machine through OpenAI Whisper and NVIDIA Parakeet. No internet connection required, no audio sent to external servers. Perfect for air-gapped environments, privacy-conscious setups, or just plain principle."},"models":{"title":"Multiple Speech Models","body":"Choose from OpenAI Whisper and NVIDIA Parakeet models based on your hardware. Running on a lightweight laptop? Use a smaller model for speed. Have a powerful workstation? Use a large model for maximum accuracy in 99+ languages."}},"scenarios":{"title":"How Linux Users Use OpenWhispr","developers":{"title":"Developers on Linux Workstations","body":"You're running Ubuntu or Arch as your daily driver. You need to write documentation, commit messages, emails, and LLM prompts throughout the day. OpenWhispr sits in the background, activated by a global hotkey. Hold it, speak, release — your text appears wherever your cursor is. Works in VS Code, terminals, browsers, Slack — anywhere."},"sysadmins":{"title":"Sysadmins and DevOps Engineers","body":"Documenting infrastructure changes, writing runbooks, and composing incident postmortems. OpenWhispr lets you dictate detailed notes while keeping your hands free for the terminal. Agent mode can format your speech as markdown documentation or structured incident reports."},"privacy":{"title":"Privacy-Focused Linux Users","body":"You chose Linux for a reason — control and privacy. OpenWhispr respects that philosophy completely. No telemetry, no cloud processing, no accounts required. Build from source if you want to verify every line. This is dictation software that aligns with why you use Linux in the first place."}},"whyChoose":{"title":"Why OpenWhispr Is the Best Dictation App for Linux","points":["One of the only quality, actively maintained dictation apps for Linux","Fully open source — build from source or use pre-built binaries","100% local processing via OpenAI Whisper and NVIDIA Parakeet","System-wide push-to-talk hotkey works in any application","AI-powered agent mode for text formatting and cleanup","99+ languages with accuracy from tiny to large models","Same app and workflow on Linux, macOS, and Windows","Free forever — no subscription required for core features"],"allUseCases":"All use cases","compareAlternatives":"Compare alternatives"},"cta":{"title":"Dictation That Respects Your OS Choice","description":"Open source voice-to-text, built for the platform that deserves better. Download OpenWhispr and start dictating on Linux today."}},"useCaseOffline":{"metadata":{"title":"Offline Dictation — Local Speech-to-Text | OpenWhispr","description":"Offline dictation software using local Whisper models. No internet, no data sent to servers. Works on planes, remote locations, and sensitive environments.","keywords":["offline voice to text","offline dictation app","transcribe without internet","local speech recognition","air gapped transcription","offline whisper","nvidia parakeet","openai whisper"]},"breadcrumb":{"home":"Home","useCases":"Use Cases","offline":"Offline Use"},"hero":{"badge":"Offline Voice-to-Text","title":"Transcribe Without Internet","subtitle":"100% offline voice-to-text powered by OpenAI Whisper and NVIDIA Parakeet running locally on your machine. No internet connection needed, no data leaves your device, ever.","viewOnGitHub":"View on GitHub"},"whyOffline":{"title":"Why Offline Matters","paragraph1":"Most voice-to-text tools send your audio to cloud servers for processing. That means they're useless on a plane, unreliable on a train, and completely dead in areas with spotty internet. Worse, your voice data — including everything you dictate — passes through third-party servers.","paragraph2":"OpenWhispr takes a fundamentally different approach. OpenAI Whisper and NVIDIA Parakeet models are downloaded once and run entirely on your computer. After initial setup, the app never needs to contact a server. Your transcription works the same whether you're connected to fiber optic or sitting in airplane mode."},"howItWorks":{"title":"How Offline Transcription Works","download":{"title":"Download Models Once","body":"Choose from OpenAI Whisper and NVIDIA Parakeet models in various sizes. Download once over WiFi, then the model lives on your machine permanently. Larger models are more accurate; smaller models are faster."},"local":{"title":"Local Processing via Whisper & Parakeet","body":"Transcription runs on your CPU/GPU using whisper.cpp and NVIDIA Parakeet. No server calls, no API keys, no internet. Just your hardware doing the work."},"zeroData":{"title":"Zero Data Transmission","body":"Your audio is processed in memory and never written to disk or transmitted anywhere. The transcribed text goes straight to your clipboard. No telemetry, no analytics on your voice data."},"platforms":{"title":"Works on All Platforms","body":"macOS, Windows, and Linux. The same offline experience on every operating system. Take your laptop anywhere and dictate without worrying about connectivity."}},"scenarios":{"title":"Where Offline Dictation Shines","flight":{"title":"On a Flight","body":"Six hours of uninterrupted flight time is perfect for writing. While cloud-based dictation tools show \"no connection\" errors, OpenWhispr works exactly as it does on the ground. Draft emails, write reports, or capture ideas — all by voice, all offline."},"remote":{"title":"Remote Locations","body":"Field researchers, journalists, and remote workers often find themselves where cell service and WiFi don't reach. OpenWhispr doesn't care. Record notes at a remote field site, transcribe interviews in a cabin, or dictate journal entries while camping. If your laptop has battery, you have dictation."},"privacy":{"title":"Privacy-Sensitive Environments","body":"Lawyers, healthcare professionals, and anyone handling sensitive information need dictation they can trust. With OpenWhispr, your voice data physically cannot reach the internet — it's processed locally and never stored. No HIPAA concerns about cloud processing, no risk of data breaches on third-party servers."}},"whyChoose":{"title":"Why Choose OpenWhispr for Offline Use","trulyOffline":{"title":"Truly Offline","body":"Not \"offline with occasional sync\" — genuinely zero internet dependency after downloading the model. Works in airplane mode, permanently."},"modelSizes":{"title":"Multiple Model Sizes","body":"From tiny (fast, lower accuracy) to large (slower, highest accuracy). Pick the right tradeoff for your hardware and needs. Download several and switch freely."},"openSource":{"title":"Free and Open Source","bodyPrefix":"No subscription fees for offline transcription. The","link":"full source code","bodySuffix":"is available on GitHub. Optional $8/mo Pro for convenience."},"languages":"99+ languages — all work offline","pushToTalk":"Push-to-talk and toggle mode"},"internalLinks":{"lead":"More about OpenWhispr:","windows":"best dictation for Windows","sep1":",","email":"email dictation","sep2":", and","accessibility":"accessibility","sep3":". Or","compare":"compare OpenWhispr","suffix":"with other dictation tools."},"cta":{"title":"Go Offline with OpenWhispr","description":"Download the app, grab a Whisper or Parakeet model, and you are set. From that point on, you never need an internet connection to transcribe again.","footer":"Free forever · 100% offline · No account required"}},"useCaseTranslators":{"metadata":{"title":"Dictation Software for Translators | OpenWhispr","description":"Dictate translations in 99+ languages with local Whisper processing. Private, offline, and works in any CAT tool. Try OpenWhispr free.","keywords":["voice dictation for translators","translation dictation software","speech to text for translation","multilingual dictation app","translator productivity tool","voice typing for interpreters","dictation for language professionals","multilingual voice to text","nvidia parakeet","openai whisper"]},"breadcrumb":{"home":"Home","useCases":"Use Cases","forTranslators":"For Translators"},"hero":{"badge":"For Translators & Language Professionals","title":"Dictate Translations at the Speed of Speech","subtitle":"Stop typing in two languages. OpenWhispr lets you speak your translations naturally in 99+ languages — with AI cleanup that formats your output for professional delivery. Private, local, and works in any CAT tool.","viewOnGitHub":"View on GitHub"},"problem":{"title":"Why Translators Need Voice Dictation","paragraph1":"Translation is mentally demanding work. You read in one language, think through meaning and nuance, then produce text in another. The last thing you need is friction at the output stage — switching keyboard layouts, hunting for special characters, fighting autocorrect that doesn't understand your target language.","paragraph2":"Most dictation tools are built for one language. They struggle with accented speech, specialized terminology, and the mixed-language workflow that translators live in every day. OpenAI Whisper and NVIDIA Parakeet change this — trained on massive multilingual datasets, they handle 99+ languages with native-level accuracy.","callout":"OpenWhispr gives you system-wide voice dictation powered by Whisper and Parakeet. Speak your translations naturally, use agent mode to clean up and format the output, and paste directly into whatever tool you work in — Trados, memoQ, a simple text editor, anything."},"solutions":{"title":"How OpenWhispr Helps Translators Work Faster","languages":{"title":"99+ Languages with Whisper & Parakeet","body":"Dictate in any source or target language with native-level accuracy. OpenAI Whisper and NVIDIA Parakeet were trained on massive multilingual datasets — they handle accents, regional dialects, and specialized vocabulary that generic dictation tools miss. No keyboard layout switching required."},"agentMode":{"title":"Agent Mode for Translation Workflows","body":"Set up custom system prompts tailored to your work: \"clean up and format as professional translation,\" \"structure as bilingual draft,\" \"remove filler words and polish for publication.\" Your raw dictation goes in, formatted translation-ready text comes out."},"catTool":{"title":"Works in Any CAT Tool","body":"Trados, memoQ, OmegaT, Memsource, Smartcat, Wordfast — or a plain text editor. OpenWhispr operates system-wide via clipboard integration. If you can paste text into it, OpenWhispr works with it. No plugins, no compatibility issues."},"privacy":{"title":"Private & Local Processing","body":"Client documents stay on your machine. With local Whisper and Parakeet processing, your audio never leaves your computer — no cloud uploads, no third-party access. Perfect for NDA-bound projects, confidential legal texts, and sensitive client work."}},"scenarios":{"title":"Real-World Translation Scenarios","literary":{"title":"Literary & Book Translation","body":"You're translating a novel and need to produce natural, flowing prose in the target language. Instead of typing sentence by sentence — constantly switching mental gears between reading and typing — you read a passage, then speak your translation aloud. Agent mode cleans up your dictation into polished prose, preserving the literary tone while removing hesitations and false starts.","note":"Common translator workflow: dictate the target-language draft naturally, then do a terminology and style pass for final QA."},"technical":{"title":"Technical & Legal Document Translation","body":"Technical manuals, legal contracts, medical reports — these documents demand precise terminology and consistent formatting. Dictate specialized terms naturally (Whisper and Parakeet handle domain-specific vocabulary well), then use agent mode with a prompt like \"format as formal legal translation, maintain paragraph structure\" to ensure consistent output. All processing stays local, so confidential documents never leave your machine."},"subtitling":{"title":"Subtitling & Localization","body":"Subtitle translation requires short, punchy text delivered quickly. Watch the video, dictate your translated subtitle lines, and paste them into your subtitle editor. Speaking is dramatically faster than typing when you need to produce hundreds of subtitle segments. Agent mode can help keep lines concise and within character limits."}},"whyChoose":{"title":"Why Translators Choose OpenWhispr","accuracy":{"title":"Multilingual Accuracy","body":"99+ languages with native-level accuracy. Whisper and Parakeet were built for multilingual speech — they handle accents, dialects, and code-switching between languages."},"confidentiality":{"title":"Client Confidentiality","body":"Local processing means client documents never leave your computer. No cloud uploads, no NDA concerns, no worrying about data privacy regulations."},"speed":{"title":"3x Faster Output","body":"Speak at 130 words per minute instead of typing 40. For translators billing by the word or working on deadlines, that speed difference is transformative."},"linkAllUseCases":"All use cases","linkCompare":"Compare alternatives"},"bottomCta":{"title":"Translate at the Speed of Speech","description":"Stop typing translations character by character. Speak naturally in any language, let AI clean up the output, and paste directly into your tools. Open source and private by design."}},"useCaseWindows":{"metadata":{"title":"Dictation Software for Windows — Free & Local | OpenWhispr","description":"Free, open-source dictation software for Windows. Runs offline on your PC, powered by OpenAI Whisper. No cloud, no subscription, no account required.","keywords":["dictation app for windows","voice to text windows","windows speech recognition alternative","whisper windows","windows voice typing","nvidia parakeet","openai whisper","free dictation software windows"]},"breadcrumb":{"home":"Home","useCases":"Use Cases","forWindows":"For Windows"},"hero":{"badge":"Windows Voice-to-Text","title":"The Best Dictation App for Windows","subtitle":"AI-powered voice-to-text that actually works. More accurate than Windows Speech Recognition, free forever, and runs entirely on your machine.","viewOnGitHub":"View on GitHub"},"problem":{"title":"Windows Dictation Deserves Better","paragraph1":"Windows built-in speech recognition hasn't meaningfully improved in years. It struggles with accents, technical vocabulary, and anything beyond simple sentences. Dragon NaturallySpeaking was the gold standard, but Nuance discontinued the consumer version — leaving millions of Windows users without a serious dictation tool.","paragraph2":"Meanwhile, the AI revolution has produced OpenAI Whisper and NVIDIA Parakeet — state-of-the-art speech recognition models. OpenWhispr brings both to your Windows PC as a native desktop app, free and open source."},"howItHelps":{"title":"How OpenWhispr Solves It","whisperParakeet":{"title":"Whisper & Parakeet Accuracy","body":"Choose from OpenAI Whisper and NVIDIA Parakeet models in various sizes. The larger models handle accents, technical jargon, and natural speech with remarkable accuracy — far beyond what Windows Speech Recognition can do."},"pushToTalk":{"title":"Push-to-Talk Hotkey","body":"Hold a key to dictate, release to transcribe. Or use toggle mode for longer sessions. Works system-wide — in any window, any app, any text field."},"agentMode":{"title":"Agent Mode (AI Cleanup)","body":"System prompts transform raw transcription into polished text. Fix grammar, format as bullet points, rewrite as professional prose — all automatically."},"worksEverywhere":{"title":"Works in Every App","body":"OpenWhispr pastes text via clipboard, so it works in Notepad, Word, VS Code, Slack, browser text fields — any app that accepts text input."}},"scenarios":{"title":"Real-World Scenarios on Windows","dragon":{"title":"Replacing Dragon NaturallySpeaking","body":"Dragon was your go-to for years, but Nuance killed the consumer version. OpenWhispr fills that gap with modern AI. OpenAI Whisper and NVIDIA Parakeet match or exceed Dragon's accuracy for most use cases, and the push-to-talk hotkey gives you the same quick-dictation workflow. Unlike Dragon, it's free and actively maintained by an open source community."},"coding":{"title":"Coding with Voice on Windows","body":"You're working in VS Code or Visual Studio and need to write comments, documentation, or commit messages without breaking your flow. Hold your hotkey, speak naturally, and OpenWhispr pastes the text right where your cursor is. Agent mode can even format your speech as proper code comments."},"multilingual":{"title":"Bilingual and Multilingual Dictation","body":"Windows Speech Recognition struggles with non-English languages. Whisper and Parakeet support 99+ languages with strong accuracy across all of them. Switch between English and another language mid-sentence — both engines handle it naturally."}},"why":{"title":"Why Choose OpenWhispr on Windows","free":{"title":"Free Forever","body":"Core features are completely free. Optional Pro ($8/user/mo) for unlimited cloud transcription, more meeting hours, and the mobile app."},"local":{"title":"100% Local Processing","body":"Your voice never leaves your machine. No cloud, no accounts, no data collection. Perfect for sensitive work."},"openSource":{"title":"Open Source","body":"Fully open source on GitHub. Inspect the code, contribute features, or fork it. No black boxes."},"highlights":{"languages":"99+ languages supported via Whisper & Parakeet","history":"Transcription history for easy reference","crossPlatform":"Cross-platform — also runs on macOS and Linux","byok":"BYOK — bring your own cloud API keys"}},"internalLinks":{"text":"Exploring your options? Compare OpenWhispr with other dictation tools, or see how it works for email dictation, accessibility, and offline use."},"bottomCta":{"title":"Get Started on Windows","description":"Download OpenWhispr and start dictating in minutes. No account required, no internet needed, no strings attached.","footer":"Free forever · Windows 10 & 11 · No account required"}},"useCaseWriters":{"breadcrumb":{"home":"Home","useCases":"Use Cases","writers":"For Writers"},"metadata":{"title":"Free Speech To Text Software For Writers - OpenWhispr","description":"OpenWhispr is free, open-source speech to text software for writers. It writes down what you said instead of drafting for you, on your own machine.","keywords":["speech to text software for writers","dictation software for writers","voice to text for authors","speech to text for novelists","dictation app for bloggers","voice typing for writing"]},"schema":{"name":"OpenWhispr","applicationCategory":"Speech to text software for writers","operatingSystem":"macOS, Windows, Linux, iOS","description":"Local-first, open-source speech to text software for writers. Your dictation is written down in your own words rather than redrafted for you, and in local mode it is transcribed on your own machine. Free on macOS, Windows and Linux.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Speech to text software for writers & authors","title":"Talk the draft out. Get your words, not an AI's.","body":"OpenWhispr writes down what you actually said, ums trimmed and nothing invented. Take the local route at setup and the manuscript never leaves your machine.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Person speaking at a desk with printed pages and a laptop"}},"trustBar":{"label":"Write into:","names":["Scrivener","Ulysses","iA Writer","Obsidian","Google Docs","Microsoft Word"]},"pain":{"label":"What you're choosing between","title":"You said the line. It came back as somebody else's.","subtitle":"You want the draft down at the speed you talk. You do not want it back sounding like everyone else's.","cards":[{"title":"The draft comes back in nobody's voice.","body":"OpenWhispr's cleanup mode is instructed to keep your wording and return the cleaned transcript. It does not compose an unsaid sentence. When you dictate a draft, the page begins with what you said."},{"title":"Dictate a line of dialogue, get an answer back.","body":"Talk to an assistant and it answers you. Say \"Where were you last night?\" into OpenWhispr and that is what lands on the page: your detective's question, not a reply to it. The cleanup model is told the speaker is never talking to it, so a question in your draft stays a question in your draft."},{"title":"You cannot see what got changed.","body":"A polished transcript is hard to trust when you cannot compare it with the words you dictated. OpenWhispr keeps both. Expand any dictation in your history and the raw transcript is sitting there word for word, with its own copy button. When cleanup changed nothing, the entry says so."},{"title":"An unfinished manuscript in someone else's account.","body":"Choose local transcription when you want your manuscript handled on your own machine. It needs no upload."}]},"features":{"label":"What lands on the page","title":"Faster drafts. Still your prose.","subtitle":"Local dictation is unlimited on every plan, runs on your own machine, and gives back the sentences you said.","items":[{"image":{"label":"Person speaking beside a laptop and printed pages"},"title":"It types what you said, at your cursor","body":"Hold the hotkey, talk, and the words land in whatever you are writing in. Scrivener, Word, Obsidian, a browser text box: it pastes straight at the cursor, so there is no plugin to install and no separate window to copy out of."},{"image":{"label":"Person using a laptop beside an open notebook"},"title":"The raw transcript, kept beside the clean one","body":"Open the history to compare the finished text with the raw transcript from a dictation. Expand the entry and copy the raw words. Prefer no cleanup? One toggle in Settings turns it off and the transcript is what lands at your cursor."},{"image":{"label":"Person speaking at a laptop beside an open book"},"title":"Your character names, spelled your way","body":"Put the names, places and invented words from your book in the custom dictionary. Local Whisper and cloud transcription use those spellings as hints. Type Aerdwyn once and those routes lean on it instead of guessing phonetically."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Your recordings on our servers","value":"None","footnote":"audio is never stored, and on the local route it never leaves your device"},{"label":"Cost per 100,000 words","value":"$$0","footnote":"local dictation is unlimited on every plan, including Free"},{"label":"Setup","value":"Your pace","footnote":"download, pick a model, choose a hotkey"}]},"howItWorks":{"label":"Download to first paragraph","title":"Talk it through once. Then edit, not rewrite.","subtitle":"No account to open, no plugin to install, nothing to upload. Choose a local model at setup and the whole thing runs on your own machine, offline.","image":{"label":"Person seated at a desk with a closed laptop and printed pages"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Skip the account if you want to. Take the local route at setup and it puts both the transcribing and the cleanup on your own machine."},{"title":"Hold the hotkey and talk.","body":"Open Scrivener, Word, Obsidian or anything else that takes text, hold your hotkey, and say the paragraph. It lands at the cursor when you let go."},{"title":"Read it back, then write.","body":"If a line reads wrong, open the history and check it against the raw transcript."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"What it will and will not do","title":"Ten reasons writers keep their own words.","subtitle":"Picking speech to text software is a decision about your voice before it is anything else. The cleanup controls are in Settings, and the prompt that governs cleanup ships with the app.","items":[{"title":"It transcribes, it does not draft.","body":"The cleanup prompt is instructed to keep your voice, wording, formality and intent, and to keep technical terms and proper nouns exactly as spoken. It removes fillers, fixes grammar and punctuation, and breaks up run-ons. It does not write you a new sentence."},{"title":"The AI only composes when you ask it to.","body":"Composing is a separate mode. It runs when you address the assistant out loud by whatever name you gave it, or when you hold the separate assistant hotkey. Say nothing to it and nothing is generated. Its own instructions forbid adding content that was not spoken or requested."},{"title":"Questions stay questions.","body":"Assistants answer you. This one is told the speaker is never talking to it, so a rhetorical question in an essay or a line of dialogue in a novel comes out on the page as you said it."},{"title":"Deliberate bad grammar survives, if you want it to.","body":"Cleanup fixes grammar and breaks up run-on sentences, which is wrong for a stream-of-consciousness passage or a character who talks in fragments. Turn cleanup off for that stretch and the raw transcript is what you get. Either way both versions are saved."},{"title":"You can turn cleanup off.","body":"One toggle in Settings, labeled Enable text cleanup. Off means no cleanup model is in the path and the transcript is what lands at your cursor."},{"title":"Your invented vocabulary goes to the transcriber.","body":"The custom dictionary is passed to the model as the spellings to use, on a local Whisper model and on cloud transcription alike, so character names, places and coinages come back closer to how you wrote them."},{"title":"Unlimited local dictation, on every plan.","body":"Local transcription is unlimited and free on Free, Pro and Business alike. A novel is not a metered event. The word allowance in the pricing table applies to cloud transcription only."},{"title":"It works offline.","body":"Take the local route at setup and there is no network call to make. A retreat with one bar of signal, or seat 24B on a red-eye. The draft still comes out."},{"title":"100+ languages, and you can switch mid-draft.","body":"Dictate in the language you think in and switch between languages without changing a setting. Useful if your dialogue is not in the language your narration is."},{"title":"It writes into whatever you already use.","body":"Dictation pastes at the cursor rather than into a proprietary editor, so Scrivener, Ulysses, iA Writer, Obsidian, Google Docs and Word all work with no plugin to install."}]},"faq":{"label":"Before you install it","title":"The questions writers ask first.","subtitle":"Straight answers on voice, privacy and cost.","items":[{"question":"Will it rewrite my prose?","answer":"It cleans, it does not compose. Cleanup is instructed to keep your voice, wording, formality and intent, and to keep proper nouns and jargon exactly as spoken. What it does change is fillers, false starts, stutters and repetitions, and it fixes grammar and punctuation and breaks up run-on sentences. If your prose runs long on purpose, turn cleanup off in Settings and you get the raw transcript. Either way both versions are saved."},{"question":"How do I check what it changed?","answer":"Open the history in the control panel and expand any dictation. Underneath the finished text you get the raw transcript, word for word, with a copy button. If cleanup made no change at all, the entry tells you so."},{"question":"Will it get my characters' and places' names right?","answer":"Add them to the custom dictionary in Settings. Those spellings are passed to the transcriber as the ones to use, on a local Whisper model and on cloud transcription alike. It is a strong hint rather than a hard rule, and the local Parakeet model does not take a dictionary at all."},{"question":"Is my unpublished manuscript private?","answer":"Choose local transcription at setup to process audio on your own machine without upload. Cloud transcription sends audio to be transcribed, then immediately discards it. Cloud backup can sync your transcriptions and notes across devices when enabled. Team-space and shared notes can sync separately."},{"question":"Is it really free for a whole book?","answer":"Yes. Local dictation is unlimited on the Free plan and every plan above it, with no word cap and no card. The 2,000 words a week you see in the pricing table is the Free tier's cloud transcription allowance, which is a different thing."},{"question":"Does it work in Scrivener, Ulysses or Word?","answer":"It works anywhere you can type. Dictation pastes at the cursor rather than living in its own editor, so there is nothing to install into your writing app and nothing to copy out afterwards. Mac, Windows and Linux, in 100+ languages."}]},"reviews":{"label":"What people tell us","title":"Local models. No WiFi. Better than typing.","subtitle":"From our Discord, GitHub and inboxes. They write code, not chapters, so read them for the tool rather than the trade.","score":{"value":"Raw transcript","label":"Available to check","body":"Cleanup is instructed to keep your wording and return only the cleaned transcript, with no commentary or answers. When a raw transcript is available, open it in history and copy it to check."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for writers","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Writers drafting a book who want unlimited, offline dictation for free"}]},{"parts":[{"text":"Writers who also want cloud transcription without managing a model"}]},{"parts":[{"text":"Editorial teams and agencies who want unlimited recordings and priority support"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited local dictation at no cost. Upgrade for unlimited Cloud transcription and workflow features."}]},{"parts":[{"text":"Unlimited cloud transcription when you would rather not run a model"}]},{"parts":[{"text":"Everything in Pro plus unlimited meeting recordings, agent mode and priority support"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"More: unlimited meeting recordings, Agent mode and chat over your data"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","Contact us about Enterprise."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Why we built it","title":"The line you got right out loud, first time. We built for that.","paragraphs":["OpenWhispr started with a simple objection: getting your own sentences onto the page faster should not mean handing them to something that rewrites them into its own voice.","The cleanup prompt ships with the app. You can read exactly how it is told to keep your wording.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"One of OpenWhispr's founders"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Nothing on the page that you did not say.","body":"Download OpenWhispr, take the local route at setup, and your next chapter gets talked out on your own machine, in your own sentences.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free to download on Mac, Windows and Linux."}}},"security":{"metadata":{"title":"Local Speech-to-Text Privacy & Security | OpenWhispr","description":"OpenWhispr keeps your voice private — local on-device processing, no audio stored, no AI training on your data, and full user controls."},"breadcrumb":{"home":"Home","security":"Security"},"hero":{"title":"Security at OpenWhispr","subtitle":"Your voice data is personal. Here's exactly how we handle it, what we store, and what controls you have."},"howItWorks":{"title":"How OpenWhispr Works","intro":"OpenWhispr is a desktop application for macOS, Windows, and Linux that converts speech to text. You choose how your audio is processed — there are three modes, and you pick the one that fits your needs.","local":{"title":"Local Models — Nothing Leaves Your Device","body":"You can run open-source Whisper models directly on your machine. When you choose this option, your audio is processed entirely on-device and never transmitted anywhere. If keeping your data off the internet is important to you, this is the option to use."},"byok":{"title":"Bring Your Own API Key","body":"You can connect your own API key from providers like OpenAI. Your audio is sent directly to that provider for transcription using your own account. You control the relationship with the provider, and their privacy policy applies."},"cloud":{"title":"OpenWhispr Cloud","body":"Our managed cloud transcription service handles processing on our end. Audio is sent to our API, transcribed, and returned. Neither we nor our providers store your audio — it is processed in real time and discarded. We use third-party AI providers behind the scenes and may switch between them, but we always ensure no provider trains on your data."},"openSource":{"label":"Open source desktop app:","body":"The OpenWhispr desktop application is fully open source and available on GitHub. You can inspect exactly how audio is captured, processed, and transmitted. The cloud API service is separate and managed by us."}},"dataStorage":{"title":"Data Processing and Storage","audio":{"title":"Audio Recordings","body":"Nobody stores your audio. With local models, it never leaves your device. With your own API key, audio goes directly to your chosen provider and is not retained. With OpenWhispr Cloud, audio is processed in real time and discarded — neither we nor our providers keep it."},"transcriptions":{"title":"Transcriptions","body":"Transcriptions are stored locally on your device. If you opt in to cloud sync, they are also stored in our database so you can access them across devices. You can disable sync at any time."},"account":{"title":"Account Data","body":"If you create an account, we store your email, name, and authentication credentials. Passwords are hashed and never stored in plain text. Payment processing is handled entirely by Stripe."},"diagnostic":{"title":"Diagnostic Data","body":"Diagnostic reporting is off by default. You must explicitly enable it in settings. When enabled, we collect anonymous performance metrics to improve the product. No audio or transcription content is ever included."}},"training":{"title":"AI Training Policy","intro":"We take a firm stance on AI training: your data is not used to train any AI models. Period.","cloud":{"label":"OpenWhispr Cloud:","body":"Every provider we use behind our cloud service is opted out of using your data for model training. We may switch between providers for quality and speed, but the no-training guarantee always applies."},"byok":{"label":"Bring your own key:","body":"If you use your own API key, your provider's training policies depend on your account settings with them. Most major providers offer training opt-outs — check your provider's documentation."},"self":{"label":"OpenWhispr itself:","body":"We do not train models on your transcriptions, audio, or any user-generated content."},"local":{"label":"Local models:","body":"No training concern at all. Your audio stays on your device and is never sent anywhere."}},"infrastructure":{"title":"Infrastructure Security","intro":"For users who opt in to cloud features, we follow industry-standard security practices:","encryption":{"title":"Encryption","body":"All data in transit is encrypted using TLS. Data at rest in our cloud database is encrypted. API connections to third-party providers use secure, authenticated endpoints."},"authentication":{"title":"Authentication","body":"We support email/password and Google OAuth sign-in. Sessions are managed with secure, short-lived tokens that refresh automatically. Credentials are never shared with third parties."},"dependencies":{"title":"Dependency Management","body":"We use automated tools to monitor for known vulnerabilities in our dependencies and apply patches promptly. Our open-source codebase means the community can audit our supply chain."}},"controls":{"title":"Your Controls","intro":"You decide how OpenWhispr handles your data. You choose your processing mode during setup, and you can change it at any time in settings.","mode":{"title":"Choose Your Processing Mode","body":"Pick local models, your own API key, or OpenWhispr Cloud. If you want your audio to never leave your device, choose local models."},"sync":{"title":"Enable or Disable Cloud Sync","body":"Keep transcriptions local-only or sync them across devices. Your choice."},"diagnostic":{"title":"Opt In to Diagnostic Reporting","body":"Diagnostic data collection is disabled by default. You can turn it on if you want to help us improve OpenWhispr."},"delete":{"title":"Delete Your Data","body":"You can delete individual transcriptions, clear all local data, or delete your account entirely from within the app."}},"providers":{"title":"Cloud Providers","cloud":{"title":"OpenWhispr Cloud","body":"When you use OpenWhispr Cloud, your audio is processed through our managed API. We use third-party AI providers behind the scenes and may change them to improve quality or speed, with advance notice per our Data Processing Addendum. All providers are opted out of training on your data."},"byok":{"title":"Bring Your Own Key","body":"When you use your own API key, your audio goes directly to your chosen provider. Their privacy policy and training settings apply based on your account with them."},"tableIntro":"Services we use across the platform:","headerProvider":"Provider","headerPurpose":"Purpose","headerOptOut":"Training Opt-Out","rows":[{"provider":"OpenAI","purpose":"Cloud transcription & AI (via OpenWhispr Cloud or BYOK)","optOut":"Opted out for our accounts"},{"provider":"Groq","purpose":"Cloud AI inference & transcription failover (via OpenWhispr Cloud or BYOK)","optOut":"Opted out for our accounts"},{"provider":"Parasail","purpose":"Cloud AI inference for agent features","optOut":"Zero data retention"},{"provider":"Mem0","purpose":"AI memory for agent features","optOut":"No training on your data"},{"provider":"Exa","purpose":"Web search for agent features","optOut":"No training on your data"},{"provider":"Neon","purpose":"Database & authentication","optOut":"N/A"},{"provider":"Vercel","purpose":"Website hosting & analytics","optOut":"N/A"},{"provider":"Stripe","purpose":"Payment processing","optOut":"N/A"},{"provider":"Resend","purpose":"Transactional email","optOut":"N/A"},{"provider":"Google","purpose":"OAuth sign-in","optOut":"N/A"},{"provider":"Slack","purpose":"Internal support communications","optOut":"N/A"}],"footnote":"We give at least 30 days' advance notice before adding or replacing providers that process customer data, per our Data Processing Addendum. The current list is published at trust.openwhispr.com, and this page is updated to reflect changes."},"vulnerability":{"title":"Vulnerability Reporting","intro":"If you discover a security vulnerability in OpenWhispr, we want to hear about it. As an open-source project, we value responsible disclosure from the community.","howToReport":"How to Report","emailPrefix":"Email","emailSuffix":"with a description of the issue.","steps":"Include steps to reproduce if possible.","ack":"We will acknowledge your report within 48 hours and keep you updated on the fix.","disclosure":"Please do not disclose vulnerabilities publicly until we have had a reasonable opportunity to address them."},"contact":{"title":"Questions?","intro":"If you have questions about our security practices or want to learn more about how we handle your data, reach out:","teamName":"OpenWhispr Team","securityLabel":"Security:","generalLabel":"General:","privacyLink":"Privacy Policy","termsLink":"Terms of Service","sourceLink":"Source Code"}},"models":{"metadata":{"title":"Whisper Model Sizes, Parakeet & More | OpenWhispr","description":"Compare Whisper model sizes — Tiny to Large-v3 — plus NVIDIA Parakeet, Groq, and OpenAI. Run locally for complete privacy or use your own API keys.","keywords":["whisper models","nvidia parakeet","speech to text models","whisper tiny base small medium large turbo","groq transcription","local AI models","voice to text providers","openai whisper","parakeet tdt","whisper.cpp","sherpa-onnx"]},"breadcrumb":{"home":"Home","models":"Models"},"badges":{"recommended":"Recommended","fastest":"Fastest"},"hero":{"titlePrefix":"Models &","titleHighlight":"Providers","subtitle":"OpenWhispr supports multiple speech-to-text engines and AI providers. Run everything locally for privacy, use cloud APIs for speed, or mix both."},"localTranscription":{"title":"Local Transcription Models","subtitle":"Run speech-to-text entirely on your machine. No internet required, no data leaves your device.","whisper":{"meta":"via whisper.cpp · 99+ languages · GGML format","headerModel":"Model","headerSize":"Size","headerSpeed":"Speed","headerQuality":"Quality","headerBestFor":"Best for","models":[{"id":"tiny","name":"Tiny","size":"75 MB","speed":"Fastest","speedLabel":"Fastest","quality":"Basic","description":"Lightweight model for quick drafts and low-powered hardware"},{"id":"base","name":"Base","size":"142 MB","speed":"Fast","speedLabel":"Fast","quality":"Good","recommended":true,"description":"Best balance of speed and accuracy for most users"},{"id":"small","name":"Small","size":"466 MB","speed":"Medium","speedLabel":"Medium","quality":"Better","description":"Improved accuracy for professional transcription"},{"id":"medium","name":"Medium","size":"1.5 GB","speed":"Slow","speedLabel":"Slow","quality":"High","description":"High accuracy for multilingual and technical content"},{"id":"large-v3","name":"Large v3","size":"3 GB","speed":"Slowest","speedLabel":"Slowest","quality":"Best","description":"Maximum accuracy across all 99+ languages"},{"id":"turbo","name":"Turbo","size":"1.6 GB","speed":"Fast","speedLabel":"Fast","quality":"Good","description":"Large-v3 architecture optimized for speed — near-large accuracy at base-level latency"}]},"parakeet":{"meta":"via sherpa-onnx · 25 languages · ONNX INT8","sotaBadge":"State-of-the-art English","languagesCount":"{count} languages","fullyLocal":"100% local","languagesList":"Parakeet languages: English, Bulgarian, Croatian, Czech, Danish, Dutch, Estonian, Finnish, French, German, Greek, Hungarian, Italian, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Russian, Slovak, Slovenian, Spanish, Swedish, Ukrainian.","models":[{"id":"parakeet-tdt-0.6b-v3","name":"Parakeet TDT 0.6B v3","size":"680 MB","speed":"Fast","speedLabel":"Fast","quality":"Excellent","recommended":true,"languages":25,"description":"State-of-the-art English accuracy with 25-language multilingual support. INT8 quantized for efficient local inference."}]}},"cloudTranscription":{"title":"Cloud Transcription Providers","subtitle":"Use your own API keys for cloud-powered transcription. Bring-your-own-key keeps you in control.","getApiKey":"Get API key","custom":{"title":"Custom Endpoint","label":"Any OpenAI-compatible API","descPrefix":"Connect Ollama, self-hosted Whisper servers, LocalAI, or any service with an OpenAI-compatible","descSuffix":"endpoint."},"providers":[{"name":"OpenAI","keyUrl":"https://platform.openai.com/api-keys","models":[{"id":"gpt-4o-mini-transcribe","name":"GPT-4o Mini Transcribe","speed":"Very Fast","speedLabel":"Very Fast","description":"Fast, accurate transcription"},{"id":"gpt-4o-transcribe","name":"GPT-4o Transcribe","speed":"Standard","speedLabel":"Standard","description":"Most accurate OpenAI transcription"},{"id":"whisper-1","name":"Whisper","speed":"Standard","speedLabel":"Standard","description":"Original Whisper API endpoint"}]},{"name":"Groq","recommended":true,"keyUrl":"https://console.groq.com/keys","models":[{"id":"whisper-large-v3-turbo","name":"Whisper Large v3 Turbo","speed":"Ultra-Fast","speedLabel":"Ultra-Fast","description":"216x real-time speed — the fastest cloud transcription available"}]},{"name":"Mistral","keyUrl":"https://console.mistral.ai/api-keys","models":[{"id":"voxtral-mini-latest","name":"Voxtral Mini","speed":"Fast","speedLabel":"Fast","description":"Multilingual transcription from Mistral AI"}]}]},"openWhisprCloud":{"title":"OpenWhispr Cloud","subtitle":"Zero-config transcription and AI processing. No API keys needed — just sign in and go.","point1":{"title":"Managed transcription","body":"We route to the fastest provider automatically"},"point2":{"title":"AI text cleanup included","body":"Formatting, punctuation, and filler removal"},"point3":{"title":"Free tier available","body":"Upgrade to Pro for unlimited usage"}},"aiReasoning":{"title":"AI Text Processing","subtitle":"After transcription, OpenWhispr can clean up your text — fixing grammar, removing filler words, and formatting output. Choose from cloud or local AI models.","cloud":{"title":"Cloud Providers (BYOK)","providers":[{"name":"OpenAI","models":["GPT-5.2","GPT-5 Mini","GPT-5 Nano","GPT-4.1","GPT-4.1 Mini","GPT-4.1 Nano"]},{"name":"Anthropic","models":["Claude Opus 4.5","Claude Sonnet 4.5","Claude Haiku 4.5"]},{"name":"Google","models":["Gemini 3 Pro","Gemini 3 Flash","Gemini 2.5 Flash Lite"]},{"name":"Groq","models":["Qwen3 32B","GPT-OSS 120B","GPT-OSS 20B","LLaMA 3.3 70B","LLaMA 3.1 8B","Mixtral 8x7B"]}]},"local":{"title":"Local Models (via llama.cpp)","footnote":"Local reasoning models run via llama.cpp with GGUF quantization. All processing stays on your machine.","models":[{"provider":"Qwen","recommended":"Qwen3 8B","others":"0.6B, 1.7B, 4B, 8B, 32B variants"},{"provider":"Mistral","recommended":"Mistral 7B Instruct","others":"Q4 and Q5 quantizations"},{"provider":"Meta LLaMA","recommended":"LLaMA 3.2 3B","others":"1B, 3B, 8B variants"},{"provider":"OpenAI OSS","recommended":"GPT-OSS 20B","others":"Open-source flagship"}]}},"howItWorks":{"title":"How It Fits Together","step1":{"title":"1. You speak","body":"Hold your hotkey and talk naturally into any app"},"step2":{"title":"2. Model transcribes","body":"Whisper, Parakeet, or a cloud provider converts speech to text"},"step3":{"title":"3. AI cleans up","body":"Optional AI processing fixes grammar, removes filler, and formats your text"}},"cta":{"title":"Try every model for free","body":"Download OpenWhispr and choose the models that work best for you. Local models are completely free. Cloud providers use your own API keys.","githubButton":"GitHub"}},"download":{"metadata":{"title":"Download OpenWhispr — Free Dictation Software","description":"Download OpenWhispr free — open-source dictation software that runs locally on macOS, Windows & Linux. Private, offline, no cloud required.","keywords":["download OpenWhispr","OpenWhispr macOS","OpenWhispr Windows","OpenWhispr Linux","voice dictation download","free voice to text app"]},"breadcrumb":{"home":"Home","download":"Download"},"referral":{"title":"Your first month is FREE!","welcome":"Welcome to OpenWhispr! Your referral discount has been applied to your account.","nextStepsTitle":"Next steps:","step1":"Download OpenWhispr using the button below","step2":"Install and sign in with your account","step3":"Upgrade to Pro - your discount will be applied automatically at checkout"},"latestLabel":"Latest","hero":{"title":"Download OpenWhispr","subtitle":"Available for macOS, Windows, Linux, and iOS. Choose the build that matches your device."},"autoDownloadNotice":"Your download should start automatically...","loading":"Loading...","recommendedLabel":"Recommended for your device:","allDownloads":{"title":"All downloads","subtitle":"Choose the installer that matches your system.","releaseNotes":"View release notes","installGuide":"Installation guide","uninstall":"Uninstall instructions","systemRequirements":"System requirements","downloadHelp":"Download help"},"platforms":{"macos":{"label":"macOS","description":"DMG builds for Apple Silicon and Intel","options":[{"label":"Mac (Apple Silicon)","meta":"DMG"},{"label":"Mac (Intel)","meta":"DMG"}]},"windows":{"label":"Windows","description":"Installer and portable build","options":[{"label":"Windows (Installer)","meta":"EXE (recommended)"},{"label":"Windows (Portable)","meta":"EXE"}]},"linux":{"label":"Linux","description":"AppImage, Debian, RPM, and tarball","options":[{"label":"Linux AppImage (x86_64)","meta":"AppImage"},{"label":"Linux DEB (amd64)","meta":"Debian package"},{"label":"Linux RPM (x86_64)","meta":"RPM package"},{"label":"Linux tar.gz (x64)","meta":"Portable tarball"}]},"ios":{"label":"iPhone & iPad","description":"Universal app on the App Store","optionLabel":"OpenWhispr for iOS","optionMeta":"App Store · Requires iOS 17"}}},"contact":{"metadata":{"title":"Contact Us | OpenWhispr","description":"Get in touch with the OpenWhispr team. Reach out for support, partnerships, press inquiries, or general questions.","keywords":["OpenWhispr contact","contact OpenWhispr","OpenWhispr support","OpenWhispr sales","OpenWhispr help"]},"breadcrumb":{"home":"Home","contact":"Contact"},"hero":{"title":"Contact us","subtitle":"We'd love to hear from you. Reach out through any of the channels below."},"cards":{"team":{"title":"Want OpenWhispr for your team?","description":"Roll out voice-to-text across your organization with volume pricing and priority support.","linkText":"Talk with our sales team"},"talk":{"title":"Talk to the team","description":"Questions, feedback, or just want to say hi? Drop us a line."},"help":{"title":"Need help?","description":"Bug reports, feature requests, or technical support — our community is here.","linkText":"Open an issue on GitHub"}}},"contactSales":{"metadata":{"title":"Enterprise AI Meeting Notes & Dictation | OpenWhispr","description":"Contact our sales team for enterprise pricing, SSO, and custom plans for private, local AI meeting notes and dictation software."},"hero":{"title":"Talk to our sales team","subtitle":"Get a custom plan for your organization with dedicated support, SSO, and team management."},"success":{"title":"We received your inquiry","body":"Our team will reach out within 24 hours.","backHome":"Back to Home"},"form":{"name":{"label":"Name","placeholder":"Your full name"},"email":{"label":"Email","placeholder":"work@company.com"},"company":{"label":"Company","placeholder":"Company name"},"companySize":{"label":"Company size","placeholder":"Select size"},"companySizeOptions":{"1-10":"1-10 employees","11-50":"11-50 employees","51-200":"51-200 employees","201-1000":"201-1,000 employees","1000+":"1,000+ employees"},"platform":{"label":"Email & calendar platform","placeholder":"What does your team use?"},"platformOptions":{"google":"Google (Gmail / Google Calendar)","microsoft":"Microsoft (Outlook / Teams)","other":"Other"},"message":{"label":"Anything else we can help answer?","placeholder":"Tell us about your use case, requirements, or any questions you have..."},"submit":"Send Inquiry","submitting":"Sending...","errors":{"nameRequired":"Name is required","emailInvalid":"Please enter a valid email address","companyRequired":"Company name is required","companySizeRequired":"Please select your company size","platformRequired":"Please select your platform","alreadySubmitted":"You've already submitted recently. We'll be in touch soon!","generic":"Something went wrong. Please try again.","network":"Network error. Please check your connection and try again."}},"features":{"title":"Enterprise features","managedApiCredits":"Managed API credits","ssoCompliance":"SSO & compliance","teamFeatures":"Team features & admin","cloudSync":"Cloud sync across devices"},"responseTime":{"value":"< 24 hours","label":"Average response time"},"emailDirect":{"prefix":"Prefer email? Reach us directly at"}},"legalPrivacy":{"metadata":{"title":"OpenWhispr Privacy Policy | Local-First, No Selling","description":"Your audio stays on your device. OpenWhispr is local-first, open source, and sells no data. See what we collect and your privacy rights."},"breadcrumb":{"home":"Home","privacy":"Privacy Policy"},"hero":{"title":"Privacy Policy","subtitle":"How OpenWhispr handles your data.","lastUpdated":"Last updated: August 12, 2026"},"overview":{"title":"Overview","paragraph1":"OpenWhispr is published and operated by Gizmo Labs Inc., a Delaware corporation (221 W 9th Street, Wilmington, Delaware 19801, USA). OpenWhispr is a desktop dictation and meeting-notes app built with privacy at the center. By default, your audio is processed on your device and your transcriptions stay on your device. Cloud features exist, but every one of them is optional and clearly marked.","paragraph2Prefix":"This policy explains exactly what data OpenWhispr collects, where it lives, who else sees it, and how to control or delete it. For a plain-English summary of how we handle your data and what we never do with it, see our","securityLink":"Security page","paragraph2Suffix":"."},"dataGlance":{"title":"Data at a Glance","intro":"A scannable summary of what we collect and where it goes. Full descriptions follow below.","headerCategory":"Category","headerWhere":"Where it lives","headerWho":"Who else sees it","headerOptional":"Optional?","rows":[{"category":"Account & billing","where":"Our cloud database","who":"Neon, Stripe","optional":"Required for paid plans"},{"category":"Voice & audio","where":"Device only by default","who":"OpenWhispr Cloud subprocessors, or your own provider (BYOK)","optional":"You pick the mode"},{"category":"Transcriptions","where":"Device SQLite","who":"No one (not synced)","optional":"—"},{"category":"Notes & agent conversations","where":"Device SQLite; cloud if synced","who":"Neon (if synced); OpenWhispr Cloud AI subprocessors, or your own (BYOK)","optional":"Sync & agent are opt-in"},{"category":"Google Calendar","where":"Device only","who":"Google","optional":"Optional connection"},{"category":"Meeting detection signals","where":"Device memory only","who":"No one","optional":"Can disable"},{"category":"Diagnostic data","where":"Our servers if enabled","who":"No one","optional":"Off by default"},{"category":"Website analytics & ads","where":"Your browser (cookies)","who":"Google (Analytics 4, Ads)","optional":"Opt-in (EEA/UK), opt-out (US)"},{"category":"B2B visitor identification","where":"Your browser (script)","who":"RB2B","optional":"US visitors only; on by default (opt-out, GPC honored)"},{"category":"Referral program","where":"Our servers","who":"No one","optional":"Optional"}]},"collect":{"title":"Information We Collect","account":{"title":"Account & billing","intro":"When you create an OpenWhispr account, we collect the minimum needed to manage it:","emailLabel":"Email/password sign-up:","emailBody":"email, name, and a hashed password (we never store your password in plain text)","googleLabel":"Google sign-in:","googleBody":"email, name, and basic profile from Google. We never receive your Google password.","paidLabel":"Paid plans:","paidBody":"billing details processed by Stripe. We store your customer ID and subscription state, never your full card number.","sessionCookies":"Session cookies for authentication and automatic session refresh","websiteCookiesPrefix":"Website cookies for analytics and advertising, subject to your cookie choices (opt-in in the EEA/UK, opt-out in the US). See our","cookiePolicyLink":"Cookie Policy","websiteCookiesSuffix":"."},"voice":{"title":"Voice, audio, and transcriptions","intro":"OpenWhispr offers three transcription modes. You pick the one that fits your needs and you can change at any time in settings.","localLabel":"Local processing (default).","localBody":"OpenWhispr can run speech-to-text fully on your device using local open-source models. When you choose this mode, no audio is transmitted anywhere.","byokLabel":"Bring your own API key.","byokBody":"If you provide your own API key for a third-party transcription provider, your audio is sent directly to that provider for processing under your account. Their privacy policy applies. We don't see, store, or proxy that audio.","cloudLabel":"OpenWhispr Cloud.","cloudBodyPrefix":"Our managed transcription service. Audio is sent to our API, processed in real time, and discarded. Neither we nor any provider behind our API trains on your data. See our","securityLink":"Security page","cloudBodySuffix":"for the current provider list and training opt-outs.","retentionLabel":"Audio Retention (Optional, off by default).","retentionBody":"If you enable this in Settings, recorded audio is saved to your device for up to 30 days so you can review or re-process it. You set the retention period; OpenWhispr deletes files automatically once they pass it. Audio Retention is local-only and is never synced to OpenWhispr Cloud or any third party.","transcribedLabel":"Transcribed text","transcribedBody":"is stored in a local SQLite database on your device. Transcriptions are not synced to OpenWhispr Cloud — only notes and agent conversations sync (see below).","dictionaryLabel":"Custom dictionary.","dictionaryBody":"If you add words or phrases (technical terms, names, jargon) to improve transcription accuracy, they're stored locally and passed to your transcription provider as a hint. They're not used for anything else."},"notes":{"title":"Notes and AI agent conversations","paragraph1":"Notes are stored on your device in a local SQLite database. A note can contain a title, body content, AI-enhanced versions (if you use the enhancement feature), and — for meeting notes — the meeting transcript and participant list pulled from your calendar invite.","syncLabel":"Cloud sync (Optional).","syncBody":"If you enable sync, your notes and agent conversation history are stored in our cloud database (Neon, encrypted at rest) so you can access them across devices. Raw audio and transcriptions don't sync — only notes and conversations do. You can disable sync at any time.","agentLabel":"The AI Agent (Optional).","agentBody":"When you address the AI agent by name, your prompt and any tool results (notes it reads, calendar events it queries, web searches, clipboard contents you ask it to handle) are sent to whichever AI provider you've configured. If you choose to run the agent against a local model on your own device, your prompts and any retrieved context never leave your machine.","searchLabel":"Local semantic search.","searchBody":"Notes are automatically indexed for semantic search using a small embedding model that runs entirely on your device. The embeddings never leave your machine."},"calendar":{"title":"Google Calendar (Optional)","intro":"Connecting Google Calendar is optional. If you do, OpenWhispr reads your calendar list and upcoming events (read-only — we never create, edit, or delete anything) and uses that data to:","use1":"Notify you when a scheduled meeting is about to start","use2":"Suggest a title for your meeting note based on the event name","use3":"Attach the calendar invite participants to that note","cached":"Upcoming events are cached locally alongside your notes and transcriptions. This data stays on your device — we don't send it to our servers, share it, or use it for ads or AI training.","disconnectPrefix":"Disconnect from OpenWhispr settings to delete the local cache, or revoke access anytime via your","permissionsLink":"Google Account permissions","disconnectSuffix":".","limitedUseLabel":"Limited Use.","limitedUsePrefix":"OpenWhispr's use and transfer of information received from Google APIs to any other app adheres to the","userDataPolicyLink":"Google API Services User Data Policy","limitedUseSuffix":", including the Limited Use requirements."},"detection":{"title":"Meeting detection signals","intro":"OpenWhispr can detect when a meeting is happening so it can offer to take a note. To do this it monitors two things on your device only:","signal1":"Whether known meeting apps (Zoom, Teams, Webex, FaceTime) are running on your computer","signal2":"Whether any application is currently using your microphone (on macOS, Windows, and Linux)","body":"OpenWhispr does not capture what those other apps are doing, what they're recording, or who's on the call. It only knows the binary fact that \"a meeting app is open\" or \"the mic is in use.\" These signals are held in memory only, never written to disk, never sent anywhere. You can disable meeting detection in Settings.","systemAudioLabel":"System audio capture (macOS 14.2+, Optional).","systemAudioBody":"On recent versions of macOS, OpenWhispr can record both your microphone and the audio coming from other apps so it can transcribe both sides of a meeting. This requires you to explicitly grant macOS's \"System Audio Recording\" permission. The captured audio is treated exactly like microphone audio: processed in real time and discarded unless you have Audio Retention enabled. This feature is macOS-only today."},"accessibility":{"title":"Accessibility-based features","intro":"The first time you use OpenWhispr to paste a transcription, it asks for your operating system's accessibility permission. We use this for two things:","autopasteLabel":"Auto-paste:","autopasteBody":"pastes your transcribed text into whatever text field you have focused, so you don't have to copy and paste manually.","correctionLabel":"Correction learning:","correctionBody":"for up to 30 seconds after a paste, OpenWhispr watches the focused text field for edits you make to the text it pasted. If you fix a misspelled word, OpenWhispr learns the correction so future transcriptions are more accurate. The captured edits stay on your device and are never sent to OpenWhispr or any third party.","revoke":"You can revoke accessibility permission in your OS settings at any time, which disables both features."},"diagnostic":{"title":"Diagnostic & technical data (Optional)","intro":"Diagnostic reporting is off by default. If you opt in, we collect anonymous performance metrics to help us improve quality and reliability:","metric1":"Audio processing metrics (duration, format)","metric2":"Speech-to-text metrics (provider, model, processing time, word count)","metric3":"LLM processing metrics (provider, model, token counts)","metric4":"Round-trip timing and error information","metric5":"Client information (client type, version, app version)","body":"Diagnostic data may be submitted anonymously without any user or session identifier attached. It is used solely to improve performance and reliability, and is never sold or shared. You can opt out at any time in your application settings."},"referral":{"title":"Referral program (Optional)","intro":"If you participate in the referral program, we collect:","item1":"Your unique referral code and the invitations you send","item2":"The email addresses of people you invite (uploaded to our servers so we can track conversions and prevent abuse)","item3":"Click analytics on your referral link (user agent, referring URL)","item4":"Whether a referral results in a sign-up or subscription","body":"We only email a recipient when you initiate the invite. Referral data is used to attribute rewards and prevent abuse — nothing else."}},"use":{"title":"How We Use Your Information","intro":"We use the data we collect for two purposes only:","delivery":{"title":"Service delivery","item1":"Provide transcription, notes, and agent features","item2":"Process payments and manage subscriptions","item3":"Deliver software updates","item4":"Provide customer support"},"improvement":{"title":"Product improvement","item1":"Analyze opt-in diagnostic logs to improve quality","item2":"Fix bugs and improve performance","item3":"Develop new features","item4":"Ensure security and stability"},"neverLabel":"We never","neverBody":"use your transcribed content, notes, or audio for advertising, marketing to third parties, or training our own AI models. Period."},"sharing":{"title":"How We Share Information","intro":"We do not sell, trade, or rent your personal information. We share information only in the limited circumstances below.","subprocessors":{"title":"Subprocessors","intro":"These are the vendors we rely on to operate OpenWhispr Cloud and our core account features. Each is governed by its own privacy policy and bound by our Data Processing Addendum. We give at least 30 days' advance notice before adding or replacing subprocessors that process customer data, and will update this table to reflect changes.","headerService":"Service","headerPurpose":"Purpose","headerDataCategory":"Data category","rows":[{"service":"Neon","purpose":"Account database & cloud sync storage","dataCategory":"Account info, synced notes & conversations"},{"service":"Stripe","purpose":"Payment processing & subscription management","dataCategory":"Billing details, customer email"},{"service":"Resend","purpose":"Transactional email delivery","dataCategory":"Email addresses, message contents"},{"service":"Vercel","purpose":"Website hosting & cookieless product analytics","dataCategory":"Web request logs, anonymized page metrics"},{"service":"Google (Tag Manager, Analytics 4, Ads)","purpose":"Website analytics & advertising measurement (consent-gated)","dataCategory":"Pseudonymous usage & device/cookie identifiers, per your region and cookie choices (see Cookie Policy)"},{"service":"RB2B","purpose":"B2B website-visitor identification for sales outreach (consent-gated, US visitors only)","dataCategory":"Device & network identifiers matched to public professional profiles (name, company, work email), per your cookie choices (see Cookie Policy)"},{"service":"Google","purpose":"OAuth sign-in; Calendar API (if connected)","dataCategory":"Email + profile; calendar list & events"},{"service":"OpenWhispr API (api.openwhispr.com)","purpose":"Our backend: streaming token broker, referral tracking, contact form, sync endpoints","dataCategory":"Varies by feature"},{"service":"OpenAI","purpose":"Cloud speech-to-text & language models behind OpenWhispr Cloud (our API key)","dataCategory":"Voice audio (transient), prompts, text"},{"service":"Groq","purpose":"Cloud AI inference & transcription failover behind OpenWhispr Cloud (our API key)","dataCategory":"Voice audio (transient), prompts, text"},{"service":"Parasail","purpose":"LLM inference behind OpenWhispr Cloud agent and reasoning features","dataCategory":"Prompts, conversation context, tool results"}]},"byok":{"title":"Third-party AI providers you connect (BYOK)","intro":"You can also connect these providers directly with your own API key. On that BYOK path your data goes straight to the provider under your account, OpenWhispr is not the data controller for what you send them, and their privacy policy and account settings apply. Some of these providers (for example OpenAI and Groq) also operate as OpenWhispr Cloud subprocessors under our own keys — see the Subprocessors table above; the BYOK path is separate and applies only when you supply your own key.","transcriptionLabel":"Cloud transcription:","transcriptionBody":"OpenAI Whisper, Deepgram, AssemblyAI, Groq, Mistral, custom endpoints","agentLabel":"AI agent / reasoning:","agentBody":"OpenAI, Anthropic, Google Gemini, custom endpoints"},"local":{"title":"Local processing (no third party involved)","body":"OpenWhispr ships with optional on-device speech-to-text and AI models. When you choose to use these, no audio, transcription, or agent prompt is transmitted to OpenWhispr or any third party — the processing happens entirely on your computer. There is no subprocessor involved because no data leaves your device."},"legal":{"title":"Legal & business transfers","body":"We may disclose information if required by law, court order, or to protect our rights and safety. In the event of a merger, acquisition, or sale, your information may be transferred to the new owner under terms consistent with this policy."}},"retention":{"title":"Data Retention & Deletion","localLabel":"Local data","localBody":"(notes, transcriptions, agent conversations) lives on your device until you delete it. Uninstalling OpenWhispr removes the app data directory.","audioLabel":"Audio Retention (if enabled)","audioBody":"auto-deletes recorded audio files after the period you set (up to 30 days).","cloudLabel":"Cloud-synced data","cloudBody":"is deleted within 30 days of account deletion.","calendarLabel":"Google Calendar local cache","calendarBody":"is deleted immediately when you disconnect your Google account from OpenWhispr settings.","diagnosticLabel":"Diagnostic data","diagnosticBody":", when opted in, is retained only as long as it is useful to investigate the relevant issue."},"security":{"title":"Security","item1":"All cloud connections use TLS 1.2 or higher.","item2":"Our cloud database (Neon) is encrypted at rest.","item3":"Account passwords are hashed with industry-standard algorithms via our identity provider.","apiKeysLabel":"API keys","apiKeysPrefix":"you enter for third-party providers are currently stored in plaintext in your local app data directory. We're migrating to OS-native secure storage —","apiKeysLink":"track progress on issue #532","apiKeysSuffix":".","diskEncryption":"Your local SQLite database is unencrypted by OpenWhispr — it relies on your operating system's disk encryption (FileVault on macOS, BitLocker on Windows, LUKS on Linux). We recommend you keep disk encryption enabled.","morePrefix":"For more on our security practices, see our","securityLink":"Security page","moreSuffix":"."},"rights":{"title":"Your Rights and Choices","intro":"You have the following rights regarding your personal information:","accessLabel":"Access & portability:","accessBody":"request a copy of your personal information and export your notes and transcriptions from Settings → Data Export.","correctionLabel":"Correction:","correctionBody":"update or correct your personal information at any time through your account settings.","deletionLabel":"Deletion:","deletionBody":"delete individual notes, clear all local data, or delete your account entirely from within the app.","optoutLabel":"Opt-out:","optoutBody":"disable cloud features, unsubscribe from communications, or delete your account at any time."},"international":{"title":"International Users","intro":"OpenWhispr is operated from the United States. By using our services, you acknowledge that your information may be processed in the US.","gdpr":{"title":"For EU users (GDPR)","body":"You have additional rights under GDPR, including the right to object to processing, restrict processing, and lodge complaints with supervisory authorities."},"ccpa":{"title":"For California users (CCPA)","body":"You have the right to know what personal information is collected, request deletion, and opt out of the sale of personal information (we don't sell personal information)."}},"changes":{"title":"Changes to This Policy","intro":"We may update this Privacy Policy from time to time. We will notify you of material changes by:","item1":"Posting the updated policy on our website","item2":"Sending email notification to registered users","item3":"Displaying a notice in the OpenWhispr application","body":"Your continued use of OpenWhispr after changes become effective constitutes acceptance of the updated policy."},"contact":{"title":"Contact Us","intro":"If you have questions, concerns, or requests regarding this Privacy Policy, reach out to us:","teamName":"OpenWhispr Team","emailLabel":"Email:","subjectLine":"Subject line: Privacy Inquiry","responsePrefix":"We aim to respond to privacy-related inquiries within 48 hours. For a plain-English summary of how we handle your data and what we never do with it, see our","securityLink":"Security page","responseSuffix":"."},"legalNotice":"The English version of this document is the authoritative version and controls if any translation differs. Localized versions are provided for convenience and should be reviewed with counsel before relying on them."},"legalTerms":{"metadata":{"title":"Terms of Service | OpenWhispr","description":"Terms governing your use of OpenWhispr software and services. Open source licensing, subscriptions, and data ownership.","keywords":["OpenWhispr terms of service","OpenWhispr terms","terms and conditions","software license","subscription terms","data ownership"]},"breadcrumb":{"home":"Home","terms":"Terms of Service"},"header":{"title":"Terms of Service","subtitle":"The agreement between you and OpenWhispr.","lastUpdated":"Last updated: July 23, 2026"},"overview":{"title":"Overview","p1":"These Terms of Service (\"Terms\") are a contract between Gizmo Labs Inc., a Delaware corporation doing business as OpenWhispr (\"OpenWhispr\", \"we\", \"us\", or \"our\"), and you (\"Customer\", \"you\", or \"your\"). They govern your use of OpenWhispr's software, services, and website.","p2":"By using OpenWhispr you accept these Terms. Our handling of your personal data is governed by our Privacy Policy, which is incorporated into these Terms by reference."},"definitions":{"title":"Definitions","intro":"For clarity, the following terms have specific meanings throughout this document:","service":"Service — the OpenWhispr desktop application, the OpenWhispr Cloud API, the openwhispr.com website, and any related software or service we provide.","customer":"Customer / you — the individual or organization using the Service.","account":"Account — your registered OpenWhispr account.","customerContent":"Customer Content — audio recordings, transcriptions, notes, prompts, files, and other content you create, upload, or process through the Service.","customerData":"Customer Data — Customer Content plus any account information about you or your organization.","subscription":"Subscription — a paid plan you've signed up for.","subprocessor":"Subprocessor — a third-party service we use to operate the Service. The current list lives in our Privacy Policy."},"service":{"title":"The Service","intro":"OpenWhispr is a desktop dictation, transcription, and meeting-notes tool with an optional AI agent and Google Calendar integration. You choose how your audio is processed:","local":"Local processing (default). Speech-to-text runs entirely on your device using local open-source models. No audio leaves your computer.","byok":"Bring your own API key. You connect your own API key from a third-party transcription or AI provider. Your audio goes directly to that provider. See §6 for details.","cloud":"OpenWhispr Cloud. Our managed transcription service. Audio is sent to our API, processed in real time, and discarded. See our Security page for the current provider list and training opt-outs.","openSource":"Open-source desktop app. The OpenWhispr desktop application is released under the MIT License. Full source is available on GitHub. The OpenWhispr Cloud API and the openwhispr.com website are proprietary and operated by us.","noSla":"No service level agreement (SLA). OpenWhispr does not currently offer an SLA for the Service. We aim to provide reliable service but make no uptime commitments. Enterprise customers who need an SLA can contact sales@openwhispr.com for an addendum."},"account":{"title":"Your Account","intro":"You can create an OpenWhispr account with email and password or by signing in with Google. Some Service features require an account.","bullet1":"One account per individual or organization","bullet2":"Keep your credentials safe. What happens under your account is your responsibility.","bullet3":"Tell us right away if you suspect unauthorized access","bullet4":"We may suspend or close accounts for serious violations of these Terms or after extended inactivity","age":"Age requirement. OpenWhispr is intended for users aged 13 and over. If you are under 18, you confirm that your parent or legal guardian has read these Terms and agrees to them on your behalf."},"acceptableUse":{"title":"Acceptable Use","intro1":"OpenWhispr's desktop application is open source under the MIT License — anyone can use, modify, and distribute it freely under that license. The acceptable use rules below apply when you use the proprietary parts of the Service (OpenWhispr Cloud, the website, our APIs), and when applicable law restricts what you may do with any version.","intro2":"Use OpenWhispr for lawful purposes only. You may not use the Service to:","bullet1":"Record any individual without their knowledge and consent where required by law (see §8 — recording-consent obligations are on you, not us)","bullet2":"Transcribe content you don't have the legal right to record or process (copyrighted broadcasts, confidential calls you're not party to, etc.)","bullet3":"Generate, distribute, or store content that is illegal, harmful, defamatory, harassing, hateful, or that infringes on others' rights","bullet4":"Create deepfakes, voice clones, or any content intended to deceive or impersonate","bullet5":"Reverse engineer, decompile, or attempt to extract the source code of the proprietary parts of the Service. The desktop app is MIT-licensed and you can read all of it on GitHub; the OpenWhispr Cloud API and website are not.","bullet6":"Use outputs generated by the OpenWhispr Cloud API to train, develop, or improve a competing artificial intelligence or machine learning product. This restriction applies only to OpenWhispr Cloud — the MIT-licensed desktop application has no such restriction.","bullet7":"Bypass rate limits, usage caps, or access controls","bullet8":"Spam, scrape, or use automated tools to access the Service in a way that disrupts other users","bullet9":"Resell or sublicense access to OpenWhispr Cloud without our written permission","outro":"If you misuse OpenWhispr, we may suspend or close your account, and in serious cases pursue legal remedies."},"byok":{"title":"Bring Your Own Key (BYOK)","p1":"OpenWhispr lets you connect third-party transcription and AI providers (such as OpenAI, Anthropic, Google Gemini, Deepgram, AssemblyAI, Groq, Mistral, or a custom endpoint) by providing your own API key. The current list of supported providers lives in our Privacy Policy and may change over time.","p2":"When you use BYOK, your audio and prompts go directly to the provider you've chosen. We don't see, store, or proxy that traffic. Your relationship is with that provider — and so is theirs with you:","bullet1":"You're bound by the third-party provider's terms in addition to ours","bullet2":"You're responsible for your own API costs, rate limits, data handling, and compliance with the underlying provider's policies","bullet3":"We're not liable for the actions, errors, downtime, or charges of any third-party provider you connect"},"aiOutputs":{"title":"AI Outputs and User Verification","p1":"The Service uses machine learning to generate transcriptions, AI-enhanced notes, agent responses, and other outputs. Like all AI systems, OpenWhispr can produce results that are inaccurate, incomplete, biased, or completely fabricated (\"hallucinations\").","p2":"You are responsible for reviewing and verifying AI-generated content before relying on or acting on it. We don't warrant the accuracy or completeness of any AI output.","p3":"AI outputs are not professional advice. Don't use them as the sole basis for legal, medical, financial, safety-critical, or other high-stakes decisions."},"customerContent":{"title":"Customer Content and Recording Consent","ownTitle":"You own your Customer Content","ownBody":"OpenWhispr claims no ownership of your audio recordings, transcriptions, notes, or any content you create through the Service. It's yours.","licenseTitle":"Limited license to operate the Service","licenseBody":"You grant us a limited license to process your Customer Content as needed to provide the Service — for example, to transcribe audio you submit, store notes you save, or sync content across your devices. This license is limited to operating the Service for you and ends when you delete the content or close your account.","noTrainingTitle":"No training on your data","noTrainingBody":"We will not use Customer Content — audio, transcriptions, notes, or AI agent prompts — to train any artificial intelligence or machine learning model. This is a contractual commitment, not just a privacy promise.","consentTitle":"Recording consent is your responsibility","consentBody":"You represent and warrant that you have the legal right to record, transcribe, and process every audio source you use OpenWhispr on. You are solely responsible for complying with applicable wiretapping, two-party consent, and recording-disclosure laws in every jurisdiction where you or any other participant is located.","indemnity":"You will indemnify OpenWhispr against any claim arising from Customer Content you create, upload, or process — including claims that you recorded someone without consent (see also §12 on Indemnification)."},"ip":{"title":"Intellectual Property","desktop":"Desktop app — released under the MIT License. Full source on GitHub.","cloud":"OpenWhispr Cloud and openwhispr.com — proprietary, licensed (not sold) to you under these Terms.","customerContent":"Customer Content — yours (see §8).","trademarks":"Trademarks — the OpenWhispr name, logo, and related marks belong to Gizmo Labs Inc. You may not use them without our written permission.","feedback":"Feedback license — if you suggest features, report bugs, or otherwise share feedback with us, we may use that feedback freely without compensation or attribution."},"billing":{"title":"Plans, Billing, and Refunds","freeTier":"Free tier — includes 2,000 words per week of OpenWhispr Cloud transcription.","paidPlans":"Paid plans — billed monthly or annually in advance. Plans auto-renew unless you cancel before the next billing date.","pricingChanges":"Pricing changes — we'll give 30 days notice before any price change takes effect.","processor":"Payment processor — Stripe handles all payments. We never see your full card number.","failedPayments":"Failed payments — if a payment fails, we may suspend the Service after a grace period.","taxes":"Taxes — you're responsible for any applicable taxes.","guarantee":"30-day money-back guarantee. If you're not satisfied with a new paid subscription, contact support@openwhispr.com within 30 days of your first payment and we'll refund you. Refunds are processed to your original payment method within 5–7 business days. The guarantee applies to first-time subscriptions only — not to renewals or custom enterprise work."},"privacy":{"title":"Privacy","body":"Our handling of your personal data is governed by our Privacy Policy, which is incorporated into these Terms by reference. The Privacy Policy describes what data OpenWhispr collects, how we use it, who else processes it, and how to control or delete it. A Data Processing Addendum is available on request — contact support@openwhispr.com."},"liability":{"title":"Disclaimers, Liability, and Indemnification","asIsTitle":"As-is disclaimer","asIsBody":"The Service is provided \"as is\" and \"as available\", without warranties of any kind, express or implied — including any warranties of merchantability, fitness for a particular purpose, or non-infringement. We don't warrant transcription accuracy, AI output accuracy (see §7), or uninterrupted availability.","betaBody":"Beta or preview features are provided as-is, may change or be removed without notice, and are excluded from any warranties or service commitments.","limitTitle":"Limitation of liability","limitBody":"To the maximum extent permitted by law, our total liability arising out of or related to these Terms or the Service is capped at the greater of (a) the fees you paid to us in the 12 months preceding the claim, or (b) USD 100. We are not liable for indirect, incidental, consequential, special, or punitive damages — including lost profits, lost data, or business interruption.","indemnityTitle":"Indemnification","indemnityByYou":"By you. You will indemnify, defend, and hold Gizmo Labs Inc. harmless against any third-party claim arising from (a) your Customer Content, (b) your misuse of the Service, (c) your violation of these Terms or any law, (d) your unauthorized recording or transcription of any individual (see §8), or (e) any third-party AI provider you connect via BYOK (see §6).","indemnityByUs":"By us. We will indemnify, defend, and hold you harmless against any third-party claim that the Service itself, when used in accordance with these Terms, infringes a US patent, copyright, or trademark. This obligation excludes claims arising from your modifications, your combination of the Service with non-OpenWhispr software, beta features, BYOK providers, or Customer Content. Our total indemnification obligation is capped at the limitation of liability above."},"termination":{"title":"Term, Termination, and Changes","terminationTitle":"Termination","terminationBody":"You can stop using OpenWhispr at any time by closing your account or uninstalling the desktop app. We may suspend or terminate your access for violations of these Terms, non-payment, or extended inactivity — with reasonable notice except in cases of serious breach.","dataTitle":"What happens to your data","dataBody":"Local data on your device stays under your control until you delete it. Uninstalling OpenWhispr removes the local app data directory. Cloud-synced data is deleted within 30 days of account deletion (matching our Privacy Policy).","surviving":"Surviving sections. The following provisions survive termination: Customer Content ownership (§8), intellectual property (§9), payment obligations (§10), disclaimers and liability limits (§12), indemnification (§12), governing law (§14), and these survival rules.","changesTermsTitle":"Changes to these Terms","changesTermsBody":"We may update these Terms from time to time. We'll notify you of material changes by email and/or an in-app notice at least 30 days before they take effect. Your continued use of OpenWhispr after the effective date means you accept the updated Terms.","changesServiceTitle":"Changes to the Service","changesServiceBody":"We may add, modify, or discontinue features at any time. For material changes that reduce functionality you depend on, we'll give reasonable notice. Beta features may change or be removed at any time without notice."},"legal":{"title":"Legal & General Provisions","governingTitle":"Governing law and venue","governingBody":"These Terms are governed by the laws of the State of Delaware, United States, without regard to its conflict-of-laws principles. Any legal action must be brought in the state or federal courts located in Delaware (the Court of Chancery for matters of corporate law; the U.S. District Court for the District of Delaware for federal matters).","disputeBody":"Before filing any legal action, you agree to first contact support@openwhispr.com to try to resolve the dispute informally.","exportTitle":"Export controls","exportBody":"You may not use the Service in violation of US export controls, OFAC sanctions, or in any embargoed country (currently Cuba, Iran, North Korea, Syria, and the Crimea, Donetsk, and Luhansk regions of Ukraine).","forceMajeureTitle":"Force majeure","forceMajeureBody":"Neither party is liable for delays or failures caused by events beyond reasonable control — natural disasters, war, pandemics, internet outages, third-party service failures, or government actions.","dmcaTitle":"DMCA","dmcaBody":"If you believe content on the Service infringes your copyright, send a DMCA takedown notice to support@openwhispr.com with the subject line \"DMCA Notice\", including the information required by 17 U.S.C. § 512(c)(3). We respond to valid notices in accordance with the DMCA.","confidentialityTitle":"Confidentiality","confidentialityBody":"Each party agrees to protect the other's non-public business information shared during the relationship and not to disclose it without permission, except as required by law.","publicityTitle":"Publicity","publicityBody":"We may identify you as a customer in marketing materials unless you opt out by emailing support@openwhispr.com.","noticesTitle":"Notices","noticesBody":"We notify you by email (to the address on your account) or through an in-app notice. You notify us by email to support@openwhispr.com.","otherTitle":"Other provisions","severability":"Severability — if any provision of these Terms is found unenforceable, the rest remains in full force.","assignment":"Assignment — you may not assign these Terms without our written consent. We may assign these Terms in connection with a merger, acquisition, or sale of all or substantially all of our assets.","waiver":"Waiver — our failure to enforce any provision doesn't waive our right to enforce it later.","entireAgreement":"Entire agreement — these Terms plus the Privacy Policy (incorporated by reference) constitute the entire agreement between you and OpenWhispr regarding the Service, and supersede any prior agreements."},"contact":{"title":"Contact Us","intro":"Questions about these Terms? Get in touch:","company":"Gizmo Labs Inc.","dba":"doing business as OpenWhispr","email":"Email: support@openwhispr.com","subject":"Subject line: Terms Inquiry","seeAlso":"See also our Privacy Policy and Security page."},"legalNotice":"The English version of this document is the authoritative version and controls if any translation differs. Localized versions are provided for convenience and should be reviewed with counsel before relying on them."},"legalCookies":{"metadata":{"title":"Cookie Policy | OpenWhispr","description":"How OpenWhispr uses cookies. Analytics and marketing cookies are off by default and load only with your consent."},"breadcrumb":{"home":"Home","cookies":"Cookie Policy"},"hero":{"title":"Cookie Policy","subtitle":"How openwhispr.com uses cookies, and how you stay in control.","lastUpdated":"Last updated: July 5, 2026"},"whatThisCovers":{"title":"What this covers","body":"This covers cookies and similar technologies on openwhispr.com. It complements our Privacy Policy. The OpenWhispr desktop app processes audio locally by default and is covered there."},"ourApproach":{"title":"Our approach","body":"OpenWhispr is privacy-first. We use Google Consent Mode (Advanced): Google's analytics and ad tags load on every page, but they only set cookies based on your region and choice (see By region below). When cookies aren't allowed yet, the tags still send Google an anonymous, cookieless signal — including your IP address and that consent was denied — used only for aggregate, non-identifying measurement (this goes to Google in the US). Our usage metrics via Vercel are cookieless and don't identify you, so they run either way. You can change your choice anytime from Cookie Settings in the footer."},"categories":{"title":"Cookie categories","headerCategory":"Category","headerWhatItDoes":"What it does","headerConsent":"Consent required?","rows":[{"category":"Strictly necessary","whatItDoes":"Sign-in, session security, your region, and remembering your cookie choice.","consent":"No — always active"},{"category":"Analytics","whatItDoes":"Aggregate usage measurement via Google Analytics 4.","consent":"Yes in the EEA/UK; on by default (opt-out) in the US"},{"category":"Marketing","whatItDoes":"Measuring our advertising via Google Ads and Meta, remembering where your visit came from, and identifying US business visitors via RB2B so we can follow up.","consent":"Yes in the EEA/UK; on by default (opt-out) in the US (GPC opts you out)"}]},"cookies":{"title":"Cookies we may set","headerCookie":"Cookie","headerProvider":"Provider","headerPurpose":"Purpose","headerType":"Type","headerDuration":"Duration","rows":[{"cookie":"ow_consent","provider":"OpenWhispr","purpose":"A copy of your cookie choices and the policy version (read before tags load)","type":"First-party","duration":"12 months"},{"cookie":"ow-geo","provider":"OpenWhispr","purpose":"Your region, to apply the right defaults and notice","type":"First-party","duration":"1 day"},{"cookie":"Session cookie","provider":"OpenWhispr (Better Auth)","purpose":"Keeps you signed in","type":"First-party","duration":"Per session"},{"cookie":"ow_attribution","provider":"OpenWhispr","purpose":"Where your visit came from (ad click ids, campaign tags, referrer), so we know which channel a signup came from. Marketing consent only.","type":"First-party","duration":"90 days"},{"cookie":"_ga, _ga_","provider":"Google (Analytics 4)","purpose":"Tells repeat visits apart so usage can be counted","type":"Third-party","duration":"Up to 24 months"},{"cookie":"_gcl_au","provider":"Google Ads","purpose":"Conversion and ad measurement","type":"Third-party","duration":"Up to 90 days"},{"cookie":"_fbc, _fbp","provider":"Meta","purpose":"Ad click and browser ids for measuring our Meta advertising (only if we run Meta ads)","type":"Third-party","duration":"Up to 90 days"},{"cookie":"RB2B identifiers","provider":"RB2B","purpose":"Matches US business visitors to their public professional profile (name, company, work email) so we can reach out. US visitors only. Marketing consent only.","type":"Third-party","duration":"Set by RB2B's script; see RB2B's privacy policy"}],"note":"Google Tag Manager sets no cookies itself. The third-party cookies above appear only after a Google, Meta, or RB2B tag fires — after you consent in the EEA/UK, or by default in opt-out regions like the US (unless you opt out). RB2B additionally fires for US visitors only, never elsewhere. Even with cookies off, Google's tags may send an anonymous, cookieless signal (including your IP and consent status) for aggregate measurement only. We save your choice in the ow_consent cookie (a slim copy, read before tags load) and in full — your choices, a random non-identifying id, time, and region — in your browser's local storage. We keep no copy on our servers, so clearing your browser asks you again."},"managing":{"title":"Managing your choices","body":"Use Cookie Settings to accept, reject, or fine-tune your choices anytime; your browser can also block or delete cookies. If your browser sends a Global Privacy Control (GPC) signal, we treat it as an opt-out from ad cookies — use Cookie Settings to also turn off analytics.","button":"Manage cookie preferences"},"byRegion":{"title":"By region","items":["EEA & UK: no analytics or ad cookies until you consent, and you can withdraw anytime. (Tags load first to send an anonymous, cookieless signal — see “Our approach” — but store nothing on your device until you agree.)","United States: analytics and ad cookies are on by default; you can opt out anytime. Use Your Privacy Choices in the footer or send a Global Privacy Control (GPC) signal — GPC turns off ad cookies; use Cookie Settings to also turn off analytics.","Elsewhere: we show a notice and you can adjust your preferences anytime."]},"changes":{"title":"Changes to this policy","body":"We'll update this page and the “Last updated” date when our cookie use changes. Questions? Email support@openwhispr.com."},"legalNotice":"The English version of this document is the authoritative version and controls if any translation differs. Localized versions are provided for convenience and should be reviewed with counsel before relying on them."},"changelog":{"metadata":{"title":"Changelog — What's New in OpenWhispr","description":"See every release, improvement, and bug fix shipped in OpenWhispr. Updated automatically from GitHub releases.","keywords":["OpenWhispr changelog","OpenWhispr releases","OpenWhispr updates","what's new OpenWhispr","OpenWhispr release notes"]},"breadcrumb":{"home":"Home","changelog":"Changelog"},"hero":{"title":"Changelog","subtitle":"Every release, improvement, and bug fix — pulled directly from our GitHub releases."},"empty":{"prefix":"No releases found. Check back soon or view releases on","githubLink":"GitHub","suffix":"."},"prerelease":"Pre-release","viewOnGitHub":"View on GitHub →"},"home":{"metadata":{"title":"OpenWhispr: Open-Source Dictation Software (Free Download)","description":"OpenWhispr is open-source dictation software you can rely on. Dictate, transcribe, and capture meeting notes with fast, accurate AI speech-to-text.","keywords":["voice to text","dictation app","speech to text","whisper AI","nvidia parakeet","open source dictation","voice dictation","AI transcription"]},"breadcrumb":{"home":"Home"},"video":{"name":"OpenWhispr Demo — AI Voice Dictation","description":"See OpenWhispr in action: press a hotkey, speak naturally, and watch your words appear instantly in any app. Powered by OpenAI Whisper and NVIDIA Parakeet."},"schema":{"name":"OpenWhispr","description":"Open source AI voice dictation. Turn your voice into text instantly with OpenAI Whisper and NVIDIA Parakeet. Private, cross-platform, and free.","offerDescription":"Free and open source core. Optional Pro plan available.","featureList":["Local speech-to-text via OpenAI Whisper and NVIDIA Parakeet","99+ language support","System-wide push-to-talk hotkey","AI agent mode for text cleanup","Fully open source","Cross-platform (macOS, Windows, Linux, iOS)","Offline capable"]},"hero":{"title":"Your voice is 3x faster than your keyboard.","subtitle":"Open source voice-to-text assistant. Private by design.\nWrite faster in every app.","availability":"Available on macOS, Windows, Linux and iOS"},"privacy":{"eyebrow":"Open source & privacy first","title":"Your data stays yours.","subtitle":"OpenWhispr is open source. Here’s how we handle your data:","principles":[{"headline":"Your voice never reaches us","description":"When you dictate locally, your words are transcribed on your own computer by open models. Nothing is sent to us or to anyone else."},{"headline":"Cloud is a choice, not a default","description":"Switch the cloud on and your audio goes out to be transcribed, then it's gone the moment the words come back. We never keep the recording."},{"headline":"We don't train on what you say","description":"Your transcriptions are not used to train models or improve the product. If that ever changes, it will be opt-in and we will ask first."},{"headline":"Don't take our word for it","description":"OpenWhispr is open source on GitHub under the MIT license. Every claim on this page is one you can go and check for yourself."}],"visuals":{"retention":"Data Retention","training":"Data Training","offByDefault":"Off by default","public":"Public","github":"GitHub","contributors":"+214 contributors"}},"socialProof":{"title":"What people are saying","description":"Real feedback from developers, writers, and teams using OpenWhispr."},"faqSection":{"title":"Common questions","description":"What people usually want to know before they download."},"cta":{"title":"Try it for free","descriptionLead":"Works on macOS, Windows, Linux and iOS.","descriptionEmphasis":"Open source and runs locally."},"faqs":[{"question":"How private is it? Can anyone see my transcriptions?","answer":"With local processing, your voice never leaves your device — it's 100% private. Even with cloud processing, audio is processed and immediately discarded. We never store, sell, or train on your data. OpenWhispr is open source so you can verify this yourself."},{"question":"Can I use my own API keys instead of OpenWhispr Cloud?","answer":"Absolutely. Bring your own keys for OpenAI, Anthropic, Google Gemini, Groq, or any compatible provider. You control your costs, your data relationship stays direct with the provider, and there are no limits beyond what your key allows."},{"question":"Is OpenWhispr really free?","answer":"Yes — the core app is open source and free forever. Local AI models run on your device with zero limits. Bring-your-own API keys also have zero limits. We also have our own built-in cloud service, OpenWhispr Cloud — every free account gets 2,000 words/week with no API keys or setup needed. Pro ($8/mo) removes that cap and adds mobile, meeting recording, and email support."},{"question":"What platforms and apps does it work with?","answer":"OpenWhispr runs on macOS, Windows, and Linux. It works with any application — Slack, VS Code, Chrome, Word, Notion, your terminal, literally anything that accepts text input. Just press your hotkey, speak, and the text appears wherever your cursor is."},{"question":"What languages are supported?","answer":"OpenWhispr supports 100+ languages including English, Spanish, French, German, Chinese, Japanese, Arabic, and many more. You can set a preferred language or use auto-detect to switch between languages on the fly."},{"question":"What's the difference between local and cloud processing?","answer":"Local processing runs AI models directly on your machine — fully private, works offline, but speed depends on your hardware. Cloud processing sends audio to a server for transcription — faster and more accurate on most machines, but requires an internet connection."},{"question":"How do I customize the hotkey?","answer":"Open Settings and you can set any key or key combination as your recording hotkey. Hold-to-record and push-to-toggle modes are both supported. On Windows, compound hotkeys like Ctrl+Shift+Space work too."},{"question":"Is it really open source? Can I contribute?","answer":"100%. OpenWhispr is MIT licensed — free for personal and commercial use. The full source code is on GitHub. We welcome contributions, bug reports, and feature requests from the community."}],"compliance":{"title":"Compliance & certifications","subtitle":"Everything your security team will ask for, already in place.","certifications":[{"status":"Compliant","title":"HIPAA","description":"Whether you use local or cloud processing, we keep your health information safe."},{"status":"Compliant","title":"AICPA SOC 2 & ISO 27001","description":"Independently verified. Whatever your security review needs to see, we already have it.","standards":["SOC 2 TYPE II","ISO 27001"]}]}},"homeBento":{"section":{"title":"What you get","subtitle":"Dictation, AI cleanup, offline mode, and 100+ languages."},"cards":{"models":{"headline":"Your machine. Your models.","description":"Run local speech-to-text models or bring your own API keys. All free."},"voiceInstructions":{"headline":"Give instructions by voice.","description":"Say things like 'clean this up' or 'draft an email to Mike' and it does exactly that."},"dictionary":{"headline":"It learns your words.","description":"Names, jargon, and terms the AI might miss — auto-learned from your corrections."},"languages":{"headline":"Speak any language.","descriptionPrefix":"100+ languages with auto-detection. Switch mid-conversation,","descriptionSuffix":"."},"apps":{"headline":"Works across all your apps.","description":"ChatGPT, Claude, Cursor, Slack, Docs — if it accepts text, it works."},"offline":{"headline":"Works offline.","description":"No wifi, no problem. OpenWhispr runs offline so you can transcribe anytime, anywhere."}},"modelPicker":{"speechToText":"Speech-to-Text","aiModels":"AI Models","cloud":"Cloud","local":"Local","custom":"Custom","endpointUrl":"Endpoint URL","apiKey":"API Key","optional":"(optional)","modelName":"Model name","model":"Model","autoDiscovered":"Auto-discovered from endpoint","openAiCompatibleNote":"Works with any OpenAI-compatible API — Ollama, LM Studio, vLLM, and more.","descriptions":{"noApiKeyNeeded":"No API key needed","mostAccurate":"Most accurate","fastAndAccurate":"Fast & accurate","originalWhisper":"Original Whisper","highAccuracy":"High accuracy","realtime216x":"216x realtime","multilingual":"Multilingual","fastest":"Fastest","fast":"Fast","balanced":"Balanced","accurate":"Accurate","best":"Best","highestQuality":"Highest quality","fastReasoning":"Fast reasoning","latestFlagship":"Latest flagship","fastAndEfficient":"Fast & efficient","context1M":"1M context","mostCapable":"Most capable","flagshipReasoning":"Flagship reasoning","ultraFast":"Ultra-fast","context131K":"131K context","throughput500":"500 T/sec","throughput280":"280 T/sec","recommended":"Recommended","compact":"Compact","lightweight":"Lightweight","powerful":"Powerful"}},"agentMode":{"youSaidLabel":"You said","youSaid":"\"hey OpenWhispr, clean up this email — I need to tell mike that the meeting is moved to thursday at 3 and also remind him about the deck\"","outputLabel":"Output","outputGreeting":"Hi Mike,","outputBody":"Quick heads up — our meeting has been moved to Thursday at 3 PM. Also, just a reminder to have the deck ready.","outputClosing":"Thanks!"},"dictionary":{"addPlaceholder":"Add a word or phrase...","autoLearns":"Auto-learns from your corrections"},"languages":{"more":"100+ more"}},"homeSocial":{"title":"What people are saying","description":"Real feedback from developers, writers, and teams using OpenWhispr.","translatedFromJapanese":"Translated from Japanese"},"pricing":{"metadata":{"title":"Pricing: Free Forever Tier Available - OpenWhispr","description":"OpenWhispr is free forever for local dictation and transcription. Upgrade anytime for cloud transcription, meeting hours and team tools.","keywords":["openwhispr pricing","openwhispr pricing plans","voice to text pricing","dictation app pricing","free dictation app","speech to text plans"],"ogImageAlt":"OpenWhispr pricing plans"},"breadcrumb":{"home":"Home","pricing":"Pricing"},"hero":{"title":"Free forever. Upgrade when you want more.","subtitle":"Unlimited local dictation is free for everyone. Paid plans add unlimited cloud transcription, more meeting recording hours, and team tools."},"billingToggle":{"ariaLabel":"Billing period","monthly":"Monthly","yearly":"Annual","savings":"2 months free"},"perUserMonth":"/user/mo","perMonthShort":"/mo","billedYearly":"$${total}/user billed annually","billedMonthly":"billed monthly","moreInfo":"More info","cta":{"download":"Download","getStarted":"Get Started","contactSales":"Contact Sales"},"tiers":{"free":{"subtitle":"For everyone","features":["## OpenWhispr Cloud","2,000 words/week transcription","5 hours of meeting recordings/month","## Always free","Unlimited local AI models {i}","Unlimited cloud transcription with your own API keys","100+ languages supported","Custom dictionary","Zero data retention","Community support"]},"pro":{"subtitle":"For individuals and teams","includesPrefix":"All Free features +","features":["Unlimited cloud transcription","20 hours of meeting recordings/month","Sync across devices","Personal API access","MCP integration in all your apps","iPhone & iPad app","Email support"]},"business":{"subtitle":"For individuals and teams","badge":"Most Popular","includesPrefix":"All Pro features +","features":["Unlimited meeting recordings","Agent mode","Chat over your data","Priority support"]},"enterprise":{"subtitle":"For advanced security & support","priceDisplay":"Custom","includesPrefix":"All Business features +","features":["Team features & admin","SSO, SAML & SCIM","Audit logs & retention controls","Dedicated support"]}},"featureInfo":{"Unlimited local AI models":{"title":"Included local models","sections":[{"label":"Speech-to-Text","items":["OpenAI Whisper (all sizes)","NVIDIA Parakeet"]},{"label":"AI Writing Assistant","items":["Qwen","Llama","Mistral","Gemma","GPT-OSS"]}]}},"compareSection":{"title":"Compare every plan","description":"Every feature, side by side. Local dictation is unlimited and free on every plan."},"enterpriseBanner":{"title":"Need SSO, compliance, or admin controls?","description":"Our Enterprise plan adds SSO & SAML, SCIM provisioning, audit logs, org-wide retention controls, custom data processing agreements, and dedicated support.","cta":"Talk to us"},"socialProof":{"title":"Loved by people who'd rather talk than type","description":"Real feedback from developers, writers, and teams using OpenWhispr."},"faqSection":{"title":"Pricing questions","description":"Straight answers about plans and billing."},"comparison":{"included":"Included","notIncluded":"Not included","selectPlanAriaLabel":"Select a plan to compare","caption":"Compare OpenWhispr plans feature by feature","plansHeader":"Plans","aboutLabel":"About {label}","groups":[{"title":"Speech-to-text","rows":[{"label":"Local AI models","tooltip":"Runs entirely on your device with no internet required — unlimited and free on every plan.","values":["Unlimited","Unlimited","Unlimited","Unlimited"]},{"label":"OpenWhispr Cloud transcription","tooltip":"Audio is sent to OpenWhispr Cloud for top accuracy and speed with no local models to download, then never stored.","values":["2,000 words/week","Unlimited","Unlimited","Unlimited"]},{"label":"Bring your own API keys","tooltip":"Use your own OpenAI, Anthropic, Gemini, or Groq keys for unlimited cloud transcription with no OpenWhispr limits.","values":["","","",""]},{"label":"Languages","values":["100+","100+","100+","100+"]},{"label":"Custom dictionary","tooltip":"Teach OpenWhispr your names, jargon, and acronyms for sharper transcripts.","values":["","","",""]}]},{"title":"Meetings & workflow","rows":[{"label":"Meeting recordings","values":["5 hrs/month","20 hrs/month","Unlimited","Unlimited"]},{"label":"Speaker labels","tooltip":"Automatically identify and label who said what — in meetings and uploaded audio.","values":["","","",""]},{"label":"Sync across devices","values":["","","",""]},{"label":"Mobile companion app","tooltip":"Dictate from your phone.","values":["","","",""]},{"label":"API & MCP access","tooltip":"Connect OpenWhispr to your own tools and agent workflows.","values":["","","",""]},{"label":"Command-line interface","tooltip":"Script OpenWhispr from your terminal. The local CLI drives the desktop app on every plan; paid plans add the remote CLI backed by the OpenWhispr Cloud API.","values":["Local","Local & remote","Local & remote","Local & remote"]},{"label":"Agent mode","tooltip":"Let AI take actions on your transcripts automatically.","values":["","","",""]},{"label":"Chat over your data","tooltip":"Ask questions across your notes, files, and transcripts.","values":["","","",""]}]},{"title":"Privacy & security","rows":[{"label":"Zero data retention","tooltip":"Cloud audio is transcribed then immediately discarded. We never store, sell, or train on your data.","values":["","","",""]},{"label":"Open source (MIT)","values":["","","",""]},{"label":"SSO & SAML","values":["","","",""]},{"label":"SCIM provisioning","tooltip":"Automatically provision and deprovision team members from your identity provider.","values":["","","",""]},{"label":"Audit logs","tooltip":"Exportable logs of account and admin activity for security and compliance reviews.","values":["","","",""]},{"label":"Org-wide retention controls","tooltip":"Admin-configurable data retention and auto-deletion policies across your organization.","values":["","","",""]},{"label":"Compliance & DPA","tooltip":"Custom data processing agreements and compliance reviews for your organization.","values":["","","",""]}]},{"title":"Team & support","rows":[{"label":"Team management & admin","values":["","","Basic","Advanced"]},{"label":"Support","values":["Community","Email","Priority","Dedicated"]},{"label":"Available on","values":["macOS · Windows · Linux","macOS · Windows · Linux · iOS","macOS · Windows · Linux · iOS","macOS · Windows · Linux · iOS"]}]}]},"faqs":[{"question":"Is the free plan really free forever?","answer":"Yes. Unlimited local dictation in 100+ languages on macOS, Windows, and Linux is $0 — no card, no trial clock. The Free plan also includes 2,000 OpenWhispr Cloud words a week. Paid plans are optional upgrades, not a countdown."},{"question":"What's the difference between local and cloud transcription?","answer":"Local transcription runs on your own device — fully private, works offline, and unlimited and free. Cloud transcription runs on OpenWhispr Cloud for top accuracy on any hardware. Free includes 2,000 cloud words a week; Pro, Business, and Enterprise make it unlimited."},{"question":"Can I pay yearly?","answer":"Yes. Pro is $8/month, or $80/year (two months free). Business is $16/user/month, or $160/user/year. Yearly billing works out cheaper per month."},{"question":"Do you offer student, military or non-profit discounts?","answer":"We do. If OpenWhispr is part of your studies, your classroom, your service, your non-profit's work, or how you get things done with an assistive tool, you qualify for 40% off for 12 months. Email support@openwhispr.com with a quick proof of status and we'll send your personal code.","linkText":"Full details in the Help Center","linkHref":"https://docs.openwhispr.com/help/account/discounts-and-promo-codes"},{"question":"Do I need a card to start?","answer":"No. Download and use the Free plan with no card and no account barrier. You only add payment details if and when you upgrade to Pro or Business."},{"question":"Can I change or cancel my plan anytime?","answer":"Yes. Upgrade, downgrade, or cancel whenever you like. If you cancel, you keep unlimited local dictation on the Free plan — your private, on-device transcription never stops working."},{"question":"Can I use my own API keys instead of paying?","answer":"Absolutely. On any plan you can bring your own OpenAI, Anthropic, Gemini, or Groq keys for unlimited cloud transcription with no OpenWhispr limits — you control your costs and your data relationship stays direct with the provider."},{"question":"Do you train AI models on my voice data?","answer":"No. Local transcription never leaves your device, and cloud audio is transcribed then immediately discarded. We never store, sell, or train on your data — and because OpenWhispr is open source, you can verify it yourself."},{"question":"How do you handle security and compliance?","answer":"Audio is encrypted in transit and processed with zero data retention on every plan. Enterprise adds SSO & SAML, SCIM provisioning, audit logs, org-wide retention controls, and custom data processing agreements. For full details, see our Trust Center at trust.openwhispr.com."},{"question":"What does the Enterprise plan add?","answer":"Enterprise adds SSO and SAML, SCIM provisioning, audit logs, org-wide data-retention controls, custom data processing agreements, team management and admin controls, and dedicated support — with managed onboarding tailored to your organization. Talk to us to set it up."}]},"blogHub":{"metadata":{"title":"Dictation Software Blog: Guides & Tips | OpenWhispr","description":"Learn voice dictation, open source speech-to-text, and AI meeting notes. Guides on Whisper, speaker diarization, and privacy-first tools.","keywords":["voice dictation blog","speech to text guides","whisper AI articles","dictation tips","voice technology blog","openwhispr blog"]},"breadcrumb":{"home":"Home","blog":"Blog"},"hero":{"title":"Blog","subtitle":"Guides, comparisons, and deep dives into voice dictation, OpenAI Whisper, NVIDIA Parakeet, and speech-to-text technology."},"readArticle":"Read article →","posts":[{"title":"HIPAA-Compliant AI Note Takers: What a BAA Actually Covers","description":"Which AI note takers sign a Business Associate Agreement, which quietly exclude their AI features from it, and why local transcription changes the analysis.","date":"August 11, 2026","tag":"Compliance"},{"title":"Google Docs Voice Typing: The Complete Guide","description":"How to use Google Docs voice typing: setup, commands, fixes, limitations, and how to dictate in any app with OpenWhispr.","date":"June 26, 2026","tag":"Guide"},{"title":"How to Take Good Meeting Notes: A Practical Guide","description":"A simple system for meeting notes that actually get used — capture decisions, action items, and open questions, clean up fast, and share them well. Includes a free copy-paste template.","date":"June 1, 2026","tag":"Guide"},{"title":"Local Speaker Diarization: How We Built a 100% Private Meeting Note Taker","description":"The full technical breakdown of OpenWhispr's local speaker diarization pipeline — pyannote segmentation, CAM++ embeddings, sherpa-onnx, and on-device voice profiles. Audio never leaves your device.","date":"April 15, 2026","tag":"Technical"},{"title":"The Best Dictation Tools for Windows in 2026","description":"A comprehensive comparison of every viable dictation and voice-to-text tool available on Windows — from built-in Win+H to AI-powered tools, honestly compared.","date":"March 15, 2026","tag":"Comparison"},{"title":"The Best Dictation Tools for Linux in 2026","description":"A comprehensive comparison of every viable dictation and voice-to-text tool available on Linux — honestly compared.","date":"February 14, 2026","tag":"Comparison"},{"title":"The Case for Local AI: Why Your Voice Data Shouldn't Leave Your Device","description":"Regulatory landscape, hardware capabilities, and the privacy case for processing voice data locally — with real breach statistics and compliance requirements.","date":"February 12, 2026","tag":"Privacy"},{"title":"Why Open Source Matters for Voice Tools","description":"Why voice software should be auditable — real privacy incidents, the open model ecosystem (OpenAI Whisper, NVIDIA Parakeet), and what to look for in open source voice tools.","date":"February 10, 2026","tag":"Thought Leadership"},{"title":"Local vs Cloud Transcription: Privacy, Speed, and Accuracy Compared","description":"An honest comparison of local and cloud-based transcription — real pricing data, privacy policies, accuracy benchmarks, and when each makes sense.","date":"February 8, 2026","tag":"Guide"},{"title":"Whisper Model Sizes Explained: Tiny vs Base vs Small vs Medium vs Large","description":"Which Whisper model should you use? Complete breakdown of all 7 models with parameters, VRAM requirements, speed benchmarks, and accuracy trade-offs.","date":"February 5, 2026","tag":"Technical"},{"title":"How Whisper AI Works: A Complete Guide","description":"A technical deep dive into the Whisper speech recognition model — architecture, training data, model sizes, and how it actually transcribes speech.","date":"February 3, 2026","tag":"Technical"}]},"blogWindowsTools":{"metadata":{"title":"The Best Dictation Tools for Windows in 2026 | OpenWhispr","description":"Best dictation software for Windows in 2026, honestly compared. OpenWhispr, Dragon, Wispr Flow and more — find your free, private, open-source pick.","keywords":["best dictation app for windows","windows voice to text","windows speech recognition","dictation software windows","voice typing windows","speech to text windows 2026","whisper windows","dragon dictation alternative","windows voice dictation"],"ogTitle":"The Best Dictation Tools for Windows in 2026","ogDescription":"A comprehensive comparison of dictation and voice-to-text tools available for Windows in 2026."},"schema":{"title":"The Best Dictation Tools for Windows in 2026","description":"A comprehensive comparison of dictation and voice-to-text tools available for Windows in 2026."},"breadcrumb":{"home":"Home","blog":"Blog","article":"Best Dictation Tools for Windows"},"layout":{"title":"The Best Dictation Tools for Windows in 2026","subtitle":"Windows has more dictation options than ever — from the built-in Win+H to AI-powered tools that rewrite as you speak. Here's every serious option in 2026, honestly compared.","date":"March 15, 2026","tag":"Blog"},"sections":[{"id":"direct-answer","title":"Direct Answer"},{"id":"fact-check","title":"Fact-Check Snapshot"},{"id":"openwhispr","title":"OpenWhispr"},{"id":"windows-voice-typing","title":"Windows Voice Typing"},{"id":"dragon-professional","title":"Dragon Professional"},{"id":"wispr-flow","title":"Wispr Flow"},{"id":"typeless","title":"Typeless"},{"id":"aqua-voice","title":"Aqua Voice"},{"id":"microsoft-365-dictation","title":"Microsoft 365 Dictation"},{"id":"talon-voice","title":"Talon Voice"},{"id":"google-docs-voice-typing","title":"Google Docs Voice Typing"},{"id":"comparison","title":"Side-by-Side Comparison"},{"id":"recommendations","title":"Our Recommendations"},{"id":"sources","title":"Sources"}],"directAnswer":{"leadStrong":"The best dictation tool for Windows in 2026 depends on what you need.","bodyBeforeLink":"For most users who want a polished, privacy-first experience with AI text cleanup,","bodyAfterLink":"offers the most complete package: push-to-talk dictation powered by whisper.cpp, system-wide hotkey support, 99+ languages, and optional AI formatting — all open source and free. For a zero-install starting point, Windows Voice Typing (Win+H) is built right in. If budget is no concern and you need the highest trained accuracy for English, Dragon Professional remains the legacy gold standard at $500–$669. Cloud-first users who want AI auto-editing should look at Wispr Flow or Typeless. And for developers who need full hands-free computing, Talon Voice is in a category of its own.","lastUpdated":"Last updated: March 2026. Tool availability and feature claims are cross-checked against at least two sources per major claim."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","item1Lead":"OpenWhispr runs local Whisper via whisper.cpp:","item1Body":"feature set and local runtime are documented publicly.","item1Link1":"openwhispr.com","item1Link2":"whisper.cpp","item2Lead":"Windows Voice Typing requires internet:","item2Body":"transcription is cloud-based via Azure Speech Services. Copilot+ PCs add local AI grammar correction (\"Fluid Dictation\") but core transcription remains cloud.","item2Link1":"Microsoft Support","item2Link2":"Windows Central","item3Lead":"Dragon Professional is still sold but in maintenance mode:","item3Body":"Microsoft closed its $19.7B acquisition of Nuance in March 2022 and has focused Nuance tech on Azure and Microsoft 365 rather than Dragon.","item3Link1":"Nuance Dragon","item3Link2":"Wikipedia","item4Lead":"Cloud-first tools trade convenience for privacy:","item4Body":"Wispr Flow, Typeless, and Aqua Voice send audio to cloud servers for processing.","item4Link1":"Wispr Flow","item4Link2":"Typeless","item4Link3":"Aqua Voice","item5Lead":"OpenWhispr positioning:","item5Body":"OpenWhispr is the only open-source, offline-capable, system-wide dictation tool with AI cleanup available on Windows.","item5Link1":"OpenWhispr","item5Link2":"Whisper upstream"},"labels":{"pros":"Pros","cons":"Cons","pricing":"Pricing:","bestFor":"Best for:"},"comparisonTable":{"title":"Side-by-Side Comparison","headerTool":"Tool","headerOpenSource":"Open Source","headerOffline":"Offline","headerLanguages":"Languages","headerAiCleanup":"AI Cleanup","headerSystemWide":"System-Wide","headerPrice":"Price","rows":[{"tool":"OpenWhispr","openSource":"full","offline":"full","languages":"99+","aiCleanup":"full","systemWide":"full","price":"Free (Pro $8/mo)"},{"tool":"Windows Voice Typing","openSource":"none","offline":"none","languages":"44","aiCleanup":"partial","systemWide":"full","price":"Free (with Windows)"},{"tool":"Dragon Professional","openSource":"none","offline":"full","languages":"6","aiCleanup":"none","systemWide":"full","price":"$$500–$669 one-time"},{"tool":"Wispr Flow","openSource":"none","offline":"none","languages":"100+","aiCleanup":"full","systemWide":"full","price":"Free tier / $15/mo"},{"tool":"Typeless","openSource":"none","offline":"none","languages":"100+","aiCleanup":"full","systemWide":"full","price":"Free tier / $12/mo"},{"tool":"Aqua Voice","openSource":"none","offline":"none","languages":"49","aiCleanup":"full","systemWide":"full","price":"Free tier / $8/mo"},{"tool":"Microsoft 365 Dictation","openSource":"none","offline":"none","languages":"~55","aiCleanup":"partial","systemWide":"none","price":"Included with M365"},{"tool":"Talon Voice","openSource":"partial","offline":"full","languages":"English","aiCleanup":"none","systemWide":"full","price":"Free / $25/mo Patreon"},{"tool":"Google Docs Voice Typing","openSource":"none","offline":"none","languages":"100+","aiCleanup":"none","systemWide":"none","price":"Free"}]},"openwhispr":{"title":"1. OpenWhispr","intro":"OpenWhispr is an open-source, cross-platform desktop dictation app built on Electron and whisper.cpp. It provides push-to-talk voice-to-text that works system-wide — hold a hotkey, speak, release, and your transcribed text appears wherever your cursor is. It also includes an AI agent mode that can clean up, format, and restructure your dictated text using LLMs (OpenAI, Anthropic, Google Gemini, or local models). On Windows, it installs as a standard .exe with NSIS installer and runs natively on x64 hardware.","pros":["Polished GUI with system-wide push-to-talk hotkey","AI agent mode for automatic text cleanup and formatting","99+ languages via Whisper and NVIDIA Parakeet models","Fully open source — build from source or use pre-built binaries","Fully offline with local models — no audio leaves your device"],"cons":["Electron-based — heavier on resources than native apps","AI cleanup features require API keys or a Pro subscription","Larger Whisper models need decent hardware for real-time performance"],"pricingValue":"Free core features, optional Pro at $8/mo","bestForValue":"Users who want a complete, private dictation experience with AI text cleanup"},"windowsVoiceTyping":{"title":"2. Windows Voice Typing (Win+H)","intro":"Windows Voice Typing is Microsoft's built-in dictation feature, accessible system-wide by pressing Win+H. It uses cloud-based Azure Speech Services for transcription and includes auto-punctuation. On Copilot+ PCs, \"Fluid Dictation\" adds local AI-powered grammar correction and filler word removal. The legacy Windows Speech Recognition feature has been replaced by Voice Access, which provides both dictation and full hands-free PC control.","pros":["Zero cost, zero setup — built into every Windows 10/11 PC","System-wide — works in any text field across the OS","Auto-punctuation and 44 languages supported","Fluid Dictation on Copilot+ PCs corrects grammar and filler words"],"cons":["Requires internet — all audio is sent to Microsoft's cloud","Commonly reported accuracy of 85–90% for clear English speech","No AI text reformatting or cleanup (except on Copilot+ PCs)","Not open source — proprietary cloud service"],"pricingValue":"Free (included with Windows)","bestForValue":"Quick, casual dictation when you don't want to install anything"},"dragon":{"title":"3. Dragon Professional v16","intro":"Dragon Professional is the legacy heavyweight of Windows dictation. Originally Dragon NaturallySpeaking, it was developed by Nuance (acquired by Microsoft in a $19.7B deal that closed in March 2022). Version 16 is the latest release and works fully offline with trained user profiles that can reach 99%+ accuracy for English. However, Dragon is effectively in maintenance mode — Microsoft has not announced a new version and has focused Nuance technology on Azure and Microsoft 365 instead.","pros":["Nuance claims up to 99% accuracy for trained English profiles","Fully offline — recognition runs entirely on-device","Deep voice command customization for workflows","Industry standard for legal and medical professionals"],"cons":["Extremely expensive — $500–$669 depending on retailer","Effectively in maintenance mode — no new development","Only 6 languages (English, German, French, Spanish, Dutch, Italian)","Windows-only, no Mac or Linux support"],"pricingValue":"$$500–$669 one-time (upgrades ~$350)","bestForValue":"Legal and medical professionals who need maximum English accuracy offline","assessmentLead":"Honest assessment:","assessmentBody":"Dragon was the gold standard for decades, but its future is uncertain. Microsoft has not invested in Dragon since the Nuance acquisition. At $500–$669, it is a significant investment in software that may not receive meaningful updates. Modern Whisper-based tools now match or exceed Dragon's accuracy for most use cases at a fraction of the cost."},"wisprFlow":{"title":"4. Wispr Flow","intro":"Wispr Flow is a cloud-based AI dictation tool available on Windows, macOS, iOS, and Android. It excels at AI post-processing — automatically removing filler words, correcting grammar, and adapting its tone to match the app you're dictating into (more casual in Slack, more formal in an email client). It supports 100+ languages with automatic detection and includes a snippet library for voice-triggered shortcuts.","pros":["Excellent AI auto-editing with per-app tone adaptation","100+ languages with automatic detection","Cross-platform (Windows, macOS, iOS, Android)","HIPAA-eligible on all plans"],"cons":["No offline mode — completely unusable without internet","All audio is sent to cloud servers for processing","Expensive at $15/month ($12/month annually)","Free tier limited to 2,000 words/week"],"pricingValue":"Free (2,000 words/week), Pro $15/mo","bestForValue":"Users who want the best AI auto-editing and don't mind cloud processing"},"typeless":{"title":"5. Typeless","intro":"Typeless is a cloud-based dictation tool available on Windows, macOS, iOS, and Android. It focuses on intelligent text transformation — removing filler words, eliminating repetition, auto-formatting lists and key points, and adapting to your personal writing style over time. It also supports real-time translation between languages and works across 60+ apps including Notion, Slack, Gmail, VS Code, and Obsidian.","pros":["Generous free tier — 4,000 words/week (~16,000/month)","Excellent AI auto-editing with personal style adaptation","Real-time translation between 100+ languages","Works across 60+ apps system-wide"],"cons":["No offline mode — requires internet for all transcription","All audio leaves your device for cloud processing","Pro plan is $30/month if not on annual billing","Not open source — proprietary cloud service"],"pricingValue":"Free (4,000 words/week), Pro $12/mo (annual) or $30/mo","bestForValue":"Users who want generous free usage with strong AI formatting"},"aquaVoice":{"title":"6. Aqua Voice","intro":"Aqua Voice is a cloud-based dictation tool purpose-built for developers and technical writers. Its proprietary Avalon transcription model is trained to handle technical terminology — variable names, acronyms, API names, and domain-specific jargon — better than general-purpose speech engines. It supports 49 languages, works system-wide on both Windows and macOS, and starts up in under 50ms.","pros":["Excellent at technical vocabulary (code, acronyms, jargon)","Context-aware formatting adapts output to target app","Very fast startup — under 50ms latency","Affordable Pro plan at $8/month"],"cons":["No offline mode — entirely cloud-based","Fewer languages (49) than competitors","Relatively new product with a smaller user base","Free tier limited to 1,000 words"],"pricingValue":"Free (1,000 words), Pro $8/mo (annual)","bestForValue":"Developers and technical writers who need accurate transcription of code and jargon"},"microsoft365":{"title":"7. Microsoft 365 Dictation","intro":"Microsoft 365 Dictation is built into Word, Excel, Outlook, PowerPoint, and OneNote on both desktop and web. It uses Microsoft's cloud-based speech recognition with auto-punctuation and supports ~55 languages (15 fully supported, ~40 in preview). Voice formatting commands (\"bold that,\" \"new line,\" \"delete last sentence\") give you basic hands-free editing within Office apps. It requires an active internet connection. A Microsoft 365 subscription is needed for desktop apps, though dictation is also available free in the web versions of Office.","pros":["Already included if you have Microsoft 365","Auto-punctuation and voice formatting commands","~55 languages (15 fully supported, ~40 in preview)","Works in both desktop and web versions of Office"],"cons":["Only works within Microsoft 365 apps — not system-wide","Requires internet — cloud-based transcription only","Desktop apps require Microsoft 365 subscription ($9.99+/mo)","No AI text cleanup or filler word removal"],"pricingValue":"Free on web, or included with M365 ($9.99+/mo) for desktop","bestForValue":"Microsoft 365 users who only need dictation inside Office apps"},"talon":{"title":"8. Talon Voice","intro":"Talon Voice is a comprehensive hands-free input system that goes far beyond dictation. It lets you control your entire computer — window management, coding, app switching, file browsing — entirely by voice, with optional eye tracking via Tobii devices. Talon includes its own built-in Conformer speech engine and also supports Dragon Professional (Windows only) as an alternative engine. It is primarily command-based rather than prose-dictation-based, making it ideal for developers and accessibility users who need to replace their keyboard and mouse entirely.","pros":["Full hands-free computer control, not just dictation","Purpose-built voice coding with language-specific commands","Eye tracking integration for mouse replacement","Invaluable for accessibility — genuinely life-changing for RSI"],"cons":["Steep learning curve — requires memorizing commands and a phonetic alphabet","Core engine is proprietary (community scripts are open source)","Primarily English — limited multilingual support","Not designed for long-form prose dictation"],"pricingValue":"Free public version, $25/mo beta via Patreon","bestForValue":"Users who need full hands-free computing, voice coding, or accessibility-driven input"},"googleDocs":{"title":"9. Google Docs Voice Typing","intro":"Google's built-in voice typing is accessible through Google Docs (Tools > Voice typing) in Chrome, Edge, or Safari. It uses Google's cloud-based speech recognition and supports 100+ languages for transcription, though voice commands for formatting work only in English. It requires no installation beyond a browser, making it the lowest-friction option — but also the most limited, since it cannot work outside of Google Docs.","pros":["Zero setup — just open your browser and start talking","Good accuracy powered by Google's speech models","100+ languages for transcription","Completely free with a Google account"],"cons":["Only works inside Google Docs (Chrome, Edge, Safari) — not system-wide","All audio is sent to Google's servers — no offline mode","No AI cleanup — you must manually speak punctuation","Not open source — proprietary Google service"],"pricingValue":"Free","bestForValue":"Quick dictation inside Google Docs when you don't need system-wide input or privacy"},"recommendations":{"title":"Our Recommendations","intro":"There is no single \"best\" dictation tool for Windows — it depends on what you value. Here is how we would break it down:","mostUsersTitle":"For most users","mostUsersStrong":"OpenWhispr","mostUsersBody":"offers the most complete package: a polished GUI, push-to-talk system-wide, AI cleanup, Whisper accuracy, and it is fully open source. It works offline by default and is the only tool here where your audio never leaves your device.","zeroInstallTitle":"For zero-install simplicity","zeroInstallStrong":"Windows Voice Typing (Win+H)","zeroInstallBody":"is already on your PC. Press the hotkey and start talking. It won't match the accuracy or features of dedicated tools, but you can't beat the convenience of something that's built in.","aiEditingTitle":"For best AI auto-editing","aiEditingStrong1":"Wispr Flow","aiEditingAnd":"and","aiEditingStrong2":"Typeless","aiEditingBody":"lead in AI post-processing — removing filler words, correcting grammar, adapting tone. The trade-off is that all audio goes to the cloud. Typeless has the more generous free tier; Wispr Flow has per-app tone adaptation.","developersTitle":"For developers","developersStrong":"Aqua Voice","developersBody":"is purpose-built for technical vocabulary. If you dictate code comments, API docs, or technical specs, its Avalon model handles jargon better than general-purpose engines.","accuracyTitle":"For maximum English accuracy offline","accuracyStrong":"Dragon Professional","accuracyBody":"still delivers the highest trained accuracy for English, but at $500–$669 with an uncertain future. Only consider it if you need specialized vocabulary profiles and are comfortable with a legacy product.","accessibilityTitle":"For accessibility and hands-free computing","accessibilityStrong":"Talon Voice","accessibilityBody":"is in a different category entirely. If you need to replace your keyboard and mouse — not just supplement them with dictation — Talon is the tool to learn. The investment in time is significant, but the payoff is transformative."},"sources":{"title":"Sources and Further Reading","openwhispr":"OpenWhispr","windowsVoiceTyping":"Windows Voice Typing (Microsoft Support)","dragon":"Dragon Professional (Nuance)","wisprFlow":"Wispr Flow","typeless":"Typeless","aquaVoice":"Aqua Voice","microsoft365":"Microsoft 365 Dictation (Microsoft Support)","talon":"Talon Voice","googleDocs":"Google Docs Voice Typing (Google Support)","whisperCpp":"whisper.cpp","openaiWhisper":"OpenAI Whisper"},"cta":{"title":"Want to Try OpenWhispr?","body":"Open source dictation for Windows, macOS, and Linux. Push-to-talk, whisper.cpp, AI cleanup, and 99+ languages — free forever.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":{"whisperModelSizes":"Whisper model sizes compared","windowsSpeech":"OpenWhispr vs Windows Speech","dragon":"OpenWhispr vs Dragon","windowsUseCase":"OpenWhispr for Windows","allComparisons":"All comparisons"}},"blogLinuxTools":{"metadata":{"title":"The Best Dictation Tools for Linux in 2026 | OpenWhispr","description":"A comprehensive comparison of dictation and voice-to-text tools available for Linux in 2026. Covers OpenWhispr, Nerd Dictation, Speech Note, and more.","keywords":["best dictation app for linux","linux voice to text","linux speech recognition","dictation software linux","voice typing linux","speech to text linux 2026","whisper linux","open source dictation linux","linux voice dictation"],"ogTitle":"The Best Dictation Tools for Linux in 2026","ogDescription":"A comprehensive comparison of dictation and voice-to-text tools available for Linux in 2026."},"article":{"title":"The Best Dictation Tools for Linux in 2026","description":"A comprehensive comparison of dictation and voice-to-text tools available for Linux in 2026."},"breadcrumb":{"home":"Home","blog":"Blog","article":"Best Dictation Tools for Linux"},"layout":{"title":"The Best Dictation Tools for Linux in 2026","subtitle":"Linux users have been underserved by dictation software for years. That's finally changing. Here's every viable option in 2026, honestly compared.","date":"February 14, 2026","tag":"Blog"},"sections":[{"id":"direct-answer","title":"Direct Answer"},{"id":"fact-check","title":"Fact-Check Snapshot"},{"id":"spectrum","title":"Dictation Spectrum"},{"id":"openwhispr","title":"OpenWhispr"},{"id":"nerd-dictation","title":"Nerd Dictation"},{"id":"speech-note","title":"Speech Note"},{"id":"simon","title":"Simon (KDE)"},{"id":"chrome-voice-typing","title":"Chrome Voice Typing"},{"id":"whisper-cpp","title":"Whisper.cpp CLI"},{"id":"talon-voice","title":"Talon Voice"},{"id":"comparison","title":"Comparison Table"},{"id":"recommendations","title":"Recommendations"},{"id":"sources","title":"Sources"},{"id":"try-openwhispr","title":"Try OpenWhispr"}],"directAnswer":{"leadStrong":"The best dictation tool for Linux in 2026 depends on what you need.","bodyBeforeLink":"For most users who want a polished, ready-to-use experience with AI text cleanup,","linkLabel":"OpenWhispr","bodyAfterLink":"offers the most complete package: push-to-talk dictation powered by whisper.cpp, system-wide hotkey support, 99+ languages, and optional AI formatting — all open source. For CLI-oriented users who want minimal dependencies, Nerd Dictation is excellent. KDE and Qt users should look at Speech Note. Developers comfortable piping audio through the terminal can use whisper.cpp directly. And if you need full hands-free computer control beyond just dictation, Talon Voice is in a category of its own.","lastUpdated":"Last updated: February 2026. Tool availability and feature claims are cross-checked against at least two sources per major claim."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","items":[{"lead":"OpenWhispr runs local Whisper via whisper.cpp:","body":"feature set and local runtime are documented publicly.","linkLabel1":"openwhispr.com","linkLabel2":"whisper.cpp"},{"lead":"Linux options vary from polished GUIs to CLI scripts:","body":"project repos and distribution pages confirm scope and maintenance status.","linkLabel1":"Nerd Dictation","linkLabel2":"Speech Note (Flathub)"},{"lead":"Cloud-first tools trade convenience for privacy:","body":"browser voice typing is easy but routes audio through provider infrastructure.","linkLabel1":"Chrome voice typing help","linkLabel2":"Google STT data logging"},{"lead":"Legacy Linux voice projects exist but may be stale:","body":"repo activity and historical release timelines matter before adoption.","linkLabel1":"Simon repo","linkLabel2":"KDE unmaintained archive"},{"lead":"OpenWhispr positioning:","body":"OpenWhispr is included here because it combines local ASR, system-wide capture, and optional post-processing in one Linux-capable app.","linkLabel1":"OpenWhispr","linkLabel2":"Whisper upstream"}]},"labels":{"pros":"Pros","cons":"Cons","pricing":"Pricing:","bestFor":"Best for:"},"tools":{"openwhispr":{"title":"1. OpenWhispr","intro":"OpenWhispr is an open-source, cross-platform desktop dictation app built on Electron and whisper.cpp. It provides push-to-talk voice-to-text that works system-wide — hold a hotkey, speak, release, and your transcribed text appears wherever your cursor is. It also includes an AI agent mode that can clean up, format, and restructure your dictated text using LLMs (OpenAI, Anthropic, Google Gemini, or local models).","pros":["Polished GUI with system-wide push-to-talk hotkey","AI agent mode for automatic text cleanup and formatting","99+ languages via Whisper models (tiny through large)","Fully open source — build from source or use pre-built binaries"],"cons":["Electron-based — heavier on resources than native apps","AI cleanup features require API keys or a Pro subscription","Larger Whisper models need decent hardware for real-time performance"],"pricingValue":"Free core features, optional Pro at $8/mo","bestForValue":"Users who want a complete, polished dictation experience with AI text cleanup"},"nerdDictation":{"title":"2. Nerd Dictation","intro":"Nerd Dictation is a single-file Python script that provides offline speech-to-text using the VOSK API. It's designed for Linux power users who are comfortable with the command line. You start and stop it manually via CLI commands, and it types into whatever window has focus using xdotool. User configuration is a Python script, making it endlessly hackable.","pros":["Extremely lightweight — single Python file, minimal dependencies","Fully offline with small VOSK language models (under 50 MB)","Infinitely customizable via Python configuration scripts","No background processes — starts and stops on demand"],"cons":["No GUI — entirely CLI-driven, not beginner-friendly","VOSK accuracy is noticeably below Whisper-based tools","Requires xdotool (X11 only — limited Wayland support)"],"pricingValue":"Free and open source","bestForValue":"CLI-savvy Linux users who want a lightweight, hackable, fully offline dictation script"},"speechNote":{"title":"3. Speech Note (Dsnote)","intro":"Speech Note is a Qt-based Linux application for note taking, reading, and translating using offline speech-to-text, text-to-speech, and machine translation. It supports multiple STT engines including Whisper (via whisper.cpp and Faster Whisper), and is available as a Flatpak on Flathub. The latest versions support GPU acceleration via Vulkan for Intel, AMD, and NVIDIA cards.","pros":["Full GUI with note-taking, TTS, and translation features built in","Supports multiple STT engines including Whisper models","GPU acceleration (Vulkan) for faster transcription","Easy install via Flatpak on most distros"],"cons":["Designed as a note-taking app, not a system-wide dictation tool","Global keyboard shortcuts exist but are less seamless than push-to-talk","Qt interface may feel less polished on GNOME-based desktops"],"pricingValue":"Free and open source","bestForValue":"Users who want an all-in-one note-taking app with built-in speech recognition and translation"},"simon":{"title":"4. Simon (KDE)","intro":"Simon is an open-source speech recognition system originally developed as part of the KDE project. It takes a unique do-it-yourself approach, allowing users to create custom language and acoustic models from scratch rather than relying on pre-trained models. However, it is important to be upfront: Simon is effectively unmaintained. The last significant release was in 2013, and despite a brief revival effort in 2017, there has been no active development in recent years.","pros":["Custom acoustic models can be trained for specific vocabulary","Fully open source with a graphical interface","Runs entirely offline"],"cons":["Effectively abandoned — no meaningful updates since 2013","Never ported to KDE Frameworks 5 or Qt5/Qt6","Recognition accuracy far below modern Whisper-based tools"],"pricingValue":"Free and open source","bestForValue":"Historical interest only — not recommended for practical use in 2026","honestAssessmentLead":"Honest assessment:","honestAssessmentBody":"Simon pioneered open-source speech recognition on Linux, but modern tools like whisper.cpp have made its approach obsolete. We include it here for completeness, but we would not recommend it for new users."},"chrome":{"title":"5. Google Chrome Voice Typing","intro":"Google's built-in voice typing is accessible to Linux users through Chrome or Google Docs (Tools > Voice typing). It uses Google's cloud-based speech recognition and supports a wide range of languages. It requires no installation beyond having Chrome — which makes it the lowest-friction option, but also the most limited.","pros":["Zero setup — just open Chrome and start talking","Good accuracy powered by Google's speech models","Supports 100+ languages"],"cons":["Only works inside Chrome/Google Docs — not system-wide","All audio is sent to Google's servers — no offline mode","Not open source, closed-source proprietary service"],"pricingValue":"Free","bestForValue":"Quick dictation inside Google Docs when you don't need system-wide input or privacy"},"whisperCpp":{"title":"6. Whisper.cpp CLI","intro":"Whisper.cpp is a C/C++ port of OpenAI's Whisper speech recognition model, optimized for CPU inference. It is not a dictation app per se — it is a command-line tool for transcribing audio files. However, many Linux users build dictation workflows around it by piping microphone input through scripts. Projects like whisper-dictation and whispertux wrap whisper.cpp in push-to-talk interfaces for exactly this purpose.","pros":["State-of-the-art Whisper accuracy, runs locally","Highly optimized C++ — fast even on CPU-only machines","99+ languages supported across all Whisper model sizes","Fully open source, very actively maintained"],"cons":["No GUI, no push-to-talk — purely a transcription engine","Requires scripting to build a real-time dictation workflow","Designed for file transcription, not live streaming dictation"],"pricingValue":"Free and open source","bestForValue":"Developers who want to build custom dictation pipelines or transcription workflows"},"talon":{"title":"7. Talon Voice","intro":"Talon Voice is a comprehensive hands-free input system that goes far beyond dictation. It lets you control your entire computer — window management, coding, app switching, file browsing — entirely by voice, with optional eye tracking via Tobii devices. Talon supports Linux on X11 (Ubuntu 18.04+, and most modern distros). However, Wayland is not supported and is not planned at the time of writing.","pros":["Full hands-free computer control, not just dictation","Purpose-built voice coding with language-specific commands","Eye tracking integration for mouse replacement","Invaluable for accessibility — genuinely life-changing for RSI"],"cons":["Steep learning curve — requires memorizing commands and a phonetic alphabet","Core engine is proprietary (community scripts are open source)","Linux support limited to X11 — no Wayland support"],"pricingValue":"Free public version, $25/mo beta via Patreon","bestForValue":"Users who need full hands-free computing, voice coding, or accessibility-driven input"}},"comparison":{"title":"Side-by-Side Comparison","headerTool":"Tool","headerOpenSource":"Open Source","headerOffline":"Offline","headerLanguages":"Languages","headerAiCleanup":"AI Cleanup","headerSystemWide":"System-Wide","headerPrice":"Price","rows":[{"tool":"OpenWhispr","openSource":"full","offline":"full","languages":"99+","aiCleanup":"full","systemWide":"full","price":"Free (Pro $8/mo)"},{"tool":"Nerd Dictation","openSource":"full","offline":"full","languages":"Varies by VOSK model","aiCleanup":"none","systemWide":"full","price":"Free"},{"tool":"Speech Note","openSource":"full","offline":"full","languages":"Varies by backend/model","aiCleanup":"none","systemWide":"partial","price":"Free"},{"tool":"Simon","openSource":"full","offline":"full","languages":"Custom","aiCleanup":"none","systemWide":"partial","price":"Free"},{"tool":"Chrome Voice Typing","openSource":"none","offline":"none","languages":"Many browser-supported languages","aiCleanup":"none","systemWide":"none","price":"Free"},{"tool":"Whisper.cpp CLI","openSource":"full","offline":"full","languages":"99+","aiCleanup":"none","systemWide":"none","price":"Free"},{"tool":"Talon Voice","openSource":"partial","offline":"full","languages":"English","aiCleanup":"none","systemWide":"full","price":"Free core / paid Patreon tiers"}]},"recommendations":{"title":"Our Recommendations","intro":"There is no single \"best\" dictation tool for Linux — it depends on what you value. Here is how we would break it down:","mostUsers":{"title":"For most users","leadBrand":"OpenWhispr","body":"offers the most complete experience: a real GUI, push-to-talk system-wide, AI cleanup, Whisper accuracy, and it is open source. It is the closest thing Linux has to what macOS and Windows users take for granted."},"cliPurists":{"title":"For CLI purists","leadBrand":"Nerd Dictation","body":"is brilliant in its simplicity. A single Python file, VOSK-powered, endlessly hackable. If you prefer configuring your tools via scripts and don't need Whisper-level accuracy, it is hard to beat."},"kdeQt":{"title":"For KDE/Qt users","leadBrand":"Speech Note","body":"integrates well with Qt-based desktops and offers translation and TTS in addition to speech recognition. It is a strong choice if you want an all-in-one note-taking tool."},"developers":{"title":"For developers","leadBrand":"Whisper.cpp CLI","body":"is the engine that powers many of these tools. If you want maximum control and are comfortable building your own pipeline, go straight to the source."},"accessibility":{"title":"For accessibility and hands-free computing","leadBrand":"Talon Voice","body":"is in a different category entirely. If you need to replace your keyboard and mouse — not just supplement them with dictation — Talon is the tool to learn. The investment in time is significant, but the payoff is transformative."}},"sources":{"title":"Sources and Further Reading","links":["OpenWhispr","OpenAI Whisper","Nerd Dictation (GitHub)","Speech Note (Flathub)","Simon (KDE GitHub mirror)","whisper.cpp","Talon Voice","Chrome voice typing help","Google Speech-to-Text data logging policy"]},"cta":{"title":"Want to Try OpenWhispr?","body":"Open source dictation for Linux, macOS, and Windows. Push-to-talk, whisper.cpp, AI cleanup, and 99+ languages — free forever.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":{"whisperModelSizes":"Whisper model sizes compared","talonCompare":"OpenWhispr vs Talon","linuxUseCase":"OpenWhispr for Linux","allComparisons":"All comparisons"}},"blogMeetingNotes":{"metadata":{"title":"How to Take Good Meeting Notes: Practical Guide | OpenWhispr","description":"Learn how to take good meeting notes that people actually use — decisions, action items, and open questions. Free template and AI notes tool included.","keywords":["how to take good meeting notes","how to take meeting notes","meeting notes template","effective meeting notes","meeting notes best practices","note taking methods","how to take notes in a meeting","meeting minutes vs notes","action items meeting","cornell note taking method","meeting note taker","meeting notes app"]},"schema":{"title":"How to Take Good Meeting Notes: A Practical Guide","description":"A simple system for meeting notes that get used: capture decisions, action items, and open questions, clean up fast, and share them well. Free template included."},"breadcrumb":{"home":"Home","blog":"Blog","article":"How to Take Good Meeting Notes"},"hero":{"title":"How to Take Good Meeting Notes: A Practical Guide","subtitle":"Stop transcribing every word. Capture the decisions, action items, and open questions that actually matter — and let the right tool do the rest.","date":"June 1, 2026","tag":"Guide"},"sections":[{"id":"why","title":"Why Notes Matter"},{"id":"before","title":"Prep in Five Minutes"},{"id":"structure","title":"Use One Structure"},{"id":"capture","title":"What to Write Down"},{"id":"methods","title":"Three Methods"},{"id":"after","title":"Clean Up After"},{"id":"tools","title":"Let a Tool Help"},{"id":"privacy","title":"Consent & Privacy"},{"id":"share","title":"Share and Store"},{"id":"mistakes","title":"Common Mistakes"},{"id":"template","title":"Free Template"},{"id":"faq","title":"FAQ"}],"intro":{"lead":"Good meeting notes capture decisions, action items, and open questions — not a word-for-word transcript.","paragraph1":"The goal is not to write down everything that was said. It is to leave the meeting with a short, shared record of what was decided, who is doing what by when, and what is still unresolved. Do that well and a 30-minute call turns into something the whole team can act on. Do it badly and you spend the next meeting re-deciding things you already settled.","paragraph2":"This guide walks through a simple system: prep in five minutes, use the same structure every time, write down only what matters, clean up right after, and share the notes where people will actually see them. It works whether you type by hand or let a meeting note taker handle the recording for you.","lastUpdated":"Last updated: June 1, 2026."},"why":{"title":"Why Good Meeting Notes Matter","paragraph1":"Meetings are expensive. Six people in a room for an hour is a working day of someone's time. The output of that hour is almost never the conversation itself — it is the decisions and the next steps. If those never get written down, the meeting mostly didn't happen.","paragraph2":"Good notes do three jobs. They give the people who attended a shared memory, so nobody relitigates a settled decision next week. They give the people who missed it a fast way to catch up without a recap call. And they turn vague intentions into tracked action items with an owner and a date. That is the difference between \"we should look into that\" and a task that actually gets done."},"prep":{"title":"Prep in Five Minutes Before the Meeting","lead":"The best notes start before the meeting does.","intro":"Five minutes of setup means you walk in already knowing what to listen for, instead of scrambling to keep up from the first sentence.","items":[{"label":"Read the agenda.","text":"It tells you what the meeting is for, which means it tells you which moments are worth capturing."},{"label":"Open your template.","text":"Start a page with the date, attendees, and a one-line purpose at the top, and your four empty buckets below (more on those next)."},{"label":"Note any open items from last time.","text":"Carrying forward unresolved questions stops them from quietly disappearing."}],"outro":"If there is no agenda, write one yourself in two lines: the goal of the meeting and the decisions you expect to make. That alone makes the conversation — and your notes — sharper."},"structure":{"title":"Use the Same Structure Every Time","lead":"A consistent format is the single biggest upgrade you can make to your notes.","intro":"When the page is already laid out, you stop spending attention on where to put things and spend it on listening instead. The same four buckets work for almost every meeting:","items":[{"label":"Decisions.","text":"What the group actually agreed to. One line each, in plain language."},{"label":"Action items.","text":"The next steps — each with an owner and a due date. A task with no name attached is a wish, not an action."},{"label":"Open questions.","text":"Anything raised but not resolved, so it gets picked up instead of forgotten."},{"label":"Notes and context.","text":"The few background details worth keeping — numbers, links, the reasoning behind a decision."}],"outro":"Reorder them however you like, but keep the set fixed. Notes that all share a shape are far easier to scan, search, and hand to someone who wasn't there."},"capture":{"title":"What to Write Down (and What to Skip)","lead":"The hardest part of note-taking is deciding what not to write.","intro":"Trying to capture everything means you capture nothing well — your head is down, transcribing, while the actual decision happens without you. Instead, listen for a few signal phrases that mark something worth keeping:","signals":[{"label":"Decisions:","text":"\"Let's go with…\", \"We've decided…\", \"We'll ship…\""},{"label":"Ownership:","text":"\"I'll take that\", \"Sam will handle…\", \"Can you own…\""},{"label":"Deadlines:","text":"\"by Friday\", \"before the launch\", \"next sprint\""},{"label":"Risks and blockers:","text":"\"the problem is…\", \"we're blocked on…\", \"that depends on…\""},{"label":"Hard-to-remember specifics:","text":"numbers, names, dates, and links you won't recall an hour later."}],"skip":"Skip the rest: small talk, people thinking out loud, tangents, and anything already written in the agenda or the slides. There is no point copying down a chart that's attached to the invite.","testTitle":"The one-week test","testBody":"Before writing something down, ask: \"Will someone need this next week?\" If yes, capture it. If no, let it go. That single question quietly filters out most of the noise."},"methods":{"title":"Three Note-Taking Methods That Work","intro":"There is no single \"right\" way to take notes — only the method that fits the meeting. Three are worth knowing:","list":[{"label":"The outline method.","text":"Nested bullets — main point, sub-points beneath it. Fast, flexible, and the right default for most meetings. It maps cleanly onto the four-bucket structure above."},{"label":"The Cornell method.","text":"Split the page into a narrow cue column, a wide notes column, and a summary strip at the bottom. Originally a study technique, it shines in one-on-ones and learning-heavy sessions where you want to review later."},{"label":"The decisions-and-actions method.","text":"Skip the narrative entirely and write only two lists: decisions made and action items owed. Brutally efficient for status updates, stand-ups, and planning meetings where outcomes are all that matter."}],"outro":"Pick by meeting type and stay consistent within it. A planning meeting and a customer interview don't need the same method — but every planning meeting should use the same one."},"after":{"title":"Clean Up Within Five Minutes","lead":"Your memory of a meeting starts decaying the moment it ends.","intro":"The fragments and shorthand that made sense during the call get fuzzy fast. Spend five minutes right after — before the next thing pulls you away — turning rough notes into something a stranger could read.","items":["Expand your shorthand while you still remember what it meant.","Check that every action item has an owner and a due date. Fill in the gaps now, not next week.","Delete the noise that crept in, and fix anything you half-wrote."],"outro":"Five minutes now is worth an hour of confused back-and-forth later. This is the step most people skip — and it is the one that separates notes people trust from notes people ignore."},"tools":{"title":"Let a Tool Do the Heavy Lifting","lead":"There is a hard limit to taking notes by hand: you can't fully listen and fully transcribe at the same time.","paragraph1":"A meeting note taker removes that trade-off. It records the call, transcribes it, and labels who said what, so you can stay present in the conversation and edit a ready-made draft afterward instead of typing live.","paragraph2Pre":"That changes your job from","capturing":"capturing","paragraph2Mid":"to","curating":"curating","paragraph2Post":": the transcript already exists, so you just pull out the decisions and action items. It is especially useful when you are the one running the meeting and can't look down at a keyboard.","paragraph3":"But before you switch one on, two things are worth getting right: everyone's consent, and knowing where your audio actually ends up. That is the next section."},"privacy":{"title":"Recording Meetings: Get Consent and Protect Privacy","lead":"A recording is the most accurate way to capture a meeting — and the fastest way to create a privacy problem if you are careless with it.","intro":"Two things matter before you hit record: getting consent, and knowing where the audio and transcript end up afterward.","consentLead":"Get consent first.","consentBody":"Recording laws vary by region — some places let a single participant consent, others require everyone on the call to agree, and rules like the EU's GDPR treat a recorded voice as personal data. You don't need to memorize the statutes; you need one habit: ask before you record, every time. Saying it out loud and noting \"recorded with consent\" in your notes keeps you covered.","scriptTitle":"A one-line consent script","scriptBody":"\"Mind if I record this so I can stay in the conversation and share notes afterward?\" — asked before you start. If anyone objects, take notes by hand instead.","audioLead":"Know where the audio goes.","audioBody":"Most cloud note takers — Otter, Fireflies, Granola — upload your meeting to their servers to process it, where the recording and transcript may be stored, used to train their AI, or exposed in a breach. For a routine internal sync that is usually an acceptable trade. For anything sensitive — HR conversations, legal or medical detail, customer data, deal terms — it is a real exposure worth thinking about before you record.","habitsIntro":"A few habits keep you on the safe side:","habits":["Match the tool to the sensitivity — keep confidential meetings off third-party servers.","Check whether the tool trains its AI on your recordings. Many do by default, and it is often opt-out, not opt-in.","Delete recordings once the notes are written, if you don't need the audio itself.","Prefer a tool that processes on-device whenever the conversation is private."],"ctaPre":"This is exactly why","ctaMid":"is built local-first — a free, open source meeting note taker that records, transcribes, and labels who said what entirely on your own device. The audio and transcript never leave your machine, so \"where does my data go?\" has a simple answer. We break down the on-device speaker labeling in","diarizationLink":"local speaker diarization","ctaMid2":", and the full trade-offs in","transcriptionLink":"local vs cloud transcription","ctaPost":"."},"share":{"title":"Share and Store Notes Where People Look","lead":"Notes nobody reads are wasted effort.","intro":"The last step is what turns a personal scribble into a team asset.","items":[{"label":"Send them within the hour,","text":"while the meeting is still fresh for everyone who was there."},{"label":"Lead with decisions and action items.","text":"Put what people need to act on at the top; tag each owner directly so nobody can miss their task."},{"label":"Store them where the team already works","text":"— the shared doc, the project tool, the channel. One findable home per project keeps notes searchable months later."}],"outro":"The test of a good note system is whether someone can answer \"what did we decide about X?\" in 30 seconds, three months from now. If your notes can do that, they are doing their job."},"mistakes":{"title":"Common Mistakes to Avoid","items":[{"label":"Transcribing everything.","text":"A wall of text helps no one. Capture outcomes, not the whole conversation."},{"label":"Action items with no owner or date.","text":"\"Someone should look into this\" never gets done. Attach a name and a deadline to every task."},{"label":"A different format every time.","text":"Inconsistent notes are slow to write and impossible to scan. Reuse one template."},{"label":"Never sharing them.","text":"Notes that live only in your notebook can't align a team. Send them out, every time."},{"label":"Trusting your memory for the cleanup.","text":"\"I'll tidy these up later\" usually means never. Do it in the five minutes after the call."}]},"template":{"title":"A Meeting Notes Template You Can Copy","intro":"Here is the whole system on one page. Copy it, keep it somewhere handy, and reuse it for every meeting:","code":"Meeting: \nDate: <date> Attendees: <names>\nPurpose: <one line — why are we here?>\n\nDECISIONS\n-\n\nACTION ITEMS\n- [ ] <task> — <owner> — <due date>\n\nOPEN QUESTIONS\n-\n\nNOTES / CONTEXT\n-\n\nNext meeting / follow-up: <date or trigger>","outro":"That is it. Prep with it, fill it during the call, clean it up for five minutes after, and send it out. The structure does most of the work for you."},"faq":{"title":"Frequently Asked Questions"},"faqs":[{"question":"How do you take good meeting notes?","answer":"Prepare a simple template before the meeting, then capture only what matters: decisions, action items with an owner and a due date, and open questions. Don't try to transcribe every word — listen for signals like \"we've decided\", \"I'll take that\", or \"by Friday\". Clean the notes up within a few minutes of the meeting ending, then share them where your team already works."},{"question":"What should meeting notes include?","answer":"Good meeting notes include the date, attendees, and purpose at the top, followed by four sections: decisions made, action items (each with an owner and a deadline), open questions, and any background context worth keeping. Everything else — small talk, thinking out loud, anything already in the deck — can be left out."},{"question":"What is the best format for meeting notes?","answer":"A consistent, reusable format beats any clever one. The most reliable structure is four buckets: Decisions, Action items, Open questions, and Notes. Using the same template every time means you spend your attention listening instead of deciding how to lay the page out, and it makes notes from different meetings easy to scan and search later."},{"question":"What is the difference between meeting notes and meeting minutes?","answer":"Meeting minutes are a formal, often required record of attendance, motions, and votes — common for boards and legal settings. Meeting notes are a lighter, practical summary aimed at helping the team remember decisions and follow through on action items. Most working meetings need good notes, not formal minutes."},{"question":"How do you take notes and pay attention at the same time?","answer":"Stop trying to write everything down. When you only capture decisions, owners, and deadlines, your hands are free most of the time and you can actually listen. A meeting note taker that records and transcribes the call for you removes the conflict entirely: you stay present during the meeting and edit a ready-made draft afterward."},{"question":"Should you record meetings to take notes?","answer":"Recording lets you stay present and get an accurate transcript instead of typing live, but always tell participants first — many places require consent. The bigger question is where the audio goes. Most cloud note takers upload your meeting to their servers. For sensitive conversations, a local-first tool like OpenWhispr records, transcribes, and labels speakers entirely on your own device, so nothing is uploaded."},{"question":"Is it legal to record a meeting for notes?","answer":"It depends on where the participants are. Some regions allow recording with just one person's consent, while others require everyone on the call to agree, and rules like the EU's GDPR treat a recorded voice as personal data. This isn't legal advice, but the safe habit everywhere is the same: tell people you're recording and get their okay before you start."},{"question":"Where does my audio go when I use an AI meeting note taker?","answer":"With most cloud note takers, your meeting audio and transcript are uploaded to the company's servers to be processed and stored, and some use that data to train their models. With a local-first tool like OpenWhispr, the recording, transcript, and speaker labels stay on your own device and are never uploaded — which matters most for sensitive meetings like HR, legal, or anything involving customer data."},{"question":"How do you take notes in a meeting you are leading?","answer":"When you are running the meeting you can't also type. Either assign a notetaker, or let a meeting note taker capture the transcript so you can focus on the room. If you must self-serve, jot only decisions and action items in short fragments, then expand them in the five minutes after the call while everything is still fresh."},{"question":"What is the best app for taking meeting notes?","answer":"The best meeting notes app is the one that captures the conversation accurately and fits your privacy needs. Cloud tools like Otter, Fireflies, and Granola are convenient but send your audio to their servers. OpenWhispr is a free, open source meeting note taker that transcribes and identifies speakers locally, so your audio never leaves your device."}],"cta":{"title":"Take Meeting Notes Without Typing a Word","body":"OpenWhispr is a free, open source meeting note taker that records, transcribes, and labels who said what — entirely on your own device. Stay present in the meeting and edit a ready-made draft afterward. Works on macOS, Windows, Linux and iOS.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":[{"href":"/blog/local-speaker-diarization","label":"Private meeting note taker"},{"href":"/blog/local-vs-cloud-transcription","label":"Local vs cloud transcription"},{"href":"/blog/best-dictation-tools-windows-2026","label":"Best dictation tools for Windows"},{"href":"/blog","label":"All blog posts"}]},"blogHowWhisper":{"metadata":{"title":"How Whisper AI Works: A Complete Guide | OpenWhispr","description":"How Whisper AI works explained: transformer architecture, 680K training hours, model sizes, and accuracy benchmarks. Open-source, runs on-device.","keywords":["how whisper AI works","whisper speech recognition","openai whisper explained","whisper model architecture","whisper AI accuracy","whisper transformer model","automatic speech recognition","whisper vs other ASR","whisper.cpp explained"]},"schema":{"title":"How Whisper AI Works: A Complete Guide","description":"A technical deep dive into OpenAI's Whisper speech recognition model — architecture, training data, accuracy benchmarks, and practical usage."},"breadcrumb":{"home":"Home","blog":"Blog","article":"How Whisper AI Works"},"hero":{"title":"How Whisper AI Works: A Complete Guide","subtitle":"A technical deep dive into OpenAI's open-source speech recognition model — from mel spectrograms to transformer decoding.","date":"February 3, 2026","tag":"Technical"},"sections":[{"id":"what-is-whisper","title":"What Is Whisper?"},{"id":"architecture","title":"Architecture"},{"id":"training-data","title":"Training Data"},{"id":"model-sizes","title":"Model Sizes"},{"id":"accuracy","title":"Accuracy"},{"id":"whisper-cpp","title":"whisper.cpp"},{"id":"comparison","title":"ASR Comparison"},{"id":"applications","title":"Applications"},{"id":"sources","title":"Sources"}],"intro":{"paragraph":"Whisper is an automatic speech recognition (ASR) model developed by OpenAI. Released in September 2022, it is a general-purpose speech recognition system trained on 680,000 hours of multilingual audio data collected from the internet. Whisper uses an encoder-decoder Transformer architecture that converts audio into log-mel spectrograms, processes them through a neural network encoder, and autoregressively decodes text tokens. It supports transcription in 99 languages, translation to English, language identification, and timestamp prediction — all within a single unified model. Unlike most commercial ASR services, Whisper's weights are <openSourceLink>fully open source</openSourceLink> under the MIT license, which has enabled a thriving ecosystem of ports, optimizations, and applications — including <whisperCppLink>whisper.cpp</whisperCppLink>, the C/C++ implementation that makes Whisper practical for real-time, on-device speech recognition.","lastUpdated":"Last updated: February 17, 2026. Quantitative claims in this article are cross-checked against at least two primary sources."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","items":[{"label":"Whisper release, training scale, and license:","body":"OpenAI released Whisper in September 2022, trained on 680,000 hours, under MIT."},{"label":"Task scope and language support:","body":"Whisper supports transcription, translation-to-English, language identification, and timestamp tokens across 99 languages."},{"label":"English benchmark context:","body":"around 3% WER is a benchmark figure on LibriSpeech test-clean for English-focused evaluation, not a universal real-world rate."},{"label":"Recent model updates:","body":"large-v3 reports lower errors than large-v2 on many languages, and turbo focuses on lower latency with smaller decoder compute."},{"label":"OpenWhispr relevance:","body":"OpenWhispr uses Whisper via whisper.cpp for local, offline-first dictation across desktop platforms."}],"sources":["<a1>arXiv paper</a1> · <a2>OpenAI Whisper repo</a2>","<a1>Whisper model card</a1> · <a2>Whisper README</a2>","<a1>large-v2 model card</a1> · <a2>benchmark details</a2>","<a1>large-v3 model card</a1> · <a2>turbo release discussion</a2>","<a1>whisper.cpp</a1> · <a2>OpenWhispr</a2>"],"sourceUrls":[["https://arxiv.org/abs/2212.04356","https://github.com/openai/whisper"],["https://github.com/openai/whisper/blob/main/model-card.md","https://github.com/openai/whisper"],["https://huggingface.co/openai/whisper-large-v2","https://github.com/openai/whisper/blob/main/model-card.md"],["https://huggingface.co/openai/whisper-large-v3","https://github.com/openai/whisper/discussions/2363"],["https://github.com/ggml-org/whisper.cpp","https://openwhispr.com"]]},"whatIs":{"title":"What Is Whisper?","paragraph1":"Whisper was introduced in the paper <paperLink>\"Robust Speech Recognition via Large-Scale Weak Supervision\"</paperLink> by Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever at OpenAI. The code and model weights were published on <githubLink>GitHub</githubLink> in September 2022 under the MIT license.","paragraph2":"Before Whisper, state-of-the-art speech recognition systems were typically trained on curated, human-labeled datasets — often limited to specific languages, accents, or domains. Whisper broke from this approach by training on a massive, diverse dataset of 680,000 hours of audio paired with transcripts scraped from the internet. The key insight was that this \"weakly supervised\" approach — using imperfect, machine-generated or user-uploaded transcripts rather than hand-labeled data — could produce a model with remarkable robustness across languages, accents, background noise, and technical vocabulary.","paragraph3":"The result is a single model that can handle multiple speech processing tasks: transcription (speech to text in the same language), translation (speech in any language to English text), language identification, and voice activity detection with timestamps. This multitask capability is baked into the model's architecture through a system of special tokens that specify which task to perform.","paragraph4":"Whisper's open-source release was significant. It democratized access to high-quality ASR, allowing developers, researchers, and companies to use a model competitive with commercial APIs — without per-minute pricing, without sending audio to the cloud, and without vendor lock-in. This catalyzed the development of tools like <whisperCppLink>whisper.cpp</whisperCppLink>, <fasterWhisperLink>Faster Whisper</fasterWhisperLink>, and dozens of applications built on top of them — including <openWhisprLink>OpenWhispr</openWhisprLink>."},"architecture":{"title":"Architecture Deep Dive","intro":"Whisper uses an encoder-decoder Transformer architecture — the same fundamental design behind models like the original Transformer (Vaswani et al., 2017) and many machine translation systems. The encoder processes audio, the decoder generates text. Here is how each stage works.","preprocessing":{"title":"Audio Preprocessing","intro":"Before any neural network processing, raw audio is converted into a visual representation of sound called a log-mel spectrogram. The process works as follows:","steps":["Audio is resampled to 16,000 Hz (16 kHz mono).","A Short-Time Fourier Transform (STFT) is computed using 25-millisecond windows with a 10-millisecond stride (hop length).","The frequency spectrum is projected onto 80 mel-frequency filter banks (128 for large-v3), which approximate human auditory perception by spacing frequency bands logarithmically.","A logarithmic scaling is applied, producing the final log-mel spectrogram.","The spectrogram is divided into 30-second chunks. Audio shorter than 30 seconds is zero-padded; longer audio is processed in sequential chunks."],"outro":"The result is a 2D representation with shape <code>(80, 3000)</code> — 80 mel channels by 3,000 time steps (30 seconds at 10ms stride). This is analogous to an image, and the encoder processes it much like a vision model processes pixel data."},"encoder":{"title":"The Encoder","intro":"The encoder converts the mel spectrogram into a sequence of learned audio representations. It consists of:","steps":["Two 1D convolution layers — The first convolution has a kernel size of 3 and padding of 1, followed by a GELU activation. The second convolution has a kernel size of 3, stride of 2, and padding of 1, which halves the time dimension. This downsamples the 3,000 time steps to 1,500 positions.","Sinusoidal positional embeddings — Unlike the decoder, the encoder uses fixed sinusoidal embeddings (not learned) to encode the position of each time step.","Transformer blocks — A stack of standard Transformer encoder layers, each containing multi-head self-attention and a feed-forward network with GELU activation, connected by residual connections and layer normalization."],"outro":"The output is a sequence of 1,500 contextualized audio feature vectors — one for every 20 milliseconds of input audio — that the decoder will attend to during text generation."},"decoder":{"title":"The Decoder","intro":"The decoder generates text tokens autoregressively — one token at a time, each conditioned on the encoded audio and all previously generated tokens. It consists of:","steps":["Token embedding layer — Converts token IDs into dense vector representations.","Learned positional embeddings — Unlike the encoder's sinusoidal embeddings, the decoder uses learned position embeddings that are trained alongside the model.","Transformer blocks with cross-attention — Each decoder layer has three sub-layers: masked self-attention (preventing the model from looking at future tokens), cross-attention to the encoder output (allowing the model to \"listen\" to the audio), and a feed-forward network."],"outro":"The final decoder output is projected onto the vocabulary to produce token probabilities. Decoding strategies include greedy search, beam search, and temperature-based sampling."},"tokens":{"title":"Special Tokens and Multitask Training","intro":"Whisper performs multiple tasks with a single model through a clever system of special tokens that serve as instructions. The decoder input follows a structured format:","startOfTranscript":"— Signals the start of the output sequence.","language":"— Specifies the language (one of 99 language tokens). Can be auto-detected or manually set.","taskJoin":"or","task":"— Specifies whether to transcribe in the original language or translate to English.","timestampsPart1":"— Controls whether to produce timestamp tokens. When timestamps are enabled, the model generates time offset tokens like","timestampsJoin":"and","timestampsPart2":"that align text to audio.","outro":"This unified token format means a single model can perform transcription, translation, language detection, and timestamped transcription — all without separate model heads or fine-tuning. The task is determined entirely by the token sequence provided to the decoder."}},"trainingData":{"title":"Training Data","paragraph1":"The original Whisper models (tiny through large-v2) were trained on 680,000 hours of audio paired with transcripts collected from the internet. This dataset is not publicly available. The key characteristic is that it is weakly supervised — the transcript labels were not hand-verified by human annotators. Instead, they came from sources like subtitles, closed captions, and other user-generated text paired with audio.","compositionTitle":"Data Composition","compositionIntro":"According to the <modelCardLink>official model card</modelCardLink>, the 680,000-hour training set breaks down as follows:","headers":{"segment":"Segment","hours":"Hours","share":"Share","description":"Description"},"rows":[{"segment":"English ASR","hours":"438,000","share":"65%","description":"English audio with English transcripts"},{"segment":"Translation (X-to-English)","hours":"126,000","share":"18%","description":"Non-English audio with English transcripts"},{"segment":"Multilingual ASR","hours":"117,000","share":"17%","description":"Non-English audio with native-language transcripts (covering 98 languages)"}],"paragraph2":"The heavy English bias (65% of training data) explains why Whisper performs significantly better on English than on lower-resource languages. For the large-v3 model released in November 2023, OpenAI expanded the training set to 1 million hours of weakly labeled audio plus an additional 4 million hours of pseudo-labeled audio generated by running Whisper large-v2 over unlabeled data — a form of self-training or knowledge distillation.","weakTitle":"Why \"Weakly Supervised\" Matters","weakParagraph1":"Most prior ASR systems trained on datasets like LibriSpeech (960 hours) or Common Voice (a few thousand hours per language) — carefully curated, human-verified datasets. Whisper's approach traded label quality for sheer scale: 680,000 hours versus the hundreds or low thousands of hours used by conventional systems. The paper demonstrated that this trade-off was worthwhile. The diversity of the internet-sourced data gave Whisper exceptional robustness to real-world conditions: background noise, overlapping speech, accents, domain-specific vocabulary, and acoustic environments that would challenge models trained on studio-quality read speech.","weakParagraph2":"However, the weak supervision also introduces a known limitation: hallucination. Because the training transcripts are imperfect, the model sometimes generates plausible-sounding text that was not actually spoken — particularly during silence or very quiet passages. This is a trade-off inherent to the approach and remains an active area of improvement."},"modelSizes":{"title":"Model Sizes","intro":"Whisper is available in multiple sizes, ranging from 39 million to 1.55 billion parameters. Smaller models are faster but less accurate; larger models are more accurate but require more compute and memory. The <code>.en</code> variants are English-only models that tend to perform better for English, especially at smaller sizes. No English-only variants exist for the large models.","headers":{"model":"Model","parameters":"Parameters","englishOnly":"English-Only","vram":"VRAM","speed":"Relative Speed"},"rows":[{"model":"tiny","parameters":"39 M","englishOnly":"tiny.en","vram":"~1 GB","speed":"~10x"},{"model":"base","parameters":"74 M","englishOnly":"base.en","vram":"~1 GB","speed":"~7x"},{"model":"small","parameters":"244 M","englishOnly":"small.en","vram":"~2 GB","speed":"~4x"},{"model":"medium","parameters":"769 M","englishOnly":"medium.en","vram":"~5 GB","speed":"~2x"},{"model":"large (v1, v2, v3)","parameters":"1,550 M","englishOnly":"--","vram":"~10 GB","speed":"1x"},{"model":"turbo (large-v3-turbo)","parameters":"809 M","englishOnly":"--","vram":"~6 GB","speed":"~8x"}],"tableNote":"Speed is relative to the large model. VRAM requirements are for GPU inference using fp16 precision via the original Python implementation. <whisperCppLink>whisper.cpp</whisperCppLink> uses significantly less memory (e.g., the large model requires ~3.9 GB RAM instead of ~10 GB VRAM).","turbo":"The turbo model (released October 2024) is a distilled version of large-v3 with a significantly smaller decoder. It retains most of large-v3's accuracy while running approximately 8x faster — making it a strong choice when speed matters. Note that turbo does not support the translation task.","large":"The large model has gone through three iterations: large-v1 (September 2022), large-v2 (December 2022), and large-v3 (November 2023). Each version improved accuracy, with large-v3 introducing 128 mel-frequency bins (up from 80) and training on a much larger dataset. For a detailed breakdown of each size and how to choose, see our <guideLink>guide to Whisper model sizes</guideLink>."},"accuracy":{"title":"How Accurate Is Whisper?","intro":"Whisper's accuracy is typically measured using Word Error Rate (WER) — the percentage of words that are incorrectly transcribed (insertions, deletions, and substitutions combined). Lower WER is better.","englishTitle":"English Benchmarks","englishParagraph":"On LibriSpeech test-clean — the most widely used English ASR benchmark, consisting of clean read speech from audiobooks — Whisper large-v2 achieves a WER of approximately 3.0%<ref1>[1]</ref1><ref2>[2]</ref2>. This is competitive with commercial ASR systems and purpose-trained models, though specialized models fine-tuned on LibriSpeech can achieve lower WER on that specific benchmark. Whisper's strength lies not in beating narrow benchmarks but in robustness — performing consistently well across a wide range of real-world conditions.","largeV3Title":"Large-v3 Improvements","largeV3Paragraph":"According to OpenAI's evaluation, Whisper large-v3 shows a 10-20% reduction in errors compared to large-v2 across languages where the model achieves below 60% error rate on the Common Voice 15 and Fleurs evaluation datasets<ref2>[2]</ref2><ref3>[3]</ref3>. The improvements are particularly notable for non-English languages, where the expanded training data (5 million total hours) makes the biggest difference.","compareTitle":"How Whisper Compares (February 2026)","compareParagraph":"Since Whisper's release, newer open-source ASR models have surpassed it on clean-speech benchmarks. NVIDIA's Parakeet TDT (0.6B parameters) achieves 1.69% WER and their Canary model reaches 1.6% WER on LibriSpeech test-clean — both significantly below Whisper large-v3's ~2.7%. However, Whisper remains the most widely deployed open-source ASR model due to its ecosystem maturity, language coverage, and the availability of optimized runtimes like whisper.cpp.","chartNote":"Note: Commercial cloud APIs (Deepgram Nova-3, Google Chirp, AssemblyAI Universal-2) do not publish LibriSpeech WER figures, so they cannot be directly compared in this chart. Their benchmarks use proprietary real-world test sets. OpenAI also released GPT-4o-transcribe in March 2025 as an API-only model (not open-source, not Whisper-based) with reportedly lower error rates — but it is cloud-only and not available for local inference.","excelsTitle":"Where Whisper Excels","excels":[{"title":"Robustness to accents and dialects","body":"— Trained on diverse internet audio, Whisper handles regional accents better than systems trained on read speech."},{"title":"Background noise tolerance","body":"— The model maintains reasonable accuracy even with music, other speakers, or ambient noise present."},{"title":"Technical vocabulary","body":"— Due to the breadth of its training data, Whisper handles domain-specific terms (medical, legal, technical) better than many general-purpose ASR systems."},{"title":"Multilingual capability","body":"— A single model supporting 99 languages, with no per-language fine-tuning required."}],"strugglesTitle":"Where Whisper Struggles","struggles":[{"title":"Hallucination","body":"— The most commonly cited issue. During silence or very quiet passages, Whisper can generate text that was never spoken. This is a consequence of the weakly supervised training approach."},{"title":"Low-resource languages","body":"— Languages with little training data (much of the 680K hours is English) have significantly higher error rates."},{"title":"Repetitive text generation","body":"— The autoregressive decoder can get \"stuck\" in loops, producing the same phrase repeatedly. Beam search with specific parameters can mitigate but not eliminate this issue."},{"title":"Not real-time by default","body":"— Whisper processes 30-second chunks, so there is inherent latency. Streaming solutions (like whisper.cpp's real-time mode) work around this but add complexity."}]},"whisperCpp":{"title":"whisper.cpp and Local Inference","paragraph1":"The original Whisper implementation requires Python, PyTorch, and a CUDA-capable GPU for reasonable performance. This makes it impractical for desktop applications, mobile devices, and edge deployments. <whisperCppLink>whisper.cpp</whisperCppLink>, created by Georgi Gerganov, solves this problem.","paragraph2":"whisper.cpp is a complete reimplementation of Whisper inference in plain C/C++ with no dependencies. It reads the same model weights but runs without Python, without PyTorch, and without a GPU (though it benefits enormously from one). Key features include:","features":[{"title":"Zero runtime memory allocations","body":"— All memory is pre-allocated, making performance predictable and suitable for embedded systems."},{"title":"Apple Silicon optimization","body":"— On Macs with M-series chips, the encoder can run on the Apple Neural Engine via Core ML, delivering more than 3x faster inference compared to CPU-only execution. Metal GPU acceleration is also supported for full GPU inference."},{"title":"Quantization support","body":"— Models can be quantized to 4-bit, 5-bit, or 8-bit integer precision, dramatically reducing memory usage and increasing speed with minimal accuracy loss."},{"title":"GPU acceleration","body":"— Supports NVIDIA CUDA (via cuBLAS), Vulkan (cross-vendor), OpenVINO (Intel), and OpenBLAS for CPU acceleration."},{"title":"Voice Activity Detection (VAD)","body":"— Built-in VAD to skip silent segments, improving both speed and transcript quality."}],"memoryTitle":"Memory Requirements (whisper.cpp)","memoryHeaders":{"model":"Model","disk":"Disk Size","ram":"RAM Required"},"memoryRows":[{"model":"tiny","disk":"75 MB","ram":"~273 MB"},{"model":"base","disk":"142 MB","ram":"~388 MB"},{"model":"small","disk":"466 MB","ram":"~852 MB"},{"model":"medium","disk":"1.5 GB","ram":"~2.1 GB"},{"model":"large","disk":"2.9 GB","ram":"~3.9 GB"}],"paragraph3":"whisper.cpp runs on macOS (Intel and Apple Silicon), Linux, Windows, iOS, Android, FreeBSD, Raspberry Pi, and even in the browser via WebAssembly. It achieves faster-than-real-time transcription on modest consumer hardware — including devices as small as the Raspberry Pi 4 and iPhone 13.","paragraph4":"This portability is what makes local, private speech recognition practical. Applications like <openWhisprLink>OpenWhispr</openWhisprLink> use whisper.cpp under the hood to provide real-time push-to-talk dictation that runs entirely on the user's device — no audio ever leaves the machine."},"comparison":{"title":"Whisper vs Other ASR Systems","intro":"Whisper exists in a broader landscape of speech recognition systems. Here is how it compares to the most common alternatives.","cards":[{"title":"Google Cloud Speech-to-Text","body":"Google's cloud ASR service. Excellent accuracy, especially on English, with real-time streaming support and speaker diarization. The primary differences: it is proprietary, requires sending audio to Google's servers, and is priced per minute of audio processed. Whisper offers comparable accuracy for general use, runs locally, and is free — but lacks built-in streaming and diarization.","tags":["Cloud-only","Proprietary","Pay-per-use"]},{"title":"Amazon Transcribe (AWS)","body":"Amazon's cloud ASR, tightly integrated with the AWS ecosystem. Offers medical transcription, call analytics, and custom vocabulary. Similar trade-offs to Google: cloud dependency, per-minute pricing, and proprietary. Whisper is a better fit for applications that need offline operation, open-source licensing, or no vendor lock-in.","tags":["Cloud-only","Proprietary","AWS integration"]},{"title":"Mozilla DeepSpeech","body":"Mozilla's open-source ASR engine based on Baidu's Deep Speech research. Was one of the first high-quality open ASR models, but Mozilla discontinued active development in 2021. Whisper has effectively superseded DeepSpeech in accuracy, multilingual support, and community momentum. DeepSpeech remains useful as a lightweight, well-understood baseline but is no longer recommended for new projects.","tags":["Open source","Discontinued","English-focused"]},{"title":"Vosk","body":"An open-source offline speech recognition toolkit that supports 20+ languages with small, fast models. Vosk is excellent for resource-constrained environments — its models are a fraction of Whisper's size and run well on low-power hardware. The trade-off is accuracy: Vosk is noticeably less accurate than Whisper, particularly on accented speech, noisy audio, and technical vocabulary.","tags":["Open source","Lightweight","Lower accuracy"]},{"title":"Faster Whisper (CTranslate2)","body":"A reimplementation of Whisper using SYSTRAN's CTranslate2 inference engine. It runs the same Whisper models with the same accuracy but is up to 4x faster than the original OpenAI implementation with lower memory usage. Faster Whisper is Python-based (unlike whisper.cpp's C/C++) and is a strong choice for server-side batch transcription. It supports CUDA GPUs and CPU inference with INT8 quantization.","tags":["Open source","Python / CTranslate2","Same Whisper accuracy"]}]},"applications":{"title":"Practical Applications","intro":"Whisper's combination of accuracy, multilingual support, and open-source availability has made it the backbone of a wide range of applications:","cards":[{"title":"Desktop Dictation","body":"Apps like <openWhisprLink>OpenWhispr</openWhisprLink> use whisper.cpp to provide system-wide push-to-talk dictation on macOS, Windows, and Linux. Audio is processed locally — nothing is sent to the cloud."},{"title":"Subtitle Generation","body":"Whisper's timestamp prediction makes it well-suited for generating subtitles and closed captions. Tools like <stableTsLink>stable-ts</stableTsLink> and <autoSubtitleLink>auto-subtitle</autoSubtitleLink> automate this workflow."},{"title":"Podcast and Meeting Transcription","body":"Whisper handles long-form audio transcription well, especially with the chunked processing approach. Many podcast editors and meeting note tools use Whisper (or Faster Whisper) for batch transcription of recordings."},{"title":"Translation and Localization","body":"Whisper's built-in translation capability (any language to English) is used for cross-lingual content access, media localization, and real-time translation prototypes."},{"title":"Accessibility","body":"Real-time captioning for deaf and hard-of-hearing users, voice interfaces for users with motor disabilities, and audio description transcription are all active areas where Whisper is deployed."},{"title":"Data Annotation and Research","body":"Researchers use Whisper to create pseudo-labeled datasets for training other models — the same approach OpenAI used to create large-v3's training data. It is also used extensively in linguistic research and corpus creation."}]},"sources":{"title":"Sources and Further Reading","items":[{"label":"[Paper]","text":"Radford, A., Kim, J. W., Xu, T., Brockman, G., McLeavey, C., & Sutskever, I. (2022). Robust Speech Recognition via Large-Scale Weak Supervision. arXiv:2212.04356.","linkText":"arxiv.org/abs/2212.04356"},{"label":"[Code]","text":"OpenAI Whisper GitHub repository.","linkText":"github.com/openai/whisper"},{"label":"[Code]","text":"whisper.cpp — C/C++ port by Georgi Gerganov.","linkText":"github.com/ggml-org/whisper.cpp"},{"label":"[Model Card]","text":"Whisper large-v2 on Hugging Face (WER benchmark: ~3.0% on LibriSpeech test-clean).","linkText":"huggingface.co/openai/whisper-large-v2"},{"label":"[Model Card]","text":"Whisper large-v3 on Hugging Face (10-20% error reduction over v2).","linkText":"huggingface.co/openai/whisper-large-v3"},{"label":"[Model Card]","text":"Whisper large-v3-turbo on Hugging Face (latency-focused distilled variant).","linkText":"huggingface.co/openai/whisper-large-v3-turbo"},{"label":"[Model Card]","text":"OpenAI Whisper model card (training data composition breakdown).","linkText":"github.com/openai/whisper/blob/main/model-card.md"},{"label":"[Release Notes]","text":"OpenAI large-v3 announcement and evaluation notes.","linkText":"github.com/openai/whisper/discussions/1762"},{"label":"[Release Notes]","text":"OpenAI turbo release discussion and speed-focused tradeoffs.","linkText":"github.com/openai/whisper/discussions/2363"},{"label":"[Code]","text":"Faster Whisper — CTranslate2-based reimplementation by SYSTRAN.","linkText":"github.com/SYSTRAN/faster-whisper"},{"label":"[Product]","text":"OpenWhispr uses Whisper through whisper.cpp for local dictation workflows.","linkText":"openwhispr.com"}],"links":["https://arxiv.org/abs/2212.04356","https://github.com/openai/whisper","https://github.com/ggml-org/whisper.cpp","https://huggingface.co/openai/whisper-large-v2","https://huggingface.co/openai/whisper-large-v3","https://huggingface.co/openai/whisper-large-v3-turbo","https://github.com/openai/whisper/blob/main/model-card.md","https://github.com/openai/whisper/discussions/1762","https://github.com/openai/whisper/discussions/2363","https://github.com/SYSTRAN/faster-whisper","https://openwhispr.com"]},"cta":{"title":"Try Whisper Locally with OpenWhispr","body":"OpenWhispr uses Whisper (via whisper.cpp) for local, private dictation on macOS, Windows, and Linux. Push-to-talk, 99+ languages, no cloud required. Try it free.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":[{"href":"/blog/whisper-model-sizes-explained","label":"Whisper model sizes compared"},{"href":"/blog/local-vs-cloud-transcription","label":"Local vs cloud transcription"},{"href":"/blog","label":"All blog posts"},{"href":"/use-cases/developers","label":"OpenWhispr for developers"},{"href":"/compare","label":"Compare dictation tools"}]},"blogVoicePrivacy":{"metadata":{"title":"Voice Data Privacy: Keep Your Audio On-Device | OpenWhispr","description":"Cloud speech tools log your voice. See why local AI transcription with OpenWhispr keeps audio on-device, private, and fully offline — free and open source.","keywords":["local AI privacy","on-device AI","voice data privacy","local speech recognition","private AI processing","edge AI voice","local LLM","AI without cloud","voice data security","nvidia parakeet"]},"schema":{"title":"The Case for Local AI: Why Your Voice Data Shouldn't Leave Your Device","description":"AI is powerful enough to run on your laptop. Here's why local processing is better for privacy, security, and control."},"breadcrumb":{"home":"Home","blog":"Blog","article":"The Case for Local AI"},"layout":{"title":"The Case for Local AI: Why Your Voice Data Shouldn't Leave Your Device","subtitle":"The hardware in your pocket is powerful enough to run speech recognition that rivals cloud services. So why are we still sending voice data to servers?","date":"February 12, 2026","tag":"Blog"},"sections":[{"id":"opening","title":"Overview"},{"id":"fact-check","title":"Fact-Check Snapshot"},{"id":"local-ai-revolution","title":"The Local AI Revolution"},{"id":"voice-sensitivity","title":"Voice Data Sensitivity"},{"id":"voice-in-the-cloud","title":"Voice in the Cloud"},{"id":"performance-gap","title":"Performance Gap"},{"id":"regulatory-landscape","title":"Regulatory Landscape"},{"id":"cloud-makes-sense","title":"When Cloud Makes Sense"},{"id":"future-of-local-ai","title":"Future of Local AI"},{"id":"what-you-can-do","title":"What You Can Do"},{"id":"sources","title":"Sources"}],"opening":{"lead":"Local AI processes data entirely on your device — no internet connection, no data leaving your machine, no trusting a third party with your information.","body1":"With models like","whisperLink":"OpenAI Whisper running efficiently via whisper.cpp","body2":"and NVIDIA Parakeet, consumer laptops can now transcribe speech with accuracy that matches or approaches cloud services. This article makes the case for why voice data, specifically, should stay on your device — and why the technology to make that happen is already here.","lastUpdated":"Last updated: February 17, 2026. Quantitative and legal claims are cross-checked against at least two primary sources."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","item1":{"lead":"Local ASR is practical on consumer hardware:","body":"Whisper + whisper.cpp and Parakeet have public model/runtime docs with hardware guidance."},"item2":{"lead":"Cloud provider logging behavior is configurable, not uniform:","body":"data handling depends on vendor defaults and opt-ins/opt-outs.","link1":"Google data logging","link2":"AWS AI opt-out policy"},"item3":{"lead":"Voice data can trigger strict compliance obligations:","body":"biometric and AI-system risk rules are codified in EU law.","link1":"GDPR official text","link2":"EU AI Act official text"},"item4":{"lead":"Breach impact is financially material:","body":"breach costs remain high, which raises the stakes of centralizing sensitive audio.","link1":"IBM data breach report","link2":"FTC voice-cloning risk alert"},"item5":{"lead":"OpenWhispr context:","body":"OpenWhispr is designed for local-first transcription with optional cloud usage.","link2":"whisper.cpp backend"}},"revolution":{"title":"The Local AI Revolution: What Changed","p1":"Five years ago, running a capable AI model on a laptop was impractical. The models were too large, the hardware too slow, and the software ecosystem barely existed. That has changed dramatically.","p2a":"The catalyst was","p2b":", Georgi Gerganov's C/C++ port of OpenAI Whisper speech recognition model. By rewriting the model inference in plain C++ with zero runtime dependencies, Gerganov made it possible to run Whisper on everything from a MacBook to a Raspberry Pi. The project supports models from the 75 MB \"tiny\" variant (needing roughly 273 MB of RAM) to the full \"large-v3\" model at 2.9 GB on disk — with integer quantization to reduce memory usage further. It runs on Mac (Intel and Apple Silicon), Linux, Windows, Android, iOS, and even in WebAssembly.","p3a":"The same approach exploded across the AI landscape. Gerganov's","p3b":"did for large language models what whisper.cpp did for speech — it brought LLaMA, Mistral, Qwen, DeepSeek, and dozens of other models to consumer hardware, with quantization from 8-bit down to as low as 1.5-bit. Tools like","p3c":"and","p3d":"wrapped these capabilities in user-friendly interfaces.","p4a":"Hardware caught up, too. Apple's M-series chips, starting with M1 in 2020, put a unified memory architecture and a dedicated Neural Engine into every Mac — making local AI inference fast enough for real-time use. Apple leaned into this with","p4b":", which processes AI tasks on-device by default and only routes to Apple's Private Cloud Compute servers when necessary — with strong guarantees that data is never stored and is used only for the immediate request.","p5":"The trend is clear: models are getting smaller and better simultaneously. Techniques like knowledge distillation (training small models to mimic large ones) and quantization (reducing numerical precision without destroying accuracy) mean that what required a data center GPU in 2023 can run on a laptop in 2026. Whisper's \"large-v3-turbo\" model, for example, is significantly faster than its predecessor while maintaining comparable accuracy."},"sensitivity":{"title":"Why Voice Data Is Uniquely Sensitive","p1":"Not all data is equally sensitive. Text can be anonymized. Browsing history can be cleared. But voice is different — it is a biometric identifier. Your voice is as unique to you as your fingerprint.","p2":"A recording of your voice can reveal far more than the words you spoke. Research in speech analysis has demonstrated that voice patterns can indicate:","identity":{"title":"Identity markers","item1":"Your unique identity (voice is used for biometric authentication by banks and security systems)","item2":"Gender, approximate age, and regional accent","item3":"Native language and socioeconomic background"},"health":{"title":"Health and emotional state","item1":"Emotional state — stress, anxiety, fatigue, and depression can be detected from vocal patterns","item2":"Neurological conditions — voice biomarkers have been studied for early detection of Parkinson's disease","item3":"Respiratory health — conditions like COVID-19 and asthma affect vocal characteristics"},"p3":"This is not speculative. Companies are actively building products that analyze voice for health screening, insurance risk assessment, and hiring decisions. The biometric richness of voice data is precisely what makes it valuable — and precisely what makes it dangerous when it leaves your control.","p4a":"Unlike a password, you cannot change your voice if it is compromised. Once a recording is on a server, you have permanently lost exclusive control over that biometric data. And data breaches are not hypothetical — IBM's 2025 Cost of a Data Breach Report found the global average cost of a breach reached","breachCost1":"$$4.88 million","p4b":"in 2024 and","breachCost2":"$$4.44 million","p4c":"in 2025","ibmCite":"IBM, 2025","p4d":"The question is not whether breaches happen, but when."},"cloud":{"title":"What Happens to Your Voice in the Cloud","p1":"When you use a cloud-based speech recognition service, your audio is transmitted to remote servers for processing. What happens after that varies by provider — and the details matter.","google":{"a":"Google's documentation states that","bold1":"by default, Cloud Speech-to-Text does not log customer audio data or transcripts","b":". However, Google offers a voluntary data logging program where users can opt in to share audio data in exchange for discounted pricing. When opted in, Google uses the data to \"improve service quality.\" Notably, logged data","bold2":"is not deleted when you delete your project","c":"— you must submit a separate deletion request","cite":"Google Cloud Docs"},"aws":{"a":"AWS AI services, including Amazon Transcribe,","bold1":"may use customer content to develop and improve AWS services","b":"unless users explicitly opt out via an organization-level policy. AWS updated this policy in 2023 after public pressure, but the default in many regions still allows data usage for model improvement","cite":"AWS Transcribe Docs"},"azure":{"a":"Azure's Speech Service","bold1":"processes audio in real-time and does not persist customer audio data by default","b":". However, users who opt in to custom model training will have their data stored. Microsoft's data handling is governed by the","cite":"Data Protection Addendum","c":", which varies by service tier and region."},"p2a":"Beyond the providers themselves, consider the infrastructure chain. Your audio may traverse multiple network hops, be processed in a data center in a different country, and potentially be accessible to subprocessors or contractors. Even with strong provider policies, data stored on US-based servers can be subject to government access under","effLink":"National Security Letters and FISA Section 702","p2b":", which do not require notification to the data subject.","callout":{"lead":"To be clear:","body":"these providers generally have strong security practices and legitimate reasons for their policies. The point is not that cloud services are malicious — it is that sending data to any third party inherently means trusting their policies, their security, and their jurisdiction. Local processing eliminates that trust requirement entirely."}},"performance":{"title":"The Performance Gap Is Closing","p1":"The traditional argument for cloud transcription was simple: cloud models were dramatically more accurate. That gap has narrowed significantly.","p2":"OpenAI Whisper large-v3, released in late 2023, achieves word error rates competitive with commercial cloud APIs across most languages. NVIDIA's Parakeet models have pushed accuracy even further, achieving some of the lowest WERs on standard English benchmarks. Independent benchmarks on English speech show these open models achieving word error rates (WER) in the 3-5% range on clean speech — comparable to or better than Google's and AWS's commercial offerings. The newer \"large-v3-turbo\" variant is significantly faster while maintaining similar accuracy, making real-time local transcription practical on modern hardware.","stat1":"Whisper model sizes (tiny to large)","stat2Value":"99+ languages","stat2":"Supported by Whisper models","stat3Value":"Real-time","stat3":"On Apple Silicon and modern x86 CPUs","p3":"Hardware acceleration has made a major difference. whisper.cpp supports ARM NEON and Apple's Accelerate framework and Metal GPU on Mac, CUDA on NVIDIA GPUs, Vulkan for cross-platform GPU inference, and even specialized backends for Ascend NPUs and Intel's OpenVINO. The GGUF model format, developed alongside llama.cpp, enables efficient quantization — reducing a model's memory footprint by 50-75% with minimal accuracy loss.","p4":"For dictation specifically — where recordings are typically 5-30 seconds long — local inference is effectively instant on any machine made in the last three years. The \"small\" Whisper model (466 MB, ~852 MB RAM) offers excellent accuracy for most languages and runs comfortably on machines with 8 GB of memory. You do not need a high-end GPU. You do not need a gaming PC. You need a reasonably modern laptop.","p5":"Apple recognized this trend early. Their on-device speech recognition, built into iOS and macOS, improved substantially with each OS release — supporting dictation without an internet connection for dozens of languages. The fact that Apple, a company with vast cloud infrastructure, is moving speech recognition on-device should tell you something about where this technology is heading."},"regulatory":{"title":"The Regulatory Landscape","p1":"Privacy regulations worldwide are catching up to the reality that voice data is sensitive. For organizations handling voice data, local processing is not just a privacy preference — it is increasingly a compliance advantage.","gdpr":{"title":"GDPR (European Union)","a":"Under the General Data Protection Regulation, voice data is classified as","bold1":"biometric data","b":"when used to uniquely identify a person — placing it in the \"special categories\" of personal data under","link":"Article 9","c":"Processing biometric data requires explicit consent or one of a narrow set of legal bases. Even when not used for identification, voice recordings are personal data subject to GDPR's data minimization principle. Local processing avoids many GDPR obligations entirely — if data never leaves the device, there is no data transfer, no third-party processing, and no cross-border compliance issues."},"aiAct":{"title":"EU AI Act","a":"The EU AI Act, which entered into force in August 2024 with provisions being phased in through 2026, classifies AI systems that process biometric data for identification purposes as","bold1":"high-risk","link":"EU AI Act","b":"High-risk systems face extensive requirements including conformity assessments, documentation obligations, and ongoing monitoring. Real-time biometric identification in public spaces is outright prohibited with narrow exceptions. While standard speech-to-text may not trigger the high-risk classification by itself, any system that could be used to identify speakers from voice — even inadvertently — faces scrutiny under this framework."},"hipaa":{"title":"HIPAA and Professional Privilege","a":"In healthcare contexts, voice recordings that contain patient information are protected under","b":"in the United States. Sending such recordings to a cloud service requires a Business Associate Agreement (BAA) and compliance with the Security Rule. Similarly, attorneys dictating notes about client matters face","bold2":"attorney-client privilege","c":"considerations — sending privileged communications to a third-party server introduces risk. Local processing sidesteps these issues: if the audio never leaves the device, there is no third-party data handler to worry about."},"ccpa":{"title":"CCPA / CPRA (California)","a":"The California Consumer Privacy Act, as amended by the California Privacy Rights Act, treats voice and audio data as regulated personal information, with additional rules when data is used for biometric identification","link1":"California AG","link2":"CPPA FAQ","b":"Businesses collecting voice data must provide additional disclosures and respect consumer rights to deletion and opt-out. Multiple other US states — including Illinois (BIPA), Texas, and Washington — have enacted their own biometric privacy laws with varying requirements and enforcement mechanisms."},"p2":"The regulatory direction is clear: voice data is being treated with increasing seriousness worldwide. For organizations that handle voice data — whether in healthcare, legal, finance, or any regulated industry — local processing offers a structurally simpler compliance posture. No data transfers to audit. No subprocessors to vet. No cross-border data flows to justify."},"cloudSense":{"title":"When Cloud Still Makes Sense","p1":"Intellectual honesty matters. Local AI is not always the right choice, and pretending otherwise would undermine the legitimate points above. Here are situations where cloud processing remains the better option:","case1":{"title":"Massive-scale real-time streaming","body":"Call centers processing thousands of concurrent audio streams need cloud infrastructure. No single device can handle that throughput."},"case2":{"title":"Speaker diarization at scale","body":"Identifying \"who said what\" in multi-speaker recordings is computationally expensive. Cloud APIs from Google and AWS handle this more reliably than most local solutions today."},"case3":{"title":"Underrepresented languages","body":"While Whisper and Parakeet support many languages, accuracy varies. Specialized cloud providers may offer better models for specific languages or dialects that Whisper handles poorly."},"case4":{"title":"Very long recordings on limited hardware","body":"Transcribing a 4-hour meeting recording on a low-end laptop may be painfully slow. Cloud APIs process hours of audio in minutes."},"p2":"The ideal approach is often hybrid: local by default, cloud by choice. Process your everyday dictation locally. Keep sensitive content on-device. And when you genuinely need cloud-scale compute or specialized features, make that a conscious decision — not an invisible default."},"future":{"title":"The Future of Local AI","p1":"The hardware trajectory strongly favors local AI. Modern chips are being designed specifically to run neural networks efficiently:","npu1":{"title":"Apple Neural Engine","body":"Apple's dedicated neural processing unit, present in every M-series and A-series chip, handles up to 38 TOPS (trillion operations per second) on M4. Apple Intelligence relies heavily on this hardware for on-device processing."},"npu2":{"title":"Qualcomm Hexagon NPU","body":"The Snapdragon X Elite series, powering Windows laptops since 2024, includes a 45 TOPS NPU designed explicitly for local AI workloads. Microsoft's Copilot+ PC initiative requires NPU capability."},"npu3":{"title":"Intel AI Boost NPU","body":"Intel's Meteor Lake Core Ultra processors introduced dedicated NPUs at around 11 TOPS, while the newer Lunar Lake generation reaches 48 TOPS. Intel has explicitly positioned these as enabling \"AI PC\" experiences."},"p2a":"On the model side, research in distillation, pruning, and efficient architectures continues to push the boundary of what runs locally. Projects like","p2b":"demonstrate that purpose-built smaller models can achieve 99% of a large model's accuracy at a fraction of the compute cost.","p3":"The convergence is unmistakable: every major chip manufacturer is adding dedicated AI hardware. Every major OS vendor is building on-device AI features. Every major model lab is publishing smaller, more efficient models. On-device AI processing is not an alternative approach — it is becoming the default. Cloud processing will remain important for heavy workloads, but for personal computing tasks like dictation, the future is local."},"actions":{"title":"What You Can Do Today","intro":"You do not need to wait for the future. Local AI is practical right now. Here are concrete steps you can take:","step1":{"title":"Use local speech recognition for dictation","a":"provides push-to-talk dictation powered by OpenAI Whisper and NVIDIA Parakeet with a full GUI. If you prefer the command line,","b":"can be used directly. Both run entirely on your machine."},"step2":{"title":"Run LLMs locally for sensitive work","a":"Tools like","b":"and","c":"let you run capable language models locally. For queries involving personal, financial, or confidential information, a local model eliminates the data exposure entirely."},"step3":{"title":"Audit your current tools","body":"Check the privacy policies of every voice-enabled service you use. Does your dictation app send audio to a server? Does your virtual assistant store recordings? Does your meeting transcription tool train on your content? The answers may surprise you."},"step4":{"title":"Prefer apps that offer local processing","body":"When choosing new tools, favor those that offer on-device processing. The capability exists — demand it. Every user who chooses local processing sends a market signal that privacy matters."}},"sources":{"title":"Sources and Further Reading","s1":"C/C++ port of OpenAI Whisper. Model sizes, hardware requirements, and supported platforms.","s2":"Original ASR model release, training methodology, and evaluation context.","s3Title":"NVIDIA Parakeet model card","s3":"Open ASR metrics and benchmark details for local speech recognition.","s4":"C/C++ LLM inference. Supports 1.5-bit to 8-bit quantization across dozens of model architectures.","s5Title":"Google Cloud Speech-to-Text Data Logging","s5":"Default no-logging policy and opt-in data program details.","s6Title":"AWS Transcribe Security Documentation","s6":"Data handling, opt-out policies, and encryption practices.","s7Title":"IBM Cost of a Data Breach Report 2024","s7":"Global breach costs, average records exposed, and industry analysis.","s8Title":"GDPR Article 9","s8":"Processing of special categories of personal data, including biometric data.","s9Title":"EU AI Act","s9":"Classification of AI systems processing biometric data as high-risk.","s10":"California privacy guidance for personal and sensitive data handling.","s11Title":"Azure Speech data privacy","s11":"Microsoft's service-level speech data handling documentation.","s12":"On-device processing and Private Cloud Compute architecture.","s13Title":"Voice biomarkers for depression detection","s13":"Low, Bentley, Ghosh (2020). Automated assessment of psychiatric disorders using speech.","s14Title":"EFF on FISA and National Security Letters","s14":"Government access to data stored by US companies."},"cta":{"title":"Your Voice Should Stay Yours","body":"OpenWhispr processes everything locally by default. No audio ever leaves your device unless you explicitly choose cloud processing.","subtext":"Open source. Powered by OpenAI Whisper and NVIDIA Parakeet. Available on macOS, Windows, Linux and iOS.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"links":{"openSource":"Why open source matters","localVsCloud":"Local vs Cloud Transcription","allArticles":"All articles","offline":"Offline use cases","medical":"Medical dictation"}},"blogDiarization":{"metadata":{"title":"Speaker Diarization: 100% Local & Private | OpenWhispr","description":"See how OpenWhispr runs speaker diarization locally — pyannote, CAM++ embeddings, sherpa-onnx. Private by design. Open source and free.","keywords":["local speaker diarization","on-device speaker diarization","private meeting transcription","private meeting note taker","local meeting notes","offline meeting transcription","open source meeting assistant","open source meeting note taker","open source AI meeting notes","audio never leaves device","private AI meeting assistant","self-hosted meeting transcription","speaker identification privacy","pyannote segmentation","cam++ speaker embedding","sherpa-onnx diarization","voice fingerprint local","who said what transcript","otter alternative privacy","otter open source alternative","granola open source alternative","fireflies alternative on-device"]},"schema":{"title":"Local Speaker Diarization: How We Built a 100% Private Meeting Note Taker","description":"The full technical breakdown of OpenWhispr's local speaker diarization — pyannote, CAM++, sherpa-onnx. Your audio never leaves your device."},"breadcrumb":{"home":"Home","blog":"Blog","article":"Local Speaker Diarization"},"hero":{"title":"Local Speaker Diarization: How We Built a 100% Private Meeting Note Taker","subtitle":"OpenWhispr is the open source meeting assistant where your audio never leaves your device. Every voice fingerprint, every speaker label, every embedding — processed and stored on your own machine.","date":"April 15, 2026","tag":"Technical"},"sections":[{"id":"what-is-diarization","title":"What Speaker Diarization Is"},{"id":"pipeline","title":"The Four-Stage Pipeline"},{"id":"vad","title":"Stage 1: Voice Activity Detection"},{"id":"segmentation","title":"Stage 2: Segmentation"},{"id":"embeddings","title":"Stage 3: Speaker Embeddings"},{"id":"clustering","title":"Stage 4: Clustering"},{"id":"live","title":"Live Diarization"},{"id":"profiles","title":"Voice Profiles"},{"id":"privacy","title":"Privacy"},{"id":"limitations","title":"Honest Limitations"},{"id":"why-sherpa","title":"Why Sherpa-ONNX"},{"id":"faq","title":"FAQ"}],"opening":{"lead":"Local speaker diarization is the process of labeling who spoke when in a meeting recording, run entirely on the user's device with no audio, embeddings, or transcripts transmitted to any external server.","body":"Meeting transcripts have a “who said what” problem. Zoom gives you a wall of text. Most AI note takers — Otter, Fireflies, Read, Granola — solve it by uploading your meeting audio to their servers, running a speaker model there, and sending back labeled segments. Your voice, and your colleagues' voices, end up as vectors in someone else's database. We didn't want that, and neither do our users. OpenWhispr is the open source meeting note taker built on the opposite assumption: audio never leaves your device, speaker fingerprints live in a local SQLite file on your disk, and nothing about a meeting gets uploaded unless you explicitly choose to. This article is the technical breakdown of the diarization side of that promise — what it is, which models we chose, how they fit together, and exactly where every byte of data lives at every step.","footnote":"Last updated: April 15, 2026. Implementation details reference the speaker diarization feature merged into the OpenWhispr desktop app on April 14, 2026. Every technical claim below is sourced from primary references or directly inspectable in the open source repository."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","items":[{"lead":"Speaker diarization answers “who spoke when.”","body":"It is evaluated against community benchmarks like NIST Rich Transcription and the pyannote.audio reference pipeline.","sources":[{"href":"https://www.nist.gov/itl/iad/mig/rich-transcription-evaluation","label":"NIST RT evaluation"},{"href":"https://arxiv.org/abs/1911.01255","label":"pyannote.audio paper (Bredin et al., 2020)"}]},{"lead":"We use pyannote-segmentation-3.0 and CAM++ from the 3D-Speaker project.","body":"Both are open-source, both are ONNX-portable, and both are directly downloaded from the sherpa-onnx model releases.","sources":[{"href":"https://huggingface.co/pyannote/segmentation-3.0","label":"pyannote-segmentation-3.0 model card"},{"href":"https://arxiv.org/abs/2303.00332","label":"CAM++ / 3D-Speaker paper (Wang et al., 2023)"}]},{"lead":"All inference runs through sherpa-onnx, a native ONNX binary — no Python runtime, no PyTorch, no GPU required.","body":"We version-pin the binary and its dependencies at build time.","sources":[{"href":"https://github.com/k2-fsa/sherpa-onnx","label":"sherpa-onnx on GitHub"},{"href":"https://github.com/k2-fsa/sherpa-onnx/releases","label":"sherpa-onnx releases"}]},{"lead":"Voice fingerprints are 512-dimensional float32 vectors stored as local SQLite BLOBs,","body":"roughly 2KB per speaker. They never leave the device.","sources":[{"href":"https://www.sqlite.org/datatype3.html","label":"SQLite datatype docs"},{"href":"https://onnxruntime.ai/","label":"ONNX Runtime"}]},{"lead":"No audio, no embeddings, and no transcripts are transmitted off-device during diarization.","body":"The only network activity in the entire feature is a one-time model download on first launch, which you can inspect yourself in the open source repo.","sources":[{"href":"https://openwhispr.com/privacy","label":"OpenWhispr privacy policy"},{"href":"https://github.com/OpenWhispr","label":"OpenWhispr on GitHub"}]}]},"whatIs":{"title":"What Speaker Diarization Actually Is","lead":"Speaker diarization is the process of labeling who spoke when in a recording — not what they said.","paragraph1":"Transcription answers what; diarization answers who. The two are complementary: a good meeting note taker needs both, and getting them working together is harder than either one alone.","paragraph2":"Diarization is genuinely hard for a handful of stubborn reasons. Two voices can overlap on the same call. Background noise bleeds in. Short utterances (“yeah,” “okay”) carry almost no speaker information. New speakers show up halfway through a meeting. And unlike transcription, diarization has to make global decisions — it cannot label segment seven without comparing it to segment two. That global context is why most commercial solutions do diarization server-side on the full recording, then ship the labeled transcript back.","paragraph3":"Transcription alone is not enough. A ten-person meeting with forty-five minutes of cross-talk, rendered as one undifferentiated wall of text, is unreadable. Meeting notes — the actual deliverable people want — require speaker attribution, because action items, decisions, and commitments are all attached to people. “Alice will ship the migration by Friday” is a note. “Will ship the migration by Friday” is noise.","calloutTitle":"Diarization is not the same as speaker recognition","calloutBody":"Diarization says “these three chunks are Speaker A; these two are Speaker B.” It does not know who Speaker A is. Speaker recognition puts a name on Speaker A by matching a voice fingerprint to a stored profile. OpenWhispr does both — diarization runs on every meeting, recognition kicks in once you have labeled someone by name. Both stay local."},"pipeline":{"title":"The Four-Stage Local Pipeline","lead":"OpenWhispr's speaker diarization pipeline has four stages: voice activity detection, segmentation, speaker embedding, and clustering — followed by a merge step that reconciles speaker segments with the transcript.","paragraph1":"Each stage is a separate ONNX model or native binary. No Python runtime, no PyTorch, no CUDA. The total on-disk footprint of the models is about 45MB.","paragraph2":"Audio capture is dual-stream from day one. The microphone is definitionally “you,” so we never spend compute trying to diarize your own voice — every mic segment gets labeled you by source, not by voice matching. System audio (everyone else on the call) is the only thing that actually needs speaker-splitting, which immediately halves the work.","paragraph3":"Once audio is captured, the four stages run in sequence. VAD drops silence so the expensive embedding stage never wastes cycles on empty frames. Segmentation cuts continuous speech into single-speaker chunks. The embedding model turns each chunk into a 512-dimensional voice fingerprint. Agglomerative clustering groups fingerprints that look alike, producing a final set of speaker IDs. A last pass merges those IDs back into the transcript by timestamp overlap, so every word ends up tagged with a speaker.","paragraph4Prefix":"The whole thing runs inside a single child process spawned from the app — no server, no listening port, no IPC to anything outside Electron. You can confirm that by reading","paragraph4Suffix":"in the open source repo."},"vad":{"title":"Stage 1: Voice Activity Detection (Silero)","lead":"Voice activity detection (VAD) is the cheap filter that separates speech from silence before the expensive stages run.","paragraph1":"Running a speaker embedding model on silence is wasted CPU and produces garbage vectors. VAD is what keeps the rest of the pipeline fast.","paragraph2Prefix":"We use","sileroLinkLabel":"Silero VAD","paragraph2Suffix":": a 2MB ONNX model, MIT-licensed, effectively the open-source industry default. Silero returns a per-frame probability that the current 32-millisecond window contains speech. On a modern CPU, the cost is negligible — about 0.1 milliseconds per 32-millisecond chunk, or roughly a third of a percent of one core.","paragraph3":"The live pipeline runs Silero continuously on system audio during a call. The thresholds we actually ship with are tuned for real meeting audio, not a clean benchmark:","list":[{"label":"Window size:","value":"512 samples (32ms at 16kHz)"},{"label":"Speech threshold:","value":"0.15 — deliberately aggressive so quiet or distant speakers are not missed"},{"label":"Silence threshold:","value":"0.08"},{"label":"Segment ends after:","value":"16 consecutive silent windows (~512ms)"},{"label":"Minimum segment for embedding:","value":"0.8 seconds"},{"label":"Live identification cadence:","value":"every 1 second once ≥1.6 seconds of speech have accumulated"}],"calloutTitle":"Honest trade-off","calloutBody":"An aggressive speech threshold of 0.15 catches soft talkers on distant mics, but it also occasionally false-triggers on loud keystrokes or room noise. The 0.8-second minimum-segment filter discards most of those false positives before they reach the embedding stage, so downstream accuracy is protected."},"segmentation":{"title":"Stage 2: Segmentation (pyannote 3.0)","lead":"Segmentation takes continuous speech and cuts it into chunks that each contain only one speaker, detecting both speaker-change points and overlap regions.","paragraph1":"It is the stage that decides where the speaker boundaries are. Everything after it assumes those boundaries are correct.","paragraph2Prefix":"We use","paragraph2Mid":", the current version of the model from the pyannote.audio team, ONNX-exported by the","paragraph2Suffix":"project. Pyannote is the de-facto academic baseline: the pyannote.audio pipeline achieves diarization error rates in the 12–15% range on standard benchmarks like AMI and CALLHOME — the same numbers commercial cloud providers like to quote.","paragraph3Prefix":"Why the ONNX port specifically? Because shipping PyTorch inside an Electron app is a non-starter — the runtime alone is roughly 1.5GB and effectively requires a GPU to be fast. The ONNX export is a single 6.6MB file that runs on CPU via","paragraph3Suffix":". Same model, tiny fraction of the weight.","paragraph4Prefix":"We invoke it through sherpa-onnx's offline diarization binary with a handful of flags. The minimum-duration settings come from the pyannote paper's recommended defaults: short speech bursts get merged into adjacent silence, and short silences do not break a continuous speaker segment. Output is a list of start/end time pairs tagged with provisional speaker IDs (","paragraph4Suffix":", etc.)."},"embeddings":{"title":"Stage 3: Speaker Embeddings (CAM++)","lead":"A speaker embedding is a fixed-length list of numbers that represents a voice the way a hash represents a file.","paragraph1":"Two recordings of the same voice produce embeddings that are close in vector space; different voices produce embeddings that are far apart. “Close” is measured by cosine similarity — a single number between -1 and 1. This is the mathematical heart of everything else in the pipeline.","paragraph2Prefix":"Our embedding model is CAM++ from the 3D-Speaker project at Alibaba's DAMO Academy, trained on the","paragraph2Mid":"speaker-verification benchmark. The specific file we ship is","paragraph2Mid2":", a roughly 28MB ONNX export that produces a 512-dimensional float32 vector per segment. The paper is","camppPaperLabel":"arXiv:2303.00332 (Wang et al., 2023)","paragraph2Suffix":".","paragraph3Prefix":"We picked CAM++ over the obvious alternative,","paragraph3Suffix":", for three concrete reasons: roughly half the parameters, lower equal-error rate on VoxCeleb1-O, and faster CPU inference. NVIDIA NeMo's MSDD model has no first-party ONNX export and leans heavily on PyTorch. Picovoice Falcon carries a commercial per-seat license. Apple's CoreML Speech framework is macOS-only. CAM++ was the only option that satisfied our constraints — cross-platform, ONNX, open license, CPU-fast.","paragraph4Prefix":"To feed CAM++ we ship our own log-mel filterbank feature extractor in","paragraph4Mid":": 80 mel bands, 25-millisecond windows, 10-millisecond hops, standard parameters for this class of model. Features flow into the ONNX model via","paragraph4Suffix":", running in the Electron main process. Out the other side comes a 512-entry Float32Array, which we write into SQLite as a 2,048-byte BLOB.","paragraph5":"The matching step itself is the simplest possible thing: cosine similarity between two vectors. Here is the literal code that decides whether two meeting segments are the same speaker:","calloutTitle":"Why 512 dimensions?","calloutBody":"512 is the current sweet spot on the VoxCeleb speaker-verification leaderboard between accuracy and storage cost. 256 dimensions start losing discriminative power on close-but-different voices; 1024 dimensions add storage and compute with diminishing returns. 512 float32s × 2KB × a thousand meetings of ten speakers is 20MB on disk. That fits comfortably in a local SQLite file."},"clustering":{"title":"Stage 4: Clustering — the “how many speakers?” problem","lead":"After embedding every segment you have N vectors. Clustering groups them into K speakers — but the hard part is that you usually don't know K up front.","paragraph1":"A sales call might have 2 speakers. An all-hands might have 8. Fixed-K clustering breaks at both ends.","paragraph2":"We use agglomerative clustering with a cosine-similarity threshold of 0.5: start with every segment as its own cluster, merge the two clusters with the highest similarity, repeat until no remaining pair exceeds the threshold. The threshold, not a target count, decides where to stop. This scales naturally from two speakers to ten without any configuration.","paragraph3Prefix":"The actual invocation lives in","paragraph3Suffix":". It is a spawn of the sherpa-onnx binary with a handful of CLI flags:","paragraph4Prefix":"Output is plain text: one line per segment, formatted as","paragraph4Mid":". We parse that into an array of objects and then walk the transcript, matching each transcript segment to the diarization segment with maximum time overlap. Mic-sourced segments are always labeled","paragraph4Suffix":" — source beats voice, every time.","paragraph5Prefix":"The 0.5 threshold is not a guess. We tuned it against our own eval harness (","paragraph5Suffix":") on real meeting recordings. Higher thresholds under-cluster — one speaker gets split into two. Lower thresholds over-cluster — two similar voices get merged. 0.5 is the sweet spot we converged on."},"live":{"title":"Live Diarization: Labels As You Speak","lead":"Batch-only diarization is accurate but invisible during a call. Live-only is immediate but noisy. OpenWhispr runs both: live labels appear during the meeting, and a full batch pass refines them after.","paragraph1Prefix":"The live path lives in","paragraph1Mid":". It taps the system audio stream at 16kHz, feeds every 32-millisecond frame into Silero VAD, and accumulates speech frames until a segment either closes (after ~512ms of silence) or hits the live-identification cadence (every 1 second once ≥1.6 seconds have accumulated). When it triggers, it extracts a CAM++ embedding, compares it against the in-memory map of active-call embeddings and stored speaker profiles, and fires an IPC event —","paragraph1Suffix":" — with the best match. The React side picks that up and updates the transcript bubble in place.","paragraph2Prefix":"Once a human has manually set a speaker name in the UI, that assignment is","lockedWord":"locked","paragraph2Mid":". The post-meeting batch pass is not allowed to overwrite it, even if its own analysis disagrees. This lock behavior lives in","paragraph2Suffix":", and it matters: the worst possible UX is letting an automated pass silently overwrite a user's correction.","paragraph3":"When the recording stops, we run the full offline diarization on the complete WAV file. Batch has full context — it sees the whole meeting, not just the last 1.6 seconds — and produces a cleaner final labeling. The batch result gets reconciled with live labels, respects every user lock, and replaces provisional guesses with the more accurate grouping.","calloutTitle":"Why hybrid beats either alone","calloutBody":"Live-only is fast but noisy on short segments. Batch-only is accurate but silent until the meeting ends. Hybrid gives you both — instant feedback during the call, and a final transcript that reflects the full-context analysis. The cost is one extra reconciliation step, which runs in the background once and is effectively free."},"profiles":{"title":"Voice Profiles: Remembering Speakers Across Meetings","lead":"Once you label a speaker “Alice” in one meeting, OpenWhispr labels her automatically in every future meeting — without you doing anything, and without her voice fingerprint ever leaving your device.","paragraph1":"The mechanism is straightforward: every speaker's centroid embedding gets stored in your local SQLite database. When new embeddings appear on a future call, we compare them by cosine similarity against every stored profile. A match above threshold auto-assigns the name. Nothing about this touches a server. The three tables that power it are plain SQLite:","matchingIntro":"Matching is three-tier, to balance confidence against false positives:","matchingTiers":[{"label":"Cosine ≥ 0.70:","body":"auto-confirm. The label appears immediately with no prompt."},{"label":"0.55 ≤ Cosine < 0.70:","body":"suggest. The UI shows “Is this Alice?” with confirm / dismiss controls, and waits for your input."},{"label":"Cosine < 0.55:","body":"stay anonymous. The segment remains Speaker N until you name them manually."}],"paragraph2Prefix":"When you confirm a match, the stored centroid updates via running mean rather than a raw overwrite. The formula is simple:","paragraph2Suffix":". This handles voice drift — a cold morning, a different mic, a bad cable — without breaking future matches. The more samples of Alice's voice the system has seen, the more stable her profile becomes.","paragraph3Prefix":"New profiles also trigger","retroactiveWord":"retroactive relabeling","paragraph3Suffix":". When you name a speaker for the first time, a background task walks every past meeting's stored embeddings, finds matches, and updates previously-unnamed speakers in historical transcripts. User-locked mappings are never touched. The result: naming Alice once retroactively labels her in every prior call she attended.","paragraph4Prefix":"If Google Calendar is connected and the meeting has attendees, the speaker-picker UI pre-populates with their names and emails. One click maps","speakerTwo":"Speaker 2","paragraph4Suffix":" and persists forever — attendees and voice profiles stay linked, so the right names show up even across devices."},"privacy":{"title":"Where Your Data Actually Lives","lead":"Audio never leaves your device. That is not a marketing line — it is a mechanical property of the code, and because OpenWhispr is open source you can verify it yourself.","intro":"Here is exactly what happens to each piece of data at each step.","tableHeaderWhat":"What","tableHeaderWhere":"Where it lives","tableHeaderLeaves":"Leaves your device?","tableRows":[{"what":"Raw meeting audio","where":"OS temp directory, deleted after diarization","leaves":"No — ever","leavesEmphasis":true},{"what":"Diarization ONNX models","where":"~/.cache/openwhispr/diarization-models/","leaves":"Downloaded once on first launch, then never again"},{"what":"Speaker embeddings (voice fingerprints)","where":"Local SQLite BLOB","leaves":"No — ever","leavesEmphasis":true},{"what":"Speaker-to-name mappings","where":"Local SQLite table","leaves":"No — ever","leavesEmphasis":true},{"what":"Transcript text","where":"Local SQLite (optional cloud sync if you turn it on)","leaves":"Only if you opt in"}],"paragraph1Prefix":"The raw PCM audio from a meeting is written to a file in your OS temp directory, handed to the locally-spawned sherpa-onnx binary, and deleted in a","paragraph1Mid":"block when diarization completes. There is no code path in OpenWhispr's diarization pipeline that opens a network socket to upload audio. Because the code is open source, you can grep","paragraph1Mid2":", and","paragraph1Mid3":"for","paragraph1Mid4":", or","paragraph1Suffix":"and confirm that yourself.","paragraph2Prefix":"IPC between Electron's main and renderer processes uses the built-in","paragraph2Suffix":"— a kernel-level pipe inside a single process, not a network socket. Speaker embeddings and voice profiles are stored as SQLite BLOBs in the same database file as your notes, with the same OS-level file permissions.","calloutTitle":"One exception, transparently","calloutBody":"The only network activity in this feature is the one-time model download on first launch — three ONNX files fetched from the sherpa-onnx GitHub releases CDN. We document this explicitly because “one network call during setup, zero after” is a very different privacy story from “always-on cloud inference,” and as an open source meeting assistant we would rather tell you than have you discover it."},"limitations":{"title":"Honest Limitations","lead":"Local diarization is good, not perfect. Here are the cases where it genuinely struggles, and what we do about each one.","items":[{"label":"Overlapping speakers.","body":"Two people talking at once is inherently lossy for single-label attribution. Pyannote 3.0 detects the overlap region, but a single speaker_id per segment is not enough to represent “both people talking.” We mark those segments as provisional and let you correct them in the UI."},{"label":"Cold-start accuracy.","body":"The first meeting with a new colleague uses only the generic VoxCeleb-trained features. Once you label them once and the profile has a few samples, subsequent meetings are noticeably more accurate."},{"label":"Short utterances.","body":"“Yeah,” “okay,” a quick laugh — anything under ~0.8 seconds cannot produce a reliable embedding. We drop those segments from the embedding stage and fall back to label propagation from the surrounding context."},{"label":"Far-field or low-SNR audio.","body":"Someone on a phone speakerphone in a noisy room will degrade gracefully but visibly. Automatic gain control and VAD tuning help, but no amount of model sophistication completely overcomes bad physics."},{"label":"CPU-only inference.","body":"A GPU would be faster, but we deliberately don't require one. In practice, throughput on a modern CPU is fine: ~100 milliseconds per embedding, and about 30 seconds of batch diarization for a 45-minute meeting on an M1 Mac."}],"outroPrefix":"We publish our eval harness (","outroSuffix":") in the open source repo so you can measure any of this yourself, on your own recordings — not our cherry-picked demo clips. That feels more honest than a glossy number in a marketing deck."},"whySherpa":{"title":"Why We Chose Sherpa-ONNX","lead":"We needed a diarization runtime that was cross-platform, Python-free, CPU-fast, and open source. Sherpa-ONNX was the only option that hit all four.","paragraph1":"Our requirements were concrete. It had to run on macOS (both Intel and Apple Silicon), Windows x64, and Linux x64 without separate code paths. It could not bundle PyTorch — the runtime alone is about 1.5GB and essentially requires a GPU to be usable at interactive latencies. It could not require CUDA. And the license had to be permissive enough to ship in a commercial-but-open-source desktop app.","evaluatedIntro":"Here is what we actually evaluated:","evaluated":[{"label":"pyannote.audio (Python):","body":"ruled out. Bundling PyTorch in an Electron app is a non-starter."},{"label":"NVIDIA NeMo MSDD:","body":"PyTorch-dependent, no first-party ONNX export, GPU-centric."},{"label":"Picovoice Falcon:","body":"commercial per-active-user license plus cloud-activation on first run."},{"label":"Apple CoreML + Speech framework:","body":"macOS-only, no cross-platform parity."},{"label":"sherpa-onnx","body":"Apache 2.0, native ONNX binary, ~35MB of models, spawn-and-read-stdout interface, actively maintained by k2-fsa. Same toolchain powers production ASR in other shipping products."}],"outroPrefix":"One honest caveat: sherpa-onnx's offline speaker diarization binary is relatively new. We version-pin to","outroSuffix":" and run our full eval harness against every update before bumping the pin. That's the right trade-off for something this load-bearing."},"faqHeading":"Frequently Asked Questions","faqs":[{"question":"Is there a private meeting note taker that doesn't upload my audio to the cloud?","answer":"Yes. OpenWhispr is a free, open source meeting note taker where your audio never leaves your device. Transcription runs locally with OpenAI Whisper or NVIDIA Parakeet, speaker diarization runs locally with pyannote segmentation and CAM++ embeddings, and voice fingerprints are stored in a local SQLite file on your own machine — not on our servers."},{"question":"Can Otter, Fireflies, or Granola see my meetings?","answer":"Yes. Cloud-based AI note takers upload your meeting audio to their servers in order to transcribe and diarize it. Speaker labels, transcripts, and voice embeddings are computed in their infrastructure and stored in their databases. If you need an alternative that keeps every byte of audio on your device, you need a local-first tool like OpenWhispr."},{"question":"What is speaker diarization?","answer":"Speaker diarization is the process of labeling who spoke when in a recording, as distinct from transcription, which answers what was said. A diarized transcript shows each utterance tagged with a speaker label — Alice, Bob, Speaker 3 — instead of one undifferentiated wall of text."},{"question":"Does local speaker diarization actually work, or is it a downgrade from the cloud?","answer":"It works. OpenWhispr's pipeline — pyannote-segmentation-3.0 plus CAM++ embeddings from the 3D-Speaker project, run via sherpa-onnx — achieves diarization error rates in the same ballpark as commercial cloud APIs on standard benchmarks. The main honest trade-offs are overlapping-speaker handling and cold-start accuracy on voices the system has never heard before."},{"question":"How much disk space do the local diarization models use?","answer":"About 45MB in total: pyannote segmentation (6.6MB) plus the CAM++ speaker embedding model (28MB) plus Silero VAD (2MB). All three are downloaded once on first launch and cached at ~/.cache/openwhispr/diarization-models/. After that they stay on disk and never re-download."},{"question":"Do I need a GPU to run OpenWhispr's local diarization?","answer":"No. Everything runs on CPU via ONNX Runtime. Modern Apple Silicon and recent x86 CPUs handle it comfortably — in practice, about 30 seconds of batch diarization for a 45-minute meeting on an M1 Mac, with live speaker labels appearing within a second or two during the call itself."},{"question":"Is OpenWhispr actually open source?","answer":"Yes. The diarization code, the transcription pipeline, and the rest of the desktop app are open source on GitHub. The privacy claims in this article are directly verifiable: grep the diarization helpers for fetch, axios, or http and you will not find any network calls in the diarization path."},{"question":"What happens when two people talk over each other on a call?","answer":"Pyannote 3.0 explicitly detects and annotates overlap regions, but single-label attribution during overlap is inherently lossy. We mark those segments as provisional in the transcript and let you correct them by clicking the speaker label. Your manual corrections are locked and are never overwritten by later automated passes."}],"cta":{"title":"Try the Private Meeting Note Taker","body":"OpenWhispr is a free, open source meeting assistant where your audio never leaves your device. Local diarization, local transcription, local voice profiles. Works on macOS, Windows, and Linux.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":[{"href":"/blog/local-vs-cloud-transcription","label":"Local vs cloud transcription"},{"href":"/blog/how-whisper-ai-works","label":"How Whisper AI works"},{"href":"/blog/local-ai-voice-data-privacy","label":"The case for local AI"},{"href":"/blog/why-open-source-voice-tools","label":"Why open source voice tools"},{"href":"/blog","label":"All blog posts"}]},"blogLocalVsCloud":{"metadata":{"title":"Local vs Cloud Speech-to-Text: Privacy, Speed & Cost","description":"Compare local vs cloud speech-to-text on privacy, accuracy, speed, and cost. See why on-device transcription keeps your audio private and free.","keywords":["local vs cloud transcription","local speech to text","cloud transcription privacy","offline transcription","local voice recognition","private speech to text","on-device transcription","cloud ASR vs local ASR","nvidia parakeet","whisper vs parakeet"]},"schema":{"title":"Local vs Cloud Transcription: Privacy, Speed, and Accuracy Compared","description":"An honest comparison of local and cloud-based speech-to-text — privacy, latency, accuracy, cost, and offline capability."},"breadcrumb":{"home":"Home","blog":"Blog","article":"Local vs Cloud Transcription"},"sections":{"factCheck":"Fact-Check Snapshot","howCloudWorks":"Cloud Transcription","howLocalWorks":"Local Transcription","privacy":"Privacy","accuracy":"Accuracy","speed":"Speed & Latency","cost":"Cost","offline":"Offline Capability","comparison":"Comparison Table","recommendations":"Recommendations","sources":"Sources"},"hero":{"title":"Local vs Cloud Transcription: Privacy, Speed, and Accuracy Compared","subtitle":"Your voice data is valuable. Here is what happens to it depending on where you process it.","date":"February 8, 2026","tag":"Comparison"},"intro":{"lead":"Local transcription processes your audio entirely on your device","body1":"using models like OpenAI Whisper and NVIDIA Parakeet, running via optimized runtimes such as whisper.cpp. No audio ever leaves your machine.","cloudLead":"Cloud transcription sends your audio to remote servers","body2":"operated by providers like Google, AWS, Microsoft, Deepgram, or AssemblyAI, where large GPU clusters process it and return a text transcript. The key tradeoffs between the two approaches come down to five factors: privacy, latency, accuracy, cost, and offline capability. This article examines each one with real numbers and real policies, so you can make an informed decision.","lastUpdated":"Last updated: February 17, 2026. Quantitative and policy claims are cross-checked against at least two primary sources."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","item1":{"lead":"Open local models are production-capable:","body":"Whisper and Parakeet have public model cards and open implementations suitable for on-device inference.","source1":"OpenAI Whisper","source2":"NVIDIA Parakeet"},"item2":{"lead":"Cloud STT is usage-priced:","body":"major providers meter by audio volume and model tier.","source1":"Google pricing","source2":"AWS pricing"},"item3":{"lead":"Provider data handling differs:","body":"logging/retention settings and opt-outs are service-specific and must be configured.","source1":"Google data logging","source2":"AWS AI opt-out policy"},"item4":{"lead":"Regulatory context matters:","body":"biometric and sensitive-data obligations vary by jurisdiction and use case.","source1":"GDPR official text","source2":"EU AI Act official text"},"item5":{"lead":"OpenWhispr deployment model:","body":"OpenWhispr defaults to local processing and also supports BYOK cloud routing when users need it.","source1":"OpenWhispr","source2":"whisper.cpp runtime"}},"howCloudWorks":{"title":"How Cloud Transcription Works","paragraph1":"When you use a cloud-based speech-to-text service, your audio follows a multi-step journey. First, your device captures raw audio from your microphone and compresses it into a format suitable for transmission — typically Opus, FLAC, or linear PCM. This compressed audio is then sent over the internet to the provider's API endpoint, usually via HTTPS or a WebSocket connection for real-time streaming.","paragraph2":"On the server side, the provider runs your audio through large neural network models hosted on GPU clusters. These models are typically trained on hundreds of thousands of hours of labeled speech data, far more than what any individual could assemble. The resulting transcript is sent back to your device.","providersIntro":"The major cloud transcription providers include:","providers":{"google":"— Broad language coverage with batch and real-time streaming options.","aws":"— Amazon's offering with automatic language identification, custom vocabulary support, and PII redaction.","azure":"— Deep integration with the Microsoft ecosystem, custom speech models, and real-time transcription.","deepgram":"— Known for speed and developer experience, with their Nova-3 model offering strong accuracy at competitive pricing.","assemblyai":"— Focused on accuracy and developer tooling, with built-in speaker diarization and content moderation."},"paragraph3":"Cloud latency is variable because it includes capture, upload, server-side inference, and response delivery. In well-connected environments it can feel fast, while unstable networks or high-latency regions can make the experience noticeably slower. Batch transcription can take from seconds to minutes depending on file length and queueing."},"howLocalWorks":{"title":"How Local Transcription Works","paragraph1":"Local transcription runs the speech recognition model directly on your device. There is no network involved — audio goes from your microphone straight into a model running on your CPU or GPU, and the text comes out the other side. The entire process happens in memory on your machine.","paragraph2a":"The breakthrough that made high-quality local transcription practical was OpenAI's release of Whisper in September 2022 — an open-source model trained on 680,000 hours of multilingual audio. More recently, NVIDIA released its Parakeet family of ASR models, which achieve state-of-the-art accuracy on English benchmarks and are available under open licenses. Shortly after, Georgi Gerganov created","paragraph2b":", a C/C++ reimplementation optimized for CPU inference, which brought Whisper's accuracy to everyday hardware without requiring a dedicated GPU.","modelTable":{"intro":"Whisper comes in several model sizes, each trading accuracy for speed:","headerModel":"Model","headerParameters":"Parameters","headerVram":"VRAM","headerSpeed":"Relative Speed","rows":[{"model":"Tiny","params":"39M","vram":"~1 GB","speed":"~10x"},{"model":"Base","params":"74M","vram":"~1 GB","speed":"~7x"},{"model":"Small","params":"244M","vram":"~2 GB","speed":"~4x"},{"model":"Medium","params":"769M","vram":"~5 GB","speed":"~2x"},{"model":"Large-v3","params":"1550M","vram":"~10 GB","speed":"1x (baseline)"},{"model":"Turbo","params":"809M","vram":"~6 GB","speed":"~8x"}]},"paragraph3":"On Apple Silicon (M1 through M4), whisper.cpp runs inference fully on the GPU via Metal, achieving real-time or faster-than-real-time speeds even with the medium model. The Turbo model — an optimized variant of large-v3 with minimal accuracy loss — runs at high throughput on modern MacBooks, which is why it is commonly used for low-latency local dictation workflows.","paragraph4a":"Other local transcription engines include","paragraph4b":"(lightweight, lower accuracy) and","paragraph4c":"(next-gen Kaldi with ONNX runtime support). However, OpenAI Whisper via whisper.cpp and NVIDIA Parakeet remain the gold standard for local transcription quality."},"privacy":{"title":"Privacy: The Core Tradeoff","paragraph1":"Privacy is the single biggest differentiator between local and cloud transcription. With local processing, the question of \"what happens to my audio data?\" has a simple answer: nothing. It stays on your device, in your RAM, and is discarded after transcription. There is no third party to trust, no policy to read, no data breach to worry about.","paragraph2":"Cloud transcription requires you to trust your provider. Here is what the major providers actually say about how they handle your audio:","google":{"body1":"Google's data logging for Speech-to-Text is off by default. When disabled, Google states that audio data is processed in memory, used only to return the transcription result, and not stored on servers. However, if you opt into data logging (or use certain features that require it), your audio and transcriptions may be stored and used to improve Google's models. Google's","link":"data logging documentation","body2":"details these distinctions.","footnote":"Data processing location can be specified by region, which matters for GDPR compliance."},"aws":{"body1":"By default, AWS may use content processed by Amazon Transcribe to develop and improve AWS AI/ML services. However, you can opt out of this by using","link":"AWS AI service opt-out policies","body2":". When opted out, AWS states that your content is not stored or used for service improvement. Audio data is encrypted in transit and at rest.","footnote":"Important: the opt-out is not the default. You have to actively configure it."},"azure":{"body1":"Azure does not use customer data to improve its base models by default. Their","link":"data privacy documentation","body2":"states that audio sent to the Speech service is processed and immediately deleted from their servers. If you create a custom model, the training data you provide is stored in the same region as the resource until you explicitly delete it.","footnote":"Azure offers disconnected container deployment for fully on-premises processing."},"regulatory":{"title":"Regulatory Considerations","gdprLead":"GDPR:","gdprBody":"Voice recordings are considered biometric data under EU law. Sending audio to US-based cloud providers raises data transfer concerns under the GDPR, even with Standard Contractual Clauses (SCCs) or the EU-US Data Privacy Framework. Local processing avoids this entirely by keeping biometric data within the data subject's own device.","hipaaLead":"HIPAA:","hipaaBody":"Healthcare providers using cloud transcription must ensure their provider offers a Business Associate Agreement (BAA). Google, AWS, and Azure all offer BAAs, but the configuration requirements are strict. Local transcription sidesteps HIPAA data handling rules because protected health information never leaves the covered entity's control."},"localAdvantage":{"title":"The Local Advantage","body":"With local transcription, there is no privacy policy to parse, no data retention schedule to trust, and no opt-out toggle to find. Your audio is processed in RAM and never written to disk (unless you choose to save a recording). This is a fundamentally different trust model — you do not need to trust anyone, because the data never leaves your machine."}},"accuracy":{"title":"Accuracy: How Close Is Local?","paragraph1":"Let us be honest: the top cloud providers still have an edge in raw accuracy for certain use cases. They train on vast proprietary datasets, offer real-time model updates, and can leverage context-specific custom vocabularies. For challenging audio — heavy accents, noisy environments, domain-specific jargon — cloud services like Google's Chirp 2 or Deepgram's Nova-3 tend to perform well.","paragraph2":"That said, the gap has narrowed dramatically. OpenAI Whisper large-v3, trained on 680,000 hours of multilingual audio, achieves competitive Word Error Rates (WER) across most common languages. NVIDIA's Parakeet models have pushed the bar even further, achieving some of the lowest WERs on standard English benchmarks. Independent benchmarks consistently show these open models performing within a few percentage points of commercial cloud services for English, and sometimes outperforming them for specific languages in the long tail.","paragraph3":"The Turbo model — a distilled variant of large-v3 with 809M parameters instead of 1.55B — retains most of the accuracy while running roughly 8x faster. For real-time dictation where you are speaking clearly into a decent microphone, the difference between Turbo and a cloud API is negligible for most users.","cloudAdvantages":{"title":"Cloud Accuracy Advantages","item1":"Custom vocabulary and domain adaptation","item2":"Continuous model improvements server-side","item3":"Better speaker diarization (who said what)","item4":"Automatic punctuation and formatting tuned per language"},"localAdvantages":{"title":"Local Accuracy Advantages","item1":"No audio compression artifacts from network transmission","item2":"Consistent, deterministic performance — no server variability","item3":"Full uncompressed 16kHz audio fed directly to the model","item4":"Whisper large-v3 and Parakeet match cloud for clear-speech dictation"},"bottomLine":"Bottom line: if you are dictating in a quiet environment with a decent microphone, models like Whisper Turbo, Whisper large-v3, or Parakeet running locally will give you results indistinguishable from cloud services for most practical purposes. Cloud pulls ahead for noisy multi-speaker scenarios, niche languages, and domain-specific vocabularies."},"speed":{"title":"Speed and Latency","paragraph1":"Latency matters most for real-time dictation, where you want to see your words appear immediately after speaking. Here, local and cloud transcription have fundamentally different latency profiles.","cloudBreakdown":{"title":"Cloud Latency Breakdown","item1":"Audio capture + encoding","item2Lead":"Network upload and download","item2Rest":"(variable by route quality)","item3Lead":"Server-side queueing + inference","item3Rest":"(variable by provider load)","summary":"User-perceived latency can vary significantly by network and region."},"localBreakdown":{"title":"Local Latency (Apple Silicon)","item1Lead":"Tiny/base:","item1Rest":"fastest response, lower quality ceiling","item2Lead":"Small:","item2Rest":"balanced speed and quality","item3Lead":"Turbo:","item3Rest":"high-quality real-time dictation candidate","item4Lead":"Large models:","item4Rest":"highest quality, more compute-heavy","summary":"No network dependency. Consistent regardless of connection."},"paragraph2":"For real-time streaming, cloud providers can return partial hypotheses quickly when connectivity is strong. But this still requires a stable low-latency connection. On a plane, in congested public Wi-Fi, or behind restrictive VPN routes, perceived latency can degrade quickly.","paragraph3":"Local transcription with whisper.cpp is not real-time streaming in the traditional sense. Most local implementations use a \"push-to-talk\" pattern: you speak, the audio is buffered, and then the model processes the complete segment. With the Turbo model on an Apple M-series chip, this processing step is fast enough that the delay feels negligible for dictation — typically under a second for short utterances.","footnote":"For long-form transcription of pre-recorded files, cloud services can parallelize across GPU clusters and often finish large batches quickly. Local throughput is bounded by your machine and model choice, so runtime can be slower for very long recordings."},"cost":{"title":"Cost: Free vs. Pay-Per-Minute","paragraph1":"Local transcription costs nothing beyond the hardware you already own. There is no API key, no billing dashboard, no per-minute charge. You download a model file (ranging from 75MB for Tiny to 3GB for Large-v3), and you are done. Forever.","paragraph2":"Cloud providers charge for usage (per-minute or per-hour, depending on vendor and model). Plans and discounts change frequently, so verify current rates before budgeting:","table":{"headerProvider":"Provider","headerBatch":"Batch / Pre-recorded","headerStreaming":"Streaming / Real-time","headerFreeTier":"Free Tier","rows":[{"provider":"Google Speech-to-Text","batch":"Usage-priced (tiered by model)","streaming":"Usage-priced (tiered by model)","free":"Trial/free quotas vary by account"},{"provider":"AWS Transcribe","batch":"Usage-priced","streaming":"Usage-priced","free":"Intro free tier (time-limited)"},{"provider":"Azure Speech","batch":"Varies by region/model","streaming":"Varies by region/model","free":"Limited free quota"},{"provider":"Deepgram","batch":"Usage-priced by model tier","streaming":"Usage-priced by model tier","free":"Credit-based trial"},{"provider":"AssemblyAI","batch":"Usage-priced by model tier","streaming":"Usage-priced by model tier","free":"Credit-based trial"},{"provider":"Local (Whisper / Parakeet)","batch":"Free","streaming":"Free","free":"Unlimited, forever"}]},"estimate":{"title":"How to Estimate Your Cloud Spend","body1a":"Use this simple formula:","formula":"annual minutes of audio x vendor rate","body1b":"If your team dictates heavily, total cost scales linearly with usage and seat count.","body2":"Local inference (Whisper/Parakeet) avoids ongoing API charges, which is why many teams keep local as the default and only route special cases to cloud."},"footnote":"Note: cloud pricing often includes additional charges for features like speaker diarization, PII redaction, or custom vocabularies. The base prices above are for standard transcription only."},"offline":{"title":"Offline Capability: When Local Wins Definitively","paragraph1":"This one is binary: cloud transcription requires internet, local transcription does not. For many people, this is the deciding factor.","scenariosIntro":"Scenarios where offline capability is not optional:","travel":{"title":"Travel and Remote Work","body":"Flights, trains, rural areas, developing regions — anywhere with spotty or nonexistent Wi-Fi. Local transcription works exactly the same at 35,000 feet as it does at your desk."},"airGapped":{"title":"Air-Gapped Environments","body":"Government facilities, defense contractors, secure research labs, and financial institutions often operate on air-gapped networks where outbound internet access is blocked entirely. Local transcription is the only option."},"fieldWork":{"title":"Field Work","body":"Journalists in conflict zones, researchers in remote field stations, healthcare workers in rural clinics — these professionals need reliable transcription in environments where cell service may be unreliable or unavailable."},"reliability":{"title":"Reliability","body":"Even in well-connected offices, cloud services experience outages. AWS, Google Cloud, and Azure have all had multi-hour incidents that affected speech APIs. Local transcription has no external dependency that can go down."}},"comparisonTable":{"title":"Side-by-Side Comparison","headerFactor":"Factor","headerLocal":"Local","headerCloud":"Cloud","headerNotes":"Notes","rows":[{"factor":"Privacy","local":"full","cloud":"none","notes":"Local: data never leaves device. Cloud: audio sent to third-party servers."},{"factor":"Accuracy (English)","local":"partial","cloud":"full","notes":"Cloud models are marginally better; Whisper large-v3 and NVIDIA Parakeet close the gap significantly."},{"factor":"Accuracy (Multilingual)","local":"partial","cloud":"full","notes":"Cloud providers have broader language coverage with custom-trained models."},{"factor":"Latency","local":"full","cloud":"partial","notes":"Local avoids network round-trips. Cloud latency depends on connection quality and region."},{"factor":"Cost","local":"full","cloud":"none","notes":"Local has no per-minute API billing. Cloud services use usage-based pricing."},{"factor":"Offline Use","local":"full","cloud":"none","notes":"Local works anywhere. Cloud requires internet."},{"factor":"Setup Ease","local":"partial","cloud":"full","notes":"Cloud is an API call. Local requires model downloads and hardware consideration."},{"factor":"Speaker Diarization","local":"none","cloud":"full","notes":"Multi-speaker identification is a strong cloud advantage."},{"factor":"Scalability","local":"none","cloud":"full","notes":"Cloud scales infinitely. Local is bound by your hardware."}]},"recommendations":{"title":"When to Choose What","intro":"Neither approach is universally better. The right choice depends on what you prioritize.","chooseLocal":{"title":"Choose Local If...","item1":"Privacy is non-negotiable (medical, legal, personal journals)","item2":"You need offline or air-gapped operation","item3":"You want zero ongoing costs for transcription","item4":"Your primary use case is dictation (single speaker, clear audio)","item5":"You are subject to GDPR, HIPAA, or similar regulations"},"chooseCloud":{"title":"Choose Cloud If...","item1":"You need speaker diarization for multi-person meetings","item2":"Maximum accuracy for difficult audio conditions","item3":"You are building an application that needs to scale to many users","item4":"You need real-time streaming transcription","item5":"Your hardware is limited and cannot run large models"},"hybrid":{"title":"The Hybrid Approach","body":"You do not have to choose one exclusively. Some applications — including OpenWhispr — support both modes: use local models like Whisper and Parakeet by default for privacy and zero cost, and optionally bring your own cloud API key (OpenAI, Deepgram, etc.) when you need the extra accuracy or features that cloud provides. This \"BYOK\" (Bring Your Own Key) model gives you the best of both worlds without locking you into either approach."}},"sourcesSection":{"title":"Sources and Further Reading","whisper":"— Original model repository with model sizes, benchmarks, and documentation.","whisperCpp":"— C/C++ port of Whisper optimized for CPU and Apple Silicon GPU inference.","parakeetLabel":"NVIDIA Parakeet RNNT model card","parakeet":"— Open ASR model metrics and usage guidance.","googlePricingLabel":"Google Cloud Speech-to-Text Pricing","googlePricing":"— Current pricing tiers for Google's speech API.","awsPricingLabel":"AWS Transcribe Pricing","awsPricing":"— Amazon's current usage-based pricing for batch and streaming.","azurePricingLabel":"Azure Speech Services Pricing","azurePricing":"— Microsoft's pricing for real-time and batch speech-to-text.","deepgramPricingLabel":"Deepgram Pricing","deepgramPricing":"— Deepgram model-tier pricing details.","assemblyaiPricingLabel":"AssemblyAI Pricing","assemblyaiPricing":"— AssemblyAI usage and model-tier pricing details.","googleLoggingLabel":"Google Speech-to-Text Data Logging","googleLogging":"— Google's documentation on how audio data is handled and stored.","awsOptOutLabel":"AWS AI Service Opt-Out Policies","awsOptOut":"— How to prevent AWS from using your content to improve their AI services.","azurePrivacyLabel":"Azure Speech Data Privacy","azurePrivacy":"— Microsoft's data privacy and security documentation for Speech Services.","whisperPaperLabel":"Robust Speech Recognition via Large-Scale Weak Supervision","whisperPaper":"— The original Whisper paper by Radford et al. (2022), detailing training methodology and WER benchmarks.","gdprLabel":"GDPR (official EU text)","gdpr":"— Legal baseline for personal/biometric data handling.","aiActLabel":"EU AI Act (official EU text)","aiAct":"— AI risk-tier obligations relevant to biometric and high-risk systems.","openwhispr":"— Example of a local-first + optional BYOK cloud workflow in one app."},"cta":{"title":"Try Both Approaches in One App","body":"OpenWhispr supports both local models (OpenAI Whisper and NVIDIA Parakeet) for privacy and zero cost, and bring-your-own cloud API keys when you need maximum accuracy. Open source, free forever.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":{"localAi":"The case for local AI","offline":"Offline transcription","allBlog":"All blog posts","allComparisons":"All comparisons"}},"blogModelSizes":{"metadata":{"title":"Whisper Model Sizes: Complete Guide | OpenWhispr","description":"Every Whisper model size compared: parameters, speed, accuracy, and RAM. Choose between tiny, small, large-v3, and turbo for local transcription.","keywords":["whisper model sizes","whisper tiny vs large","whisper model comparison","which whisper model to use","whisper large v3","whisper model accuracy","whisper model speed","whisper model parameters","whisper.cpp model size"]},"schema":{"title":"Whisper Model Sizes Explained: Tiny vs Base vs Small vs Medium vs Large","description":"A practical guide to choosing the right Whisper model. Parameters, speed, accuracy, and hardware requirements compared."},"breadcrumb":{"home":"Home","blog":"Blog","article":"Whisper Model Sizes Explained"},"layout":{"title":"Whisper Model Sizes Explained: Tiny vs Base vs Small vs Medium vs Large","subtitle":"Every Whisper model, compared. Parameters, speed, accuracy, and which one you should actually use.","date":"February 5, 2026","tag":"Blog"},"sectionsToc":{"factCheck":"Fact-Check Snapshot","comparisonTable":"Model Comparison","tiny":"Tiny (39M)","base":"Base (74M)","small":"Small (244M)","medium":"Medium (769M)","large":"Large (1.55B)","turbo":"Turbo (809M)","englishVsMultilingual":"English vs Multilingual","hardware":"Hardware Requirements","recommendations":"Recommendations","openwhispr":"OpenWhispr","sources":"Sources"},"cards":{"resources":"Resources","speed":"Speed","accuracy":"Accuracy","bestFor":"Best for:"},"intro":{"lead":"OpenAI's Whisper comes in 9 model variants across 5 size categories:","body1":"tiny (39M parameters), base (74M), small (244M), medium (769M), and large (1.55B). Each size has a multilingual version and an English-only version (except large and turbo, which are multilingual only). There is also a turbo model (809M) — a distilled version of large-v3 that is nearly as accurate but dramatically faster. Larger models are more accurate but slower and require more memory.","body2":"For the whisper.cpp quantized (GGML) format used by most local transcription apps, model files range from 75 MB (tiny) to 2.9 GB (large). Memory usage at runtime ranges from around 273 MB for tiny to 3.9 GB for large. The right model depends on your hardware, your language, and how much accuracy you need.","lastUpdated":"Last updated: February 17, 2026. Parameter, memory, and benchmark claims are cross-checked against at least two primary sources."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","item1":{"lead":"Official model families and parameter counts:","body":"tiny/base/small/medium/large plus turbo are documented by OpenAI.","link1":"OpenAI Whisper README","link2":"Whisper model card"},"item2":{"lead":"GGML/GGUF disk and runtime memory context:","body":"local deployment footprints come from whisper.cpp conversion/runtime docs.","link1":"whisper.cpp","link2":"whisper.cpp memory notes"},"item3":{"lead":"Benchmark caveat:","body":"WER varies heavily by dataset and language; LibriSpeech figures are not a universal quality score.","link1":"Whisper paper","link2":"large-v2 model card"},"item4":{"lead":"Turbo model positioning:","body":"turbo is a speed-focused variant with accuracy tradeoffs documented in release notes/model cards.","link1":"turbo model card","link2":"turbo release discussion"},"item5":{"lead":"OpenWhispr usage context:","body":"OpenWhispr exposes multiple Whisper sizes so users can tune speed/accuracy on-device.","link1":"OpenWhispr","link2":"whisper.cpp backend"}},"table":{"title":"The Complete Model Comparison","intro":"This table covers every official Whisper model. \"Relative Speed\" is relative to large (1x baseline). VRAM figures are for PyTorch inference on GPU. The GGML columns show file sizes and RAM usage for whisper.cpp, which is what most desktop apps use.","headers":{"model":"Model","params":"Params","englishOnly":"English-Only","vram":"VRAM (GPU)","ggmlDisk":"GGML Disk","ram":"RAM (whisper.cpp)","speed":"Speed","englishWer":"English WER","multilingualWer":"Multilingual WER"},"notAvailable":"N/A","notesLead":"Notes:","notesBody":"WER (Word Error Rate) figures are approximate averages from the Whisper paper, HuggingFace model cards (LibriSpeech test sets for English), and OpenAI's published benchmarks. Exact WER varies significantly by dataset, language, audio quality, and evaluation methodology. Lower is better. Speed is relative to large (1x = slowest)."},"tiny":{"title":"Tiny (39M Parameters)","para1":"The tiny model is the smallest and fastest Whisper variant. At just 39 million parameters and a GGML file size of 75 MB, it can run comfortably on almost any hardware — including Raspberry Pi, older laptops, and low-end mobile devices. It transcribes at roughly 10x real-time speed relative to the large model, making it the best choice when latency matters more than perfect accuracy.","resources":{"l1":"GGML: 75 MiB on disk","l2":"RAM: ~273 MB runtime","l3":"VRAM: ~1 GB (GPU mode)"},"speed":{"l1":"~10x faster than large","l2":"Near-instant on Apple Silicon","l3":"Real-time on most CPUs"},"accuracy":{"l1":"English WER: ~7.6%","l2":"Multilingual WER: ~12%","l3":"Struggles with accents/noise"},"para2":"On the tiny.en variant (English-only), LibriSpeech test-clean WER is around 5.6%, rising to ~14.9% on test-other (noisy/accented audio). The multilingual tiny model performs slightly worse on English but supports all 99 Whisper languages.","para3":"Tiny is ideal for quick notes, low-stakes dictation, live captioning prototypes, and embedded or edge devices where compute is limited. It is the default starting model for many whisper.cpp-based tools because it downloads in seconds and runs instantly.","bestFor":"Fastest possible transcription, low-resource hardware, Raspberry Pi, real-time applications"},"base":{"title":"Base (74M Parameters)","para1":"The base model doubles the parameter count of tiny (74M vs 39M) while remaining very lightweight. At 142 MiB (GGML) and ~388 MB of RAM at runtime, it still fits comfortably on virtually any modern device. The accuracy improvement over tiny is noticeable — especially on noisy audio and accented speech.","resources":{"l1":"GGML: 142 MiB on disk","l2":"RAM: ~388 MB runtime","l3":"VRAM: ~1 GB (GPU mode)"},"speed":{"l1":"~7x faster than large","l2":"Still very fast on CPU","l3":"Sub-second on Apple Silicon"},"accuracy":{"l1":"English WER: ~5.0%","l2":"Multilingual WER: ~10%","l3":"Better on noisy audio than tiny"},"para2":"The base.en variant achieves around 4.3% WER on LibriSpeech test-clean and ~12.8% on test-other. This is a meaningful improvement over tiny.en (5.6% and 14.9% respectively), particularly on challenging audio.","para3":"Base is a good choice for low-power devices where tiny is not quite accurate enough. It is also popular for real-time transcription pipelines where you want a balance of speed and quality without going above 500 MB of RAM.","bestFor":"Low-power devices that need better accuracy than tiny, real-time streaming, budget hardware"},"small":{"title":"Small (244M Parameters)","para1":"Small is where Whisper starts to feel genuinely accurate. At 244M parameters, it represents a 3.3x jump from base and delivers a significant accuracy improvement — especially for non-English languages and noisy recordings. The GGML file is 466 MiB, and runtime RAM usage is around 852 MB. It runs at about 4x the speed of large.","resources":{"l1":"GGML: 466 MiB on disk","l2":"RAM: ~852 MB runtime","l3":"VRAM: ~2 GB (GPU mode)"},"speed":{"l1":"~4x faster than large","l2":"Fast on modern laptops","l3":"Real-time on Apple Silicon"},"accuracy":{"l1":"English WER: ~3.4%","l2":"Multilingual WER: ~7%","l3":"Handles accents well"},"para2":"The small.en variant reaches about 3.0% WER on LibriSpeech test-clean — a remarkable result for a model that fits in under 500 MB. For many English-only use cases, small.en provides accuracy that is close to medium at a fraction of the resource cost.","para3":"Small is often recommended as the default model for most users. It runs well on any modern laptop (including MacBook Air M1 with 8 GB RAM), delivers good accuracy across languages, and transcribes fast enough for real-time dictation workflows.","recommendationLead":"Our recommendation:","recommendationBody":"If you are not sure which model to start with, start with small. It is the best balance of speed, accuracy, and resource usage for the majority of hardware and use cases.","bestFor":"Most users on modern laptops, default recommended model, good multilingual support"},"medium":{"title":"Medium (769M Parameters)","para1":"Medium is where diminishing returns begin to set in — but the accuracy gains over small are still meaningful, especially for multilingual transcription, technical vocabulary, and challenging audio conditions. At 769M parameters, it requires 1.5 GiB of disk space (GGML) and about 2.1 GB of RAM at runtime.","resources":{"l1":"GGML: 1.5 GiB on disk","l2":"RAM: ~2.1 GB runtime","l3":"VRAM: ~5 GB (GPU mode)"},"speed":{"l1":"~2x faster than large","l2":"Comfortable on 8 GB+ machines","l3":"Needs decent hardware"},"accuracy":{"l1":"English WER: ~2.9%","l2":"Multilingual WER: ~5%","l3":"Strong on technical content"},"para2":"The medium.en model achieves around 3.0% WER on LibriSpeech test-clean and approximately 7.5% on test-other. However, across broader benchmarks — particularly those with diverse accents, background noise, and domain-specific vocabulary — medium consistently outperforms small.","para3":"Medium is a strong choice for professional transcription workflows where accuracy matters but you do not have the hardware (or patience) for large. It runs well on machines with 8 GB or more of RAM and benefits significantly from GPU acceleration.","bestFor":"Professional use, multilingual transcription, technical content, machines with 8 GB+ RAM"},"large":{"title":"Large (1,550M Parameters)","para1":{"part1":"The large model is the most accurate Whisper variant, with 1.55 billion parameters. There are three versions:","part2":"(the original release),","part3":"(trained for 2.5x more epochs with added regularization), and","part4":"(trained on more data with 128 Mel frequency bins instead of 80). Large-v3 is the current recommended version for maximum accuracy."},"resources":{"l1":"GGML: 2.9 GiB on disk","l2":"RAM: ~3.9 GB runtime","l3":"VRAM: ~10 GB (GPU mode)"},"speed":{"l1":"1x (baseline — slowest)","l2":"GPU strongly recommended","l3":"CPU: may be slower than real-time"},"accuracy":{"l1":"English WER: ~2.4% (v3)","l2":"Multilingual WER: ~3.5% (v3)","l3":"Best across all languages"},"versionsTitle":"large-v2 vs large-v3","v2":{"l1":"Trained for 2.5x more epochs than large-v1","l2":"Added regularization for better generalization","l3":"LibriSpeech test-clean WER: ~3.0%","l4":"More stable on some audio types"},"v3":{"l1":"128 Mel bins (vs 80) — finer frequency resolution","l2":"Trained on 1M hours labeled + 4M hours pseudo-labeled audio","l3":"10-20% error reduction over large-v2 across languages","l4":"Added Cantonese language support"},"whichLead":"Which large model to use:","whichBody":"For new projects, use large-v3. It is strictly better than large-v2 on average across languages. However, some users have reported that large-v2 produces fewer hallucinations (generating text not present in the audio) in specific edge cases with very quiet or silent audio segments. If you encounter hallucination issues with large-v3, try large-v2 as a fallback.","bestFor":"Maximum accuracy, professional/medical/legal transcription, low-resource languages, batch processing"},"turbo":{"title":"Turbo (809M Parameters)","para1":"The turbo model is a distilled version of large-v3 that dramatically reduces inference time while retaining most of the accuracy. OpenAI achieved this by pruning the decoder from 32 layers down to just 4 layers — cutting the parameter count from 1,550M to 809M. The encoder (which does the heavy lifting) remains identical to large-v3.","resources":{"l1":"GGML: ~1.6 GiB on disk","l2":"RAM: ~2.3 GB runtime","l3":"VRAM: ~6 GB (GPU mode)"},"speed":{"l1":"~8x faster than large","l2":"Near-large accuracy, near-tiny speed","l3":"Great on GPU with torch.compile"},"accuracy":{"l1":"English WER: ~2.5%","l2":"Multilingual WER: ~3.7%","l3":"Minor quality loss vs large-v3"},"para2":"Turbo is the best choice when you want near-large-v3 accuracy but cannot afford the latency of the full large model. On GPU with optimizations like torch.compile, turbo can achieve up to 4.5x speed improvement over standard inference, making it extremely fast.","limitationLead":"Important limitation:","limitationBody":"The turbo model was not trained for translation tasks. If you need to translate speech from one language to English, use medium or large-v3 instead.","bestFor":"High-accuracy real-time transcription, latency-sensitive apps, when large-v3 is too slow"},"englishVsMultilingual":{"title":"English-Only vs Multilingual Models","para1":{"part1":"For the tiny, base, small, and medium sizes, Whisper offers two variants: a multilingual model and an English-only (","part2":") model. The large and turbo models are multilingual only — there are no English-only versions."},"enTitle":"When to use .en (English-only)","en":{"l1":"You only transcribe English audio","l2":"Better English accuracy at tiny/base size levels","l3":"Slightly faster — no language detection overhead","l4":"Same file size as multilingual equivalent"},"multiTitle":"When to use multilingual","multi":{"l1":"You transcribe non-English audio","l2":"Your audio contains mixed languages","l3":"You need speech translation (to English)","l4":"Required if using large or turbo (no .en variant exists)"},"para2":"The performance gap between .en and multilingual models is most significant at smaller sizes. For tiny and base, the .en variants are noticeably better at English. By the time you reach small and medium, the difference narrows considerably. At the large level, there is only a multilingual model — and it performs excellently on English regardless.","ruleLead":"Rule of thumb:","ruleBody":"If you exclusively use English and are choosing tiny or base, pick the .en variant. For small or larger, the multilingual version works great for English and gives you flexibility for other languages."},"hardware":{"title":"Hardware Requirements: What Runs Where","intro":"The hardware you need depends heavily on the model size and whether you are using whisper.cpp (CPU/Metal/CUDA) or the PyTorch implementation (GPU-focused). Here is what to expect on common hardware classes.","m1Air":{"title":"MacBook Air M1 (8 GB RAM)","r1Label":"tiny/base:","r1":"Instant. Faster than real-time.","r2Label":"small:","r2":"Fast. Real-time or better with Metal.","r3Label":"medium:","r3":"Usable. May be slightly slower than real-time on CPU. Metal helps significantly.","r4Label":"large:","r4":"Possible but tight on RAM. Expect slower than real-time.","r5Label":"turbo:","r5":"Good fit. Near real-time with Metal acceleration."},"mbpPro":{"title":"MacBook Pro M2/M3 (16-36 GB RAM)","r1Label":"tiny/base/small:","r1":"Instant. All well within hardware limits.","r2Label":"medium:","r2":"Fast. Comfortable with Metal.","r3Label":"large:","r3":"Good. Real-time or near it with Metal.","r4Label":"turbo:","r4":"Excellent. Fast real-time transcription."},"midRange":{"title":"Mid-Range Windows/Linux Laptop (8 GB RAM, integrated GPU)","r1Label":"tiny/base:","r1":"Fast. CPU-only is fine.","r2Label":"small:","r2":"Good. Comfortable on most modern CPUs.","r3Label":"medium:","r3":"Workable. May be 2-3x slower than real-time on CPU.","r4Label":"large:","r4":"Challenging. CPU-only will be slow.","r5Label":"turbo:","r5":"Workable. Benefits from Vulkan if available."},"nvidia":{"title":"Desktop with NVIDIA GPU (RTX 3060+, 8 GB+ VRAM)","r1Label":"All models:","r1":"Fast. GPU handles everything comfortably.","r2Label":"large:","r2":"Needs 10 GB VRAM (RTX 3060 12GB or better).","r3Label":"turbo:","r3":"Fastest option with CUDA. Significantly faster than large.","r4":"Use CUDA backend in whisper.cpp for best performance."},"backendsTitle":"whisper.cpp Acceleration Backends","backendsIntro":"whisper.cpp — the C/C++ port that most desktop apps use — supports several hardware acceleration backends that can dramatically improve performance:","metal":{"title":"Metal (Apple Silicon)","body":"Full GPU inference on M1/M2/M3/M4 Macs. Runs the entire model on the GPU with zero memory copies. This is the fastest backend for macOS users and what OpenWhispr uses on Mac."},"coreml":{"title":"Core ML (Apple)","body":"Uses Apple's neural engine for inference. Can be faster than Metal for some model sizes on newer chips, and more power-efficient. Requires converting models to Core ML format first."},"cuda":{"title":"CUDA (NVIDIA)","body":"GPU acceleration for NVIDIA cards. Fastest backend for users with dedicated NVIDIA GPUs. Requires CUDA toolkit. Best choice for large and turbo models on desktop."},"vulkan":{"title":"Vulkan (Cross-vendor GPU)","body":"Works with Intel, AMD, and NVIDIA GPUs. Good fallback for non-NVIDIA systems. Available on Linux and Windows. Performance varies by GPU vendor and driver."},"speedTitle":"Approximate Transcription Speed","speedIntro":"Real-time factor (RTF) for a 60-second audio clip. RTF < 1.0 means faster than real-time. These are rough estimates — actual performance depends on audio content, model quantization, and system load.","speedTable":{"model":"Model","m1":"M1 MacBook Air","m3":"M3 Pro MacBook","intel":"Intel i7 (CPU)","rtx":"RTX 3060 (CUDA)"},"readingLead":"Reading the table:","readingBody":"0.1x means transcribing 60 seconds of audio takes about 6 seconds. 1.0x = real-time. Values above 1.0x mean slower than real-time. All figures use whisper.cpp with default settings and the Metal (Mac) or CUDA (NVIDIA) backend where applicable."},"recommendations":{"title":"Which Model Should You Use?","intro":"Here are concrete recommendations based on common scenarios. When in doubt, start with small and move up or down based on your experience.","fastest":{"q":"\"I want the fastest possible\"","part1":"Use","part2":"or","part3":". Transcribes in milliseconds on modern hardware. Accuracy is rough but usable for quick notes and low-stakes dictation."},"laptop":{"q":"\"I want good accuracy on a laptop\"","part1":"Use","part2":". Best balance of speed and accuracy for modern laptops. If your machine has 16 GB+ RAM,","part3":"is also an excellent choice.","tag":"small or medium"},"accuracy":{"q":"\"I want the best accuracy\"","part1":"Use","part2":". It is the most accurate Whisper model overall. If speed matters too,","part3":"gets you 95% of the accuracy at 8x the speed.","tag":"large-v3 or turbo"},"englishOnly":{"q":"\"I only use English\"","part1":"Use the","part2":"variant of your chosen size for the best English accuracy. For tiny and base, the .en models are noticeably better. For small+, the difference is minimal.","tag":"small.en or medium.en"},"multiLang":{"q":"\"I use multiple languages\"","part1":"Use the multilingual variant.","part2":"or","part3":"for balanced performance.","part4":"for best multilingual accuracy.","tag":"small, medium, or large-v3"},"limited":{"q":"\"I have very limited hardware\"","part1":"Use","part2":"or","part3":". Both run on 1 GB of RAM and even work on Raspberry Pi and other single-board computers. Base offers a meaningful accuracy bump over tiny with minimal added cost.","tag":"tiny or base"}},"openwhispr":{"title":"How OpenWhispr Handles Model Selection","intro":"is an open-source desktop dictation app that uses whisper.cpp under the hood. It lets you download and switch between any Whisper model directly from the settings — no command-line required. Models are downloaded once and stored locally. You can change models at any time to balance speed and accuracy for your hardware.","c1":{"title":"Download any model","body":"Pick from tiny through large-v3 in settings. OpenWhispr downloads the GGML version automatically."},"c2":{"title":"Switch instantly","body":"Change models at any time. No restart needed. Test different sizes to find what works best on your hardware."},"c3":{"title":"Hardware optimized","body":"Uses Metal on Apple Silicon, CUDA on NVIDIA, and optimized CPU inference elsewhere. All local, all private."},"suggestionLead":"Our suggestion:","suggestionPart1":"Start with","suggestionPart2":". If it feels slow, drop down to base or tiny. If you want better accuracy and your hardware can handle it, move up to medium or large-v3. OpenWhispr makes it easy to experiment."},"sources":{"title":"Sources","s1":{"link":"Radford et al. \"Robust Speech Recognition via Large-Scale Weak Supervision\" (2022)","desc":"— The original Whisper paper with full benchmark results."},"s2":{"link":"OpenAI Whisper GitHub Repository","desc":"— Official model table, parameter counts, VRAM requirements, and relative speed figures."},"s3":{"link":"OpenAI Whisper Model Card","desc":"— Official training data composition and evaluation caveats."},"s4":{"link":"whisper.cpp GitHub Repository","desc":"— GGML model sizes, RAM usage, and hardware acceleration details."},"s5":{"link":"Whisper large-v2 Model Card (HuggingFace)","desc":"— LibriSpeech benchmark context used in many WER comparisons."},"s6":{"link":"Whisper large-v3 Model Card (HuggingFace)","desc":"— Training details, architecture changes from large-v2, and multilingual improvements."},"s7":{"link":"Whisper large-v3-turbo Model Card (HuggingFace)","desc":"— Turbo architecture details, decoder pruning specifics, and performance tradeoffs."},"s8":{"link":"OpenAI Whisper Turbo Announcement (GitHub Discussion #2363)","desc":"— Official turbo release notes and cross-language performance comparisons."},"s9":{"link":"OpenWhispr","desc":"— Practical local deployment context for choosing model sizes."}},"cta":{"title":"Try Different Whisper Models in OpenWhispr","body":"Open source, local-first dictation. Download any Whisper model, switch between sizes with one click, and find the perfect balance for your hardware.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · Open source forever"},"internalLinks":{"howWhisperWorks":"How Whisper AI works","linux":"Dictation for Linux","allPosts":"All blog posts","developers":"OpenWhispr for developers","localVsCloud":"Local vs Cloud Transcription"}},"blogOpenSource":{"metadata":{"title":"Why Open Source Matters for Voice Tools | OpenWhispr","description":"Voice tools process some of the most personal data imaginable — your actual voice. Here's why open source is the only approach that earns real trust.","keywords":["open source voice recognition","open source dictation","open source speech to text","why open source matters","voice privacy open source","open source AI tools","transparent voice software","open source whisper","nvidia parakeet"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"Why Open Source Matters"},"schema":{"title":"Why Open Source Matters for Voice Tools","description":"Voice tools process some of the most personal data imaginable. Here's why open source is the only approach that earns real trust."},"layout":{"title":"Why Open Source Matters for Voice Tools","subtitle":"Your voice is biometric data. The software that processes it should be auditable.","date":"February 10, 2026","tag":"Blog"},"sections":[{"id":"fact-check-snapshot","title":"Fact-Check Snapshot"},{"id":"voice-data-is-different","title":"Voice Data Is Different"},{"id":"the-trust-problem","title":"The Trust Problem"},{"id":"what-open-source-gives-you","title":"What Open Source Gives You"},{"id":"the-open-model-effect","title":"The Open Model Effect"},{"id":"open-source-polish","title":"Open Source Polish"},{"id":"business-model","title":"Business Model"},{"id":"what-to-look-for","title":"What to Look For"},{"id":"sources","title":"Sources"}],"opening":{"lead":"Open source voice tools let you verify exactly what happens to your audio.","body":"With proprietary voice software, you are trusting a company's privacy policy. With open source, you can read the code, build it yourself, and confirm that your voice data never leaves your device. This distinction matters more for voice tools than almost any other category of software, because voice data is uniquely personal — it is biometric, it is hard to anonymize, and once it is recorded, it cannot be unheard.","lastUpdated":"Last updated: February 17, 2026. Incident and policy claims are cross-checked against at least two primary sources."},"factCheck":{"title":"Fact-Check Snapshot (Dual Sources)","voiceData":{"lead":"Voice data can be identifying and high-risk:","body":"legal and policy frameworks increasingly treat biometric processing as sensitive.","source1":"GDPR official text","source2":"EU AI Act official text"},"assistants":{"lead":"Major assistants have faced voice-data controversies:","body":"Amazon, Google, and Apple incidents are documented by regulators and/or major news outlets.","source1":"FTC Amazon case","source2":"Reuters Siri settlement"},"cloning":{"lead":"Voice-cloning abuse is an active consumer risk:","body":"US regulators and cybersecurity guidance both flag impersonation risk.","source1":"FTC alert","source2":"IBM risk context"},"openModel":{"lead":"Open model ecosystem is real:","body":"Whisper and downstream runtimes enable auditable local deployments.","source1":"OpenAI Whisper","source2":"whisper.cpp"},"positioning":{"lead":"OpenWhispr positioning:","body":"OpenWhispr is built as an open, local-first voice workflow on top of open ASR.","source1":"OpenWhispr","source2":"Whisper upstream"}},"voiceData":{"title":"Voice Data Is Different","paragraph1":"Text is text. You can strip metadata, anonymize names, and aggregate it into statistics. Voice is none of those things.","paragraph2":"Your voice is a biometric identifier. Like a fingerprint, it is unique to you — voice authentication systems used by banks and government agencies depend on this fact. But that same uniqueness makes voice recordings inherently identifying. A leaked database of text transcriptions is a privacy incident. A leaked database of voice recordings is a biometric breach.","reveal":{"title":"What voice recordings can reveal","items":["Identity — voice biometrics can uniquely identify a speaker","Health conditions — vocal biomarkers can indicate Parkinson's, depression, respiratory illness","Emotional state — stress, anger, fatigue are detectable in vocal patterns","Demographics — age, gender, accent, native language, regional origin"]},"anonymize":{"title":"Why anonymization is harder","items":["You can redact words from a transcript, but you cannot redact a voice from a recording without destroying it","Voice cloning technology means leaked audio can be used to generate deepfakes of your voice","The EU AI Act classifies biometric identification systems — including voice — as high-risk AI, subject to strict compliance requirements"]},"paragraph3a":"Voice cloning has gone from a research curiosity to a commodity service. With just a few seconds of audio, modern voice synthesis models can produce convincing replicas of a person's voice. This is not a hypothetical risk — the","ftcLink":"FTC has warned about AI-powered voice cloning scams","paragraph3b":"since 2023. When the software processing your voice is a black box, you have no way to know whether recordings are stored, transmitted, or used for model training."},"trustProblem":{"title":"The Trust Problem with Proprietary Voice Tools","intro":"The argument for open source is not theoretical. Every major voice assistant has been caught handling audio in ways their users did not expect.","alexa":{"title":"Amazon Alexa","body1":"In April 2019, Bloomberg","reportedLink":"reported","body2":"that Amazon employed thousands of workers worldwide to listen to Alexa voice recordings captured in customers' homes and offices. Workers transcribed, annotated, and reviewed audio clips to improve the speech recognition system. Some reviewers reported hearing recordings they found distressing, including what sounded like a sexual assault. Amazon's response acknowledged the practice but emphasized it applied to \"an extremely small sample.\" The company later added opt-out controls, but the recordings had already been collected under a privacy policy most users never read closely. In 2023, Amazon","settlementLink":"paid $25 million","body3":"to settle FTC charges that it violated children's privacy by retaining kids' Alexa voice recordings indefinitely."},"google":{"title":"Google Assistant","body1":"In July 2019, Belgian broadcaster VRT NWS","obtainedLink":"obtained over 1,000 Google Assistant recordings","body2":"from a contractor. The recordings included bedroom conversations, business calls, and interactions with children — many captured by accidental activations when no one had said \"OK Google.\" Google confirmed that human reviewers listened to about 0.2% of all voice recordings and suspended the practice in Europe after the leak. A subsequent investigation by Hamburg's data protection authority led to a temporary EU-wide ban on the practice."},"siri":{"title":"Apple Siri","body1":"In July 2019, a whistleblower revealed to The Guardian that Apple contractors","heardLink":"regularly heard confidential medical information, drug deals, and sexual encounters","body2":"while grading Siri recordings. Apple had not disclosed the human review program in its privacy documentation. The company apologized, suspended the program globally, and made it opt-in — but the recordings that had already been reviewed were collected without informed consent. Apple later settled a","lawsuitLink":"$$95 million class-action lawsuit","body3":"over Siri privacy violations in January 2025."},"pattern":"The pattern is consistent: a company says \"we take your privacy seriously,\" while the actual data handling practices are only discovered through leaks, whistleblowers, or regulatory action. These were not small companies cutting corners. They were Apple, Google, and Amazon — three of the most technically sophisticated and well-resourced organizations on the planet.","quote":"\"We don't store your data\" is a claim that requires trust. Open source is a claim that requires only reading comprehension."},"whatOpenSource":{"title":"What Open Source Actually Gives You","intro":"Open source is not a marketing label. It is a set of concrete properties that change the trust model between you and the software you use.","properties":[{"title":"Auditability","body":"Anyone can read the source code and verify exactly how audio data is handled. Does the app send recordings to a server? Does it store audio on disk? Does it phone home with telemetry? You do not need to trust a privacy policy — you can check."},{"title":"Reproducibility","body":"You can build the application from source and compare it to the distributed binary. This closes the gap between \"we published the code\" and \"the app you downloaded actually runs this code.\" Reproducible builds are the gold standard of software trust."},{"title":"Community Review","body":"Security researchers, privacy advocates, and experienced developers can independently review the code. Vulnerabilities are found and patched in the open. This is not a theoretical benefit — it is the reason Linux, OpenSSL, and the entire internet infrastructure run on open source."},{"title":"Fork Rights","body":"If the maintainers make decisions you disagree with — adding telemetry, removing features, selling to an acquirer — the community can fork the project and continue development independently. This is not just insurance; it is a structural incentive for maintainers to act in users' interests."},{"title":"No Lock-in","body":"Your workflows are not hostage to a company's business decisions. If a proprietary dictation tool gets acquired, pivots, or shuts down, your investment in learning and configuring it is lost. Open source tools persist as long as anyone cares enough to run them."},{"title":"Longevity","body":"Companies disappear. Open source projects endure. Apache HTTP Server has been running since 1995. PostgreSQL since 1996. GCC since 1987. Dragon NaturallySpeaking was the leading dictation product for decades — then Nuance was acquired by Microsoft, and the standalone product was effectively discontinued. Open source does not have this failure mode."}]},"openModel":{"title":"The Open Model Effect: From Whisper to Parakeet","paragraph1a":"In September 2022, OpenAI","releasedLink":"released Whisper","paragraph1b":"— a general-purpose speech recognition model trained on 680,000 hours of multilingual audio — and open-sourced it under the MIT license. It was a watershed moment for voice tools.","paragraph2":"Before Whisper, high-quality automatic speech recognition (ASR) was expensive and proprietary. Google Cloud Speech-to-Text, Amazon Transcribe, and Microsoft Azure Speech Services all charged per minute of audio and required sending recordings to their servers. The best offline options — CMU Sphinx, Kaldi, VOSK — were functional but noticeably less accurate than the cloud APIs.","paragraph3":"OpenAI Whisper changed that equation overnight. For the first time, a model that matched or exceeded commercial API accuracy was available for anyone to download and run locally, at no cost, with no data leaving their machine. NVIDIA followed a similar path with its Parakeet family of ASR models — achieving state-of-the-art English accuracy and releasing them under open licenses. The trend is clear: the best speech recognition models are increasingly open source.","ecosystem":{"title":"The ecosystem that followed","items":[{"name":"whisper.cpp","desc":"— Georgi Gerganov's C/C++ port, optimized for CPU inference. Runs Whisper on everything from laptops to Raspberry Pis. Over 46,000 GitHub stars."},{"name":"faster-whisper","desc":"— CTranslate2-based reimplementation that achieves up to 4x faster inference than the original with the same accuracy."},{"name":"WhisperX","desc":"— Adds forced alignment and speaker diarization to Whisper, enabling word-level timestamps and multi-speaker transcription."},{"name":"Apps built on open models","desc":"— Dozens of desktop, mobile, and web applications now use OpenAI Whisper or NVIDIA Parakeet as their speech recognition engine, including OpenWhispr, MacWhisper, Buzz, and Vibe."}]},"paragraph4":"The release of OpenAI Whisper is one of the clearest demonstrations of how open source creates ecosystems. A single open model spawned optimized runtimes, new capabilities, and dozens of products that would never have existed if the model remained behind an API paywall. NVIDIA Parakeet continued this pattern, proving that multiple organizations see value in open-sourcing high-quality ASR. Today, any developer can build a voice tool with state-of-the-art accuracy without paying per minute, without sending audio to a third party, and without asking permission."},"polish":{"title":"Open Source Does Not Mean Less Polished","intro":"There is a persistent assumption that open source means rougher edges. That you trade polish for freedom. This was arguably true in 2005. It is not true in 2026.","apps":[{"name":"Firefox","desc":"A mainstream browser used by hundreds of millions"},{"name":"VS Code","desc":"The most popular code editor in the world"},{"name":"Signal","desc":"End-to-end encrypted messaging with a polished UX"},{"name":"Blender","desc":"Oscar-winning 3D animation and VFX tool"},{"name":"VLC","desc":"Plays everything, runs everywhere, no ads"},{"name":"Linux","desc":"Runs the majority of the world's servers and smartphones"}],"paragraph1":"The quality of open source software has improved dramatically because the incentive structures have changed. Companies now build businesses on open source cores (Red Hat, Elastic, GitLab). Well-funded teams maintain projects full-time. Community contributions catch bugs and add features that small teams would never have bandwidth for.","paragraph2":"To be fair about the tradeoffs: open source voice tools sometimes have smaller teams and slower release cycles than well-funded proprietary competitors. Documentation can be uneven. But these are practical challenges, not inherent limitations. And the gap has been closing quickly — especially in the voice tool space, where open models like OpenAI Whisper and NVIDIA Parakeet give open source projects access to the same foundational model quality as any commercial product."},"businessModel":{"title":"How Open Source Voice Tools Sustain Themselves","intro":"A reasonable question: if the software is free, how does anyone get paid to work on it?","paragraph1a":"The answer is that \"open source\" and \"free\" are not synonyms. The most sustainable open source projects separate the core engine (free, open, auditable) from convenience features that justify a paid tier. This is called the","openCoreModel":"open core model","paragraph1b":", and it works because it aligns incentives: the company earns revenue by making the product better, not by locking users in or monetizing their data.","openCore":{"title":"Why open core works well for privacy-focused tools","points":["The core product works locally and offline — no data collection required for the free tier","Cloud features (like managed transcription APIs) provide genuine convenience worth paying for","Community contributions improve the product for everyone, including paying customers","Users who cannot pay still benefit, and they contribute bug reports, translations, and code"]},"paragraph2":"This is the model OpenWhispr uses: the open source desktop app handles transcription locally using models like OpenAI Whisper and NVIDIA Parakeet at no cost. An optional Pro plan adds cloud-powered transcription and AI text cleanup for users who want faster results or do not want to run models on their own hardware. The important part is that the choice is yours — the local, private, fully functional version is always free."},"whatToLookFor":{"title":"What to Look For in an Open Source Voice Tool","intro":"Not all \"open source\" claims are equal. Here is a practical checklist for evaluating voice tools.","checklist":[{"question":"Is the full source code available?","detail":"Some projects open-source a library but keep the application proprietary. Check that the actual product you use — the desktop app, the mobile app — has its source published."},{"question":"Can you build from source?","detail":"Published source code that does not compile is not meaningfully open. Look for build instructions, CI pipelines, and community reports of successful builds."},{"question":"What is the license? (MIT, Apache, GPL, AGPL?)","detail":"Permissive licenses (MIT, Apache 2.0) allow maximum flexibility. Copyleft licenses (GPL, AGPL) ensure derivatives stay open. Both are legitimate — just know what you are getting."},{"question":"Does it work offline and locally?","detail":"Open source code that requires a cloud connection to function still sends your data to a server. For voice tools, offline-capable local processing is the privacy baseline."},{"question":"Is it actively maintained?","detail":"Check the commit history, issue tracker, and release cadence. An abandoned open source project may have unpatched security vulnerabilities."},{"question":"Is there a community? (Issues, PRs, discussions)","detail":"A healthy open source project has more than one contributor. Look for community engagement — pull requests from external contributors, issue discussions, and responsive maintainers."}]},"sources":{"title":"Sources and Further Reading","bloomberg":"Bloomberg: \"Is Anyone Listening to You on Alexa? A Global Team Reviews Audio\" (April 2019)","vrt":"VRT NWS: \"Google employees are eavesdropping, even in Flemish living rooms\" (July 2019)","guardian":"The Guardian: \"Apple contractors 'regularly hear confidential details' on Siri recordings\" (July 2019)","reuters":"Reuters: \"Apple agrees to pay $95 million to settle Siri privacy lawsuit\" (January 2025)","ftcAmazon":"FTC: Amazon charged with violating children's privacy law by keeping kids' Alexa recordings (May 2023)","ftcCloning":"FTC: Scammers use AI voice cloning to enhance family emergency schemes (March 2023)","openaiWhisper":"OpenAI: Introducing Whisper (September 2022)","whisperGithub":"OpenAI Whisper GitHub repository (models, license, docs)","radford":"Radford et al. (2022): Robust Speech Recognition via Large-Scale Weak Supervision","gdpr":"GDPR official text (EU 2016/679)","aiAct":"EU AI Act official text (EU 2024/1689)","whisperCpp":"whisper.cpp — C/C++ port of OpenAI Whisper","fasterWhisper":"faster-whisper — CTranslate2-based Whisper implementation","openwhispr":"OpenWhispr — open, local-first dictation product context"},"cta":{"title":"OpenWhispr Is Open Source","body":"We built OpenWhispr in the open because we believe voice software should be transparent. Don't take our word for it — check the code yourself.","starButton":"Star us on GitHub","footnote":"No account required · Works offline · MIT-licensed · Open source forever"},"internalLinks":{"localAi":"The case for local AI","localVsCloud":"Local vs cloud transcription","allPosts":"All blog posts","compare":"Compare OpenWhispr","useCases":"Use cases"}},"blogGoogleDocsVoiceTyping":{"metadata":{"title":"Google Docs Voice Typing: The Complete Guide | OpenWhispr","description":"How to turn on voice typing in Google Docs, every voice command, why it stops working, and how to dictate in any app, even offline.","keywords":["google docs voice typing","voice typing google docs","voice to text google docs","google docs dictation","turn on voice typing google docs","google docs voice commands","google docs voice typing not working","google docs voice typing languages"]},"schema":{"title":"Google Docs Voice Typing: The Complete Guide","description":"How to turn on voice typing in Google Docs, every voice command, why it stops working, and how to dictate in any app, even offline."},"breadcrumb":{"home":"Home","blog":"Blog","article":"Google Docs Voice Typing"},"layout":{"title":"Google Docs Voice Typing: The Complete Guide","subtitle":"How to turn on voice typing in Google Docs, the main voice commands, why it sometimes stops working, and how to dictate everywhere else on your computer.","date":"June 26, 2026","tag":"Guide"},"sections":[{"id":"quick","title":"Quick Answer"},{"id":"enable","title":"How to Turn It On"},{"id":"commands","title":"Voice Commands"},{"id":"languages","title":"Languages"},{"id":"mobile","title":"On Phone and Tablet"},{"id":"troubleshooting","title":"Not Working?"},{"id":"limits","title":"The Limitations"},{"id":"alternative","title":"Voice Typing Anywhere"},{"id":"faq","title":"FAQ"}],"intro":{"lead":"To turn on voice typing in Google Docs, open a document in Chrome, go to Tools > Voice typing, or press Ctrl+Shift+S on Windows or Cmd+Shift+S on a Mac, then click the microphone and start speaking.","body":"It is free, built in, and useful for a quick first draft. The trade-off is that it only runs in Chrome on desktop, needs an internet connection, and only types inside Google Docs.","summary":"Below you will find setup steps, the command categories that matter, supported language notes, troubleshooting fixes, privacy limitations, and a practical option when you need dictation outside Docs."},"enable":{"title":"How to Turn On Voice Typing in Google Docs","intro":"Voice typing is part of Google Docs on desktop and is designed for Google Chrome. Here is the full process:","steps":["Open a document at docs.google.com in Chrome.","In the menu bar, click Tools > Voice typing. Shortcut: Ctrl+Shift+S on Windows, Cmd+Shift+S on Mac.","A microphone box appears on the left side of the page. Click it to start.","The first time, your browser asks for microphone permission. Click Allow.","Speak naturally. Your words appear at the cursor as you talk.","Click the microphone again when you want to stop."],"tipLead":"Tip:","tipBody":"Voice typing also works in the speaker notes field of Google Slides with the same Tools menu and microphone control."},"commands":{"title":"Voice Commands: Punctuation, Formatting, and Editing","intro":"You can speak punctuation, line breaks, and editing commands. The important catch: formatting and editing commands only work when the document language is set to English.","groups":[{"title":"Punctuation","items":["\"period\"","\"comma\"","\"question mark\"","\"exclamation point\"","\"new line\" / \"new paragraph\""]},{"title":"Formatting","items":["\"bold\" / \"italics\"","\"underline\"","\"apply heading 1\"","\"create bulleted list\"","\"align center\""]},{"title":"Editing and moving","items":["\"select [word]\"","\"delete last word\"","\"go to end of paragraph\"","\"next paragraph\"","\"stop listening\""]}]},"languages":{"title":"Which Languages Are Supported?","paragraph1":"Voice typing handles dictation in about a hundred languages and dialects. Pick yours from the dropdown on the microphone box. Basic punctuation works broadly across languages.","paragraph2":"Spoken formatting and editing commands are the exception. Commands like bold, select paragraph, or apply heading work only in English, so non-English dictation usually means formatting with the keyboard or mouse."},"mobile":{"title":"Voice Typing on Phone and Tablet","intro":"The Google Docs mobile app does not include the desktop Voice typing tool. On phones and tablets, dictate the same way you would in any app: tap the microphone on your keyboard.","items":["Android: tap the microphone on Gboard or your preferred keyboard while the cursor is in the document.","iPhone and iPad: tap the dictation key on the iOS keyboard."],"outro":"That is your device dictation, not the Google Docs desktop feature, so accuracy, commands, and behavior will not match the Chrome version."},"troubleshooting":{"title":"Voice Typing Not Working? Try This","items":[{"problem":"The Voice typing button is missing or grayed out","fix":"Use Google Chrome on desktop. Voice typing is not available in the mobile app, and other browsers are not officially supported."},{"problem":"It will not pick up my voice","fix":"Check that docs.google.com has microphone permission in Chrome, and make sure another app is not holding the microphone."},{"problem":"It keeps stopping on its own","fix":"Voice typing can stop after a pause or when you click away from the tab. Click the microphone again and keep the Docs tab focused."},{"problem":"Nothing happens at all","fix":"You may be offline. Voice typing is cloud-based and needs an internet connection. Reconnect and reload the document."}]},"limits":{"title":"What Google Docs Voice Typing Cannot Do","intro":"It is a genuinely useful free feature, but it is built for one job: dictating inside a Google Doc. These limits matter if you dictate often:","items":[{"title":"Chrome only.","body":"No reliable support in Safari, Firefox, or Edge."},{"title":"Online only.","body":"Audio is processed in Google's cloud, so no connection means no dictation."},{"title":"Docs only.","body":"It will not type into Gmail, Slack, your browser bar, a code editor, or any other app."},{"title":"Cloud privacy.","body":"Your voice is sent to Google for transcription, which is not ideal for sensitive material."}]},"alternative":{"title":"Want Voice Typing in Every App? Use OpenWhispr","introBefore":"If Docs is the only place you dictate, voice typing does the job. But when you want to talk into an email, a Slack message, a code comment, a form, or your notes app,","introAfter":"covers what Google Docs does not:","points":["One push-to-talk hotkey types into any app or text field, system-wide","Runs 100% on your device with OpenAI Whisper and NVIDIA Parakeet, fully offline","Your audio never leaves your computer","Works in any browser and in desktop apps, not just Chrome","Open source, so anyone can inspect what it does","Agent mode can clean up and reformat what you dictated"],"outro":"Think of it as Google Docs voice typing for your whole computer, without the cloud dependency.","viewOnGitHub":"View on GitHub"},"faqTitle":"Frequently Asked Questions","faqs":[{"question":"How do I turn on voice typing in Google Docs?","answer":"Open a document in Chrome, then go to Tools > Voice typing, or press Ctrl+Shift+S on Windows or Cmd+Shift+S on Mac. Click the microphone box, allow microphone access when prompted, and start speaking."},{"question":"Why is my Google Docs voice typing not working?","answer":"The usual causes are using a browser other than Chrome, missing microphone permission, another app holding the microphone, a paused or unfocused Docs tab, or no internet connection."},{"question":"Does Google Docs voice typing work offline?","answer":"No. Google Docs voice typing is cloud-based and needs an internet connection. If you need offline dictation, use a local tool such as OpenWhispr."},{"question":"What languages does Google Docs voice typing support?","answer":"It supports dictation in about a hundred languages and dialects, but spoken formatting and editing commands work only when the document language is English."},{"question":"Does Google Docs voice typing work on iPhone or Android?","answer":"Not as the desktop Voice typing tool. In the mobile app, use your phone keyboard microphone instead, such as Gboard on Android or the iOS dictation key."},{"question":"Is Google Docs voice typing private?","answer":"Audio is sent to Google for transcription, so it is convenient but not local-only. For confidential notes, legal work, patient information, or unreleased business content, local dictation is a better fit."},{"question":"Can I use voice typing outside of Google Docs?","answer":"Google's built-in voice typing only works inside Google Docs and speaker notes in Google Slides. OpenWhispr provides a system-wide hotkey that dictates into any app."}],"links":{"items":[{"href":"/use-cases","label":"All use cases"},{"href":"/compare/apple-dictation","label":"vs Apple Dictation"},{"href":"/pricing","label":"Pricing"}]}},"useCaseMedical":{"breadcrumb":{"home":"Home","useCases":"Use Cases","medical":"For Clinicians"},"metadata":{"title":"Free Medical Dictation Software For Clinicians - OpenWhispr","description":"OpenWhispr is free, open-source medical dictation software. Dictate the visit, get the SOAP note drafted, and keep patient audio on your machine.","keywords":["medical dictation software","clinical dictation software","physician dictation","HIPAA compliant dictation software","dictation software for doctors","SOAP note dictation","offline medical transcription"]},"schema":{"name":"OpenWhispr","applicationCategory":"Medical dictation software","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source medical dictation software. Patient audio is transcribed on the clinician's own device and never leaves it. Free for unlimited local dictation.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Medical dictation software for doctors & clinicians","title":"Dictate the visit once. Get the note drafted.","body":"OpenWhispr drafts the SOAP note, the referral letter or the discharge summary from what you dictate, on your own computer. Free for every clinician, and it keeps working on a locked-down hospital network.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Clinician walking through a sunlit hospital corridor"}},"trustBar":{"label":"Dictate into:","names":["Epic","Oracle Health","athenahealth","eClinicalWorks","NextGen","Meditech"]},"pain":{"label":"Cloud or local. Choose for the visit.","title":"Cloud for speed. Local for sensitive or offline work.","subtitle":"Cloud transcription provides top accuracy and speed without a local model download. Local models process entirely on-device and need no internet.","cards":[{"title":"The whole encounter, uploaded.","body":"An ambient scribe can send the whole encounter to the cloud, including things the patient asked you not to put in the chart. OpenWhispr processes what you dictate on your device."},{"title":"Pajama-time charting.","body":"A full clinic, then an hour of notes at the kitchen table. OpenWhispr drafts the note from your dictation while the visit is fresh, so you review and sign instead of type."},{"title":"One more vendor for compliance to vet.","body":"Each cloud scribe adds a BAA, a risk assessment and another data flow to map. In local mode, no transcription vendor sits in the data path."},{"title":"\"Trust us\" is not a compliance posture.","body":"OpenWhispr's code is public on GitHub. Give your security team the repository, then choose the cloud or local path that fits your review."}]},"features":{"label":"What you get","title":"Faster charting. Same confidentiality.","subtitle":"Medical dictation software often trades privacy for speed or bills per provider every month. OpenWhispr keeps the speed without sending patient audio away or charging per clinician for local mode.","items":[{"image":{"label":"Clinician speaking beside a laptop in a medical practice"},"title":"Your mic, or the telehealth call, transcribed on-device","body":"Dictate the visit into your mic, or record a telehealth consult straight from your speakers on macOS 14.2 or later, Windows and most Linux desktops. It's transcribed locally and never uploaded. You decide what's kept and when it's deleted."},{"image":{"label":"Clinician reviewing a printed document at a hospital workstation"},"title":"SOAP notes and letters, drafted for you","body":"Save your own SOAP, referral and discharge prompts as one-click actions, then run whichever one the visit needs. You review every word before it's signed, then paste it into the chart."},{"image":{"label":"Doctor speaking while working on a laptop in an examination room"},"title":"Dictate into the chart field you're already in","body":"Press a hotkey and dictate wherever the cursor already is: the HPI box, a reply to a portal message, a refill note in your inbox. Clean, punctuated text, processed locally."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Patient audio sent to us","value":"None","footnote":"in local mode, transcription never leaves your device"},{"label":"Cost per clinician","value":"$$0","footnote":"unlimited local dictation, the whole practice, forever"},{"label":"Local dictation is","value":"Free","footnote":"unlimited on every plan"}]},"howItWorks":{"label":"How it works","title":"The note drafted before the next patient is roomed.","subtitle":"Most medical dictation is a per-provider cloud subscription or an outsourced transcription service. OpenWhispr lives on your machine, with no account or credit card required.","image":{"label":"Doctor reaching for a closed laptop in an examination room"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model during setup."},{"title":"Dictate the visit.","body":"Speak your findings, plan and follow-up, and it's transcribed on your device."},{"title":"Review, sign, delete.","body":"Edit the draft note, paste it into the chart, and delete the audio. You control all of it."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons clinicians choose OpenWhispr","subtitle":"Choosing medical dictation software is a patient-confidentiality decision before it's anything else. Here are the facts, and your compliance officer can verify every one in the repo or at trust.openwhispr.com.","items":[{"title":"Audio never leaves your device","body":"Dictation, and note drafts with a local model, are processed on your machine. The open-source code shows there's no upload step."},{"title":"Open source, auditable by anyone","body":"Your IT or compliance team can read every line on GitHub. No faith required."},{"title":"Retention is yours, not a vendor's","body":"Audio and transcripts are files on your machine, not rows in someone else's retention schedule. Delete the audio the moment the note is signed."},{"title":"Optional Corti clinical transcription","body":"For specialist terminology, Corti Transcribe is available as a clinical-grade medical transcription provider. Choose a local model for note drafting that stays on-device."},{"title":"No bot joins the telehealth call","body":"Nothing appears in the participant list, so you never have to explain a third party to the patient before you can begin."},{"title":"Works fully offline","body":"Ward round, rural clinic, a locked-down hospital network. No signal required, so it never stalls mid-dictation."},{"title":"Dictate into any EHR","body":"A system-wide hotkey drops clean text wherever your cursor is: Epic, athenahealth, a referral letter, an email."},{"title":"Free for the whole practice","body":"Unlimited local dictation at $0, every clinician in the group, forever. No per-provider licence to renew, and you can test it on your own clinic list before anyone signs anything."},{"title":"SOC 2 Type 2 audit complete. HIPAA attestation. ISO 27001-aligned.","body":"Your security review can start at trust.openwhispr.com. Enterprise includes custom data processing agreements and compliance reviews."},{"title":"100+ languages","body":"Dictate the history in the language the visit actually happened in, transcribed locally."}]},"faq":{"label":"Before you download","title":"The questions clinicians ask first.","subtitle":"Straight answers on privacy, compliance and accuracy. If yours isn't here, ask us in Discord.","items":[{"question":"Does patient audio ever leave my device?","answer":"Not in local mode. Pick a local model during setup and transcription runs on your machine; the open-source code shows there is no upload step. Cloud transcription remains optional."},{"question":"Is it HIPAA compliant?","answer":"Compliance depends on your workflow, not the tool alone. OpenWhispr's SOC 2 Type 2 audit is complete, with the final report available under NDA once issued; it has a HIPAA attestation and an ISO 27001-aligned information-security programme. For PHI, local-only and bring-your-own-key modes route content around OpenWhispr Cloud. Cloud voice audio is processed in real time and not persisted after transcription. Enterprise includes custom data processing agreements and compliance reviews."},{"question":"Can it handle medical terminology?","answer":"Yes. It runs Whisper-class models trained on a wide range of accents and 100+ languages. The custom dictionary feeds the words that trip everything else up to both the transcriber and the AI cleanup: metoprolol succinate ER, cholecystectomy, the abbreviations your specialty writes and nobody else reads."},{"question":"How is this different from an ambient AI scribe?","answer":"An ambient scribe listens to the whole encounter, sends it to a vendor's cloud, and bills per provider every month. Medical dictation software is the other shape: you say what belongs in the note, a local model drafts it from your own prompt, and it stays on your machine at $0. For a telehealth consult it captures your speakers directly, so nothing joins the call and nothing is uploaded."},{"question":"Does it draft the note, or just transcribe?","answer":"Both. Run a local model (Llama, Qwen, Gemma) or connect your own AI key, then save your own SOAP, referral and discharge prompts as one-click actions and run whichever one the visit needs. A local model keeps the PHI on your machine; your own key means the model and data path are ones you already trust."},{"question":"Is it actually free?","answer":"Yes. Unlimited local dictation, 100+ languages, all three desktop platforms: $0 forever, for every clinician. Pro ($8/month) adds more cloud transcription and mobile dictation if you want them."}]},"reviews":{"label":"What users say","title":"Built by people who don't want to be recorded either.","subtitle":"From our Discord, GitHub and inboxes, the same instinct that matters in the exam room.","score":{"value":"0","label":"Patient recordings on a server","body":"In local mode, audio stays on your device and is never stored on a server."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for doctors & clinicians","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Clinicians who want free, private, offline dictation and are fine with a quick local setup"}]},{"parts":[{"text":"Clinicians who need unlimited cloud transcription, cross-device sync and mobile access"}]},{"parts":[{"text":"Practices and teams that need unlimited meeting recordings, speaker labels and Agent mode"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Free offline dictation, with patient audio staying on your device"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models or keys"}]},{"parts":[{"text":"Unlimited meeting recordings, speaker labels and workflow features for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"Some conversations stay in the exam room. We built for those.","paragraphs":["OpenWhispr started with a simple objection: what a patient tells you in the exam room shouldn't have to pass through a stranger's server to become a note.","Our code is public on GitHub. In local mode, transcription runs on the device with no internet required.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Sign the last note before you leave the building.","body":"Download OpenWhispr. Dictate the visit, draft the note, then paste it into the chart while it is still fresh. Choose local models when processing needs to stay on your computer.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free unlimited local dictation on Mac, Windows & Linux."}}},"useCaseMac":{"metadata":{"title":"Mac Dictation — Offline AI Voice-to-Text | OpenWhispr","description":"Mac dictation powered by Whisper & Parakeet — runs locally on Apple Silicon, works offline, open source, and free. The Dragon Dictate replacement.","keywords":["mac dictation","best dictation software for mac","dragon dictate for mac","apple dictation alternative","apple silicon dictation","mac speech to text"]},"breadcrumb":{"home":"Home","useCases":"Use Cases","mac":"Mac"},"hero":{"badge":"Mac Dictation","title":"The Modern AI Mac Dictation App","subtitle":"A real upgrade over built-in Apple Dictation and a replacement for discontinued Dragon Dictate for Mac. Whisper and Parakeet accuracy runs locally on Apple Silicon, with offline privacy and open-source transparency.","viewOnGitHub":"View on GitHub"},"problem":{"title":"Mac Dictation Has a Gap, and OpenWhispr Fills It","paragraph1":"For years, serious dictation on a Mac meant Dragon Dictate. Then Nuance discontinued it, leaving Mac users with built-in Apple Dictation: convenient, but limited in accuracy, formatting, and control.","paragraph2":"AI speech recognition has improved dramatically. OpenAI Whisper and NVIDIA Parakeet now run well on Apple Silicon, and OpenWhispr brings those models to a native desktop workflow.","callout":"The result is modern Mac speech-to-text with a system-wide hotkey, stronger accuracy, local processing, and no cloud dependency for core transcription."},"benefits":{"title":"Why OpenWhispr Is the Best Dictation Software for Mac","items":[{"title":"Whisper and Parakeet Accuracy","body":"Choose from multiple local speech models. Larger models handle accents, technical vocabulary, and natural speech better than basic built-in dictation."},{"title":"Built for Apple Silicon","body":"OpenWhispr runs well on M1, M2, M3, and M4 Macs. Intel Macs are supported too; smaller models are best for speed there."},{"title":"System-Wide Push-to-Talk","body":"Hold a global hotkey to dictate anywhere on macOS: Notes, Mail, Messages, Slack, Safari, Chrome, Pages, Word, Xcode, or your terminal."},{"title":"Agent Mode AI Cleanup","body":"Convert rough dictation into polished text, bullet points, emails, or structured notes using prompts you control."},{"title":"Private and Offline","body":"Audio is transcribed on-device. Disconnect from Wi-Fi and core dictation still works."},{"title":"Open Source and Auditable","body":"The full codebase is public on GitHub, so you can inspect how the app handles data."}]},"appleComparison":{"title":"OpenWhispr vs Apple Dictation","intro":"Apple Dictation is fine for quick, casual use. If you dictate often, the differences add up.","columns":[{"title":"OpenWhispr","items":["Whisper and Parakeet models with size you choose","Agent mode reformats dictation into clean text","Always local and works fully offline","Open source and auditable","Local transcription history"]},{"title":"Apple Dictation","items":["Single built-in engine with no model choice","No structured AI reformatting","Convenient but weaker on hard speech","Closed source","No dedicated transcription history"]}]},"scenarios":{"title":"Real-World Mac Dictation Scenarios","items":[{"title":"Replacing Dragon Dictate for Mac","body":"If you relied on Dragon for years, OpenWhispr gives you a modern push-to-talk flow that works in any app, while running locally and staying actively maintained."},{"title":"Writing Email and Notes Faster","body":"Draft a reply in Mail, capture thoughts in Notes, or send a Slack message by voice. Agent mode can turn a rough first pass into a clean message."},{"title":"Coding with Voice in Xcode and VS Code","body":"Dictate comments, documentation, and commit messages without leaving your development flow."}]},"setup":{"title":"Quick Setup and Which Model Suits Your Mac","stepsTitle":"Getting started in minutes","steps":["Download OpenWhispr for macOS and open it.","Grant microphone and accessibility permissions when prompted.","Pick a speech model; it downloads and runs locally.","Set your push-to-talk hotkey, then dictate in any app."],"modelTitle":"Choosing a model","modelTips":["Apple Silicon M1-M4: run larger Whisper or Parakeet models for top accuracy with good speed.","Older or Intel Macs: choose a smaller model for fast, lightweight transcription.","Need near-instant results: smaller models return text faster; bigger models trade speed for accuracy."],"noteLead":"Privacy note:","noteBody":"Core transcription always runs locally on your Mac. Agent-mode cleanup is optional and only uses a cloud LLM if you configure one."},"toolComparison":{"title":"How It Compares to Other Mac Dictation Tools","items":[{"title":"vs MacWhisper","body":"MacWhisper is excellent for transcribing existing audio files. OpenWhispr is built for live, system-wide dictation into any app."},{"title":"vs SuperWhisper","body":"OpenWhispr stands out with both Whisper and Parakeet models, agent-mode cleanup, cross-platform support, and a fully open-source codebase."},{"title":"vs Apple Dictation","body":"OpenWhispr adds model choice, stronger AI transcription, structured cleanup, local history, and open-source transparency."}]},"faqTitle":"Mac Dictation FAQ","faqs":[{"question":"What is the best dictation software for Mac?","answer":"For accuracy and privacy, OpenWhispr is a strong choice. It runs Whisper and Parakeet locally, works system-wide, and keeps core transcription on your Mac."},{"question":"Is there a Dragon Dictate replacement for Mac?","answer":"Yes. OpenWhispr fills much of the gap left by Dragon Dictate for Mac with modern local AI speech recognition and a push-to-talk workflow."},{"question":"How is OpenWhispr better than Apple Dictation?","answer":"It offers stronger model-based transcription, agent-mode cleanup, model-size choice, local history, and open-source auditability."},{"question":"Does Mac dictation work offline with OpenWhispr?","answer":"Yes. Core transcription runs on your Mac with no internet connection."},{"question":"Does OpenWhispr run on Apple Silicon?","answer":"Yes. It runs well on M1, M2, M3, and M4 Macs, and also supports Intel Macs."},{"question":"Does it work in any app?","answer":"Yes. Hold the global hotkey and your transcribed text appears wherever the cursor is focused."}],"links":{"items":[{"href":"/pricing","label":"Pricing"},{"href":"/use-cases","label":"All use cases"},{"href":"/use-cases/windows","label":"Windows dictation"},{"href":"/compare/apple-dictation","label":"Compare Apple Dictation"},{"href":"/compare/macwhisper","label":"Compare MacWhisper"}]},"cta":{"title":"Get the Best Dictation App for Mac","description":"Download OpenWhispr and start dictating on your Mac in minutes. Runs on Apple Silicon, works offline, and is open source.","footer":"Free to start · Modern macOS · No account required"}},"authGate":{"providers":{"google":"Google","microsoft":"Microsoft","apple":"Apple"},"continueWithProvider":"Continue with {provider}","or":"or","fields":{"email":"Email address","password":"Password","fullName":"Full name","passwordHint":"At least 8 characters"},"actions":{"checking":"Checking...","continueWithEmail":"Continue with Email","back":"Back","signingIn":"Signing in...","signIn":"Sign In","createAccount":"Create account","tryAgain":"Try again","sending":"Sending...","sendResetLink":"Send reset link","creatingAccount":"Creating account...","resendVerificationEmail":"Resend verification email","useDifferentEmail":"Use a different email","useSso":"Use single sign-on (SSO)"},"status":{"completingSignIn":"Completing sign in...","redirecting":"Redirecting..."},"errors":{"generic":"Something went wrong. Please try again.","authenticationTimedOut":"Authentication timed out. Please try again.","sessionExpired":"Session expired. Please try signing up again.","waitingForVerification":"Still waiting for verification. Check your inbox or resend the email.","unableToCheckEmail":"Unable to check this email. Please try again.","network":"Network error. Please check your connection.","networkRetry":"Network error. Please check your connection and try again.","incorrectPassword":"Incorrect password. Please try again.","accountExists":"An account with this email already exists. Please sign in.","createAccountFailed":"Failed to create account. Please try again.","resendFailed":"Failed to resend. Please try again."},"signin":{"title":"Welcome back","forgotPassword":"Forgot password?","newHere":"New here?"},"forgot":{"title":"Reset your password","description":"Enter your email and we'll send you a reset link","emailSentTitle":"Check your email","emailSentBody":"We sent a password reset link to","didNotReceive":"Didn't receive the email? Check your spam folder or try again."},"signup":{"title":"Create your account","haveAccount":"Have an account?"},"verifyEmail":{"title":"Check your email","message":"We sent a verification link to"},"legal":{"line":"By continuing, you agree to our <termsLink>Terms of Service</termsLink> and <privacyLink>Privacy Policy</privacyLink>."}},"getStarted":{"title":"Get started","subtitle":"Download OpenWhispr, free and open source.","processing":"Starting your download...","skip":"Continue without an account","verificationMessage":"Verify your email to start your download. We sent a link to","metadata":{"title":"Get Started | OpenWhispr","description":"Create your OpenWhispr account and get started with private AI voice dictation."}},"checkout":{"title":"Get started with OpenWhispr","processing":"Redirecting to checkout...","period":{"month":"/mo","year":"/yr"},"plan":{"proMonthly":"Pro · Monthly","proAnnual":"Pro · Annual","businessMonthly":"Business · Monthly","businessAnnual":"Business · Annual"},"badge":{"twoMonthsFree":"2 months free"},"metadata":{"title":"Checkout | OpenWhispr","description":"Complete your OpenWhispr subscription purchase securely."}},"checkoutSuccess":{"appRedirect":{"title":"Redirecting to OpenWhispr...","body":"You can close this tab","fallback":"Tap here if the app didn't open"},"success":{"title":"Welcome to Pro!","body":"Your subscription is now active"},"download":{"autoNotice":"Your download should start automatically...","fallback":"Click here if it didn't start","open":"Download OpenWhispr","all":"See all downloads"},"linuxOptions":[{"label":"Linux AppImage (x86_64)","meta":"AppImage"},{"label":"Linux DEB (amd64)","meta":"Debian package"},{"label":"Linux RPM (x86_64)","meta":"RPM package"},{"label":"Linux tar.gz (x64)","meta":"Portable tarball"}],"metadata":{"title":"Checkout Successful | OpenWhispr","description":"Your OpenWhispr checkout is complete and your subscription is ready."}},"upgrade":{"expiredTitle":"Link expired","redirecting":"Redirecting to checkout...","errors":{"missingToken":"Missing upgrade token. Please check your email link.","invalidLink":"This link is no longer valid. You can upgrade from within the app.","generic":"Something went wrong. Please try again or upgrade from within the app."},"metadata":{"title":"Upgrade | OpenWhispr","description":"Upgrade your OpenWhispr plan to unlock more voice dictation."}},"authCallback":{"failedTitle":"Sign-in failed","redirecting":"Redirecting to OpenWhispr","closeTab":"You can close this tab","desktopFallback":"Click here if OpenWhispr didn't open","mobileFallback":"Tap here if the app didn't open","backToSignIn":"Back to sign-in","errors":{"account_not_linked":"An account with this email already exists with a different sign-in method.","state_mismatch":"Sign-in expired before completing. Please try again.","please_restart_the_process":"Sign-in expired before completing. Please try again.","invalid_code":"Sign-in failed. Please try again.","unable_to_get_user_info":"Couldn't read your profile from the provider. Please try again.","noSession":"Could not retrieve session. Please try signing in again.","default":"Sign-in failed. Please try again.","sso_required":"This workspace requires single sign-on. Sign in with your company SSO instead.","account_mismatch":"You were signed in as a different account than the one you entered."},"metadata":{"desktopCallback":{"title":"Redirecting... | OpenWhispr","description":"Completing your OpenWhispr desktop sign-in and redirecting you to the app."},"mobileCallback":{"title":"Redirecting... | OpenWhispr","description":"Completing your OpenWhispr mobile sign-in and redirecting you to the app."},"error":{"title":"Sign-in error | OpenWhispr","description":"We couldn't complete your OpenWhispr sign-in. Review the error and try again."}},"mismatch":{"switchAccount":"Sign in as a different account","continueAnyway":"Continue with the signed-in account"}},"notFound":{"title":"Oops! Page not found","subtitle":"Looks like this page got lost in transcription!","quip":"Don't worry, we think faster than we type...","description":"The page you're looking for might have been moved, deleted, or is temporarily unavailable.","home":"Take me home","back":"Go back","autoRedirectLabel":"Auto-redirect:","autoRedirectBody":"Taking you home in {seconds} seconds..."},"verifyEmail":{"loading":"Verifying your email...","errors":{"noToken":"No verification token found. Please check your email link.","failed":"Verification failed. The link may have expired.","network":"Network error. Please check your connection and try again."},"resend":{"signInFirst":"Please sign in first, then request a new link.","failed":"Could not resend. Please try again.","sent":"Verification email sent! Check your inbox.","network":"Network error. Please try again."},"success":{"title":"Email verified!","desktop":"If you're not redirected automatically, click below.","browser":"You can close this tab and return to your signup to continue.","openApp":"Open OpenWhispr"},"actions":{"sending":"Sending...","requestNewLink":"Request a new link"},"metadata":{"title":"Verify Email | OpenWhispr","description":"Verify your email address to finish setting up your OpenWhispr account."}},"resetPassword":{"loading":"Loading...","errors":{"passwordMismatch":"Passwords do not match.","passwordTooShort":"Password must be at least 8 characters.","network":"Network error. Please check your connection and try again."},"invalid":{"title":"Invalid reset link","description":"This password reset link is invalid or has expired.","cta":"Go to Get Started"},"success":{"title":"Password updated","description":"You can now sign in with your new password.","cta":"Continue to sign in"},"form":{"title":"Create new password","description":"Enter your new password below","newPassword":"New password","confirmPassword":"Confirm new password","passwordHint":"At least 8 characters","resetting":"Resetting...","submit":"Reset password"},"metadata":{"title":"Reset Password | OpenWhispr","description":"Choose a new password for your OpenWhispr account."}},"consent":{"banner":{"ariaLabel":"Cookie consent","title":"We value your privacy","eu":"We use cookies to measure usage and our marketing. Analytics and marketing cookies stay off until you allow them.","us":"We and our partners use cookies for analytics and advertising. You can opt out anytime.","row":"We use cookies to understand usage and improve OpenWhispr.","gpc":"We detected a Global Privacy Control signal and have opted you out of ad cookies.","policyLine":"See our <cookieLink>Cookie Policy</cookieLink> and <privacyLink>Privacy Policy</privacyLink>.","rejectAll":"Reject all","customize":"Customize","acceptAll":"Accept all","privacyChoices":"Your privacy choices","cookieSettings":"Cookie settings","gotIt":"Got it","updated":"We've updated our cookie policy since your last visit. Your existing choices remain in effect until you change them."},"settings":{"title":"Cookie preferences","description":"Choose which cookies OpenWhispr may use. You can change this anytime from the footer.","necessary":{"title":"Strictly necessary","description":"Required for sign-in and security. Always active.","ariaLabel":"Strictly necessary cookies (always on)"},"analytics":{"title":"Analytics","description":"Helps us understand usage (Google Analytics 4).","ariaLabel":"Analytics cookies"},"marketing":{"title":"Marketing","description":"Measures our ads and campaigns (Google Ads, Meta).","gpcDescription":"Off — your browser's Global Privacy Control signal opts you out.","ariaLabel":"Marketing cookies"},"rejectAll":"Reject all","acceptAll":"Accept all","save":"Save preferences"},"trigger":{"privacyChoices":"Your Privacy Choices","cookieSettings":"Cookie Settings"}},"referral":{"someone":"Someone","loading":"Loading...","invitedBy":"Invited by {name}","invited":"You're invited","title":"Unlock 1 free month of OpenWhispr Pro","claimPrompt":"Sign up to claim your free month of Pro","continueWithGoogle":"Continue with Google","or":"or","processing":"Setting up your account...","alreadyHaveAccount":"Already have an account?","downloadOpenWhispr":"Download OpenWhispr","fields":{"email":"Email address","password":"Password","fullName":"Full name","passwordMin":"Password (min. 8 characters)"},"actions":{"checking":"Checking...","continueWithEmail":"Continue with Email","signingIn":"Signing in...","signInClaim":"Sign in & claim offer","useDifferentEmail":"Use a different email","creatingAccount":"Creating account...","createClaim":"Create account & claim offer","useSso":"Use single sign-on (SSO)"},"signin":{"as":"Signing in as"},"signup":{"for":"Creating account for"},"legal":{"line":"By continuing, you agree to our <termsLink>Terms</termsLink> and <privacyLink>Privacy Policy</privacyLink>."},"errors":{"network":"Network error. Please try again.","incorrectPassword":"Incorrect password. Please try again.","accountExists":"Account already exists. Please sign in.","createAccountFailed":"Failed to create account.","signInTimedOut":"Sign-in timed out. Please try again.","signInFailed":"Failed to sign in.","ssoEmailRequired":"Enter your work email above to use single sign-on.","ssoFailed":"Single sign-on failed."},"spectrogram":{"friend":"a friend","signalId":"Signal ID","scanAriaLabel":"Scan audio signature","scan":"Scan","stop":"Stop","replay":"Replay","unlocked":"Unlocked","freeMonth":"1 free month of Pro","signature":"{name}'s audio signature","scanning":"Scanning..."},"continueWithProvider":"Continue with {provider}","providers":{"google":"Google","microsoft":"Microsoft","apple":"Apple"},"metadata":{"title":"Referral | OpenWhispr","description":"Join OpenWhispr through a referral link and start using private AI voice dictation."}},"demo":{"ariaLabel":"Product demo","tabpanelFallback":"Demo","tabs":[{"label":"Speech-to-Text"},{"label":"Meeting Notes"},{"label":"AI Chat"},{"label":"Audio Upload"}],"panelLabels":["Speech-to-Text demo","Meeting Notes demo","AI Chat demo","Audio Upload demo","Custom Dictionary demo","Integrations demo","History demo"],"sidebar":{"search":"Search","settings":"Settings","userName":"Alex Chen","userEmail":"alex@company.com","nav":{"home":"Home","notes":"Notes","upload":"Upload","dictionary":"Dictionary","integrations":"Integrations"}},"speech":{"dictationText":"Hey team, I just finished reviewing the Q2 roadmap and I think we should prioritize the voice pipeline before the new integrations.","withoutOpenWhispr":"Without OpenWhispr","slowWpm":"~40 WPM","messageApp":"Slack — #product","placeholder":"Type a message...","withOpenWhispr":"With OpenWhispr","fastWpm":"~150 WPM","faster":"3x faster"},"history":{"transcriptions":{"title":"Transcriptions","items":[{"time":"9:15 AM","text":"The quarterly metrics look promising. Revenue is up 12% month over month and user retention has improved significantly."},{"time":"4:30 PM","text":"Let's schedule the design review for next week. I want the whole team to see the new components before we ship."}]},"upcomingMeetings":{"title":"Upcoming Meetings","items":[{"time":"11:00 AM - 11:30 AM","title":"Design Review"},{"time":"2:00 PM - 2:45 PM","title":"Sprint Planning"},{"time":"4:00 PM - 4:30 PM","title":"1:1 with Sarah"}]}},"notes":{"rawNotes":"- ship voice pipeline by march\n- talk to design team re: new UI\n- budget Q: ask finance","meetingTitle":"Sprint Planning","title":"Product Sync","recording":"Recording","you":"You","date":"Mar 11, 2026","foldersLabel":"Folders","notesLabel":"Notes","viewTabs":{"transcript":"Transcript","notes":"Notes","enhanced":"Enhanced"},"sections":[{"heading":"Decisions","items":["Voice pipeline is the top priority — ship by end of March before any other workstream","UI redesign deferred until pipeline lands, don't want to split focus","Dictionary feature approved: custom words for medical, legal, and technical terms"]},{"heading":"Action Items","items":["Sarah: frontend migration to new component library by end of sprint","Alex: API refactor + latency benchmarks (target sub-200ms, currently ~280ms)","Gabriel: follow up with finance on Q2 budget, need headcount approval by Friday","Design review for new editor UI — next Tuesday 2pm, full team"]},{"heading":"Open Questions","items":["Real-time collaboration on notes — CRDT vs OT approach? No timeline yet","Third-party dependency audit needed before public open source release","How to handle offline sync for mobile — local-first or queue-based?"]}],"done":"Done","primaryAction":"Sync Template","enhanceNotes":"Enhance Notes","actions":[{"name":"Sync Template","description":"Format using team template"},{"name":"Meeting Summary","description":"Summarize decisions & actions"},{"name":"Email Draft","description":"Draft follow-up email"}],"folders":[{"name":"Personal","count":5,"isActive":true},{"name":"Meetings","count":3,"isActive":false},{"name":"Work","count":2,"isActive":false}],"noteList":[{"title":"Product Sync","preview":"Action items from weekly sync...","time":"10:30 AM"},{"title":"Voice pipeline design","preview":"Architecture notes for v2...","time":"9:15 AM"},{"title":"Weekly standup","preview":"Sprint progress and blockers...","time":"Yesterday"}]},"aiChat":{"title":"AI Chat","listening":"Listening...","askAnything":"Ask anything about your notes","speakQuestion":"Speak your question","voicePoweredQa":"Voice-powered Q&A over your data","userMessage":"Summarize what we discussed in today's meetings","response":"Based on today's discussions:\n\n• Voice pipeline — ship by end of March\n• UI redesign deferred post-pipeline\n• Follow up with finance on Q2 budget","releaseToFinish":"Release {hotkey} to finish","holdToSpeak":"Hold {hotkey} to speak"},"upload":{"noteTitle":"Design System Migration Planning","transcript":"The team agreed to move forward with the new design system. Sarah will lead the frontend migration and Alex will handle the API refactor by end of sprint.","now":"Now","preview":"Team aligned on migrating to the new design system...","dateMeta":"Mar 12, 2026 · 847 words","structuredIntro":"Team aligned on migrating to the new component-based design system. Frontend migration begins this sprint with Sarah leading the component audit. Alex will refactor the API layer in parallel.","sections":[{"heading":"Key Decisions","items":["Adopt new design tokens across all product surfaces","Frontend migration handled component-by-component","API refactor scoped to auth and data-fetching layers"]},{"heading":"Action Items","items":["Sarah — Component audit and migration plan (due Fri)","Alex — Scope API refactor; draft RFC by Wednesday","Jamie — Update CI pipeline for new token format","Team — Review Figma library v2 before next standup"]},{"heading":"Timeline","items":["Sprint 14: Audit + migration plan + API RFC","Sprint 15: Component migration, design tokens","Sprint 16: Integration testing, deprecate legacy"]}],"dropPrompt":"Drop audio file or browse","formats":"MP3, WAV, M4A, WebM, OGG, FLAC, AAC","fileName":"meeting-recording.mp3","fileMeta":"12:04 duration · 18.3 MB","folder":"Personal","transcribe":"Transcribe","transcribing":"Transcribing locally...","complete":"Transcription Complete","openNote":"Open Note","newUpload":"New Upload"},"dictionary":{"title":"Custom Dictionary","clearAll":"Clear all","placeholder":"Add words or phrases...","hint":"Words are sent as hints to improve transcription accuracy. Separate multiple words with commas."},"integrations":{"title":"Integrations","subtitle":"Connect your favorite tools","calendarDescription":"Auto-join meetings and transcribe in real time","connected":"Connected","comingSoon":"More integrations coming soon"}},"blogComponents":{"author":{"defaultRole":"Engineering","updated":"Updated {date}"},"layout":{"defaultTag":"Blog","tableOfContents":"Table of contents","onThisPage":"On this page"},"diarizationPipeline":{"title":"The Local Diarization Pipeline","steps":[{"label":"Audio Capture","sub":"Mic + system, 16kHz"},{"label":"VAD","sub":"Silero, 2MB"},{"label":"Segmentation","sub":"pyannote 3.0"},{"label":"Embedding","sub":"CAM++, 512-dim"},{"label":"Clustering","sub":"Agglomerative, 0.5"},{"label":"Labels","sub":"Match profiles in SQLite"}],"whyLocal":"Why it stays local:","bullets":["Every stage runs on your device via a single sherpa-onnx binary — no Python, no cloud.","~45MB of ONNX models downloaded once, then cached at ~/.cache/openwhispr/diarization-models/.","Embeddings are 512-dim float32 vectors stored as SQLite BLOBs on your local disk."]},"linuxSpectrum":{"title":"Linux Dictation Landscape (2026)","left":"Polished UX","right":"DIY control"},"windowsSpectrum":{"title":"Windows Dictation Landscape (2026)","left":"AI-powered","right":"Specialized"},"localVsCloud":{"title":"Local vs Cloud: Key Tradeoffs","local":{"title":"Local (OpenWhispr)","points":[{"label":"Privacy","detail":"Audio stays on device"},{"label":"Latency","detail":"No network round trip"},{"label":"Cost","detail":"No per-minute API billing"},{"label":"Offline","detail":"Works without internet"},{"label":"Control","detail":"Deterministic local behavior"}]},"cloud":{"title":"Cloud APIs","points":[{"label":"Privacy","detail":"Policy + config dependent"},{"label":"Latency","detail":"Network and region sensitive"},{"label":"Cost","detail":"Usage-based billing"},{"label":"Offline","detail":"Unavailable"},{"label":"Scale","detail":"Easy for large batch workloads"}]},"recommendation":"Best practical pattern: local-first, route edge cases to cloud only when needed."},"openSourceTrust":{"title":"Why Open Source Improves Voice-Data Trust","layers":[{"title":"Auditability","detail":"Inspect data flows directly in code"},{"title":"Reproducibility","detail":"Build from source and verify binaries"},{"title":"Fork rights","detail":"Users keep control if maintainers change direction"},{"title":"No hard lock-in","detail":"Migrate stacks without losing workflows"}]},"voicePrivacyFlow":{"title":"Voice Data Path: Local vs Cloud Risk Surface","microphone":"Microphone","localInference":"Local Inference","localText":"Local Text","zeroExposure":"Zero exposure","cloudApi":"Cloud API","remoteText":"Remote Text","networkExposure":"Network exposure","defaultPath":"OpenWhispr default: local path first, optional cloud only when explicitly configured."},"whisperArchitecture":{"title":"How Whisper Processes Audio","steps":["Audio Input","Log-Mel Features","Encoder","Decoder","Text"],"whyItMatters":"Why it matters for OpenWhispr:","bullets":["Runs locally via whisper.cpp — raw audio stays on-device.","Model size controls the speed/accuracy tradeoff.","Same pipeline powers dictation across macOS, Windows, and Linux."]},"werChart":{"title":"Word Error Rate: Whisper vs Open-Source Competitors","subtitle":"LibriSpeech test-clean (English read speech) — lower is better","werLabel":"WER","turboNote":"8x faster","otherOpenSource":"Other open-source","sources":"Sources: Hugging Face model cards, NVIDIA model cards, Open ASR Leaderboard (Feb 2026). English-clean read speech only — real-world accuracy varies by audio quality, accent, and domain. Commercial APIs omitted as they do not publish LibriSpeech benchmarks."},"modelTradeoff":{"title":"Whisper Model Sizes: Speed vs Accuracy","subtitle":"Bubble size reflects relative parameter count","ariaLabel":"Scatter chart of Whisper model speed vs accuracy tradeoffs","fasterAxis":"Faster →","accuracyAxis":"More accurate →","recommendation":"Recommendation: start with small, adjust based on hardware."},"openAsrWerChart":{"averageWer":"average WER","title":"English Accuracy: Average Word Error Rate","caption":"Multi-dataset averages from the Hugging Face Open ASR Leaderboard benchmarks (lower is better). Nemotron shown in its streaming configuration.","sources":"Sources: NVIDIA model cards on Hugging Face and the Open ASR Leaderboard, July 2026. WER = percentage of words transcribed incorrectly."},"streamingAsrFlow":{"title":"The Live Streaming Path in OpenWhispr"}},"compareFathom":{"metadata":{"title":"Fathom Alternative — Local, Private & Open Source","description":"Looking for a Fathom alternative? OpenWhispr runs voice-to-text and meeting notes locally — no bot, no cloud, no account needed. Free and open source.","keywords":["fathom alternative","fathom ai alternative","private fathom alternative","open source fathom alternative","local meeting notes","no bot meeting notes"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","fathom":"vs Fathom"},"hero":{"title":"The Private, Open-Source Fathom Alternative","description":"OpenWhispr is a Fathom alternative for people who want <voice>voice-to-text</voice> and <meetings>meeting notes</meetings> that run 100% on their own device. No bot joins your calls, your audio never leaves your machine, and the whole app is open source.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","fathomIntro":"is a polished cloud meeting notetaker with a famously generous free tier. Its bot joins your Zoom, Teams, and Meet calls, records and transcribes them, and produces clean summaries, highlights, and action items with very little setup. For automated meeting recaps, it is one of the easiest tools to start with.","openWhisprIntro":"optimizes for something Fathom does not: keeping your audio on your own machine. It is a voice-to-text tool plus on-device meeting notes built on OpenAI Whisper and NVIDIA Parakeet. No bot joins the call, no account is required, and nothing is uploaded. Press a hotkey and your words appear in any app — or record a meeting locally and transcribe it on-device.","chooseFathom":"Choose Fathom if: you want the easiest automated meeting summaries with a strong free tier and don't mind cloud processing.","chooseOpenWhispr":"Choose OpenWhispr if: you want private, local voice-to-text and meeting notes that never upload your audio to anyone."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"100% local — meeting audio stays on your device"},"competitor":{"status":"none","details":"Cloud-based — recordings stored on servers"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"MIT licensed, auditable on GitHub"},"competitor":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Works Offline","openWhispr":{"status":"full","details":"Local models run with no internet"},"competitor":{"status":"none","details":"Requires an internet connection"}},{"feature":"Meeting Capture","openWhispr":{"status":"partial","details":"Records system audio locally — no bot joins"},"competitor":{"status":"full","details":"A bot auto-joins Zoom, Teams, and Meet calls"}},{"feature":"General Dictation","openWhispr":{"status":"full","details":"System-wide voice typing in any app"},"competitor":{"status":"none","details":"Meeting-only; not a dictation tool"}},{"feature":"Transcription & Languages","openWhispr":{"status":"full","details":"Whisper + Parakeet, 100+ languages"},"competitor":{"status":"full","details":"Cloud transcription with AI summaries"}},{"feature":"Free Tier","openWhispr":{"status":"full","details":"Unlimited local dictation, free forever"},"competitor":{"status":"full","details":"Generous free plan for individuals"}},{"feature":"Team & CRM Features","openWhispr":{"status":"partial","details":"API/MCP access; basic team on Business"},"competitor":{"status":"full","details":"CRM sync and team features on paid plans"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","items":[{"title":"Local by default:","body":"Transcription runs on your machine. No bot in the participant list, no audio uploaded, nothing stored on a third-party server."},{"title":"Dictation everywhere:","body":"A system-wide hotkey types your voice into any app — email, docs, your editor, an AI chat. Fathom is meeting-only."},{"title":"Open source:","body":"MIT licensed and auditable — you can verify exactly how your audio is handled."}]},"whyFathom":{"title":"Why Fathom?","items":[{"title":"Effortless meeting recaps:","body":"Its bot joins and summarizes calls automatically, with highlights and action items ready the moment the meeting ends."},{"title":"Generous free tier:","body":"One of the most capable free plans among cloud notetakers for individual users."},{"title":"Polished, low-setup UX:","body":"Clean summaries and CRM sync that most teams can adopt with almost no configuration."}]},"faq":{"title":"Frequently Asked Questions"},"faqs":[{"question":"What is the best Fathom alternative?","answer":"It depends on what you need. If you want a private, open-source alternative for everyday voice-to-text and on-device meeting notes, OpenWhispr is a strong choice — it runs speech-to-text 100% locally using OpenAI Whisper and NVIDIA Parakeet, works system-wide via a push-to-talk hotkey, and never sends your audio to the cloud. If you specifically want the easiest automated meeting summaries with a generous free tier, Fathom itself is excellent at that."},{"question":"Is there a free Fathom alternative?","answer":"Yes. OpenWhispr is free and open source. Unlimited local dictation in 100+ languages works offline at no cost on macOS, Windows, and Linux. An optional Pro plan ($8/mo) adds unlimited cloud transcription, more meeting recording hours, mobile, and device sync — but you never need it for core local voice-to-text."},{"question":"Is there a local or private alternative to Fathom?","answer":"OpenWhispr is built for exactly this. Core transcription happens on your own machine — your audio never leaves the device and works fully offline — and because the app is open source, you can audit precisely how your data is handled. There is no bot in your calls and nothing is uploaded, which matters most for sensitive meetings like HR, legal, or anything with customer data."},{"question":"Does OpenWhispr join meetings with a bot like Fathom?","answer":"No — and that is the point. Fathom sends a bot that appears in your call and uploads the audio to its cloud. OpenWhispr instead captures the meeting's system audio directly on your device, so no bot joins the participant list and nothing is uploaded. You still disclose that you are recording, but the conversation stays entirely on your machine."},{"question":"How is OpenWhispr different from Fathom?","answer":"Fathom is a cloud meeting notetaker focused on effortless, automated summaries with a strong free tier. OpenWhispr is a privacy-first voice-to-text tool plus on-device meeting notes: press a hotkey and speak into any app, or record and transcribe a meeting locally. They solve different problems — many people use a cloud notetaker for routine calls while using OpenWhispr for private, everyday dictation and sensitive meetings."}],"cta":{"title":"A Fathom alternative that never uploads your audio","description":"Private, open-source voice-to-text and on-device meeting notes. Free on macOS, Windows, and Linux.","viewOnGitHub":"View on GitHub"}},"featureDictationSoftware":{"metadata":{"title":"Dictation Software (Download for Free) | OpenWhispr","description":"OpenWhispr is free dictation software that turns speech into text in any app. Works offline, runs locally and keeps your voice on your device.","keywords":["dictation software","voice dictation software","free dictation software","speech to text software","voice typing","dictation app","offline dictation","local dictation software"]},"breadcrumb":{"home":"Home","dictationSoftware":"Dictation Software"},"structuredData":{"description":"Open-source dictation software. Turn speech into text in any app — run it privately on your device or on OpenWhispr Cloud for top accuracy — in 100+ languages.","offerDescription":"Free to start. Unlimited local dictation, with optional Pro and Business plans.","featureList":["System-wide push-to-talk dictation in any app","Local, offline speech to text plus OpenWhispr Cloud for top accuracy","100+ languages with built-in translation","OpenAI Whisper and NVIDIA Parakeet models","Custom dictionary for names, jargon and acronyms","Built-in AI writing assistant to reformat and clean up dictation","Open source on macOS, Windows and Linux"]},"hero":{"badge":"Free to start · Open source","title":"Free Dictation Software for Every App You Use","description":"Press a hotkey, speak, and clean, punctuated text appears anywhere — email, docs, Slack, your editor or an AI chat. Run it privately on your device, or switch on OpenWhispr Cloud for top accuracy on any machine. Either way, your audio is never stored.","github":"View on GitHub","platforms":"Free to start on macOS, Windows & Linux · No card, no account"},"trustBar":{"title":"Dictate into every app you already use"},"painPoints":{"eyebrow":"Why people switch","title":"Most dictation software makes you trade something away.","description":"Speed for privacy, privacy for accuracy, or your whole workflow for one company’s cloud. OpenWhispr is built so you don’t have to choose.","items":[{"title":"Your ideas outrun your fingers.","body":"You speak several times faster than you type. OpenWhispr turns speech into clean text at the speed you think, so email, prompts and drafts stop getting squeezed through a keyboard."},{"title":"You can’t see what most voice tools do with your voice.","body":"Most dictation software is closed and cloud-only. OpenWhispr is open source, and whether you run it locally or on OpenWhispr Cloud, your audio is transcribed and never stored, sold, or used for training — verifiable in the code."},{"title":"Your setup shouldn’t be locked to one app or OS.","body":"Closed tools trap your workflow in their cloud and their platform. OpenWhispr runs on Mac, Windows and Linux, works in every app, and lets you bring your own AI keys — your workflow stays yours."},{"title":"Built-in dictation isn’t enough.","body":"Apple Dictation and Windows Speech are inaccurate, online-dependent and locked to one OS. OpenWhispr runs Whisper-class voice recognition on all three platforms, with a custom dictionary for your names and jargon."}]},"features":{"eyebrow":"What you get","title":"One hotkey between your voice and every app.","description":"Dictation software that works wherever your cursor is — dictate emails, prompt AI, and capture ideas with no destination app to copy and paste from.","items":[{"title":"Dictate into any app","body":"Press the hotkey, speak, release. Punctuated text appears wherever your cursor is — email, docs, Slack, your code editor. Nothing to copy and paste."},{"title":"Talk to AI instead of typing at it","body":"Speak long, context-rich prompts into ChatGPT, Claude or Cursor. Dictation is the fastest way to give AI the detail it needs for a better answer."},{"title":"Accurate in your words","body":"Teach OpenWhispr your names, jargon and acronyms with a custom dictionary, then pick the engine that fits — run locally, or switch on OpenWhispr Cloud for top accuracy on any machine."}]},"stats":{"items":[{"value":"100+","label":"Dictate in","foot":"languages, with translation built in"},{"value":"0","label":"Recordings stored","foot":"local audio never leaves your device; cloud audio is transcribed, then discarded"},{"value":"5 min","label":"Setup takes","foot":"from download to your first dictated sentence"}]},"howItWorks":{"eyebrow":"How it works","title":"Talk to your computer in three steps.","description":"No cloud account, no signup. Just software on your machine, running in about five minutes.","items":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model, or use OpenWhispr Cloud — no account, no card."},{"title":"Set your hotkey.","body":"One shortcut — hold-to-record or push-to-toggle — working system-wide across every app."},{"title":"Press. Speak. Done.","body":"Clean, punctuated text lands wherever your cursor is, in 100+ languages."}]},"reasons":{"eyebrow":"No convincing needed","title":"10 reasons people make OpenWhispr their dictation software","description":"Switching is a trust decision. Every claim here is verifiable in the open-source repo.","items":[{"title":"Works in every app","body":"A system-wide hotkey. If you can place a cursor there, you can dictate there."},{"title":"Private by design, and auditable","body":"Open source with zero data retention — local audio stays on your device, and cloud audio is transcribed then discarded."},{"title":"Local or cloud — your call","body":"Run on-device for full privacy, or switch on OpenWhispr Cloud for top accuracy and speed on any machine."},{"title":"Whisper-class and Parakeet accuracy","body":"The leading open speech-to-text models, plus a cloud option that keeps up on modest hardware."},{"title":"A custom dictionary for your words","body":"Teach it names, jargon and acronyms so your transcripts come out right the first time."},{"title":"100+ languages plus translation","body":"Speak one language, output another — set a language or auto-detect on the fly."},{"title":"Bring your own AI keys","body":"Use your own OpenAI, Anthropic, Gemini or Groq keys for unlimited cloud transcription, no markup."},{"title":"Clean up as you speak","body":"A built-in AI writing assistant reformats and polishes raw dictation — locally or in the cloud."},{"title":"Mac, Windows & Linux","body":"One of the only dictation apps that runs natively on all three, with the same features everywhere."},{"title":"Easier on your hands","body":"If typing is slow, tiring or painful — RSI, fatigue, long days — your voice does the work instead."}]},"faq":{"eyebrow":"Before you download","title":"Common questions about OpenWhispr dictation software","items":[{"question":"Does OpenWhispr work in every app?","answer":"Yes. Unlike voice-to-text software that lives in one app, the hotkey works system-wide — email, docs, Slack, browsers, code editors and AI chat windows. Wherever you can type, you can dictate."},{"question":"How accurate is the speech to text, accents included?","answer":"It runs Whisper-class models and NVIDIA Parakeet, trained on a huge range of accents and 100+ languages. Pick a larger local model, add a custom dictionary for your jargon, or switch on OpenWhispr Cloud for top accuracy on any machine."},{"question":"What is the difference between local and cloud dictation?","answer":"Local dictation runs on your own device — fully private, works offline, and unlimited. OpenWhispr Cloud runs on our servers for top accuracy and speed on any hardware, with no models to download, and audio is transcribed then never stored. Free includes 2,000 cloud words a week; Pro makes it unlimited."},{"question":"Is it free? What do I pay for?","answer":"The app is open source and free — unlimited local dictation in 100+ languages, a custom dictionary, plus 2,000 OpenWhispr Cloud words a week and 5 hours of meeting recording. Pro ($8/month) unlocks unlimited cloud transcription, 20 meeting hours, mobile and device sync."},{"question":"Does my voice ever leave my device?","answer":"In local mode, never — and the open-source code proves there is no upload step. If you use OpenWhispr Cloud or your own API keys, audio is sent only to transcribe it, then discarded. We never store, sell, or train on your data."},{"question":"Does it work as dictation software for Mac, Windows and PC?","answer":"Yes. OpenWhispr is one of the only open-source dictation apps that runs natively on macOS, Windows and Linux. Mac dictation, Windows dictation and Linux all use the same engine and the same features."},{"question":"Can it learn my vocabulary and jargon?","answer":"Yes. The custom dictionary teaches OpenWhispr your names, technical terms and acronyms so specialized language transcribes correctly the first time — included on every plan, local or cloud."},{"question":"Is this AI dictation? Can it format and clean up text?","answer":"Yes. Beyond raw transcription, a built-in AI writing assistant punctuates, reformats and rewrites what you said — using local models or the cloud, with agent mode to automate actions on your transcripts."},{"question":"How is it different from voice recognition software like Dragon or built-in dictation?","answer":"Dragon is powerful but expensive and Windows-first; Apple and Windows dictation are online-dependent and locked to one OS. OpenWhispr runs Whisper-class voice recognition on all three platforms, is open source, and lets you choose local or cloud."},{"question":"What do I need to run it, and does it work offline?","answer":"A reasonably modern Mac, Windows or Linux machine. Smaller local models run on modest hardware, and once installed, local dictation works fully offline. Prefer not to manage models? OpenWhispr Cloud needs nothing to download."}]},"testimonials":{"eyebrow":"What users say","title":"People who stopped typing."},"pricing":{"eyebrow":"Pricing","title":"Start free. Upgrade for cloud speed and more.","description":"Unlimited local dictation is free forever. Pro and Business add unlimited OpenWhispr Cloud, meeting hours, mobile, and team features."},"openSource":{"eyebrow":"Built in the open","title":"Your voice, handled the way you would handle it.","description":"OpenWhispr started with a simple objection: most dictation software listens to everything you say from behind closed source. So we built ours in the open — every release and every line of code is on GitHub, where 5,000+ engineers have starred, forked and audited the project. Local or cloud, we never store, sell, or train on your data, and you can prove it by reading the code.","github":"Star us on GitHub"},"finalCta":{"title":"Your next sentence doesn’t need a keyboard.","description":"Download OpenWhispr and dictate into any app on your computer — in 100+ languages, private by design, five minutes from now.","footer":"Free to start on Mac, Windows & Linux · No card, no account"},"demo":{"dictationText":"Hey team, I just finished reviewing the Q2 roadmap and I think we should prioritize the voice pipeline before the new integrations.","without":"Without OpenWhispr","typingSpeed":"~40 WPM","windowTitle":"Slack — #product","messagePlaceholder":"Type a message…","with":"With OpenWhispr","dictationSpeed":"~150 WPM","faster":"3x faster","transcribing":"Transcribing…","listening":"Listening…","hotkeyPrompt":"Press your hotkey and speak…","indicator":"Hold the hotkey, speak, release — text lands in the active app."}},"compareFireflies":{"metadata":{"title":"Fireflies.ai Alternative — Private, Local, Open Source","description":"A private, open-source Fireflies.ai alternative. Voice-to-text and meeting notes run 100% locally — no bot joins your calls, no audio uploaded.","keywords":["fireflies alternative","fireflies.ai alternative","private fireflies alternative","open source fireflies alternative","local meeting notes","no bot meeting notes"]},"breadcrumb":{"home":"Home","comparisons":"Comparisons","fireflies":"vs Fireflies.ai"},"hero":{"title":"The Private, Open-Source Fireflies.ai Alternative","description":"OpenWhispr is a Fireflies.ai alternative for people who want <voice>voice-to-text</voice> and <meetings>meeting notes</meetings> that run 100% on their own device. No bot joins your calls, your audio never leaves your machine, and the whole app is open source.","viewOnGitHub":"View on GitHub"},"quickTake":{"title":"The Quick Take","firefliesIntro":"is a cloud meeting assistant. Its bot joins your Zoom, Teams, and Meet calls, records and transcribes them, generates summaries and action items, and layers on conversation intelligence and CRM integrations for sales and revenue teams. It is genuinely good at that job.","openWhisprIntro":"is built around a different priority: privacy and ownership. It is a voice-to-text tool plus on-device meeting notes that runs entirely on your computer using OpenAI Whisper and NVIDIA Parakeet. No bot joins the call, no account is required, and nothing is uploaded. Press a hotkey and your words appear in any app — or record a meeting's audio locally and transcribe it on-device.","chooseFireflies":"Choose Fireflies.ai if: you want an automated bot that joins calls with team analytics and CRM sync.","chooseOpenWhispr":"Choose OpenWhispr if: you want private, local voice-to-text and meeting notes that never upload your audio to anyone."},"comparisonTable":{"title":"Feature-by-Feature Comparison","headerFeature":"Feature"},"comparisonData":[{"feature":"Privacy & Data Control","openWhispr":{"status":"full","details":"100% local — meeting audio stays on your device"},"competitor":{"status":"none","details":"Cloud-based — audio and transcripts stored on servers"}},{"feature":"Open Source","openWhispr":{"status":"full","details":"MIT licensed, auditable on GitHub"},"competitor":{"status":"none","details":"Proprietary, closed source"}},{"feature":"Works Offline","openWhispr":{"status":"full","details":"Local models run with no internet"},"competitor":{"status":"none","details":"Requires an internet connection"}},{"feature":"Meeting Capture","openWhispr":{"status":"partial","details":"Records system audio locally — no bot joins"},"competitor":{"status":"full","details":"A bot auto-joins Zoom, Teams, and Meet calls"}},{"feature":"General Dictation","openWhispr":{"status":"full","details":"System-wide voice typing in any app"},"competitor":{"status":"none","details":"Meeting-focused; not a dictation tool"}},{"feature":"Transcription & Languages","openWhispr":{"status":"full","details":"Whisper + Parakeet, 100+ languages"},"competitor":{"status":"full","details":"Cloud transcription with AI search"}},{"feature":"CRM & Team Intelligence","openWhispr":{"status":"partial","details":"API/MCP access; basic team on Business"},"competitor":{"status":"full","details":"Salesforce/HubSpot sync, conversation analytics"}},{"feature":"Free Tier","openWhispr":{"status":"full","details":"Unlimited local dictation, free forever"},"competitor":{"status":"partial","details":"Free tier with limits — see their site"}}],"whyOpenWhispr":{"title":"Why OpenWhispr?","items":[{"title":"Local by default:","body":"Transcription runs on your machine. No bot in the participant list, no audio uploaded, nothing stored on a third-party server."},{"title":"Dictation everywhere:","body":"A system-wide hotkey types your voice into any app — email, docs, your editor, an AI chat. Fireflies is meeting-only."},{"title":"Open source:","body":"MIT licensed and auditable — you can verify exactly how your audio is handled."}]},"whyFireflies":{"title":"Why Fireflies.ai?","items":[{"title":"Automated call-joining:","body":"Its bot joins scheduled calls on its own and captures them without you lifting a finger."},{"title":"CRM & revenue intelligence:","body":"Deep Salesforce/HubSpot sync and conversation analytics built for sales teams."},{"title":"Shared team workspace:","body":"Searchable, collaborative transcripts across an organization out of the box."}]},"faqs":[{"question":"What is the best Fireflies.ai alternative?","answer":"It depends on what you need. If you want a private, open-source alternative for everyday voice-to-text and on-device meeting notes, OpenWhispr is a strong choice — it runs speech-to-text 100% locally using OpenAI Whisper and NVIDIA Parakeet, works system-wide via a push-to-talk hotkey, and never sends your audio to the cloud. If you specifically need an automated bot that joins calls with team analytics and CRM integrations, Fireflies.ai itself remains excellent at that."},{"question":"Is there a free Fireflies.ai alternative?","answer":"Yes. OpenWhispr is free and open source. Unlimited local dictation in 100+ languages works offline at no cost on macOS, Windows, and Linux. An optional Pro plan ($8/mo) adds unlimited cloud transcription, more meeting recording hours, mobile, and device sync — but you never need it for core local voice-to-text."},{"question":"Is there a local or private alternative to Fireflies.ai?","answer":"OpenWhispr is built for exactly this. Core transcription happens on your own machine — your audio never leaves the device and works fully offline — and because the app is open source, you can audit precisely how your data is handled. There is no bot in your calls and nothing is uploaded, which matters most for sensitive meetings like HR, legal, or anything with customer data."},{"question":"Does OpenWhispr join meetings with a bot like Fireflies?","answer":"No — and that is the point. Fireflies sends a bot that appears in your call and uploads the audio to its cloud. OpenWhispr instead captures the meeting's system audio directly on your device, so no bot joins the participant list and nothing is uploaded. You still disclose that you are recording, but the conversation stays entirely on your machine."},{"question":"How is OpenWhispr different from Fireflies.ai?","answer":"Fireflies.ai is a cloud meeting assistant focused on automated call-joining, conversation intelligence, and CRM sync for sales and revenue teams. OpenWhispr is a privacy-first voice-to-text tool plus on-device meeting notes: press a hotkey and speak into any app, or record and transcribe a meeting locally. They solve different problems — many people keep a bot tool for collaborative sales calls while using OpenWhispr for private, everyday dictation and sensitive meetings."}],"faq":{"title":"Frequently Asked Questions"},"cta":{"title":"A Fireflies.ai alternative that never uploads your audio","description":"Private, open-source voice-to-text and on-device meeting notes. Free on macOS, Windows, and Linux.","viewOnGitHub":"View on GitHub"}},"unsubscribe":{"metadata":{"title":"Unsubscribe — OpenWhispr","description":"Manage your OpenWhispr marketing email subscription.","keywords":["OpenWhispr unsubscribe","email preferences","marketing email"]},"breadcrumb":{"home":"Home","unsubscribe":"Unsubscribe"},"invalid":{"title":"This link isn't valid","description":"The unsubscribe link is missing or damaged. Try the link from a newer email, or ask us to remove you directly.","contact":"Email <email>support@openwhispr.com</email> and we'll take care of it."},"success":{"title":"You're unsubscribed","description":"<email>{emailAddress}</email> won't receive marketing email from OpenWhispr anymore. This took effect immediately.","serviceMessages":"You'll still get service messages — receipts, security alerts, and password resets — because those are part of your account."},"confirm":{"title":"Unsubscribe from marketing email?","description":"<email>{emailAddress}</email> will stop receiving product news and offers from OpenWhispr. Service messages about your account aren't affected.","button":"Unsubscribe"}},"legalDpa":{"metadata":{"title":"Data Processing Addendum | OpenWhispr","description":"OpenWhispr's Data Processing Addendum (Gizmo Labs Inc. as processor) — self-executing, with a no-AI-training commitment, SCC and UK Addendum transfer clauses, and a product subprocessor annex."},"breadcrumb":{"home":"Home","dpa":"Data Processing Addendum"},"authoritativeNotice":"This translation is provided for convenience only. If it differs from the English version, the English version governs.","header":{"title":"Data Processing Addendum","subtitle":"How Gizmo Labs Inc. (OpenWhispr) processes personal data on behalf of its customers. Customer content is never used to train AI models (Section 3).","effective":"Effective July 17, 2026 · Version 1.2, updated August 21, 2026"},"preamble":{"title":"About this Addendum","body":"$16"},"definitions":{"title":"1. Definitions","body":"Capitalised terms not defined here have the meaning given in Data Protection Law. \"Data Protection Law\" means all laws applicable to the processing of Personal Data under the Agreement, including the EU GDPR (Regulation (EU) 2016/679), the UK GDPR and Data Protection Act 2018, the Swiss FADP, and applicable US state privacy laws (including the CCPA/CPRA). \"SCCs\" means the Standard Contractual Clauses in Commission Implementing Decision (EU) 2021/914. \"UK Addendum\" means the ICO's International Data Transfer Addendum to the EU SCCs. \"Restricted Transfer\" means a transfer of Personal Data to a third country lacking an adequacy decision. \"Sub-processor\" means a third party engaged by OpenWhispr to process Personal Data on the Controller's behalf."},"scope":{"title":"2. Scope and roles","body":"$17"},"noAiTraining":{"title":"3. No AI training on Customer Data","body":"OpenWhispr will not use Customer Personal Data — including voice audio, transcriptions, notes, prompts and AI-agent conversations — to train, retrain or improve any artificial intelligence or machine learning model, whether OpenWhispr's own or a third party's, and does not permit its Sub-processors to do so. The model providers listed in Annex 3 process such content solely to provide the Services, with training disabled or opted out for OpenWhispr's accounts. This restriction survives termination of the Agreement."},"processorObligations":{"title":"4. Processor obligations","intro":"OpenWhispr will:","items":["process Personal Data only on documented instructions from the Customer (the Agreement, this DPA, and the Customer's configuration of the Services constitute such instructions), including for transfers, unless required by law to which OpenWhispr is subject — in which case OpenWhispr informs the Customer of that legal requirement before processing, unless legally prohibited;","ensure persons authorised to process Personal Data are bound by confidentiality obligations;","implement the technical and organisational security measures set out in Annex 2 and keep them under periodic review;","assist the Customer, taking into account the nature of processing, with data-subject requests, security, breach notification, data protection impact assessments and prior consultation (Sections 6–9);","at the Customer's choice, delete or return all Personal Data at the end of the Services and delete existing copies within 30 days, unless law to which OpenWhispr is subject requires retention;","make available information reasonably necessary to demonstrate compliance and allow and contribute to audits per Section 10; and","immediately inform the Customer if, in OpenWhispr's opinion, an instruction infringes Data Protection Law."]},"subprocessing":{"title":"5. Sub-processing","body":"The Customer grants OpenWhispr general written authorisation to engage Sub-processors. The current list is set out in Annex 3 and published on the OpenWhispr trust center at <trustLink>trust.openwhispr.com</trustLink> (Subprocessors section). OpenWhispr imposes data protection obligations on each Sub-processor equivalent to those in this DPA (including transfer obligations) and remains liable for their performance. OpenWhispr gives the Customer at least 30 days'' prior written notice of intended additions or replacements (by email to the Customer's account owner, and by updating the published list), during which the Customer may object on reasonable, documented data-protection grounds. If the parties cannot resolve an objection, the Customer may terminate the affected Services and receive a pro-rata refund of any prepaid fees for the terminated Services."},"security":{"title":"6. Security","body":"OpenWhispr maintains the measures in Annex 2, appropriate to the risk, taking account of the state of the art and the nature of the Personal Data. Voice audio processed through the cloud transcription path is processed in real time and not persisted by OpenWhispr after transcription."},"breach":{"title":"7. Personal data breach","body":"OpenWhispr notifies the Customer without undue delay, and in any event within 72 hours, of becoming aware of a Personal Data breach affecting the Customer's Personal Data, providing the information the Customer reasonably needs to meet its own notification obligations, and cooperates on remediation."},"dataSubjectRequests":{"title":"8. Data-subject requests","body":"Taking into account the nature of the processing, OpenWhispr assists the Customer by appropriate technical and organisational measures in responding to requests to exercise data-subject rights (access, rectification, erasure, restriction, portability, objection). End users can self-serve export and deletion in-product; requests received directly by OpenWhispr that identify the Customer's organisation are forwarded to the Customer without undue delay."},"dpias":{"title":"9. DPIAs and prior consultation","body":"OpenWhispr provides reasonable assistance with data protection impact assessments and prior consultation with supervisory authorities, to the extent the required information is available to OpenWhispr."},"audit":{"title":"10. Audit","body":"OpenWhispr makes available on request its current compliance documentation (SOC 2 Type 2 report for the period 1 May 2026 to 31 July 2026, available under NDA; GDPR and HIPAA attestations issued via its compliance platform, 2026; ISO/IEC 27001:2022 certification, certificate KCISM202607012) via its trust center, subject to confidentiality. Where Data Protection Law grants the Customer a mandatory audit right that cannot be satisfied by such reports, OpenWhispr allows an audit by the Customer or an independent auditor not more than once in any 12-month period, and additionally following a confirmed Personal Data breach affecting the Customer's Personal Data, in each case on at least 30 days' notice, subject to confidentiality and OpenWhispr's security policies. The Customer bears its own costs of any such audit. For Restricted Transfers, this Section applies without prejudice to the audit rights in the SCCs."},"customerObligations":{"title":"11. Customer obligations and warranties","body":"The Customer warrants that its instructions and its provision of Personal Data to OpenWhispr comply with Data Protection Law, including having a lawful basis, providing required notices to data subjects, and — where end users may dictate or process special categories of personal data (for example health information) — ensuring an appropriate legal basis and any required agreements are in place (see Section 14 for HIPAA)."},"transfers":{"title":"12. International transfers","intro":"This Section governs transfers of Personal Data originating from the EEA, the UK or Switzerland to a country outside those areas, including any Restricted Transfer. OpenWhispr and its Sub-processors process Personal Data primarily in the United States (per-Sub-processor locations: Annex 3). The parties rely on the first applicable mechanism below for each transfer:","mechanisms":["Adequacy. Where the destination benefits from an adequacy decision, the transfer proceeds on that basis.","Standard Contractual Clauses. For any Restricted Transfer from the EEA, the parties enter into the SCCs, Module Two (Controller → Processor) with the Customer as data exporter and OpenWhispr as data importer, which are incorporated into this DPA by reference and completed as follows: Clause 7 (docking) not included; Clause 9 Option 2 (general authorisation, 30 days' notice); Clause 11 optional redress body not included; Clause 13 / Annex I.C competent supervisory authority: the Irish Data Protection Commission; Clause 17 governing law: the law of Ireland; Clause 18 forum: the courts of Ireland; Annexes 1–3 of this DPA populate SCC Annexes I–III. Onward transfers to Sub-processors are governed by Clauses 8.8 and 9(b) of the Module Two SCCs (same-obligations flow-down).","UK and Swiss transfers. For UK Personal Data the parties enter into the UK Addendum to the EU SCCs, completed using the details in the Annexes; for Swiss Personal Data the SCCs apply as modified by the FADP."],"futureFramework":"Should OpenWhispr self-certify under the EU–US Data Privacy Framework in the future, transfers may additionally rely on that certification.","supplementaryMeasures":"Where a transfer impact assessment identifies that supplementary measures are required, the parties implement them, and OpenWhispr supports the Customer's assessment. OpenWhispr agrees to the obligations in the SCCs, including notifying the Customer of any inability to comply."},"governmentRequests":{"title":"13. Government access requests","body":"If OpenWhispr receives a legally binding request from a public authority for the Customer's Personal Data, it will, to the extent legally permitted: notify the Customer promptly; review the legality of the request and challenge it where there are reasonable grounds; and disclose only the minimum amount of data legally required. OpenWhispr keeps records of such requests for the Customer's review."},"hipaa":{"title":"14. HIPAA","body":"This DPA is not a Business Associate Agreement. Where the Customer is a Covered Entity or Business Associate under HIPAA, end users must not process protected health information through the cloud Services unless a separate Business Associate Agreement is in place between the parties — OpenWhispr offers a BAA to Customers on the Business plan and above, on request via support@openwhispr.com. Where executed, that BAA governs PHI in place of this DPA to the extent of any conflict."},"usStatePrivacy":{"title":"15. US state privacy laws","body":"To the extent the CCPA/CPRA or another US state privacy law applies, OpenWhispr acts as a \"service provider\"/\"processor\": it does not sell or share Personal Data, does not retain, use or disclose it other than to provide the Services or as permitted by law, and certifies that it understands and will comply with these restrictions."},"liability":{"title":"16. Liability and term","body":"Liability for data protection matters is as set out in the Agreement, subject to any caps therein. This DPA continues for as long as OpenWhispr processes Personal Data for the Customer."},"signatures":{"title":"Signatures (optional — for customers requiring an executed copy)","nameRole":"Name & role","signature":"Signature","date":"Date","customer":"For the Customer (Controller)","processor":"For Gizmo Labs Inc. (Processor)","executive":"Gabriel Stein, CEO"},"annex1":{"title":"Annex 1 — Details of processing","parties":"A. Parties (SCC Annex I.A). Data exporter: the Customer (controller) — name, address and contact person as set out in the Agreement or Order Form. Data importer: Gizmo Labs Inc. (processor), 221 W 9th Street, Wilmington, Delaware 19801, USA, contact: support@openwhispr.com; activities: provision of the Services described below.","descriptionTitle":"B. Description of processing (SCC Annex I.B).","field":"Field","detail":"Detail","rows":[{"field":"Subject-matter","detail":"Provision of the OpenWhispr voice-dictation and AI note-taking services (desktop applications and API)"},{"field":"Duration","detail":"Term of the Agreement, plus the post-termination deletion/return period under Section 4(e)"},{"field":"Nature & purpose","detail":"Real-time speech-to-text transcription; AI-assisted text generation, formatting and reasoning; note storage, search and sharing; account management; billing; transactional email"},{"field":"Categories of Personal Data","detail":"Account and identity data (name, email, authentication identifiers); voice audio (processed in real time in the cloud path, not persisted by OpenWhispr); transcribed text and notes (may contain any personal data the end user dictates); conversation history with AI features; usage and device metadata; billing data (handled by the payment Sub-processor)"},{"field":"Special category data","detail":"Possible incidentally: dictated content is user-directed free text and may contain health or other special-category data. Safeguards: encryption in transit and at rest, no audio persistence, local-only and BYOK modes, BAA required for PHI (Section 14)"},{"field":"Categories of data subjects","detail":"The Customer's end users (employees, contractors); persons mentioned in dictated content or notes"},{"field":"Frequency","detail":"Continuous, for the duration of the Services"}]},"annex2":{"title":"Annex 2 — Technical and organisational measures","intro":"Summary of OpenWhispr's security programme (current detail: OpenWhispr trust center — <trustLink>trust.openwhispr.com</trustLink>, Security Controls section):","bullets":["Encryption of Personal Data in transit (TLS 1.2+) and at rest.","Cloud voice audio processed in real time and not persisted by OpenWhispr after transcription (transcription/inference Sub-processors handle audio transiently under their agreements); local-only and bring-your-own-key modes route content around OpenWhispr's cloud entirely.","Access control on least-privilege with MFA and logging/monitoring across production systems; session and login monitoring with monthly review.","Vulnerability and patch management; secure development practices.","Resilience and backup with documented recovery objectives (RTO ≤ 4h / RPO ≤ 1h for the primary datastore) and contingency-plan testing on an annual cadence.","Personnel confidentiality obligations and security training.","Sub-processor governance: contractual flow-down (Section 5), public listing, and BAAs executed with the primary content-storing Sub-processors (database; cloud transcription/LLM), with coverage of the remaining HIPAA-path Sub-processors in progress.","Independently examined: SOC 2 Type 2 report issued for the period 1 May 2026 to 31 July 2026, covering Security, Confidentiality and Availability, available under NDA. Compliance attestations: GDPR and HIPAA (2026, via OpenWhispr's compliance platform). Information security management system certified to ISO/IEC 27001:2022 (certificate KCISM202607012, issued by KVQA Certification Services, accredited by the Standards Council of Canada; initial registration 27 July 2026, valid to 27 July 2027)."]},"annex3":{"title":"Annex 3 — Sub-processors","intro":"Authorised Sub-processors as of 2026-07-17 (live list: OpenWhispr trust center, <trustLink>trust.openwhispr.com</trustLink>, Subprocessors section). Legal entity names and processing locations verified 2026-07-17 against each vendor's terms/DPA. \"US (primary)\" means the vendor's terms name the United States as the primary processing location while permitting processing in other countries by their own sub-processors. \"Clause 9(b) flow-down\" means the same-obligations Sub-processor contract under Clauses 8.8 and 9(b) of the Module Two SCCs, where a Restricted Transfer applies.","headers":{"subprocessor":"Sub-processor","service":"Service","dataCategories":"Data categories","country":"Country","safeguard":"Onward-transfer safeguard"},"rows":[{"subprocessor":"Neon, LLC (an affiliate of Databricks, Inc.)","service":"Primary database and authentication store","dataCategories":"Account data, notes, transcribed text, conversation history","country":"US (AWS US region, fixed at project creation)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Vercel Inc.","service":"Application hosting and delivery","dataCategories":"All data in transit through the Services","country":"US (primary)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"OpenAI OpCo, LLC","service":"Cloud speech-to-text, language models, embeddings (training opted out)","dataCategories":"Voice audio (transient), prompts, text","country":"US (primary)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Parasail, Inc.","service":"Cloud AI inference (default reasoning; zero-data-retention inference)","dataCategories":"Prompts, text (in transit)","country":"US (primary)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Groq, Inc.","service":"Cloud AI inference and transcription failover","dataCategories":"Voice audio (transient), prompts, text (in transit)","country":"US (primary)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Embedchain, Inc. d/b/a Mem0","service":"AI memory for agent features","dataCategories":"Conversation content keyed to user ID","country":"US","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Exa Labs, Inc.","service":"Web search for agent features","dataCategories":"Search query text","country":"US (AWS)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Stripe, LLC","service":"Payment processing","dataCategories":"Billing and payment data","country":"US (primary)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Plus Five Five, Inc. (d/b/a Resend)","service":"Transactional email","dataCategories":"Email address, message content (incl. shared-note titles)","country":"US (AWS us-east-1)","safeguard":"Clause 9(b) flow-down"},{"subprocessor":"Slack Technologies, LLC (Salesforce)","service":"Internal operations and support communications","dataCategories":"Support-related data (policy: no PHI)","country":"US (default residency)","safeguard":"Clause 9(b) flow-down"}],"footnote":"This annex lists the Sub-processors that process Customer Personal Data within the product Services. Google (OAuth sign-in and the optional Calendar integration) is not a Sub-processor: sign-in is provided by Google as an independent controller, and calendar data is cached only on the user's device. OpenWhispr's broader vendor transparency list (including marketing-site and operational vendors such as RB2B, lemlist and GitHub, for which Gizmo Labs acts as controller or which do not receive Customer Personal Data) is published on the OpenWhispr trust center."}},"blogBestNoteTakers":{"metadata":{"title":"Best AI Note Takers 2026 (Free & Private) | OpenWhispr","description":"Compare the best AI note takers of 2026 — Fathom, Otter, Granola, Fireflies, and OpenWhispr, the free, private, open-source option that runs locally.","keywords":["best ai note taker","best ai note takers","best free ai note taker","open source ai note taker","best ai note taker for privacy","ai note taker comparison","private ai note taker","best ai meeting notes"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"The Best AI Note Takers"},"schema":{"title":"The Best AI Note Takers in 2026 (Including Free and Open-Source Options)","description":"Compare the best AI note takers of 2026 — Fathom, Otter, Granola, Fireflies, and OpenWhispr, the free, private, open-source option that runs locally."},"faqs":[{"question":"What is the best AI note taker?","answer":"There is no single best one — it depends on your priority. For automated meeting summaries with the least setup, Fathom and Otter are excellent. For sales teams that need CRM sync and conversation intelligence, Fireflies is strong. For privacy — keeping audio on your own device — OpenWhispr is the standout because it transcribes locally, is open source, and never sends a bot into your call."},{"question":"What is the best free AI note taker?","answer":"Fathom has one of the most generous free cloud tiers for individuals. If your priority is a free tool that also keeps your data private, OpenWhispr is free and open source with unlimited local transcription — no per-month cap, no card, and nothing uploaded."},{"question":"Is there an open-source AI note taker?","answer":"Yes. OpenWhispr is MIT-licensed and fully open source. Its speech-to-text runs on your own device using OpenAI Whisper and NVIDIA Parakeet, so you can audit exactly how your audio is handled. Meetily is another open-source, local-first option worth a look."},{"question":"What is the best AI note taker for privacy?","answer":"The most private option is one that transcribes on your own device and never uploads your audio. OpenWhispr does this by default — no bot joins the call and the recording stays on your machine — which makes it the safest choice for sensitive meetings like HR, legal, or anything involving customer data. Most cloud note-takers, by contrast, upload and store your meetings on their servers."},{"question":"Do AI note takers need a bot to join the meeting?","answer":"Most do — Otter, Fireflies, and Fathom send a bot that appears in your participant list. A few don't: Granola captures your computer's audio without a bot, and OpenWhispr does the same while also keeping transcription entirely on your device."},{"question":"Which AI note taker works offline?","answer":"Cloud tools like Otter, Fireflies, and Fathom need an internet connection because transcription happens on their servers. OpenWhispr can run fully offline: its local models transcribe on your machine with no connection required, which is useful on flights, in secure facilities, or anywhere you simply don't want audio leaving the device."}],"sections":[{"id":"how-to-choose","title":"How to Choose"},{"id":"comparison","title":"Quick Comparison"},{"id":"picks","title":"The Picks"},{"id":"faq","title":"FAQ"}],"hero":{"title":"The Best AI Note Takers in 2026 (Including Free and Open-Source Options)","subtitle":"Automated meeting summaries, or private on-device notes? A fair, honest comparison of the top AI note takers — what each one is genuinely best at, and where it falls short.","date":"July 6, 2026","tag":"Comparison"},"opening":{"lead":"There is no single best AI note taker — the right one depends on whether you value automation, price, or privacy.","body":"For hands-off cloud summaries, Fathom and Otter are excellent. For sales teams, Fireflies. For keeping your audio on your own device, OpenWhispr. Below is an honest breakdown of each, including where they beat the others.","updated":"Last updated: July 6, 2026. OpenWhispr is on this list — we've tried to be fair about where other tools win."},"howToChoose":{"title":"How to Choose an AI Note Taker","intro":"Almost every AI note taker records a call, transcribes it, and summarizes the result. The real differences come down to three questions:","criteria":[{"lead":"Where does your audio go?","body":"Cloud tools upload and store your meeting on their servers. Local tools keep it on your device. For HR, legal, medical, or customer conversations, this is the deciding factor."},{"lead":"Does a bot join the call?","body":"Most send a visible bot into your participant list. A few capture audio without one, which is less intrusive and avoids awkward ‘who is that?’ moments."},{"lead":"What does the free tier actually cover?","body":"‘Free’ ranges from a few meetings a month to genuinely unlimited. Check the caps before you commit."}]},"comparison":{"title":"Quick Comparison","headers":{"tool":"Tool","bestFor":"Best for","runs":"Runs","bot":"Bot?","freeTier":"Free tier"}},"picks":{"title":"The Best AI Note Takers, One by One","keepInMind":"Keep in mind:","openWhisprLink":"See OpenWhispr","compareLink":"OpenWhispr vs {name}","tools":[{"name":"OpenWhispr","bestFor":"Privacy & open source","runs":"Local","bot":"No bot","freeTier":"Unlimited local, free","blurb":"Full disclosure: this is our tool — so here is the honest case. OpenWhispr transcribes on your own device using OpenAI Whisper and NVIDIA Parakeet, is open source (MIT), and captures meeting audio without sending a bot into the call. It doubles as a system-wide dictation tool for typing with your voice in any app.","pros":["Audio never leaves your device; nothing uploaded","Open source and auditable, on macOS, Windows and Linux","Unlimited local transcription, free forever"],"caveat":"It won't auto-join your scheduled calls with a bot — you start recording yourself — and the polished cloud summaries of a dedicated notetaker take a bit more setup.","href":"/meeting-notes-software"},{"name":"Fathom","bestFor":"Best free cloud notetaker","runs":"Cloud","bot":"Bot","freeTier":"Generous free plan","blurb":"Fathom is the easiest way to get automated meeting recaps, with one of the most generous free tiers among cloud notetakers. Its bot joins your call, records it, and produces clean summaries, highlights, and action items with almost no setup.","pros":["Excellent free tier for individuals","Very low-friction summaries and highlights","Solid Zoom, Teams, and Meet coverage"],"caveat":"It is cloud-based: a bot joins the call and your meeting is uploaded to Fathom's servers.","href":"/compare/fathom"},{"name":"Otter.ai","bestFor":"Live collaboration","runs":"Cloud","bot":"Bot","freeTier":"300 min/mo (historically)","blurb":"Otter is a mature meeting-transcription platform with strong live transcription, sharing, and integrations. It's a good fit for teams that want a shared, searchable record of calls and collaborative highlights.","pros":["Real-time live transcription","Team sharing and workspace features","Broad integrations"],"caveat":"Cloud-based with a bot, and the free tier is capped — heavy users hit limits quickly.","href":"/compare/otter"},{"name":"Fireflies.ai","bestFor":"Sales & CRM","runs":"Cloud","bot":"Bot","freeTier":"Free tier with limits","blurb":"Fireflies leans into revenue teams: its bot joins calls, transcribes them, and layers on conversation intelligence with deep Salesforce and HubSpot sync. If your notes need to feed a sales pipeline, it's a strong pick.","pros":["Conversation intelligence and analytics","Deep CRM integrations","Searchable team transcript library"],"caveat":"Cloud-based with a bot; the most useful features sit on paid, per-seat plans.","href":"/compare/fireflies"},{"name":"Granola","bestFor":"Bot-free cloud notepad","runs":"Cloud","bot":"No bot","freeTier":"Free tier","blurb":"Granola takes a nicer approach than most: it captures your computer's audio without a bot in the participant list and pairs it with a clean notepad and calendar sync. It's a favorite among Mac users who want tidy AI meeting notes.","pros":["No bot joins your call","Clean, fast note-taking UX","Calendar sync"],"caveat":"Processing still happens in the cloud, so your meeting audio leaves your device even though no bot appears.","href":"/compare/granola"},{"name":"Notta","bestFor":"Many languages","runs":"Cloud","bot":"Bot","freeTier":"Free tier with limits","blurb":"Notta is a cloud transcription and meeting tool with especially broad language support, plus file uploads and recording. A good option when your meetings span multiple languages.","pros":["Wide language coverage","Transcribe live calls or uploaded files","Simple exports"],"caveat":"Cloud-based; longer recordings and higher accuracy sit behind paid tiers.","href":null},{"name":"tl;dv","bestFor":"Recording & clips","runs":"Cloud","bot":"Bot","freeTier":"Free recording tier","blurb":"tl;dv is built around recording and sharing meeting moments — timestamped clips, highlights, and handoffs — with a usable free tier. Handy when you share meeting snippets with a team.","pros":["Clip and share key moments","Free recording tier","Zoom, Meet, and Teams support"],"caveat":"Cloud-based with a bot; your recordings live on tl;dv's servers.","href":null}]},"faq":{"title":"Frequently Asked Questions"},"cta":{"title":"Want AI notes that never leave your device?","body":"OpenWhispr is the free, open-source AI note taker that records and transcribes on your own machine — no bot, no cloud upload. Free on macOS, Windows, and Linux.","github":"View on GitHub"}},"blogZoomRecording":{"metadata":{"title":"How to Record a Zoom Meeting (Host or Not) | OpenWhispr","description":"Learn how to record a Zoom meeting as a host or participant — locally, in the cloud, or with no bot. Get a private transcript on your own device.","keywords":["how to record a zoom meeting","record zoom meeting","how to record a zoom meeting as a participant","record zoom meeting without permission","how to record a zoom meeting with audio","where do zoom recordings go","record zoom meeting locally","how to get a transcript of a zoom meeting","zoom meeting transcript","is it legal to record a zoom meeting","record zoom meeting on mac","record zoom meeting on windows"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"How to Record a Zoom Meeting"},"schema":{"title":"How to Record a Zoom Meeting (With or Without Being the Host)","description":"Learn how to record a Zoom meeting as a host or participant — locally, in the cloud, or with no bot. Get a private transcript on your own device."},"faqs":[{"question":"Can you record a Zoom meeting if you are not the host?","answer":"Yes, in two ways. The host can grant you recording permission during the call (Participants → hover your name → More → Allow to Record), which unlocks Zoom's built-in recorder for you. If they won't or can't, you can record your own screen and system audio with a separate recorder on your computer — that captures everything you can see and hear, no host permission required. Either way, tell the other participants you are recording."},{"question":"Where do Zoom recordings get saved?","answer":"Local recordings save to your own computer — by default in a Zoom folder under your Documents, as an MP4 video plus a separate M4A audio file. Cloud recordings (paid plans only) save to Zoom's servers, where you get a shareable link and, on eligible plans, an automatic transcript. If privacy matters, local recording keeps the file on your device."},{"question":"Is it legal to record a Zoom meeting?","answer":"It depends on where the participants are. Some places allow recording with just one person's consent, while others require everyone on the call to agree, and rules like the EU's GDPR treat a recorded voice as personal data. This is not legal advice, but the safe habit everywhere is the same: tell people you are recording and get their okay before you start."},{"question":"How do I record a Zoom meeting with the audio?","answer":"Zoom's built-in recorder captures the meeting audio automatically alongside the video. If you record with a separate screen recorder instead, make sure system audio (not just your microphone) is enabled, or you will end up with silent video. A local voice tool like OpenWhispr captures the system audio directly, so you always get the full conversation."},{"question":"Can you record a Zoom meeting without the host knowing?","answer":"When you use Zoom's built-in recorder, Zoom shows everyone a 'Recording' banner, so the host and participants are notified. A separate screen or audio recorder won't trigger that banner — but recording people secretly can be illegal depending on your location and is a trust risk regardless. Always disclose that you're recording; the point of recording a meeting is to have an accurate shared record, not to hide one."},{"question":"How do I get a transcript of a Zoom meeting?","answer":"On paid plans, Zoom can generate an audio transcript for cloud recordings automatically. For a local recording — or any Zoom plan — you can transcribe the saved audio file yourself. OpenWhispr turns your recording into an accurate, searchable transcript on your own device in 100+ languages, then lets AI summarize it into notes and action items."},{"question":"Can I record a Zoom meeting on my phone?","answer":"If you are the host on a paid plan, you can record to the cloud from the Zoom mobile app. As a participant, use your phone's built-in screen recorder (Control Center on iPhone, Screen Record on Android) and enable microphone/internal audio. Then transcribe the file afterward if you want notes."},{"question":"What is the most private way to record a Zoom meeting?","answer":"Record locally so the file never leaves your device, and avoid cloud note-taker bots that upload your meeting to a third party. OpenWhispr records the system audio and transcribes it entirely on your own machine — no bot joins the call, and nothing is uploaded — which is the safest option for sensitive meetings like HR, legal, or anything with customer data."}],"sections":[{"id":"host","title":"Record as the Host"},{"id":"participant","title":"If You're Not the Host"},{"id":"local","title":"Record Locally, No Bot"},{"id":"mobile","title":"Record on Mobile"},{"id":"transcript","title":"Get a Transcript"},{"id":"legal","title":"Consent & Legality"},{"id":"storage","title":"Where It's Stored"},{"id":"troubleshoot","title":"Troubleshooting"},{"id":"faq","title":"FAQ"}],"hero":{"title":"How to Record a Zoom Meeting (With or Without Being the Host)","subtitle":"Record a Zoom call with audio — as the host, as a participant, in the cloud or locally on your own device — then turn it into a searchable transcript. Plus how to stay on the right side of consent.","date":"July 6, 2026","tag":"Guide"},"opening":{"lead":"To record a Zoom meeting, click the Record button in the toolbar if you are the host or have been given permission.","body":"If you are a participant without permission, you have two options: ask the host to enable recording for you, or record your own screen and system audio with a separate tool on your computer — which needs no permission and captures the whole call. Whatever method you use, tell the other participants first.","scope":"This guide covers every path: recording as the host (local vs cloud), recording as a participant, recording privately on your own device without a bot, doing it on mobile, turning the recording into a transcript, the consent rules to know, and what to do when the Record button is greyed out.","updated":"Last updated: July 6, 2026."},"host":{"title":"Record as the Host (Zoom's Built-in Recorder)","intro":"If you started the meeting, recording takes one click. In the meeting toolbar, press Record. On a paid plan you'll be asked to choose between two destinations:","options":[{"title":"Record on this computer","body":"saves an MP4 video and an M4A audio file to your own machine when the meeting ends. Available on every plan, including free."},{"title":"Record to the cloud","body":"uploads the recording to Zoom's servers and gives you a shareable link plus, on eligible plans, an automatic transcript. Paid plans only."}],"outroPrefix":"Zoom shows every participant a “Recording” indicator the moment you start, so the call is always disclosed. When the meeting ends, a local recording is converted and dropped into a","outroSuffix":"folder inside your Documents by default."},"participant":{"title":"If You Are Not the Host","intro":"Participants can't use Zoom's recorder unless the host allows it. You have two routes:","options":[{"title":"Ask for permission.","body":"The host opens the Participants panel, hovers your name, clicks More, and chooses Allow to Record. Zoom's Record button then appears for you."},{"title":"Record your screen and system audio.","body":"A separate recorder on your computer captures whatever is on screen and coming through your speakers — no host permission required. This is the reliable fallback when the host can't or won't enable recording."}],"warningPrefix":"The catch with option two: a plain screen recorder often captures only your microphone, leaving the other voices out. Make sure you enable","systemAudio":"system audio","warningSuffix":", or use a tool built to record the meeting audio directly — covered next."},"local":{"title":"Record Locally, With No Bot and No Cloud","cloudPrefix":"Most cloud note-takers — Otter, Fireflies, Fathom — join your call with a","bot":"bot","cloudSuffix":"and upload the audio to their servers, which puts a stranger's bot in the participant list and your conversation on someone else's cloud. Even bot-free options still send your meeting off-device to process it — a non-starter for sensitive meetings.","approachPrefix":"takes the opposite approach. It captures the meeting's system audio directly on your machine —","noBot":"no bot joins the call","approachSuffix":", and the audio never leaves your device. It works as a participant, on any Zoom plan, on macOS, Windows and Linux. You still disclose that you're recording, but nothing is uploaded and no third party ever sees the conversation.","bullets":["Records the full call audio, not just your microphone.","No bot in the participant list, and no cloud upload.","Transcribes and summarizes on-device, in 100+ languages."],"github":"View on GitHub"},"mobile":{"title":"Record a Zoom Meeting on Your Phone","paragraph1":"If you host on a paid plan, the Zoom mobile app can record straight to the cloud — tap More then Record to the Cloud. As a participant, the simplest route is your phone's own screen recorder: the iPhone Control Center recorder or Android's Screen Record. Turn on microphone/internal audio so the call is captured, not just muted video.","paragraph2":"Phone recordings are just video files, so you can transcribe them the same way as any other recording once the call ends."},"transcript":{"title":"Turn the Recording Into a Transcript and Notes","paragraph1":"A recording is only useful if you can read and search it. On paid plans, Zoom can auto-generate a transcript for cloud recordings. For a local recording — or on the free plan — you transcribe the saved file yourself.","paragraph2Prefix":"Drop the MP4 or M4A into","paragraph2Suffix":"and it produces an accurate, timestamped transcript on your own device, then lets AI summarize the call into decisions, action items and follow-ups. Because it runs locally, even a confidential recording never gets uploaded to transcribe it."},"legal":{"title":"Consent and Legality","paragraph1":"Whether you can legally record depends on where the participants are. Some places allow recording with just one person's consent; others require everyone on the call to agree, and rules like the EU's GDPR treat a recorded voice as personal data. This isn't legal advice, but the safe habit everywhere is the same: say you're recording and get a clear okay before you start.","paragraph2":"Zoom's built-in recorder discloses automatically with its on-screen banner. A separate screen or local recorder won't show that banner, so the responsibility to disclose is on you. Recording people secretly is a legal and trust risk — and unnecessary, since the whole point is a shared record everyone can rely on."},"storage":{"title":"Where Your Recording Ends Up","localLabel":"Local recordings","localBody":"stay on your computer — an MP4 plus an M4A audio file, in a Zoom folder under Documents by default. Nothing leaves your machine unless you share it.","cloudLabel":"Cloud recordings","cloudBody":"live on Zoom's servers with a shareable link, subject to your plan's storage limits and your organization's retention settings. And","botLabel":"bot-based note-takers","botBody":"upload the whole meeting to a third party, where some providers may use it to train their models. For HR, legal, medical, or anything with customer data, recording locally and transcribing on-device keeps the conversation entirely under your control."},"troubleshoot":{"title":"Troubleshooting: Record Button Missing or Greyed Out","items":[{"title":"You're not the host and lack permission.","body":"Ask the host to allow recording, or fall back to a local screen/audio recorder."},{"title":"Recording is disabled at the account level.","body":"On managed or company accounts, an admin may have turned recording off in the Zoom web settings — you'll need them to enable it."},{"title":"Cloud recording is missing.","body":"Cloud recording requires a paid plan; the free plan supports local recording only."},{"title":"The recording has no sound.","body":"Your screen recorder is capturing only the mic. Enable system audio, or record with a tool built to capture the call audio."}]},"faq":{"title":"Frequently Asked Questions"},"closing":{"title":"Record and transcribe your meetings privately","body":"OpenWhispr captures your Zoom call audio and turns it into notes on your own device — no bot in the call, nothing uploaded. Free on macOS, Windows and Linux.","github":"View on GitHub"}},"blogStreamingStt":{"metadata":{"title":"Streaming Speech-to-Text on CPU: Local & Live | OpenWhispr","description":"How OpenWhispr runs NVIDIA Nemotron streaming ASR on your CPU — cache-aware streaming over a local WebSocket, live partial results, no GPU, no cloud.","keywords":["streaming speech to text","local streaming asr","nemotron asr","cache aware streaming asr","sherpa onnx","real time transcription local","live dictation preview","streaming asr electron","on device speech recognition"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"Streaming Speech-to-Text on Your CPU"},"schema":{"title":"Streaming Speech-to-Text on Your CPU: Shipping NVIDIA Nemotron in a Desktop App","description":"How OpenWhispr runs NVIDIA Nemotron streaming ASR on CPU — cache-aware streaming over a local WebSocket, live partials, no GPU, no cloud."},"faqs":[{"question":"What is streaming speech-to-text?","answer":"Streaming (or 'online') speech-to-text transcribes audio while it's still being spoken, emitting partial results within a fixed latency budget — typically 80 milliseconds to about a second. Batch ('offline') models like Whisper instead wait for the recording to finish and process it all at once. Streaming is what makes live captions, voice agents, and see-it-as-you-speak dictation previews possible."},{"question":"Can I run streaming speech-to-text locally without a GPU?","answer":"Yes. NVIDIA's Nemotron streaming models are 600M parameters, ship as INT8-quantized ONNX files around 650MB, and run in real time on a modern CPU via the sherpa-onnx runtime — a native binary with no Python or PyTorch. OpenWhispr ships this setup on macOS, Windows, and Linux."},{"question":"What is cache-aware streaming ASR?","answer":"It's an architecture (NVIDIA's cache-aware FastConformer) where the model keeps its internal encoder state between audio chunks. Each new chunk is processed once, using cached context from everything said so far, instead of re-encoding overlapping windows of audio. That's what makes low-latency streaming cheap enough to run on a laptop CPU."},{"question":"Why does the live preview text change after it appears?","answer":"Streaming models revise their hypothesis as more audio arrives — a word boundary or a homophone can only be resolved by later context, so partial results are replaced in place. This is by design. When you stop, OpenWhispr flushes the model's tail and commits the settled streamed text as the transcript; if the flush is interrupted or truncated, it automatically re-decodes the full recording instead, so an unstable stream never becomes your pasted text."},{"question":"Is streaming transcription less accurate than batch?","answer":"Slightly, and honestly so: a streaming model can't see future audio. Nemotron's English streaming model averages 6.93% word error rate at 1.12-second chunks on the Open ASR Leaderboard datasets — remarkably close to batch models, but still behind the 5.91% of NVIDIA's best English batch model. Choosing a streaming model means accepting that small gap in exchange for live text and an instant transcript at stop; if maximum accuracy matters more, pick an offline Parakeet model instead."},{"question":"Which languages does streaming support in OpenWhispr?","answer":"Two Nemotron models are available: English-only, and Nemotron 3.5 with 15 transcription-ready languages (English, Spanish, French, Italian, Portuguese, Dutch, German, Turkish, Russian, Arabic, Hindi, Japanese, Korean, Vietnamese, Ukrainian) with automatic language detection. Offline models like Parakeet and Whisper cover more languages but preview via buffered chunks instead of a live stream."}],"sections":[{"id":"the-problem","title":"The Problem: Chunked Previews"},{"id":"cache-aware","title":"What Cache-Aware Streaming Is"},{"id":"architecture","title":"The Architecture"},{"id":"gotchas","title":"What Broke Along the Way"},{"id":"single-pass","title":"Committing the Stream"},{"id":"footprint","title":"Performance & Footprint"},{"id":"limitations","title":"Honest Limitations"},{"id":"model-choice","title":"Choosing a Model"},{"id":"faq","title":"FAQ"}],"hero":{"title":"Streaming Speech-to-Text on Your CPU: Shipping NVIDIA Nemotron in a Desktop App","subtitle":"How OpenWhispr shows live text as you speak — a cache-aware streaming model over a local WebSocket, no GPU and no cloud — and the engineering it took to make that reliable.","date":"July 18, 2026","tag":"Technical"},"opening":{"lead":"Streaming speech-to-text transcribes audio while you're still talking, instead of waiting for the recording to end. As of version 1.7.6, OpenWhispr does this entirely on-device: NVIDIA's Nemotron streaming models (600M parameters, INT8-quantized) run on your CPU through a local sherpa-onnx WebSocket server, the dictation preview updates live as partial results arrive, and the streamed text is committed as your transcript the moment you stop.","body":"No GPU, no Python environment, no audio leaving your machine.","guidePrefix":"This is the engineering write-up: why our old live preview felt laggy, what cache-aware streaming actually is in plain terms, how the pipeline fits together inside an Electron app, the version-pinning bug that mis-decoded everything, and how we went from re-decoding every recording to committing the stream itself. If you want the model-versus-model comparison instead, that's in","comparisonLink":"Parakeet vs Whisper vs Nemotron","guideSuffix":".","updated":"Last updated July 18, 2026. Implementation details reference the streaming transcription feature shipped in OpenWhispr 1.7.6; everything described here is inspectable in the open-source repository."},"factCheck":{"title":"Fact-Check Snapshot (Primary Sources)","items":[{"lead":"Nemotron's cache-aware FastConformer processes only new audio while reusing cached encoder context,","body":"with runtime-configurable chunk sizes from 80ms to 1.12s.","sources":[{"href":"https://huggingface.co/nvidia/nemotron-speech-streaming-en-0.6b","label":"Nemotron Streaming EN model card"},{"href":"https://huggingface.co/blog/nvidia/nemotron-speech-asr-scaling-voice-agents","label":"NVIDIA on cache-aware streaming"}]},{"lead":"The English model averages 6.93% WER at 1.12s chunks while streaming;","body":"the multilingual 3.5 model covers 40 language-locales with automatic detection.","sources":[{"href":"https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b","label":"Nemotron 3.5 ASR model card"}]},{"lead":"All inference runs through sherpa-onnx, a native ONNX runtime with no Python dependency,","body":"using its online (streaming) WebSocket server. We pin and bundle version 1.13.4.","sources":[{"href":"https://github.com/k2-fsa/sherpa-onnx","label":"sherpa-onnx on GitHub"}]},{"lead":"The models are 632MB (English) and 650MB (multilingual) on disk as INT8 ONNX,","body":"downloaded once and cached locally.","sources":[{"href":"https://github.com/OpenWhispr/openwhispr","label":"OpenWhispr model registry (open source)"},{"href":"https://openwhispr.com/models","label":"OpenWhispr models page"}]}]},"problem":{"title":"The Problem: Live Preview Built on a Batch Model","paragraphs":[{"body":"OpenWhispr has always had a live transcription preview — a small window that shows what the model heard while you dictate. Before 1.7.6 it worked the only way you can make a batch model feel live: buffer about 1.5 seconds of microphone audio, run a full transcription on that chunk, append the text, repeat."},{"body":"That approach has three problems, and every one of them is structural. First, latency: your words can't appear faster than the buffer fills, so the preview always trails you by a second and a half plus inference time. Second, boundaries: a word that straddles two chunks gets cut in half, and neither chunk can fix it, because chunks are transcribed independently. Third, context: chunk five doesn't know what you said in chunk four — so “their” versus “there” is a coin flip the model re-tosses every 1.5 seconds."},{"body":"You can shrink the chunks to feel faster, but then accuracy collapses — less context per chunk, more boundaries. You can overlap chunks to fix boundaries, but then you're transcribing the same audio twice. What you actually want is a model that was trained to transcribe a stream. That's what Nemotron is."}]},"cacheAware":{"title":"What Cache-Aware Streaming Actually Means","paragraphs":[{"lead":"A cache-aware streaming model keeps its internal state between chunks.","body":"When the next slice of audio arrives, the encoder doesn't re-read everything you said — it processes only the new frames, attending to a cache of the context it already computed. It's the same trick that makes chatbots fast (KV caching), applied to a speech encoder."},{"body":"Two consequences fall out of this design. Compute stops scaling with recording length — each chunk costs the same whether you're ten seconds or ten minutes in, which is what makes CPU-only streaming viable. And context carries forward — the model resolves “their/there” using everything said so far, and revises earlier words in its running hypothesis when later audio disambiguates them. That revision is why streaming text visibly “settles” a beat after you speak: it's not a bug, it's the model changing its mind with better evidence."},{"body":"The latency budget is a knob, not a constant: Nemotron exposes chunk sizes from 80ms (fastest, least context per step) up to 1.12s (highest accuracy — 6.93% average WER on the leaderboard datasets, within about a point of the best batch models). The training recipe is NVIDIA's cache-aware FastConformer with a transducer decoder; their"}],"engineeringPostLink":"engineering post","engineeringPostSuffix":"covers the theory in depth."},"architecture":{"title":"The Architecture Inside OpenWhispr","paragraphs":[{"body":"The pipeline is deliberately boring. An AudioWorklet captures microphone PCM at 16kHz. Frames are converted to the float32 wire format and pushed over a WebSocket to a sherpa-onnx online server — a native binary OpenWhispr spawns locally and talks to over 127.0.0.1, one persistent connection per recording. The server runs the Nemotron model with INT8 weights and answers with JSON partial results, which the preview window renders by replacing its text in place rather than appending."},{"body":"Why a separate server process instead of in-process inference? Isolation. Native inference runtimes can crash — a bad allocation in an ONNX kernel shouldn't take the whole desktop app down with it. A sidecar process means a crashed runtime is a reconnect, not a lost dictation. It also keeps native code out of Electron's main process entirely: the app talks protocol, the sidecar does math."},{"body":"The same pattern runs our"}],"diarizationLink":"local speaker diarization","diarizationSuffix":"— one sherpa-onnx toolchain, multiple jobs, all on-device."},"gotchas":{"title":"What Broke Along the Way","items":[{"label":"The runtime version mattered more than we expected.","body":"Nemotron models require sherpa-onnx 1.13.4 or newer for accurate decoding — on the older runtime we'd bundled for Parakeet, they didn't fail loudly, they transcribed badly. The fix was mundane (upgrade and re-pin the bundled binaries on every platform), but the lesson generalizes: with model runtimes, “it runs” and “it's correct” are separate tests, and only the second one counts."},{"label":"Streams fail, so the old path stayed.","body":"A port can be taken, a model file can be missing, a server can die mid-recording. If the persistent stream can't start, the preview falls back to the buffered-chunk path automatically — slower, but never blank. Offline models (Parakeet, Whisper) keep the chunked preview by design; each model in our registry declares its runtime as online or offline, and the app picks the right preview path per model."},{"label":"Partials needed different rendering.","body":"The chunked path appends text; a streaming hypothesis revises itself, so the UI replaces the whole preview per partial. Append a revising hypothesis and you get stuttering duplicates — the one-line version of a real UX bug we fixed early."}]},"singlePass":{"title":"Committing the Stream: From Two Passes to One","paragraphs":[{"body":"Our first version treated streaming as a preview, not the product: when you stopped dictating, the app re-decoded the complete recording from scratch and pasted that. Safe — the final pass sees full context — but it meant every dictation paid for two decodes, and the text you'd already watched settle on screen made you wait for a second opinion that almost always agreed with it."},{"body":"So in the same release we made the stream the source of truth. The dictation now runs over the persistent stream whether or not the preview window is open, and when you stop, the app flushes the model's tail and commits the streamed text directly — no second decode. End-of-dictation latency drops to one tail flush, and per-dictation CPU is roughly halved."},{"body":"The safety net didn't go away — it became the fallback. The stop flush is truncation-aware: it extends its deadline while results are still arriving, and if anything is off — a dropped connection, a truncated result — the app discards the stream and runs the classic record-then-transcribe pass over the full recording instead. You only ever get the fast path when it's also the correct one."}]},"footprint":{"title":"Performance & Footprint","items":[{"label":"Models:","body":"Nemotron Speech Streaming EN 0.6B (632MB) and Nemotron 3.5 ASR Streaming 0.6B (650MB, 15 transcription-ready languages, auto-detect) — INT8 ONNX, downloaded once, cached locally."},{"label":"Runtime:","body":"sherpa-onnx 1.13.4, a self-contained native binary. No Python, no PyTorch, no CUDA required."},{"label":"Hardware:","body":"real-time streaming on a modern laptop CPU — Apple Silicon and recent x86 both comfortably keep up with speech."},{"label":"Feel:","body":"partial text lands well under a second behind your voice; at stop, one tail flush commits the transcript — no second decode."}],"privacyPrefix":"Nothing about this path touches the internet. The app talks to the server over 127.0.0.1 (on Windows, a scoped firewall rule additionally closes the port to the network), the model files live in your local cache, and the audio never leaves the machine — the same privacy posture as the rest of","privacyLink":"our local transcription stack","privacySuffix":"."},"limitations":{"title":"Honest Limitations","items":[{"label":"Streaming trades a little accuracy for immediacy.","body":"~6.9% streaming versus 5.91% for NVIDIA's best English batch model on the same benchmarks — and since the committed transcript now comes from the stream, that's the trade you're accepting. If you only care about the end result, an offline Parakeet model is still the accuracy pick."},{"label":"Language coverage is narrower than Whisper.","body":"15 transcription-ready languages in the multilingual model versus Whisper's 99. Streaming in Mandarin or Thai isn't here yet."},{"label":"Partial text moves.","body":"The hypothesis revises itself as context arrives. We think watching it settle is a feature; if you find it distracting, the preview can be turned off and dictation works exactly as before."},{"label":"It's another 650MB on disk","body":"if you already have an offline model — the price of a second engine tuned for a different job."}]},"modelChoice":{"title":"Choosing a Model in OpenWhispr","wantHeader":"You want","pickHeader":"Pick","rows":[{"want":"Live English preview while dictating","pick":"Nemotron Speech Streaming EN 0.6B"},{"want":"Live preview in Spanish, Japanese, Arabic, Hindi…","pick":"Nemotron 3.5 ASR Streaming 0.6B"},{"want":"Maximum English accuracy, don't care about live text","pick":"Parakeet Unified EN 0.6B"},{"want":"A language outside the NVIDIA set","pick":"Whisper (turbo or large)"}],"catalogPrefix":"All of these are a dropdown in Settings — the full catalog with sizes and trade-offs is on the","catalogLink":"models page","catalogSuffix":". OpenWhispr is free, open source, and available for macOS, Windows, and Linux.","githubCta":"Star on GitHub"},"faq":{"title":"Frequently Asked Questions"}},"blogParakeetNemotron":{"metadata":{"title":"Parakeet vs Whisper vs Nemotron: Best Local STT 2026 | OpenWhispr","description":"Honest comparison of the top local speech-to-text models: accuracy (WER), speed, languages, and streaming. Real numbers from primary sources.","keywords":["parakeet vs whisper","nemotron asr","best local speech to text","whisper alternative","nvidia parakeet","local speech to text models","open source speech to text","best speech to text model 2026","nemotron vs whisper","offline speech recognition"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"Parakeet vs Whisper vs Nemotron"},"schema":{"title":"Parakeet vs Whisper vs Nemotron: The Best Local Speech-to-Text in 2026","description":"Honest comparison of the top local speech-to-text models: accuracy (WER), speed, languages, and streaming — with numbers from primary sources."},"faqs":[{"question":"What is the best local speech-to-text model in 2026?","answer":"For English and major European languages, NVIDIA Parakeet TDT 0.6B v3 offers the best overall balance: it averages 6.34% word error rate on the Open ASR Leaderboard benchmarks — ahead of Whisper large-v3's roughly 7.4% — at a quarter of the size and with dramatically higher throughput on CPU. If you need live text while you speak, NVIDIA Nemotron's streaming models are the best choice. If you need a language the NVIDIA models don't cover, use Whisper."},{"question":"Is Parakeet more accurate than Whisper?","answer":"On the Open ASR Leaderboard's English benchmarks, yes — Parakeet TDT 0.6B v3 averages 6.34% WER versus about 7.4% for Whisper large-v3, and the English-only Parakeet Unified model reaches 5.91%. The honest caveat: a ~1-point WER difference is hard to notice in everyday dictation. Parakeet's clearer wins are speed, efficiency, and the fact that it doesn't hallucinate text during silence, a known Whisper failure mode."},{"question":"Do I need a GPU to run these models locally?","answer":"No. All three run on a modern CPU. Parakeet and Nemotron use INT8-quantized ONNX models (roughly 630–680MB on disk) designed for CPU inference. Whisper runs on CPU via whisper.cpp, though its large 3GB model benefits from GPU acceleration — OpenWhispr supports Metal on Apple Silicon, CUDA on NVIDIA, and Vulkan on AMD/Intel GPUs, with automatic CPU fallback."},{"question":"What languages does Nemotron ASR support?","answer":"NVIDIA's Nemotron 3.5 ASR model covers 40 language-locales in a single 600M-parameter model with automatic language detection. In OpenWhispr, the 15 transcription-ready languages are enabled: English, Spanish, French, Italian, Portuguese, Dutch, German, Turkish, Russian, Arabic, Hindi, Japanese, Korean, Vietnamese, and Ukrainian. The English-only Nemotron streaming model is also available."},{"question":"Is Whisper obsolete now?","answer":"No. Whisper still covers 99 languages — far more than Parakeet's 25 or Nemotron's 40 locales — and it remains the safest choice for Asian languages outside Japanese, Korean, and Vietnamese, for low-resource languages, and for heavily mixed-language audio. It's also the most battle-tested model with the largest ecosystem. What has changed is that Whisper no longer holds the accuracy or speed crown for English."},{"question":"Which of these models does OpenWhispr support?","answer":"All of them. OpenWhispr ships six Whisper sizes via whisper.cpp, both Parakeet models and both Nemotron streaming models via sherpa-onnx, and lets you switch between them in Settings. Everything runs on your device — audio never leaves your machine — and the app is free and open source."}],"sections":[{"id":"contenders","title":"The Three Contenders"},{"id":"accuracy","title":"Accuracy"},{"id":"speed","title":"Speed & Efficiency"},{"id":"languages","title":"Language Coverage"},{"id":"streaming","title":"Streaming: The Real Differentiator"},{"id":"hardware","title":"Hardware Requirements"},{"id":"which-to-choose","title":"Which One Should You Use?"},{"id":"try-them","title":"Try All Three"},{"id":"faq","title":"FAQ"}],"hero":{"title":"Parakeet vs Whisper vs Nemotron: The Best Local Speech-to-Text in 2026","subtitle":"Three open models can now turn speech into text entirely on your own computer. Here's how they actually compare on accuracy, speed, languages, and streaming — with honest numbers from primary sources.","date":"July 18, 2026","tag":"Comparison"},"opening":{"lead":"For English and major European languages, NVIDIA Parakeet TDT 0.6B v3 is the best local speech-to-text model in 2026: it beats Whisper large-v3 on measured accuracy at a quarter of the size, and runs dramatically faster on a plain CPU. If you want text that appears live while you speak, NVIDIA's new Nemotron streaming models are the first local models built for exactly that. And if you speak one of the many languages the NVIDIA models don't cover, OpenAI's Whisper — with 99 languages — is still the one to use.","paragraph2":"The surprise of 2026 is that Whisper no longer holds the open-model accuracy crown it wore for three years. But “which model is best” has become a genuinely three-way question, because the models are good at different things: Parakeet optimizes accuracy-per-watt, Nemotron optimizes latency, and Whisper optimizes breadth. This article compares them with numbers from the models' own cards and the public leaderboard — no vibes, and no pretending the differences are bigger than they are.","disclosurePrefix":"Last updated July 18, 2026. We build","disclosureSuffix":", an open-source dictation app that ships all three model families, so we benchmark and support each of them in production. That's also our bias — we don't sell any one of these models, and the honest trade-offs below are the same ones we explain to our own users."},"factCheck":{"title":"Fact-Check Snapshot (Primary Sources)","items":[{"lead":"Parakeet TDT 0.6B v3 averages 6.34% word error rate across the eight Open ASR Leaderboard benchmarks, with a throughput of 3,332× real time.","body":"600M parameters, 25 languages, CC-BY-4.0 license.","sources":[{"href":"https://huggingface.co/nvidia/parakeet-tdt-0.6b-v3","label":"Parakeet TDT v3 model card"},{"href":"https://huggingface.co/spaces/hf-audio/open_asr_leaderboard","label":"Open ASR Leaderboard"}]},{"lead":"Whisper large-v3 averages roughly 7.4% on the same leaderboard and supports 99 languages","body":"with 1.55B parameters under an MIT license.","sources":[{"href":"https://github.com/openai/whisper","label":"OpenAI Whisper repository"},{"href":"https://huggingface.co/openai/whisper-large-v3","label":"Whisper large-v3 model card"}]},{"lead":"Nemotron Speech Streaming EN 0.6B averages 6.93% WER while streaming","body":"(1.12-second chunks), reaches 2.32% on LibriSpeech test-clean, and was trained on 530,000 hours of audio.","sources":[{"href":"https://huggingface.co/nvidia/nemotron-speech-streaming-en-0.6b","label":"Nemotron Streaming EN model card"}]},{"lead":"Nemotron 3.5 ASR covers 40 language-locales in one 600M streaming model","body":"with automatic language detection, released June 4, 2026 under the OpenMDW license.","sources":[{"href":"https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b","label":"Nemotron 3.5 ASR model card"}]},{"lead":"All disk sizes and in-app behavior referenced below are verifiable","body":"in OpenWhispr's open-source model registry.","sources":[{"href":"https://github.com/OpenWhispr/openwhispr","label":"OpenWhispr on GitHub"},{"href":"https://openwhispr.com/models","label":"OpenWhispr models page"}]}]},"contenders":{"title":"The Three Contenders","intro":"All three are free, open-weight models you can download and run offline. They differ in architecture, and the architecture explains almost every practical difference you'll notice.","items":[{"title":"OpenAI Whisper (2022)","body":"The model that made open speech recognition mainstream. An encoder-decoder transformer trained on 680,000 hours of audio, available in six sizes from 75MB (tiny) to 3GB (large-v3), covering 99 languages. It generates text one token at a time, like a language model — which is also why it's the slowest of the three. MIT licensed."},{"title":"NVIDIA Parakeet (2024–2025)","body":"A 600M-parameter transducer (TDT) model built for batch transcription. Instead of generating tokens autoregressively, it emits text in a single pass over the audio — much faster, and it doesn't invent text during silence. Two variants matter: the multilingual v3 (25 languages, 680MB as INT8 ONNX) and the English-only Unified model (631MB), which currently posts state-of-the-art English accuracy. CC-BY-4.0 licensed."},{"title":"NVIDIA Nemotron ASR (2026)","body":"The newest family, built for streaming: a cache-aware FastConformer transducer that transcribes audio as it arrives instead of waiting for the recording to end. The English model shipped in January 2026; the 40-locale multilingual 3.5 model followed in June. Both are 600M parameters (632–650MB as INT8 ONNX) and openly licensed (OpenMDW)."}]},"accuracy":{"title":"Accuracy: Parakeet Wins, But the Gap Is Smaller Than It Looks","definitionPrefix":"Speech models are measured by","definitionTerm":"word error rate (WER)","definitionSuffix":" — the percentage of words the model gets wrong. Lower is better. The standard reference is the Hugging Face Open ASR Leaderboard, which averages WER across eight diverse English datasets (meetings, earnings calls, TED talks, audiobooks, and more), so a single easy dataset can't flatter a model.","analysis":"Read the chart honestly: every model here is production-grade, and the whole spread is about 1.5 points. In day-to-day dictation, a 6.3% and a 7.4% model feel similar on clean audio — the differences show up on hard audio (accents, noise, jargon). What makes the NVIDIA numbers notable is what you pay for them: Parakeet reaches its accuracy at 600M parameters versus Whisper's 1.55B, and Nemotron posts its 6.93% while streaming, without seeing the future audio at all.","calloutTitle":"Whisper's known weakness: hallucination on silence","calloutBody":"Because Whisper's decoder works like a language model, it can generate fluent, entirely invented sentences during silence or noise — a failure mode documented across the ecosystem. Transducer models like Parakeet and Nemotron are structurally resistant to this: no audio evidence, no tokens. For dictation, where recordings start and end with silence, that matters more than a benchmark point."},"speed":{"title":"Speed & Efficiency: Not Even Close","paragraph1":"On the leaderboard's hardware, Parakeet TDT v3 transcribes at 3,332× real time — an hour of audio in about a second. Whisper large-v3 posts well under a tenth of that. The reason is architectural, and it's worth understanding in plain terms: Whisper writes its transcript one token at a time, each step waiting on the last, while a transducer reads the audio once and emits text as it goes. No loop, no per-word round trips.","paragraph2Prefix":"On your laptop the absolute numbers are smaller but the ratio holds: a dictated paragraph that takes Whisper large several seconds to process lands near-instantly with Parakeet — on CPU, with no GPU involved. Independent benchmarks also report the NVIDIA transducers using roughly","energyLink":"8–10× less energy per hour of audio","paragraph2Suffix":"than Whisper at comparable quality — relevant if you dictate all day on battery.","paragraph3Prefix":"Whisper's answer to this is turbo (1.6GB), a distilled large-v3 that trades a little accuracy for roughly 8× Whisper-large speed. It's the right Whisper for dictation — but it narrows the gap rather than closing it. For the full Whisper size-by-size breakdown, see our","guideLink":"Whisper model sizes guide","paragraph3Suffix":"."},"languages":{"title":"Language Coverage: Whisper's Moat","headers":{"model":"Model","languages":"Languages","notes":"Notes"},"rows":[{"model":"Whisper large-v3","languages":"99","notes":"Broadest coverage of any open model, including low-resource languages"},{"model":"Nemotron 3.5 ASR","languages":"40 locales","notes":"Auto language detection; 15 transcription-ready languages including Japanese, Korean, Arabic, and Hindi"},{"model":"Parakeet TDT v3","languages":"25","notes":"European languages plus English and Russian; auto detection"},{"model":"Parakeet Unified / Nemotron EN","languages":"1","notes":"English-only, in exchange for the best accuracy in their class"}],"rule":"The rule of thumb is simple. English or a major European language: any of the three, so pick on speed. Japanese, Korean, Vietnamese, Arabic, Hindi: Nemotron 3.5 now covers them with streaming — a real first for local models. Mandarin, Thai, Indonesian, Swahili, or heavy code-switching between languages: Whisper is still the only serious option, and that's unlikely to change soon."},"streaming":{"title":"Streaming: The Real Differentiator","lead":"Streaming is where Nemotron isn't just incrementally better — it's doing something the other two structurally can't.","paragraph1":"Whisper processes audio in 30-second windows and re-encodes everything it sees, so “live” Whisper is really the same batch model run over and over on overlapping audio — expensive, laggy, and jittery at the seams. Nemotron's cache-aware architecture was trained for the streaming case: it processes only the newest chunk of audio, carries its internal state forward, and emits partial text within a configurable latency budget of 80 milliseconds to 1.12 seconds.","paragraph2Prefix":"The efficiency difference is dramatic at scale — NVIDIA reports","sourceLink":"17× more concurrent streams","paragraph2Middle":"than its previous-generation streaming model on the same GPU — but the part you actually feel on a laptop is simpler: text appears as you speak, from a model running on your own CPU. We wrote a full engineering breakdown of what it took to ship that in a desktop app in","deepDiveLink":"our streaming ASR deep dive","paragraph2Suffix":".","paragraph3":"If you don't care about live text — you dictate, stop, and want the final transcript — streaming buys you nothing, and Parakeet's batch models are the better pick. Streaming exists for live captions, voice agents, and the see-it-as-you-say-it dictation preview."},"hardware":{"title":"Hardware: All Three Run on a Normal Laptop","paragraph1":"None of these models needs a GPU. The NVIDIA models ship as INT8-quantized ONNX files (631–680MB) that run comfortably on any recent CPU — Apple Silicon or x86 — via the sherpa-onnx runtime, a native binary with no Python and no PyTorch. Whisper runs everywhere through whisper.cpp; its small and turbo sizes are fine on CPU, while large-v3 (3GB, ~10GB RAM) genuinely benefits from acceleration.","paragraph2":"If you do have a GPU, use it for Whisper: Metal is built in on Apple Silicon, and OpenWhispr offers one-click CUDA (NVIDIA) and Vulkan (AMD and Intel) runtimes with automatic CPU fallback if the GPU runtime fails. The NVIDIA transducers are fast enough on CPU that GPU acceleration is beside the point for dictation."},"decision":{"title":"Which One Should You Use?","headers":{"situation":"Your situation","use":"Use this"},"rows":[{"situation":"English dictation, want the best accuracy","use":"Parakeet Unified EN 0.6B"},{"situation":"European languages, fast batch transcription","use":"Parakeet TDT 0.6B v3"},{"situation":"Live text as you speak (English)","use":"Nemotron Speech Streaming EN"},{"situation":"Live text in Spanish, Japanese, Korean, Arabic, Hindi…","use":"Nemotron 3.5 ASR Streaming"},{"situation":"A language the NVIDIA models don't cover","use":"Whisper large-v3 or turbo"},{"situation":"Old or low-RAM hardware, quick notes","use":"Whisper tiny or base"}],"privacyPrefix":"And if privacy is the reason you're going local in the first place, all three qualify equally: the audio is processed on your machine and never uploaded. We've written more about that trade-off in","privacyLink":"local vs cloud transcription","privacySuffix":"."},"cta":{"title":"Try All Three in Two Minutes","paragraphPrefix":"The most honest way to choose is to dictate the same paragraph into each model. OpenWhispr ships every model in this article — six Whisper sizes, both Parakeets, both Nemotrons — behind a single settings dropdown, free and open source, on macOS, Windows, and Linux. Download a model, press the hotkey, talk. Our","modelsLink":"models page","paragraphSuffix":"lists everything with sizes and trade-offs.","github":"Star on GitHub"},"faqSection":{"title":"Frequently Asked Questions"}},"blogHipaaNoteTakers":{"metadata":{"title":"HIPAA-Compliant AI Note Takers & Dictation (2026) | OpenWhispr","description":"Which AI note takers sign a BAA, and which quietly exclude the AI features from it. A plain comparison for 2026, plus where OpenWhispr stands on SOC 2 Type II and ISO 27001.","keywords":["hipaa compliant ai note taker","hipaa compliant note taking app","hipaa compliant ai scribe","hipaa compliant transcription","hipaa compliant dictation software","hipaa compliant speech to text","is otter ai hipaa compliant","hipaa compliant ai note taker for therapists","best ai meeting note taker soc 2 hipaa compliant","hipaa compliant meeting transcription"]},"breadcrumb":{"home":"Home","blog":"Blog","article":"HIPAA-Compliant AI Note Takers"},"schema":{"title":"HIPAA-Compliant AI Note Takers and Dictation Software in 2026","description":"Which AI note takers sign a Business Associate Agreement, which exclude their AI features from it, what to ask a vendor, and how local transcription changes the analysis."},"hero":{"title":"HIPAA-Compliant AI Note Takers: What a BAA Actually Covers","subtitle":"Most tools that say “HIPAA compliant” will sign a BAA, then leave the AI features out of it. Here's how to check before patient data reaches a vendor, and where we stand on SOC 2 Type II, ISO 27001, and HIPAA.","date":"August 11, 2026","tag":"Compliance"},"sections":[{"id":"where-we-stand","title":"Where we stand: SOC 2, ISO 27001, HIPAA"},{"id":"what-hipaa-requires","title":"What HIPAA actually requires"},{"id":"ai-exclusion","title":"The trap: when the BAA excludes the AI"},{"id":"comparison","title":"Which tools sign a BAA"},{"id":"checklist","title":"Five questions to ask a vendor"},{"id":"local-processing","title":"Why local processing changes the math"},{"id":"frameworks","title":"SOC 2 and ISO 27001 are not HIPAA"},{"id":"faq","title":"Frequently asked questions"}],"opening":{"lead":"There is no such thing as a HIPAA-certified AI note taker.","body":"No agency certifies software as HIPAA compliant. When a vendor's site says “HIPAA certified,” that line came from their marketing team, not an auditor. The question that actually matters is narrower and much easier to check: will they sign a Business Associate Agreement, and does it cover the feature you want to point at patient data?","thesis":"That second half is where teams get caught. Several large platforms will sign a BAA for their core product and leave the AI features out of it. Which are, of course, the features you came for.","updated":"Last updated August 2026. Vendor terms move quickly, so confirm current coverage with the vendor before you send them anything."},"whereWeStand":{"title":"Where we stand: SOC 2, ISO 27001, HIPAA","disclosure":"We build OpenWhispr, and it is one of the tools in the table below. That is a conflict of interest, so here is our position up front — read the rest of the page with it in mind.","body":"We are a small team. We went through SOC 2 Type II, ISO 27001, and a HIPAA readiness assessment earlier than a company our size usually would, because “trust us” is not an answer when someone asks where their patient audio goes. Concretely, this is what exists today:","items":[{"lead":"SOC 2 Type II and ISO 27001.","body":"Both assessed by an independent firm, with 66 controls monitored continuously and 13 subprocessors listed publicly. Current status is in our Trust Center."},{"lead":"HIPAA, with a BAA.","body":"Available on request from the Business plan up. We are not “HIPAA certified,” because no such certificate exists, and you should be skeptical of anyone who claims otherwise."},{"lead":"A third-party penetration test.","body":"Run by an external firm. We share the report with customers on request."},{"lead":"A way to report problems.","body":"security@openwhispr.com, acknowledged within 48 hours. Publishing an address only helps if someone answers it."},{"lead":"Local mode, which matters more than all of the above.","body":"Pick a local Whisper model and your audio is transcribed on your own machine and never sent to us. There is nothing for a BAA to cover, because we never receive the recording."}],"limitsLabel":"What this does not mean.","limitsBody":"An audit is a floor, not a guarantee. It says a company has controls and follows them; it says nothing about whether a particular feature fits your workflow. OpenWhispr dictates into any EHR field through a global hotkey, so Epic, Cerner, athenahealth, and web charts all work with no integration. What it does not do is write back through an API: there is no FHIR or HL7 integration, no awareness of which patient's chart is open, and no coded or structured data. It is not a clinically validated ambient scribe either. If you need notes filed against a specific encounter with coding suggestions, a dedicated scribe product will fit better than we do.","trustLine":"The SOC 2 report, ISO 27001 certificate, subprocessor list, and monitored controls are published in our <trustLink>Trust Center</trustLink>, and how we handle data is written out on the <securityLink>security page</securityLink>."},"whatHipaaRequires":{"title":"What HIPAA actually requires of an AI note taker","intro":"If a vendor creates, receives, stores, or transmits protected health information (PHI) for you, HIPAA calls them a business associate. That triggers a short and specific list:","items":[{"lead":"A signed BAA.","body":"This is the contract that makes the vendor legally accountable for the PHI you hand over. Without one, sending them PHI is itself a violation, however good their encryption is."},{"lead":"Security Rule safeguards.","body":"Access controls, audit logging, encryption in transit and at rest, and a written incident response process."},{"lead":"Breach notification.","body":"They have to tell you about a breach of unsecured PHI without unreasonable delay, so you can meet your own deadlines."},{"lead":"Minimum necessary use.","body":"They can only use PHI the way the BAA allows. That rules out secondary uses, and training AI models on your patient data is the one to watch."},{"lead":"Subcontractor flow-down.","body":"If your audio gets routed to a third-party model provider, that provider is a subcontractor and needs its own BAA. Ask who is in the chain."}],"mythLabel":"The certification myth.","mythBody":"HIPAA has no certifying body, so “HIPAA certified” is not a claim you can verify. The honest phrasings are “HIPAA compliant,” “we support HIPAA compliance,” or “we sign BAAs.” A vendor advertising a certificate is either shortening “we passed a third-party readiness assessment” or has misunderstood the rule. Either way, ask for the BAA, not the badge."},"aiExclusion":{"title":"The trap: a BAA that excludes the AI","intro":"The common failure in 2026 is not a vendor refusing to sign a BAA. It is a vendor signing one that quietly does not cover the AI. Two of the biggest platforms work this way:","zoomLabel":"Zoom.","zoomBody":"Zoom offers a BAA on Zoom for Healthcare and eligible Business and Enterprise plans, covering Meetings, Webinars, Phone, Team Chat, and Rooms. But turning on HIPAA mode disables a set of AI Companion features, and only some of what remains, such as smart recording and meeting summary, sits inside the BAA. The compliant setup is the less capable one.","zoomSource":"Zoom's HIPAA documentation","notionLabel":"Notion.","notionBody":"Notion will sign a BAA, but only for Enterprise customers above a minimum seat count, and Notion AI is excluded from that coverage on every plan. You can store PHI in a Notion database. You cannot point Notion AI at it.","notionSource":"Notion's HIPAA help page","takeaway":"The pattern repeats because of how these products get built. Compliance reviews the platform, the AI ships later on a different legal footing, and the carve-out never comes up in the sales call. So ask for the BAA's scope in writing, feature by feature, rather than a yes or no about the company."},"comparison":{"title":"Which AI note takers will sign a BAA","intro":"Status as of August 2026, from each vendor's published documentation. The last column is the one that matters most and the one pricing pages answer least often.","headers":{"tool":"Tool","baa":"Signs a BAA?","plan":"Required plan","aiCovered":"AI features covered?"},"vendors":[{"name":"OpenWhispr","baa":"Yes, on request","plan":"Business plan and above","aiCovered":"Yes, or no PHI leaves the device","note":"A BAA is available on request from the Business plan up, a lower bar than the Enterprise-only gate Otter and Fathom apply. In local mode the question does not arise: transcription runs on your machine with open-source Whisper models and the audio is never sent to us. Gizmo Labs, which builds OpenWhispr, holds SOC 2 Type II and ISO 27001 and runs HIPAA and GDPR programs.","href":null},{"name":"Otter.ai","baa":"Yes","plan":"Enterprise only","aiCovered":"Yes, under the BAA","note":"Otter announced HIPAA compliance in July 2025 after an independent assessment and signs a BAA covering the use and disclosure of PHI. Only Enterprise qualifies. Basic, Pro, and Business customers cannot get one, so the compliant tier is a real jump in price.","href":"/compare/otter"},{"name":"Fathom","baa":"Yes","plan":"Enterprise only","aiCovered":"Yes, under the BAA","note":"HIPAA compliance backed by a BAA at the Enterprise level. Fathom's model providers — Anthropic, OpenAI, and Google — are contractually barred from training on Fathom user data, which is the subcontractor question answered properly.","href":"/compare/fathom"},{"name":"Fireflies.ai","baa":"Yes, on request","plan":"Healthcare tier","aiCovered":"Yes, under the BAA","note":"Fireflies launched a healthcare offering in September 2025 with a HIPAA-compliant tier. The BAA is not automatic: default plans are not covered and you have to ask. Fireflies says it has signed BAAs with its own vendors barring ePHI training and storage.","href":"/compare/fireflies"},{"name":"Zoom AI Companion","baa":"Partial","plan":"Healthcare / eligible plans","aiCovered":"Only some features","note":"The BAA covers Zoom's core products, but HIPAA mode switches off a range of AI Companion features and only a subset stays in scope. Check feature by feature before relying on it for clinical documentation.","href":null},{"name":"Granola","baa":"No","plan":"—","aiCovered":"No","note":"Granola does not currently offer HIPAA compliance or a BAA. Good product, but not an option for PHI until that changes.","href":"/compare/granola"},{"name":"Notion AI","baa":"Partial","plan":"Enterprise, 100+ seats","aiCovered":"No, AI is excluded","note":"Notion Enterprise signs a BAA covering pages, databases, wikis, and file uploads, but Notion AI sits outside it on every plan. Fine for storing documentation, unusable as an AI scribe for PHI.","href":null}],"compareLink":"Compare OpenWhispr and {name}","disclaimer":"Compiled from vendor documentation in August 2026. Plan names, tiers, and BAA scope change often, so treat this as a starting point for your own diligence rather than a substitute for it. Confirm current terms with the vendor and your compliance counsel."},"checklist":{"title":"Five questions to ask before you record PHI","intro":"Most vendor security pages answer the easy questions and skip the ones that decide whether you can use the product at all. Put these five in writing:","questions":[{"lead":"Will you sign a BAA, and on which plan?","body":"If the answer is “Enterprise only,” get the price before you get attached. The compliant tier is often several times the number you were first quoted."},{"lead":"Does the BAA name the AI features?","body":"Not “is the product covered” but whether transcription, summarization, and any AI chat or agent features are each in scope. This is exactly where Zoom and Notion diverge from their own marketing."},{"lead":"Which subprocessors receive PHI?","body":"If audio goes to OpenAI, Anthropic, Google, or a hosted speech service, each is a subcontractor that needs its own BAA and a no-training commitment. Ask for the list, not a reassurance."},{"lead":"Is our data used for training or product improvement?","body":"“We don't sell your data” is a different and much weaker promise. You want a contractual ban on training and on humans reviewing your recordings."},{"lead":"How long is audio kept, and can we set it to zero?","body":"Real-time processing with immediate deletion is not the same as ninety-day retention with a delete button. Retention you cannot configure is retention you will have to defend in an audit."}]},"localProcessing":{"title":"Why local processing changes the math","intro":"Every question above exists because PHI is leaving your organization and landing on someone else's servers. When the audio never leaves the device, most of the analysis stops applying:","points":["No disclosure to a business associate, because the vendor never receives the PHI.","No subprocessor chain to audit, because no third-party model provider is in the path.","No retention policy to negotiate, because there is no server-side copy.","No exposure to a vendor changing its policy later, because the model runs on hardware you control.","Works with no network access, which matters in facilities where clinical machines are deliberately offline."],"tradeoff":"The trade-off is real and worth stating. Local models need a reasonably modern machine, the largest and most accurate ones are slower on older hardware, and you give up the convenience of a meeting bot that joins calls on your behalf. For a lot of clinical dictation that is a good trade. For high-volume ambient scribing across a large practice, it may not be.","closing":"This is the design OpenWhispr is built around, and it is the reason we think the local option matters more than any certificate we hold."},"frameworks":{"title":"SOC 2 and ISO 27001 are not HIPAA","intro":"These three get bundled together in procurement conversations, and they answer genuinely different questions. Knowing which is which saves time on both sides of a security review:","items":[{"lead":"SOC 2 is a report, not a certificate.","body":"An independent auditor produces it against the Trust Services Criteria. Type I looks at how controls are designed on one date. Type II tests whether they actually worked over a period, usually three to twelve months, and it is the one worth asking for. Anyone calling themselves “SOC 2 certified” is describing a document that does not work that way."},{"lead":"ISO 27001 really is a certification.","body":"An accredited body certifies an organization's information security management system against the standard, on a three-year cycle with surveillance audits. Ask for the certificate number, who issued it, and above all the scope statement, which tells you what is actually covered."},{"lead":"HIPAA is a law, with no certificate at all.","body":"Compliance is an ongoing obligation you demonstrate through safeguards, policies, risk analysis, and signed BAAs. SOC 2 and ISO 27001 are good evidence that a vendor runs a serious security program. Neither one replaces a BAA."}]},"faq":{"title":"Frequently asked questions","legalNote":"This is general information about how vendors are set up, not legal advice. Your own HIPAA obligations depend on your organization, your risk analysis, and your counsel."},"faqs":[{"question":"Is Otter.ai HIPAA compliant?","answer":"Yes, but only on the Enterprise plan. Otter announced HIPAA compliance in July 2025 after an independent assessment and will sign a BAA covering the use and disclosure of PHI. Basic, Pro, and Business customers cannot get a BAA, so Otter is not HIPAA compliant on those tiers. Contact Otter's sales team to start the process."},{"question":"Is Fathom HIPAA compliant?","answer":"Yes, at the Enterprise level, backed by a BAA. Fathom's AI providers — Anthropic, OpenAI, and Google — are contractually barred from training on Fathom user data. Lower tiers are not covered, so confirm your plan qualifies before recording anything clinical."},{"question":"Is Notion HIPAA compliant?","answer":"Partly. Notion signs a BAA for Enterprise customers above a minimum seat count, covering pages, databases, wikis, and file uploads. Notion AI is excluded from that coverage on every plan, so you cannot use Notion's AI features on protected health information even as an Enterprise customer with a signed BAA."},{"question":"Is Zoom AI Companion HIPAA compliant?","answer":"Only in part. Zoom signs a BAA for Zoom for Healthcare and eligible Business and Enterprise plans, but enabling HIPAA mode disables several AI Companion features. Some, such as smart recording and meeting summary, are in scope; others are not. Confirm the specific features you need are covered before using it for clinical documentation."},{"question":"Is Fireflies HIPAA compliant?","answer":"Yes, on its healthcare tier. Fireflies launched a HIPAA-compliant offering in September 2025, but you have to request the BAA explicitly, as default plans are not covered. Fireflies also reports signing BAAs with its own subprocessors barring them from training on or storing ePHI."},{"question":"Is Granola HIPAA compliant?","answer":"No. Granola does not currently offer HIPAA compliance or a Business Associate Agreement, so it should not be used to record or process protected health information."},{"question":"What is the best HIPAA-compliant AI note taker for therapists?","answer":"It depends on whether you need a meeting bot. Solo and small practices are often best served by on-device dictation, where session audio never leaves the laptop and no BAA is needed for that step. If you need a bot that joins telehealth calls, Otter Enterprise, Fathom Enterprise, and the Fireflies healthcare tier all sign BAAs. Whichever you pick, check the retention setting and the subprocessor list, since therapy notes are among the most sensitive records HIPAA covers."},{"question":"Do I need a BAA if transcription runs locally on my computer?","answer":"Generally no, for that step. If audio is transcribed entirely on-device and never sent to the vendor, the vendor is not creating, receiving, storing, or transmitting PHI for you, so it is not acting as a business associate in that flow. You still need your own safeguards: device encryption, access controls, and a policy for where the resulting notes live. Confirm the analysis with your compliance counsel."},{"question":"Does SOC 2 or ISO 27001 mean a tool is HIPAA compliant?","answer":"No. SOC 2 is an auditor's report on security controls and ISO 27001 certifies an information security management system. Both are meaningful evidence of a serious security program, but neither creates the contractual relationship HIPAA requires. Only a signed BAA does that."},{"question":"What is the difference between HIPAA compliant and HIPAA certified?","answer":"“HIPAA certified” is not a real status, because no government agency or accreditation body certifies software against HIPAA. Vendors using the phrase usually mean they completed a third-party readiness assessment. What you can verify is whether they sign a BAA, what that BAA covers, and which safeguards they have documented."},{"question":"Is there a free HIPAA-compliant AI note taker?","answer":"OpenWhispr's local mode is free and open source, and because the audio never leaves your device there is no disclosure to cover. If you want our cloud instead, a BAA is available from the Business plan up. Most cloud-based note takers tie their BAA to Enterprise pricing, so a free cloud tier is almost never usable with PHI. Check the plan requirement before assuming the free option works."},{"question":"Can AI note takers train on patient data?","answer":"Not under a properly drafted BAA. The agreement should explicitly bar the vendor and every subprocessor from using PHI for model training or product improvement. Ask for that language in writing. A privacy-policy line about not selling data is a much weaker promise than a contractual ban on training."}],"cta":{"title":"Dictation and meeting notes that stay on your machine","body":"OpenWhispr runs open-source Whisper models locally, so your audio never has to leave your device. A BAA is available from the Business plan up if you do need our cloud. Free, open source, and available for macOS, Windows, and Linux.","github":"Star on GitHub"}},"useCaseLawyers":{"breadcrumb":{"home":"Home","useCases":"Use Cases","lawyers":"For Lawyers"},"metadata":{"title":"Free Dictation Software For Lawyers & Attorneys - OpenWhispr","description":"Dictation software for lawyers. OpenWhispr offers Cloud transcription or local processing for sensitive and offline work.","keywords":["legal dictation software","lawyer dictation","attorney dictation software","legal transcription","dictation for law firms","offline legal dictation"]},"schema":{"name":"OpenWhispr","applicationCategory":"Legal dictation software","operatingSystem":"macOS, Windows, Linux, iOS","description":"Open-source dictation software for lawyers. Choose Cloud transcription without a local model download, or local processing for sensitive and offline work.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Dictation software for lawyers & attorneys","title":"Speak the letter. It lands in Word. No typist ever hears the file.","body":"Dictate a letter, a file note or a memo and the clean text drops in wherever your cursor already is: Word, Clio, an email. Choose Cloud for fast transcription or local for highly sensitive and offline work.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Man speaking beside an open laptop on a sunny terrace"}},"trustBar":{"label":"Dictate into any app, including:","names":["Clio","iManage","NetDocuments","MyCase","Smokeball","Microsoft Word"]},"pain":{"label":"What you're choosing between","title":"Saying the letter takes two minutes. Getting it typed is the rest of the afternoon.","subtitle":"Every route from your voice to the file should suit the matter, not keep you typing.","cards":[{"title":"A stranger typing up privileged notes.","body":"With a local model and no Cloud fallback, the dictation is transcribed on the machine you spoke into. No outsourced transcription service handles it."},{"title":"Sensitive matters need a local option.","body":"Choose local models when the matter is highly sensitive or you will work without signal. Your device processes the recording."},{"title":"An hour of typing is an hour you can't bill.","body":"Typing up your own dictation is the slowest way to get a letter out, and you do it after the thinking is already done. Dictate it while the matter is fresh and you review and send instead of type."},{"title":"A tool that only fits one desk.","body":"OpenWhispr runs on macOS, Windows, and Linux. Start with the Free plan, then upgrade when the workflow calls for it."}]},"features":{"label":"From your voice to the file","title":"From your voice to a first draft.","subtitle":"Get the letter out while the matter is fresh. Choose the transcription path that fits the work.","items":[{"image":{"label":"Woman speaking while using a laptop at a desk"},"title":"Cloud transcription without a local model download","body":"Speak a letter, a memo or a file note into your mic. Cloud gives you fast transcription without a local model download, then immediately discards the transcription audio."},{"image":{"label":"Man reviewing a printed document beside his laptop"},"title":"Letters and memos, drafted for you","body":"Tell it what you want and use local AI models for device-only drafting. You read every word before it leaves your desk."},{"image":{"label":"Man speaking and pointing toward a laptop at a table"},"title":"Dictate into Word without an add-in","body":"One hotkey, and the clean punctuated text pastes itself where you're already typing: Word, an email, Clio, your document system. There's nothing for IT to install into any of them."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Privileged audio sent to us","value":"None","footnote":"with a local transcription model the recording never leaves your computer"},{"label":"Cost per lawyer","value":"$$0","footnote":"unlimited local dictation in the Free plan"},{"label":"Processing options","value":"2","footnote":"Cloud transcription or local models"}]},"howItWorks":{"label":"Download to first letter","title":"Three steps from download to draft.","subtitle":"Choose Cloud transcription or download a local model.","image":{"label":"Man standing beside a laptop in a glass-walled office"},"steps":[{"title":"Download it free.","body":"Mac, Windows or Linux. Setup offers OpenWhispr Cloud or local models. Pick local for privileged work and it downloads them once."},{"title":"Dictate the letter.","body":"Speak it the way you'd say it to your assistant, and it comes back as text."},{"title":"Read it, paste it, delete the audio.","body":"Edit the draft, drop it into Word or your document system, and delete the recording. Every one of those is yours to control."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"Why firms keep it","title":"Ten reasons your IT or security team can check for themselves.","subtitle":"Choosing dictation software for lawyers is a confidentiality decision before it's anything else. Most of these are facts about the code, so IT or security can check them instead of taking them from us.","items":[{"title":"The recording stays on your machine","body":"Choose a local transcription model for device-only processing. An optional Cloud fallback can send audio if you turn it on."},{"title":"Free for the whole firm","body":"Unlimited local dictation at $0 for every desk, including your assistant and paralegals."},{"title":"Nothing goes to a typing pool","body":"With local transcription and no Cloud fallback, no outsourced transcription service handles the recording."},{"title":"Works with the wifi off","body":"A courthouse hallway, a deposition, a client's office, a flight. Local models need no signal, so nothing stalls mid-sentence."},{"title":"Dictate into Word or any case system","body":"One system-wide hotkey drops clean text where your cursor already is: Word, Clio, iManage, an email. On a Mac it uses the accessibility permission you grant at setup."},{"title":"Letters and memos, not just transcripts","body":"Ask for a case memo and a local AI model can draft it on your machine. Local AI models are included on every plan."},{"title":"You control storage and deletion","body":"Recordings are files on your machine. Audio is deleted after 30 days out of the box, and Settings moves that or switches it off."},{"title":"Nothing to take on faith","body":"Your IT or security team can read every line of the app on GitHub before a partner installs it."},{"title":"Party names and citations, spelled your way","body":"Put party names, case names, statutes, and citations in the custom dictionary to guide transcription and cleanup."},{"title":"100+ languages","body":"Dictate in 100+ languages."}]},"faq":{"label":"Before you install it","title":"Eight questions law firms ask before they install anything.","subtitle":"Straight answers on confidentiality, cost, accuracy, and what IT will say. More questions? Ask in Discord.","items":[{"question":"Does the audio ever leave my device?","answer":"With local transcription, audio is transcribed on your device. Turn off the optional Cloud fallback when every attempt must stay local. Cloud sends audio only to transcribe it, then immediately discards it."},{"question":"Is this consistent with my duty of confidentiality?","answer":"Confidentiality is your professional duty and no tool discharges it for you. Local processing is the deliberate choice when a highly sensitive matter calls for device-only work. Cloud transcription discards audio immediately after transcription. This is not legal advice, and the rules of your bar or law society still apply."},{"question":"Does dictation software for lawyers get case names and citations right?","answer":"Add party names, case names, Latin, statutes, and citations to the custom dictionary. OpenWhispr passes that list to transcription and cleanup to guide the text."},{"question":"Will IT let me install it on the firm laptop?","answer":"It is a desktop app, not an add-in for Word or your document system. On a Mac, grant accessibility permission for auto-paste. Without it, copy the text from your clipboard instead."},{"question":"Will it work in a courthouse or on a flight with no signal?","answer":"Yes, once the models are on your machine. Local models run entirely offline after the first download. Dictate in the courthouse hallway, after a deposition, or on the flight home, then file the note when you are back in range."},{"question":"Does it draft the letter, or just transcribe?","answer":"Both. Dictation turns speech into text, then an AI model can shape a client letter or case memo. Transcription and AI cleanup are separate settings, so choose a local AI model when the draft must stay on-device."},{"question":"Do I need a separate legal dictation tool?","answer":"OpenWhispr runs on macOS, Windows, and Linux. Try the Free plan with a local model, then move to Cloud transcription when it fits your workflow."},{"question":"Is it actually free?","answer":"Free includes unlimited local dictation and local AI models, plus 2,000 Cloud words a week. Pro adds unlimited Cloud transcription, 20 hours of meeting recordings, sync, mobile, and API/MCP access. Business adds unlimited meeting recordings, Agent mode, and chat over data."}]},"reviews":{"label":"What people tell us","title":"Built by people who wouldn't hand their audio to a stranger either.","subtitle":"From our Discord, GitHub and inboxes. None of them is a lawyer, so read them for the tool, not the practice.","score":{"value":"0","label":"Audio stored on our servers with local transcription","body":"With a local transcription model the audio never leaves your device, so there's nothing on our side to lose, breach or hand over."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free to start.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for lawyers","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Lawyers who want local dictation and 2,000 Cloud words a week"}]},{"parts":[{"text":"Lawyers who want unlimited Cloud transcription plus mobile, sync, and API/MCP access"}]},{"parts":[{"text":"Firms documenting at volume with Agent mode and chat over data"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited local dictation plus 2,000 Cloud words a week"}]},{"parts":[{"text":"Unlimited Cloud transcription plus mobile, sync, and API/MCP access"}]},{"parts":[{"text":"Unlimited meeting recordings plus Agent mode and chat over data"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","Enterprise adds SSO, audit logs, retention controls, and dedicated support."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Why we built it","title":"A client tells you things they'd tell nobody else.","paragraphs":["OpenWhispr started with a simple objection: confidential dictation should have a local option.","We build in public. The app is open source, so your firm can inspect how it works before anyone installs it.","Community support is available in Discord. Paid plans add email or priority support."],"image":{"label":"One of OpenWhispr's founders"}},"finalCta":{"tagline":"$$0 a seat · open source · no add-in","title":"Speak the next letter while the matter is fresh.","body":"Download OpenWhispr, choose Cloud transcription or local processing, and dictate the next letter.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free on macOS, Windows, and Linux. Paid plans add Cloud transcription, mobile, and workflow tools."}}},"useCaseLegalDocuments":{"breadcrumb":{"home":"Home","useCases":"Use Cases","legalDocuments":"Legal Documents"},"metadata":{"title":"Free AI Dictation For Legal Documents - OpenWhispr","description":"AI dictation for legal documents: talk a clause, a brief section or a motion through and OpenWhispr's agent drafts it. Free forever, and it runs locally.","keywords":["ai dictation for legal documents","dictate a contract","brief and motion dictation software","discovery response drafting tool","deposition summary dictation","paralegal and legal ops dictation"]},"schema":{"name":"OpenWhispr","applicationCategory":"Legal document drafting software","operatingSystem":"macOS, Windows, Linux","description":"OpenWhispr is an open-source dictation app for the people who produce legal documents: paralegals, legal assistants and lawyers. They dictate contracts, briefs, motions, discovery responses and deposition summaries, and the AI agent redrafts them into structured text, either in place in the document or in the app's assistant panel.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"AI dictation software for legal documents","title":"Stop typing first drafts. Talk the section through.","body":"Talk a contract clause, a motion section or a discovery response through, then tell the AI agent what to make of it: headings, numbered points, a summary. Pick a local model and it's drafted on your own machine, free and unlimited on every plan.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Woman speaking beside a laptop at an office desk"}},"trustBar":{"label":"Dictate:","names":["Contract clauses","Motion sections","Discovery responses","Deposition summaries","Case memos","Engagement letters"]},"pain":{"label":"Why dictating a contract stalls","title":"Talking it through was never the slow part. Shaping it is.","subtitle":"Speak an engagement letter or a case chronology and you'll get every word back in one unbroken block. Cutting that block into clauses, headings and numbered points is what eats the afternoon, and plain dictation hands that part straight back to you.","cards":[{"title":"One block of speech, not a clause in it.","body":"Dictation cleanup hands back what you said, tidied. It will break the speech into paragraphs and number a list you clearly dictated. What it won't do is impose a heading or restructure a deposition summary into sections. That job belongs to the AI agent, and you have to ask it for it."},{"title":"The motion is the sensitive part, not the audio.","body":"A transcript is words. A motion is the argument you're about to run and the position your client will take. Choose local processing when you want the composing to run on your machine, or cloud tools when they fit your workflow."},{"title":"Paralegals and legal ops draft it too. Per-seat pricing adds up.","body":"An associate dictates the contract, a paralegal checks the cites, an assistant cleans it up, legal ops files it. Charge per seat and the cost follows the draft across four desks. OpenWhispr charges nothing for a seat on Free, so each of them can use it."},{"title":"Party names and statutes come back mangled.","body":"A general-purpose model has never seen your matter. Put the case names, the parties, the Latin and the citations your practice uses in the custom dictionary. A dictation vocabulary only fixes what was heard; this one also reaches the model that drafts, so it sees those names before it writes a word. It moves the odds rather than guaranteeing them, so read what carries legal weight before the document goes out."}]},"features":{"label":"From microphone to document","title":"You dictate a section. You get back something worth editing.","subtitle":"A transcript isn't a document. Dictation, cleanup and the AI agent each hand you something different.","items":[{"image":{"label":"Woman speaking while working on a laptop"},"title":"Say it once, read it back punctuated","body":"Talk an interrogatory response through the way you'd say it out loud, and it comes back transcribed and punctuated with the filler and the false starts taken out. Address the agent instead and it works from the same dictation, handling the disfluencies itself."},{"image":{"label":"Woman reading a printed document beside her laptop"},"title":"Headings and numbered points, not a wall of speech","body":"Cleanup breaks your speech into paragraphs and numbers a list you clearly dictated, and stops there. Ask the AI agent instead, on its own hotkey or by name mid-dictation, and it composes what you asked for: the headings, the numbered points and the lists you talked through."},{"image":{"label":"Man speaking while reviewing a printed document beside a laptop"},"title":"Nothing to install into Word or your DMS","body":"One global hotkey copies the text and pastes it at your cursor, so the clause lands in Word, a matter record, an email or a web form the same way. Auto-paste is on by default, so you edit it where it lands."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Audio kept on your own machine by default","value":"30 days","footnote":"Audio Retention ships on at 30 days. Settings takes it from 1 day to 90, or switches it off"},{"label":"Languages you can dictate a matter in","value":"100+","footnote":"a German framework agreement comes back as German text"},{"label":"Cost to add the paralegal and legal ops","value":"$$0/month","footnote":"no card details, and no per-seat licence to buy"}]},"howItWorks":{"label":"How a contract gets drafted","title":"Talk through the motion, then edit what comes back.","subtitle":"Three steps from talking the argument through to editing it in the document.","image":{"label":"Man speaking beside a laptop in a book-lined office"},"steps":[{"title":"Install it and choose where it runs.","body":"Setup puts a local model and OpenWhispr Cloud in front of you, with Cloud badged Recommended and local warning you it takes a few extra steps. Going local means pulling down a speech model of 75 MB to 3 GB and a separate local LLM of 0.25 GB to 19.8 GB first, depending which two you pick."},{"title":"Talk the clause through, ums and all.","body":"Say the deposition summary the way you'd explain it to a colleague, ums and all. Cleanup hands the same words back with the filler and the false starts stripped out."},{"title":"Ask the agent for headings and numbered points.","body":"Highlight the rough passage, press the agent hotkey and tell it what you want, and it rewrites that passage in place in the document. With nothing highlighted the draft arrives in the app's assistant panel instead, ready to copy across. Either way it's plain markdown text rather than Word styles, so you shape it in the document itself."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"Ten claims you can check","title":"What OpenWhispr does to a legal document, and where it stops.","subtitle":"What it does, what it won't do, and where the document ends up. No asterisks.","items":[{"title":"The setting that would upload a contract arrives switched off","body":"Pick a local model at setup and the audio, the transcript and the composed section all stay on the machine you spoke into. Two deliberate acts send content to us: switching on paid cloud sync, which arrives off and stores your notes, transcripts, agent conversations and your custom dictionary, party names included; and sharing a note or moving it into a team space."},{"title":"Cleanup keeps your wording and won't add a sentence you didn't say","body":"Cleanup is told to keep your voice, your wording and your intent, and never to add a sentence you didn't say. It still fixes grammar and punctuation, splits run-on sentences, drops false starts, and writes spoken numbers and dates out in standard form. That's an instruction to a model rather than a guarantee, so read an indemnity back before it goes out."},{"title":"The agent turns talk into a numbered section","body":"Say its name mid-dictation or press its hotkey, tell it what you want, and it composes that from what you talked through: headings, numbered points, a summary of the argument. Highlight a passage first and it rewrites that passage in place; with nothing highlighted the draft lands in the assistant panel. Pick local at setup and it composes on the LLM you downloaded, on your own machine. Styling stays a Word job."},{"title":"One hotkey, and no add-in for IT to approve","body":"The text goes to the clipboard and pastes itself where your cursor is. On macOS, without Accessibility permission it can copy for you but not paste."},{"title":"Know where the ceiling is before you hit it","body":"For selection edits, select 6,000 characters or fewer."},{"title":"Dictate it into a note now, shape it later","body":"Save the case chronology as a note while it's fresh, and a note action turns it into clean markdown when you come back to it."},{"title":"Recordings are files, so deleting one really deletes it","body":"Recordings are audio files on your machine you can open or delete yourself; notes and transcripts sit in a local database the app deletes rows from. Audio Retention ships on at 30 days, and Settings takes it from 1 day to 90 or switches it off. Transcripts are the other way round, kept until you set a window of your own, so set one alongside the audio. OpenWhispr doesn't encrypt that local data itself, so the folder wants the care you'd give any other matter folder."},{"title":"Send procurement the repo instead of a policy PDF","body":"OpenWhispr is open source. When IT asks what happens to a contract you dictated, they can inspect the code rather than take our word for it."},{"title":"The paralegal and the legal ops lead get the same hotkey","body":"Legal assistants and legal ops draft these documents alongside lawyers. Free is a download rather than a licence, so the person who assembles the interrogatory response has the same hotkey as the person who signs it."},{"title":"Dictate the deal in the language it was done in","body":"Whisper-class models cover 100+ languages, so the deal you did in Spanish comes back transcribed in Spanish."}]},"faq":{"label":"Before the first document","title":"What actually comes back when you dictate a contract.","subtitle":"Anything we've missed, bring it to our Discord, where the people who build OpenWhispr answer.","items":[{"question":"Which legal documents can AI dictation draft: a contract clause, a motion section, a discovery response?","answer":"All three, plus case memos, deposition summaries and file notes. Worth knowing which part does what: plain dictation hands back your own words, tidied up. The AI agent, on its own hotkey or called by name mid-sentence, drafts them into a structured section when you ask it for one. Highlight a passage first and it rewrites that passage in place; with nothing highlighted the draft arrives in the assistant panel to copy across. What you get is markdown text, so ## Heading arrives in Word as those literal characters and not a Word style. Pleading captions and court-ready tables aren't its job either, so the formatting your jurisdiction wants still happens in Word."},{"question":"Who else ends up with a copy of an unsigned contract?","answer":"Nobody, as long as you picked a local model at setup and left the cloud alone. Whisper-class transcription runs on your machine, and the composing runs on the LLM you downloaded, served by llama.cpp at 127.0.0.1, so the clause never travels. Two things send content to us, and both are deliberate acts: cloud sync, a paid setting that stays off until you switch it on, which stores your notes, your dictation transcripts, your agent conversations and your custom dictionary, party names included, in our cloud database; and sharing a note or putting it in a team space, which uploads that note on any signed-in account. One more thing if you sign in: the assistant panel can run a web search, which sends the query the model composed, not your document, to us. Stay signed out and it is never offered."},{"question":"Can it write a whole brief in one go?","answer":"For a brief, work section by section: facts, then argument, then relief."},{"question":"Will party names and citations survive a deposition summary?","answer":"Load the custom dictionary first. It holds the words your matters actually turn on: party names, case names, statutes, the Latin, the citations your practice uses. It's a hint the model sees rather than a find-and-replace, so it moves the odds a long way in your favour without promising a clean sweep. It reaches local Whisper transcription, the cleanup pass, the AI agent and translation, across 100+ languages."},{"question":"Can a paralegal take over the contract after I've dictated it?","answer":"Yes. The hotkey copies the text and pastes it wherever the cursor is, so the words land in the file the two of you already share: Word, the matter record, an email. Auto-paste is on by default. On macOS it needs Accessibility permission, or it falls back to leaving the text on your clipboard."},{"question":"What does it cost to run a matter's documents through it?","answer":"Nothing on the local path, where dictation and drafting are unlimited on every plan, Free included, and neither one is metered. The cloud side has two separate meters. On Free you get 2,000 words a week of cloud speech-to-text, and 5 hours a month of cloud meeting recording. Pro at $8 a month makes cloud speech-to-text unlimited and takes meeting recordings to 20 hours a month. Business at $16 per user makes meeting recordings unlimited and adds team features."}]},"reviews":{"label":"Straight from our users","title":"Rambling in, editable text out.","subtitle":"Nobody was asked for these. One of them stumbles mid-sentence and still gets a clean paragraph back, which is what dictating a clause actually sounds like.","score":{"value":"0","label":"Client documents sitting on our servers","body":"In local mode, signed out, with cloud sync off and nothing shared, nothing reaches us. Turn cloud sync on, a paid setting that arrives switched off, and your notes, dictation transcripts, agent conversations and custom dictionary are stored in our cloud database."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free to get started.","Upgrade when you want more."],"caption":"What each plan covers for legal document work","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Paralegals and legal teams drafting contracts, briefs and motions, who don't mind downloading a speech model and an LLM first"}]},{"parts":[{"text":"Legal assistants who'd rather dictate through OpenWhispr Cloud than manage model downloads, at the desk and on an iPhone"}]},{"parts":[{"text":"Legal ops teams handling document work across a lot of matters, and recording calls all month"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Free plan: $0."}]},{"parts":[{"text":"The cloud speech-to-text meter comes off, and device sync and the iPhone app come with it"}]},{"parts":[{"text":"Cloud meeting recordings go unlimited and team features unlock, at $16 per user per month"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud and does not require local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Read the code yourself","title":"A contract shouldn't need a round trip to get written.","paragraphs":["Turning a transcript into clauses and numbered points takes an AI model. OpenWhispr's AI agent can do that work with a model you downloaded yourself, served by the llama.cpp binary that ships inside the app, on 127.0.0.1.","You don't have to take our word for that. OpenWhispr is open source, so your risk team can inspect the code rather than read a policy about it.","Break something at 11pm the night before a filing deadline and you're talking to the engineers who wrote it, in Discord."],"image":{"label":"The people who build it"}},"finalCta":{"tagline":"$$0 a seat · No card details","title":"Say the clause out loud. Get a numbered section back.","body":"Download it free, pick a local model at setup, then hand the AI agent your rough dictation. Highlight the passage and it rewrites the clause or the case memo in place, as plain markdown text. No card details, no seat to buy.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Available on macOS, Windows, and Linux. Get community support in Discord."}}},"useCaseVeterinary":{"breadcrumb":{"home":"Home","useCases":"Use Cases","veterinary":"For Vets"},"metadata":{"title":"Free Offline Veterinary Dictation Software - OpenWhispr","description":"OpenWhispr is free, open-source veterinary dictation software that turns spoken findings into SOAP notes on your device, even on a farm call with no WiFi.","keywords":["veterinary dictation software","vet dictation app","veterinary transcription","SOAP note dictation","veterinary practice software","offline vet dictation"]},"schema":{"name":"OpenWhispr","applicationCategory":"Veterinary dictation software","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source veterinary dictation software. Dictated findings are transcribed on the vet's own device and never leave it. Free for unlimited local dictation, and works fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Veterinary dictation software for vets & vet techs","title":"Say it once. Get the SOAP note.","body":"OpenWhispr turns that dictation into a SOAP note, a discharge or a client email on your own computer. No per-DVM subscription, and it still works in a barn with no signal.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Veterinarian speaking beside a dog and laptop in a clinic"}},"trustBar":{"label":"Dictate into:","names":["ezyVet","Cornerstone","Avimark","Covetrus Pulse","Shepherd","Microsoft Word"]},"pain":{"label":"The documentation squeeze","title":"Documentation shouldn't cost your evenings and $99 a vet a month.","subtitle":"You want the note written for you. You don't want a subscription per DVM, a recorder in the exam room, or a tool that dies the moment you leave WiFi.","cards":[{"title":"Charting follows you home.","body":"Turn dictated text into a structured note with a custom action."},{"title":"A per-DVM subscription is another monthly line item.","body":"The Free plan includes unlimited local AI models and 100+ language support at $0."},{"title":"No signal, no scribe.","body":"Barn calls, farm visits, a back room with one bar. A cloud scribe can stall exactly where you are, leaving you to rebuild the note from memory that night. OpenWhispr runs on your device, fully offline."},{"title":"A recorder in the euthanasia room.","body":"Some conversations shouldn't be uploaded to a stranger's server to become a note. You dictate the findings afterwards, and with a local model everything is processed on your machine. Nothing sits on the table while you talk a client through it."}]},"features":{"label":"What you get","title":"Records closed before you lock up. No per-DVM bill.","subtitle":"An AI scribe listens to the room so you never have to say it. OpenWhispr waits for you to dictate, then a local model drafts the note from a prompt you saved yourself, on your own laptop.","items":[{"image":{"label":"Veterinarian speaking beside a dog on an exam table and a laptop"},"title":"Dictated at the treatment table, transcribed on your laptop","body":"Signalment, history, exam findings, assessment and plan, straight into the mic. Pick a local model and it's transcribed on your laptop, never uploaded, and you decide what's kept and when it goes."},{"image":{"label":"Veterinarian handing paperwork to a dog owner at a clinic doorway"},"title":"One dictation, two documents","body":"Save your own SOAP, discharge and client-email prompts as one-click actions, then run whichever one the case needs. The same dictation becomes the record for the file and the instructions the client reads in the car park. You check both before either is filed."},{"image":{"label":"Veterinarian using a laptop beside a dog in a rural clinic"},"title":"Types into the PIMS field you're already in","body":"Press a hotkey and dictate into the Cornerstone note field, an ezyVet consult tab, a discharge template or a referral email to the specialist. Clean, punctuated text, processed locally."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Client conversations uploaded","value":"None","footnote":"in local mode, not the estimate conversation, not the goodbye"},{"label":"Cost per DVM, per tech","value":"$$0","footnote":"unlimited local dictation, including the relief vet who works Saturdays"},{"label":"Setup takes","value":"No fixed time","footnote":"Install the app, then choose a local model or OpenWhispr Cloud."}]},"howItWorks":{"label":"How it works","title":"Dictate on the walk back to treatment.","subtitle":"This is one app on the practice laptop, with no PIMS migration or IT project.","image":{"label":"Veterinarian carrying a tablet past a dog in a treatment room"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model during setup."},{"title":"Dictate the case.","body":"Signalment, presenting complaint, exam findings, assessment and plan. In the room, in the treatment area, or in the truck between farms. It's transcribed on your device, and one click turns it into the draft."},{"title":"Review, paste, close the record.","body":"Fix what it got wrong, drop it into the patient's record in ezyVet or Cornerstone, and delete the audio. All of it happens on your laptop."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons to choose OpenWhispr","subtitle":"In a three-vet practice this is a budget decision the owner signs off on, not an IT project. Here are the facts, and the practice manager can hand the repo link to whoever does your IT.","items":[{"title":"Choose local when dictation must stay on your device","body":"Dictation, and note drafts with a local model, are processed on your machine. The open-source code shows there's no upload step."},{"title":"Every DVM and every tech, $0","body":"Unlimited local dictation, forever. Scribes charge per DVM and leave the techs writing records for free anyway."},{"title":"Works in the barn and in the truck","body":"The local model runs with the laptop in airplane mode. Dictate in the truck, file the note when you're back in range."},{"title":"Dictate into the PIMS you already run","body":"A system-wide hotkey drops clean text wherever your cursor is: ezyVet, Cornerstone, Avimark, a discharge template, a referral email. No integration to build, and no reason to change PIMS."},{"title":"One dictation, the SOAP note and the discharge","body":"Save the prompt once. A local model, or your own AI key, turns the same dictation into the record for the file and the instructions the client takes home."},{"title":"No recorder in the room","body":"You dictate afterwards. Nothing sits on the table during the estimate, and nothing is running when you put a hand on a client's shoulder."},{"title":"Choose how long local records stay on the practice laptop","body":"When you enable local audio or transcript retention, files stay on this device and are deleted after the retention period."},{"title":"Read the code before the owner signs anything","body":"Every line is on GitHub. The practice manager can send the link to whoever does your IT and have an answer the same week."},{"title":"Knows cephalexin from Cerenia","body":"The custom dictionary learns the drug names, breeds and procedures you correct, and you add the abbreviations your clinic already writes: OHE, TPR, BAR, QAR."},{"title":"100+ languages","body":"Take the history from a client who would rather explain it in Spanish, transcribed on your laptop."}]},"faq":{"label":"Before you download","title":"The questions vets ask first.","subtitle":"Straight answers on cost, terminology and what happens when you leave WiFi. If yours isn't here, ask us in Discord.","items":[{"question":"Do you record the appointment, like a scribe does?","answer":"No. Nothing joins the consult and nothing sits on the table. You dictate afterwards, at the treatment table or in the truck, and in local mode that dictation is transcribed on your own laptop with no upload step. The code is public, so you can check that instead of taking our word for it."},{"question":"Do vets even need this? HIPAA doesn't cover animals.","answer":"Right. HIPAA covers human patients, not pets. But the record is still confidential, and the audio carries the client's name, their address and the hardest conversation of your day. Local-first keeps all of it off a vendor's server."},{"question":"Does veterinary dictation software get drug names and breeds right?","answer":"Yes, once you teach it yours. It runs Whisper- and Parakeet-class models trained on a wide range of accents and languages, and the custom dictionary learns the terms you correct: Apoquel, Cerenia, brachycephalic, Bichon Frise, OHE, plus whatever abbreviations your clinic already writes."},{"question":"Will it work on a farm call with no signal?","answer":"Yes. Local models run entirely offline. Dictate in the barn or the truck, and file the note once you're back in range. Cloud scribes can't do this, because they stall the moment you leave WiFi."},{"question":"Can it write the discharge as well as the SOAP note?","answer":"Yes, from the same dictation. Run a local model (Llama, Qwen, Gemma) or connect your own AI key, then save your own SOAP, discharge and client-letter prompts as one-click actions and run whichever one the case needs. A local model keeps it all on your machine; your own key means the model and data path are ones you already trust."},{"question":"Is it actually free?","answer":"Yes. The Free plan includes unlimited local AI models, 100+ languages, and desktop availability on macOS, Windows and Linux at $0. Pro is $8/month and adds unlimited OpenWhispr Cloud transcription, sync across devices, and the iPhone and iPad app."}]},"reviews":{"label":"What users say","title":"Built by people who wouldn't put a recorder in the room either.","subtitle":"Three OpenWhispr users describe how they use it.","score":{"value":"0","label":"Consults on a server","body":"You dictate afterwards, so there is no appointment recording to begin with. In local mode the dictation is never uploaded either."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for veterinarians","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Vets and techs who want free, private, offline dictation and are fine with a quick local setup"}]},{"parts":[{"text":"Vets who want easy, reliable cloud transcription"}]},{"parts":[{"text":"Practices and teams documenting at volume"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Free offline dictation, with consult audio staying on your device"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models or keys"}]},{"parts":[{"text":"Unlimited recordings plus advanced workflow features for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","Enterprise includes team features and admin controls, plus dedicated support."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"The goodbye appointment doesn't belong on a vendor's server.","paragraphs":["OpenWhispr started with a simple objection: the three conversations that take the most out of you, the diagnosis, the estimate and the goodbye, shouldn't have to pass through a stranger's server to become a note.","We build in public. The code is on GitHub. When you choose local processing, dictation stays on your device. A practice can verify that in the code.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Last appointment at six. Records closed at six.","body":"Download OpenWhispr and dictate the next case on the walk back to treatment. Choose local models when dictation needs to stay on your laptop and work offline.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Unlimited local AI models are free on macOS, Windows and Linux."}}},"useCaseJournalists":{"breadcrumb":{"home":"Home","useCases":"Use Cases","journalists":"For Journalists"},"metadata":{"title":"Free AI Speech To Text Tool For Journalists - OpenWhispr","description":"Free, open-source AI speech to text for journalists. In local mode the interview is transcribed on your laptop, never uploaded. No per-minute bill.","keywords":["ai speech to text tool for journalists","transcription software for journalists","interview transcription software","offline interview transcription","private transcription for reporters","local speech to text","free interview transcription"]},"schema":{"name":"OpenWhispr","applicationCategory":"Journalism transcription software","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source speech to text software for journalists. In local mode interview audio is transcribed on the reporter's own device and never leaves it. Free for unlimited local transcription.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"AI speech to text for journalists & reporters","title":"Drop in the tape. Get the whole transcript.","body":"Get the whole transcript without uploading the tape. In local mode, OpenWhispr transcribes on your machine, free and open source, with no per-minute bill or signal once the model is downloaded.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Person typing on a laptop beside a field recorder in an open vehicle"}},"trustBar":{"label":"Paste your quotes into:","names":["Google Docs","Microsoft Word","WordPress","Arc XP","Slack","Substack"]},"pain":{"label":"What an upload actually costs you","title":"Source protection shouldn't end at the upload button.","subtitle":"You want the tape transcribed on deadline. You don't want it living in someone else's account to get there.","cards":[{"title":"Keep source audio on your device in local mode.","body":"Upload an interview and a vendor holds the recording. Some services route the audio to human transcriptionists. In local mode OpenWhispr transcribes on your device, so there is no server for the tape to sit on."},{"title":"Your tape becomes someone else's record.","body":"Every upload leaves unpublished material in an account you don't administer, on a retention schedule you didn't set. In local mode there is no upload and no account, so there is nothing on our side for anyone to subpoena."},{"title":"A quote you can't check.","body":"Speech to text mishears names and crosstalk, and every model will occasionally write a sentence nobody said. A summary will not show you where that happened. Record the interview in the app and the transcript is stamped line by line. To check a quote before you file, compare it with the original recording you keep."},{"title":"\"Trust us\" is not source protection.","body":"Closed tools ask you to take a privacy policy on faith. OpenWhispr's code is public on GitHub under the MIT license. Your security desk, or any developer you trust, can read exactly what it does and what it doesn't."}]},"features":{"label":"What you get","title":"Filed faster. The tape stays yours.","subtitle":"Transcription tools usually trade source protection for speed, or bill you by the minute. This does neither. In local mode the interview is transcribed on your machine, at no cost per minute.","items":[{"image":{"label":"Two people talking across a table with an open laptop"},"title":"Interview files transcribed on-device","body":"Drop in an MP3, WAV, M4A or half a dozen other formats, one file or a whole batch. In local mode it's transcribed on your machine, with no length limit and no per-minute bill."},{"image":{"label":"Person reviewing printed pages beside a laptop and phone"},"title":"The whole transcript, never a summary","body":"You get the full transcript, not a digest. Record the interview in the app and every line is stamped on your device, ready to export as .txt, .md, .json or .srt. On an upload, speaker detection stamps each turn locally; per-line stamps need your own Groq or Mistral key, or OpenAI set to whisper-1."},{"image":{"label":"Two people talking at a desk with a field recorder and notebook"},"title":"Live transcript from the room","body":"Put your laptop on the table and it transcribes an in-person interview as you talk, on any platform. On a remote interview it captures your computer's own audio, so no bot joins and nothing shows in the participant list. That capture needs Windows, macOS 14.2 or later, or a Linux desktop whose audio portal allows it. Recording is its own setting, so point that one at a local model too."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Source audio sent to us","value":"None","footnote":"in local mode, your audio is transcribed on your own device"},{"label":"Cost per hour of tape","value":"$$0","footnote":"unlimited local transcription, every reporter, forever"},{"label":"Setup takes","value":"5 min","footnote":"from download to your first transcript, plus the one-time model download"}]},"howItWorks":{"label":"How it works","title":"Transcribed before you write the lede.","subtitle":"No account or credit card required. Choose a local model in Settings, then drop in the tape or record in the app.","image":{"label":"Person at a desk with a laptop, notebook, and audio recorder"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Skip the account and setup hands you a local model. Either way, point uploads and recording at one in Settings."},{"title":"Drop in the tape.","body":"MP3, WAV, M4A, FLAC and more, one file or a batch. In local mode it's transcribed on your device, however long the interview ran."},{"title":"Check it, then file.","body":"Read the transcript in the app and check any line against the audio. A recorded interview exports to .txt, .md, .json or .srt with every line stamped."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"On the record","title":"10 reasons to choose OpenWhispr","subtitle":"Picking an AI speech to text tool is a source-protection decision before it's anything else. Here are the facts, and the ones about your audio you can check in the code yourself.","items":[{"title":"Local mode keeps source audio on your device","body":"Transcription runs on your machine, with no upload step."},{"title":"No bot joins a remote interview","body":"It records your computer's own audio, on Windows, macOS 14.2 or later, or a Linux desktop that allows it. Nothing appears in the participant list, and on a local model no vendor gets a copy of the call."},{"title":"Offline, once the model is downloaded","body":"Courthouse steps, a jail visit, a rural county meeting. Once the model is downloaded, no signal required."},{"title":"No import limits, no per-minute bill","body":"Local transcription is unlimited at $0, on every plan."},{"title":"Your security desk can read the code","body":"MIT licensed, every line on GitHub. A newsroom security review can read exactly what happens to the audio, before you put a source's tape through it."},{"title":"Recorded interviews, stamped line by line","body":"Export to .txt, .md, .json or .srt with every line stamped, so you can point standards or the lawyers at the exact moment."},{"title":"No length limit on your own machine","body":"A ninety-minute interview costs the same as a nine-minute one, and takes as long as your laptop needs."},{"title":"Who said what, on every plan","body":"Two voices in one interview, labelled. Speaker detection runs on your device, so it works on a local model with nothing uploaded."},{"title":"Nothing rewrites your quotes","body":"An uploaded interview is never rewritten. The note is the transcript as the model heard it. In dictation the untouched text sits beside the cleaned-up version, and you can switch cleanup off entirely."},{"title":"100+ languages","body":"Interview a source in their own language, not their second one. In local mode, transcription stays on your machine."}]},"faq":{"label":"Before you download","title":"The questions reporters ask first.","subtitle":"Straight answers, including where it falls short. Anything missing, ask in Discord.","items":[{"question":"Does my source's audio ever leave my device?","answer":"Not in local mode. Transcription runs on your machine and the open-source code proves there's no upload step. Cloud transcription is the default, so switch to a local model in Settings before your first interview. Dictation, recordings and uploaded files each carry their own transcription setting, so change the one you use for tape. If you sign in, point the AI cleanup model at a local one too, or switch cleanup off: that setting writes the note title and it defaults to the cloud."},{"question":"Can it transcribe a file I already recorded?","answer":"Yes. Drag in an MP3, WAV, M4A, FLAC, AAC, OGG, Opus or WebM file, pick it from a file dialog, or queue a batch. On a local model there's no length limit and no per-minute charge. Local video files aren't supported, but you can paste a YouTube or video link and it pulls the audio, up to six hours or 500 MB."},{"question":"Is it accurate enough to quote from?","answer":"It runs Whisper-class and NVIDIA Parakeet models, and no speech to text is perfect on accents, names or crosstalk. That's why you get the whole transcript, never a summary: record the interview in the app and it's stamped line by line, so you can check the quote against the tape before you file. Treat it as a first pass you verify, not a transcript you publish unread."},{"question":"Can it tell my source's voice from mine?","answer":"Yes. On a local model the speaker detection runs on your device too, with nothing uploaded. It separates the voices in an in-person interview and labels who said what, on every plan; the transcript is live as you talk and the labels land when you stop. A recording expects two speakers and goes up to eight, enough for a panel. On an uploaded file, switch labels on before you transcribe and it works out the count for you."},{"question":"If we get a subpoena, what do you have?","answer":"In local mode, nothing. There is no upload and no account, so there is no tape and no transcript on our side to hand anyone. What exists is on your laptop, under your own retention and source-protection rules. That is also why offline works: once the model is downloaded, a courthouse or a jail visit needs no connection."},{"question":"Is it actually free?","answer":"Yes. Unlimited local transcription, speaker labels, 100+ languages, all three desktop platforms, $0 forever. Pro ($8/month) adds unlimited OpenWhispr Cloud speech-to-text and mobile dictation. Business ($16/user/month) adds unlimited cloud meeting recordings and priority support."}]},"reviews":{"label":"What users say","title":"Different jobs. The same instinct about their own audio.","subtitle":"From our Discord, GitHub and inboxes. None of them is a reporter, so read them for the tool, not the beat.","score":{"value":"0","label":"Interviews on a vendor's server","body":"In local mode the interview never leaves your device. There is no copy on our side, because nothing is sent to us."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for journalists","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Reporters who want free, offline transcription and don't mind setting up a local model"}]},{"parts":[{"text":"Reporters who'd rather not manage a local model and are fine with cloud transcription"}]},{"parts":[{"text":"Desks that transcribe at volume, or need unlimited cloud meeting recordings"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"The tape never leaves your laptop, at $0"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models or keys"}]},{"parts":[{"text":"Unlimited cloud meeting recordings and priority support for a desk"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"Some conversations only happen off the record. We built for those.","paragraphs":["OpenWhispr started with a simple objection: a tape of a source who asked for confidentiality shouldn't end up on a stranger's server just so you can read back what they said.","We build in public. The code is on GitHub. When we say the audio stays local, that isn't just a promise in a privacy policy. Anyone on your desk can inspect the code and verify what happens to the audio.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Your tape stays between you, your source and your laptop.","body":"Download OpenWhispr, point it at a local model, and your next interview is transcribed on your machine and nowhere else.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, we'll fix it with you in Discord. Free forever on Mac, Windows & Linux."}}},"useCaseContentCreators":{"breadcrumb":{"home":"Home","useCases":"Use Cases","contentCreators":"For Content Creators"},"metadata":{"title":"Free Dictation Software For Content Creators - OpenWhispr","description":"OpenWhispr is free, open-source dictation software for content creators. Paste your own video link or an audio file, get the transcript, talk out the rest.","keywords":["dictation software for content creators","dictation for youtubers","youtube video transcript software","transcribe your own youtube video","voice to text for creators","speech to text for video scripts"]},"schema":{"name":"OpenWhispr","applicationCategory":"Dictation software for content creators","operatingSystem":"macOS, Windows, Linux, iOS","description":"Local-first, open-source transcription and dictation software for content creators. Paste the link to a video you posted yourself, or drop in an audio file, and get the transcript in a note you can search and quote. Then talk out the title, the description and the captions and the clean text lands wherever the cursor already is. With a local model chosen the audio is transcribed on the creator's own device. Free for unlimited local transcription and dictation on macOS, Windows and Linux, and it works fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Dictation software for content creators","title":"Paste your own video link. Get the transcript.","body":"Get a transcript of your own YouTube video or audio file on your laptop for $0. Then use the same hotkey to dictate the title and description.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"A person speaks to a camera in a home studio with a ring light and laptop nearby"}},"trustBar":{"label":"Dictate into any app, including:","names":["YouTube Studio","CapCut","Notion","Patreon","Buffer","Discord"]},"pain":{"label":"What you're choosing between","title":"Everything you said is in the video. None of it is in text.","subtitle":"A transcript is what turns one upload into chapters, a newsletter and a thread. Getting one means paying per minute, or sitting through your own video with a keyboard.","cards":[{"title":"Your back catalogue is locked inside the videos.","body":"Turning last year's uploads into a newsletter means sitting through them again. Paste the link to one of your own videos and OpenWhispr pulls the audio down and transcribes it. Paste a list of your links, up to 50 at a time, and it works through the queue."},{"title":"The interview is on your disk and nowhere else.","body":"The raw recording, the voice memo, the podcast cut that never went up. Drop the audio file in and it comes back as a transcript in a note you can search. mp3, wav, m4a, flac, aac, opus, ogg and webm, straight off the disk."},{"title":"The description box, still blank at midnight.","body":"The export is done and the title, the description and the pinned comment are not. Press a hotkey in the tab you already have open and say them. Clean, punctuated text lands where the cursor is."},{"title":"Another subscription before the channel pays for itself.","body":"Editor, thumbnails, scheduler, transcripts, newsletter. Five tools billing one channel that does not cover them yet. Unlimited local transcription and dictation on OpenWhispr is $0, on every machine you own, and the free plan is not a trial."}]},"features":{"label":"From the upload to the words around it","title":"The transcript is there before you sit down to write.","subtitle":"Pick local models and both the transcription and the clean-up run on your own computer, unlimited, at no cost, in 100+ languages.","items":[{"image":{"label":"A person works at a laptop in a home editing setup with audio equipment nearby"},"title":"A transcript out of a video you already posted","body":"Paste the link to a video you posted yourself, or a list of them one per line, and OpenWhispr downloads the audio and transcribes it into a note you can search and quote. Up to 50 links at a time, 500 MB and six hours a video. Playlists and live streams are out."},{"image":{"label":"A person speaks beside a camera and laptop at a kitchen table"},"title":"The audio file that never got posted","body":"Interviews, voice memos, the podcast cut, the raw take. Drop in mp3, wav, m4a, flac, aac, opus, ogg or webm and it comes back as a transcript. On a local model there is no per-minute bill, and a file already on your disk has no size limit at all."},{"image":{"label":"A person speaks while using a laptop beside handwritten notes"},"title":"One hotkey, every tab you publish from","body":"Talk into YouTube Studio, your script doc, Notion, an email to a sponsor. The clean text pastes itself where you were already typing, with no plugin in any of them. On a Mac you grant accessibility once, and that is the part that lets it paste at your cursor."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Sent to our servers","value":"None","footnote":"with a local transcription model the audio is transcribed on your own computer"},{"label":"Cost per creator","value":"$$0","footnote":"unlimited local transcription and unlimited local dictation, on every machine you own"},{"label":"Setup takes","value":"A few extra steps","footnote":"download, pick a model, paste a link or set a hotkey"}]},"howItWorks":{"label":"Download to first transcript","title":"Three steps from your posted video to searchable text.","subtitle":"Nothing to connect to the platforms you publish on.","image":{"label":"A person speaks into a microphone at a desk with a laptop and camera gear nearby"},"steps":[{"title":"Download it free.","body":"Mac, Windows or Linux, whatever the edit rig runs. Setup offers OpenWhispr Cloud or local models, and picking local downloads them once."},{"title":"Paste the link, or drop the file.","body":"Your own video's URL, a list of them one per line, or an audio file off your disk. OpenWhispr pulls the audio down and transcribes it into a note."},{"title":"Read it, use it.","body":"Search it, quote it, cut it into chapters and a newsletter. Then hit the hotkey in YouTube Studio and talk the title and the description out."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"Why creators keep it","title":"10 reasons content creators choose OpenWhispr","subtitle":"Every one is in the app or in the repo. Check before you download anything.","items":[{"title":"Paste your own video link, get the transcript.","body":"Give OpenWhispr the YouTube URL of a video you posted, a direct audio URL, or an audio file from your disk. It transcribes the audio and can queue up to 50 links; playlists and live streams are not supported."},{"title":"It types into everything you publish from.","body":"Use the system-wide hotkey in YouTube Studio, your script doc, newsletter or sponsor email. On a Mac, one accessibility permission lets OpenWhispr paste at the cursor."},{"title":"Unlimited local transcription costs nothing.","body":"The free plan is not a trial. With a local model there is no per-minute meter on a file, no word cap on dictation, and no card to enter."},{"title":"Your niche words come out spelled right.","body":"The custom dictionary takes your channel name, your recurring guests and the vocabulary only your audience uses. Dictation hands that list to the transcription model as a hint, and to the clean-up model after it."},{"title":"The ums and the false starts come out.","body":"The clean-up model strips filler, stutters and restarts, and is told to keep your voice, your wording, your formality and your intent, with your jargon and names exactly as spoken."},{"title":"100+ languages.","body":"Dictate the description in the language your audience actually comments in. Same hotkey, same model, no separate tool."},{"title":"Location shoots with no signal.","body":"A rented studio, a hotel, a plane, a shoot in the middle of nowhere. Local models need no network at all."},{"title":"Nothing is uploaded when you keep it local.","body":"Pick a local transcription model and the audio is transcribed on your own machine, so none of it reaches us. Point the clean-up at a local model too and the text stays put as well."},{"title":"The code is public.","body":"OpenWhispr is open source on GitHub, where 5,000+ engineers have starred it. Every privacy claim on this page is a file you can open and read."},{"title":"Mac, Windows and Linux, one free download.","body":"Whatever you edit on. The same app, the same free price, and a hotkey you choose yourself."}]},"faq":{"label":"Before you install it","title":"The questions creators ask before they download it","subtitle":"Short answers, and where in the app or the repo you can check them.","items":[{"question":"Can I get a transcript of a video I already published?","answer":"Paste the link to your own video. OpenWhispr accepts a YouTube URL or a direct audio file URL, downloads the audio and transcribes it into a note. Paste a list of your links, one per line and up to 50 at a time, and it works through them as a queue. Playlists and live streams are not supported, and the download caps at 500 MB and six hours a video. On a local model that cap is the only one. On OpenWhispr Cloud the free plan transcribes files up to 25 MB, and past that it is a paid plan."},{"question":"Can I upload the video file itself?","answer":"Not the video file. The upload panel takes audio: mp3, wav, m4a, webm, ogg, oga, flac, aac and opus. For something you already posted, paste its link instead and OpenWhispr pulls the audio down for you. For a cut that is not posted anywhere yet, export the audio out of your editor and drop that in."},{"question":"Is my audio uploaded anywhere?","answer":"Not if you pick a local transcription model: the audio is transcribed on your own computer and nothing is sent. One thing worth knowing, because it is two settings and not one. Transcription and the AI clean-up are chosen separately, and if you are signed in the clean-up still defaults to OpenWhispr Cloud. On that default the transcript text, never the audio, is sent to us to be tidied. Set the clean-up model to a local one too and neither leaves the machine."},{"question":"Is it really free, or is the free plan a trial?","answer":"Free is free, with no card. Unlimited local dictation and unlimited local transcription are $0 on every plan, forever. Pro ($8/month) makes OpenWhispr Cloud transcription unlimited, and Business ($16/month) adds unlimited cloud meeting recordings and agent mode. Nothing on the free plan expires."},{"question":"Will it get my channel name and my niche jargon right?","answer":"Add them to the custom dictionary once. Your channel name, your recurring guests, the in-jokes and the product names your audience uses. Dictation passes that list to the transcription model as a hint, and the clean-up model gets the same list. One exception worth knowing: the local Parakeet model does not take the hint, so use a local Whisper model if the dictionary matters."},{"question":"Does it work on a shoot with no WiFi?","answer":"Yes, once the local models are downloaded. Local transcription and a local AI model both run on your own machine with no network, so a hotel, a plane or a location with no signal is the same as your desk. Cloud transcription is the option that needs a connection."}]},"reviews":{"label":"What people tell us","title":"They tried it on their own machine and stopped typing.","subtitle":"From our Discord, GitHub and inbox, word for word. None of them runs a channel, so read them for the tool.","score":{"value":"0","label":"Plugins to install in YouTube Studio","body":"The hotkey works the same in a browser tab, your edit suite and your newsletter draft. Nothing to connect, nothing to log into, no integration to keep alive."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for content creators","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"One-person channels that want unlimited local transcription and dictation at no cost and will set up a local model"}]},{"parts":[{"text":"Creators who would rather pay $8 than look after a local model"}]},{"parts":[{"text":"Channels where an editor, a writer and a producer all touch the same content"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Transcripts of your own videos, and everything you write around them, for nothing, with the audio staying on your machine"}]},{"parts":[{"text":"Cloud transcription that never runs out, with no models or API keys to look after"}]},{"parts":[{"text":"Unlimited cloud meeting recordings and agent mode for everyone on the channel"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Why we built it","title":"The fastest way to sound like yourself is to say it out loud.","paragraphs":["OpenWhispr started with an objection: talking is how people are fluent and typing is how they get stiff, and the software that turns one into the other kept routing everything through somebody else's server.","We build in public. Every release and every line of the app is on GitHub. So when we say the dictation can stay on your machine, you do not have to take our word for it. You can open the file and read it.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"One of OpenWhispr's founders"}},"finalCta":{"tagline":"$$0 · open source · no plugin to install","title":"Get the transcript, then write around it.","body":"Download OpenWhispr, pick a local model at setup, and paste the link to the video you posted this morning. No card, no per-minute bill, no second subscription.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free forever on Mac, Windows & Linux."}}},"useCasePodcasters":{"breadcrumb":{"home":"Home","useCases":"Use Cases","podcasters":"For Podcasters"},"metadata":{"title":"Best Speech-to-Text Software For Podcasting - OpenWhispr","description":"OpenWhispr is free, open-source speech-to-text for podcasters. Transcribe unlimited episode audio on your own machine and never pay per minute again.","keywords":["speech to text for podcasting","podcast transcription software","free podcast transcription","dictation software for podcasters","podcast show notes software","offline podcast transcription","open source transcription software"]},"schema":{"name":"OpenWhispr","applicationCategory":"Podcast transcription software","operatingSystem":"macOS, Windows, Linux, iOS","description":"Local-first, open-source speech-to-text for podcasters. With a local model chosen, episodes and show notes are transcribed on the creator's own device. Free for unlimited local transcription and dictation, and works fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Speech-to-text software for podcasters","title":"Record the episode. Get the transcript. Pay nothing per minute.","body":"Turn each recorded episode into a transcript, then dictate show notes, descriptions and posts into any app. Choose local transcription for offline work, or Cloud for high accuracy and speed without downloading models.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Person speaking into a studio microphone beside a laptop and handwritten notes"}},"trustBar":{"label":"Write show notes in:","names":["Notion","Buzzsprout","Transistor","Spotify for Creators","Substack","YouTube Studio"]},"pain":{"label":"The problem with per-minute transcription","title":"Transcription shouldn't cost more than your hosting.","subtitle":"A transcript and show notes are part of shipping many episodes. Many tools bill by the minute or cap usage by the month.","cards":[{"title":"Paying by the minute to transcribe your own voice.","body":"A weekly one-hour show, metered per minute, turns into a real monthly bill that never ends. OpenWhispr transcribes on your machine: unlimited hours, $0, and the Free plan is not a trial."},{"title":"Show notes take longer than the episode.","body":"You talked for an hour, then typed for two. Press a hotkey and speak the summary, the guest bio and the links straight into your host or your docs. Clean, punctuated text, at speaking speed."},{"title":"Auto-captions butcher your guests' names.","body":"A transcript that mangles your guest's name or your show's title is a correction job. Add the name to OpenWhispr's dictionary to improve how it appears in dictated notes and AI-drafted summaries."},{"title":"Pulling one quote means scrubbing the whole episode.","body":"Clips, quote graphics and newsletters all start with finding the line. With a transcript on your machine you search text, copy the quote and move on. No scrubbing."}]},"features":{"label":"What you get","title":"Transcripts, show notes, summaries. No meter on any of them.","subtitle":"Make transcripts and show notes on your own machine in 100+ languages, without adding another monthly transcription bill.","items":[{"image":{"label":"Person monitoring audio at a microphone with a laptop and recording interface"},"title":"Unlimited local transcription","body":"Import the episode you already cut, and it is transcribed on your device with local Whisper-class models. No per-minute fees or monthly hour caps for local transcription, with nothing uploaded in local mode."},{"image":{"label":"Person speaking into a desktop microphone while working on a laptop"},"title":"Show notes at speaking speed","body":"Press a hotkey and speak your summary, description or newsletter straight into Notion, Buzzsprout, Substack, wherever your cursor is."},{"image":{"label":"Two people recording with microphones while one reviews notes beside a laptop"},"title":"A dictionary that knows your show","body":"Add your show's title, recurring guests and niche vocabulary to the custom dictionary. It improves how those names and terms are handled in dictation and AI-drafted notes."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Unreleased audio uploaded","value":"None","footnote":"in local mode the episode never leaves your machine"},{"label":"Cost per episode","value":"$$0","footnote":"unlimited local transcription and dictation, every episode, forever"},{"label":"Transcription mode","value":"Local or Cloud","footnote":"Local runs on your device. Cloud prioritizes accuracy and speed with no model download."}]},"howItWorks":{"label":"How it works","title":"A transcript and show notes before you close the editor.","subtitle":"Choose local transcription when you need offline work or want audio to stay on your machine. Choose Cloud for high accuracy and speed without downloading local models.","image":{"label":"Person holding headphones beside a closed laptop and handwritten notes"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model during setup."},{"title":"Record or dictate.","body":"Record a session for a local transcript, or press the hotkey and speak your show notes into any app."},{"title":"Review, paste, publish.","body":"Edit the text, paste it into your host or your docs, and ship the episode. You control all of it."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons podcasters choose OpenWhispr","subtitle":"Transcription cost recurs with every episode. These are the practical details to compare.","items":[{"title":"Unlimited local transcription","body":"No per-minute fees or monthly hour caps for local transcription on the Free plan."},{"title":"Free for your whole show","body":"Unlimited local dictation and transcription at $0: you, your co-host, your editor. No per-seat licence."},{"title":"Dictate straight into your host","body":"A system-wide hotkey drops clean text wherever your cursor is: Buzzsprout, Notion, Substack, YouTube Studio."},{"title":"Summaries drafted, not just transcribed","body":"Connect your own AI key, or run a local model, to turn a transcript into an episode summary or description."},{"title":"A dictionary for your guests' names","body":"Teach it the show title, recurring guests and niche terms once. The custom dictionary improves how it handles them in dictated notes and drafted summaries."},{"title":"Works fully offline","body":"Transcribe on the flight home from the conference. Local models never need a signal."},{"title":"Pull quotes without scrubbing","body":"The transcript is text on your machine. Search it, copy the line, make the clip."},{"title":"Unreleased audio stays on your machine","body":"In local mode nothing is uploaded. Embargoed interviews and unshipped episodes never touch a server."},{"title":"Open source, auditable by anyone","body":"OpenWhispr is open source under the MIT License."},{"title":"100+ languages","body":"Record in the language your audience actually speaks, transcribed locally."}]},"faq":{"label":"Before you download","title":"The questions podcasters ask first.","subtitle":"Straight answers on cost, accuracy and privacy. If yours isn't here, ask us in Discord.","items":[{"question":"What's the best speech-to-text software for podcasting?","answer":"OpenWhispr is a strong option if you want to transcribe recorded episodes locally without per-minute fees. It can turn imported audio into a transcript and lets you dictate show notes wherever your cursor is. Its open-source code makes the workflow inspectable. Trying it costs a download."},{"question":"Is it really free?","answer":"Yes. Local dictation and transcription are included at $0 on macOS, Windows and Linux. The Free plan also includes 2,000 words a week of OpenWhispr Cloud transcription and 5 hours a month of meeting recordings. Pro is $8/month with unlimited Cloud transcription and 20 meeting-recording hours. Business includes unlimited meeting recordings."},{"question":"Will it get my guests' names and niche terms right?","answer":"OpenWhispr supports 100+ languages. Its custom dictionary lets you add your show's title, guest names and niche vocabulary to improve how they appear in dictation and AI-drafted show notes."},{"question":"Does it transcribe my recorded episodes, or only what I say live?","answer":"Both. Import an episode you've already recorded, or record a session in the app, and the transcript is made on your device. Press the hotkey to dictate show notes into any app, and connect your own AI key, or a local model, to draft summaries from the transcript."},{"question":"Does it work offline?","answer":"Yes, once the models are downloaded at setup. After that they run entirely offline. Transcribe an interview on the plane and publish when you land."},{"question":"Do unreleased episodes stay private?","answer":"In local mode, the audio stays on your machine. OpenWhispr is open source under the MIT License. Cloud transcription is optional: choose local for offline work or Cloud for high accuracy and speed without downloading models."}]},"reviews":{"label":"What users say","title":"Built for people who'd rather talk than type.","subtitle":"What people say about OpenWhispr.","score":{"value":"0","label":"Invoices for your own words","body":"Local transcription runs on your machine. There is no meter, no hour cap, and no bill at the end of the month."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for podcasters","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Get unlimited Cloud transcription"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Podcasters who want free, unlimited local transcription and are fine with a quick local setup"}]},{"parts":[{"text":"Podcasters who want easy cloud transcription and mobile dictation"}]},{"parts":[{"text":"Networks and teams producing shows at volume"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited local transcription, with your audio staying on your machine"}]},{"parts":[{"text":"Unlimited cloud speech-to-text without managing local models or keys"}]},{"parts":[{"text":"Unlimited cloud recordings plus advanced workflow features for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"Your words are the show. You shouldn't rent them back.","paragraphs":["OpenWhispr started with a simple objection: turning your own voice into text shouldn't cost by the minute or pass through someone else's server.","We build in public on GitHub. Local transcription is not just a privacy-policy promise. It is a property of the code you can inspect.","And if anything breaks the night before an episode drops, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Stop paying by the minute for your own voice.","body":"Download OpenWhispr, pick a local model at setup, and your next episode gets transcribed, summarized and shipped on your own computer.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free desktop app for macOS, Windows and Linux."}}},"useCaseProductTeams":{"breadcrumb":{"home":"Home","useCases":"Use Cases","productTeams":"For Product Teams"},"metadata":{"title":"Free AI Meeting Notes For Product Teams, No Bot - OpenWhispr","description":"OpenWhispr is free, open-source AI meeting notes for product teams. No bot joins the call, and with a local model the roadmap never leaves your machine.","keywords":["AI meeting notes","meeting notes for product teams","local meeting transcription","private meeting notes","bot-free notetaker","offline meeting notes"]},"schema":{"name":"OpenWhispr","applicationCategory":"Meeting notes software","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source AI meeting notes for product teams. With a local model the call is transcribed on the user's own machine with no bot joining, and written up as decisions, action items and owners. Local meeting notes are unlimited and free, and work fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"AI meeting notes for product teams","title":"Run the meeting. Get the decisions written up.","body":"Get a clean meeting write-up with decisions and next steps. OpenWhispr captures the call on your own computer; pick a local model and the recording never reaches a server, ours or anyone's.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Two people listen to a colleague during a meeting around a table"}},"trustBar":{"label":"Paste the notes into:","names":["Linear","Jira","Notion","Confluence","Slack","Productboard"]},"pain":{"label":"The trade-offs product teams notice","title":"Your roadmap shouldn't live on a notetaker's server.","subtitle":"You want the decisions written down. You don't want a bot in every call, or your strategy sitting on a transcription vendor's cloud.","cards":[{"title":"A bot shows up to every meeting.","body":"Meeting bots can make a sensitive discovery call feel different. OpenWhispr captures your computer's audio, so it does not join the call."},{"title":"Your strategy on a vendor's server.","body":"Roadmap, pricing, competitive analysis, reorg talk: product meetings can cover sensitive work. Choose a local model when you want transcription to run on your device."},{"title":"Someone stops contributing to take notes.","body":"The PM types instead of facilitating, and the decision still gets lost. OpenWhispr writes up the decisions, action items and owners, so everyone stays in the discussion."},{"title":"Per-seat pricing for the whole org.","body":"Choose the mode and plan that fit the team. Local transcription is unlimited on every plan; Free includes 5 hours of meeting recordings a month, Pro adds unlimited Cloud transcription and 20 recording hours, and Business includes unlimited meeting recordings."}]},"features":{"label":"What you get","title":"AI meeting notes for product teams, made on your own laptop.","subtitle":"Pick local models and the roadmap discussion is transcribed and summarized on your machine.","items":[{"image":{"label":"A person gestures while speaking to a colleague across a table"},"title":"Captures the call from your computer, no bot","body":"Listens to your computer's audio on Zoom, Meet, Teams or a Slack huddle, or to the mic for a standup in the room. Transcribed on your device with a local model, and nothing joins the call."},{"image":{"label":"Two colleagues review a laptop together during a group discussion"},"title":"Decisions, action items and owners","body":"Turns your rough notes and the transcript into a structured write-up: what was decided, what happens next, who owns it. You review before it's shared."},{"image":{"label":"A person gestures while working on a laptop at a table"},"title":"Paste it where the work lives","body":"Copy the clean write-up into Linear, Jira, Notion or Slack. It's text, so there's nothing to connect to your tracker."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Roadmap calls sent to us","value":"None","footnote":"pick a local model at setup and there is no upload step in the code"},{"label":"Cost per person on the call","value":"$$0","footnote":"unlimited local meeting notes, the whole org, forever"},{"label":"Setup takes","value":"5 min","footnote":"from download to your first meeting written up"}]},"howItWorks":{"label":"How it works","title":"Leave the call with the decisions already written down.","subtitle":"Download the app, choose a model, then start the call.","image":{"label":"A person holding a tablet speaks in an open office"},"steps":[{"title":"Download for free.","body":"macOS 14.2+, Windows or Linux. Pick a local model during setup."},{"title":"Start the meeting.","body":"Open Zoom, Meet or Teams as usual. OpenWhispr captures your computer's audio and transcribes it on your device."},{"title":"Review, paste, move on.","body":"Hit Generate Notes, edit the write-up and paste it where the work lives. With a local model, transcription runs on your device."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons product teams choose OpenWhispr","subtitle":"Choosing a notetaker is a security decision before it's a note-taking one. Here are the facts, and your security team can verify the privacy ones line by line in the repo.","items":[{"title":"The roadmap call never reaches a server","body":"Pick a local model and transcription runs on your machine. The open-source code shows no upload step on that path."},{"title":"No bot in the call","body":"It captures your computer's audio, so nothing appears in the participant list. Your customer never sees a notetaker."},{"title":"Every seat is free, including the quiet ones","body":"Unlimited local meeting notes at $0, for everyone who ever sits in a call. Only the cloud side is metered."},{"title":"Works with any meeting app","body":"Zoom, Meet, Teams, Slack huddles, Webex, a standup in the room. It captures audio, not a specific app."},{"title":"Decisions and owners, not just transcripts","body":"The write-up is structured with decisions, action items and follow-ups. Where an owner is clear in the transcript, it uses that speaker label. You don't write the prompt."},{"title":"No meeting-app integration to approve","body":"No meeting-app plugin or bot account to approve. It's a desktop app that listens to your computer's audio."},{"title":"The recording is deleted after the meeting","body":"Temporary diarization audio is removed after processing. You control the recordings and transcripts saved in the app."},{"title":"Read the code before you pick the tool","body":"OpenWhispr is open source. Your security team can read the code on GitHub."},{"title":"No signal, same write-up","body":"A standup in a room with bad WiFi, a customer site, a plane. Local models need no signal."},{"title":"Who said what, worked out on your machine","body":"Speaker labels are computed on your own device, free on every plan. The write-up says who owns cutting scope, rather than merely that someone said it."}]},"faq":{"label":"Before you download","title":"The questions product teams ask first.","subtitle":"Straight answers on bots, your data and cost. If yours isn't here, ask us in Discord.","items":[{"question":"Does the meeting audio ever leave my device?","answer":"Not in local mode. Pick a local model during setup and transcription runs on your machine with no upload step. Cloud transcription is optional, and we never train on your data."},{"question":"Does a bot join the call?","answer":"No. OpenWhispr captures your computer's audio, so nothing appears in the participant list. Zoom, Meet, Teams, a Slack huddle or a standup in the room all work the same way because it listens instead of joining. With a local model, transcription runs on your device. Tell people you're recording, same as you would with a bot."},{"question":"Will it get our feature names and acronyms right?","answer":"Add them to the custom dictionary, free on every plan. The write-up is told to use those exact spellings, so codenames and feature names come out right in the notes you paste."},{"question":"Do the AI meeting notes capture decisions and owners, not just a transcript?","answer":"Yes. Jot rough notes during the call, then run the write-up when it ends. AI combines them with the transcript into decisions, next steps and action items. Where an owner is clear, it uses that speaker label. You review before anything is shared."},{"question":"Will it work for an in-person standup with no call?","answer":"Yes. It records from your mic, and local models run entirely offline. Capture the standup, then paste the write-up into Linear or Notion when you're done."},{"question":"Is it actually free?","answer":"Yes. Unlimited local meeting notes, 100+ languages, all three desktop platforms: $0 forever, for every seat. Only the cloud side is metered: cloud meeting recordings run 5 hours a month free, 20 on Pro ($8/month), unlimited on Business ($16/user/month)."}]},"reviews":{"label":"What users say","title":"Built by people who don't want a bot in their meetings either.","subtitle":"From our Discord, GitHub and inboxes, the same instinct that matters when the roadmap's on screen.","score":{"value":"0","label":"Roadmap calls on a vendor's server","body":"In local mode, transcription runs on your device. Cloud transcription is optional, and its audio is discarded after transcription."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for product teams","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Product teams who want private meeting notes and are fine with a quick local setup"}]},{"parts":[{"text":"Product teams who want easy, reliable cloud transcription"}]},{"parts":[{"text":"Product orgs in calls all day"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Free local meeting notes, with the audio staying on your device"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models or keys"}]},{"parts":[{"text":"Unlimited recordings plus advanced workflow features for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for top accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","tooltip":"Cloud meeting recordings. Recording and transcribing meetings with a local model is unlimited on every plan.","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"Some meetings decide the roadmap. We built for those.","paragraphs":["OpenWhispr started with a simple objection: the meeting where you decide the roadmap shouldn't have to pass through a stranger's server to become a write-up.","OpenWhispr is open source. Your security team can inspect how local transcription works before putting it in roadmap meetings.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Keep the call on your laptop.","body":"Download OpenWhispr, pick a local model, and your next meeting gets transcribed, written up and ready to paste on your computer.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, we'll fix it with you in Discord. Free forever on Mac, Windows & Linux."}}},"useCaseStudents":{"breadcrumb":{"home":"Home","useCases":"Use Cases","students":"For Students"},"metadata":{"title":"Free Speech-to-Text Software For Students - OpenWhispr","description":"OpenWhispr is free, open-source speech-to-text software for students. Record the lecture, get the notes written up on your own laptop, at $0 with no card.","keywords":["speech to text software for students","free dictation software for students","lecture transcription app","record lectures and get notes","speech to text for university students","free speech to text for students","student lecture notes app"]},"schema":{"name":"OpenWhispr","applicationCategory":"Speech-to-text software for students","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source speech-to-text software for students. Lectures are recorded and transcribed on the student's own laptop, written up as structured notes with a local AI model, and essays are dictated the same way. Unlimited local transcription is free, and it works fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Speech-to-text software for students","title":"Hit record in the lecture. Get the notes written up.","body":"Turn a lecture into readable notes on your laptop. OpenWhispr records the room or the stream, and one tap on Generate Notes turns the transcript into headings and key points for $0 with a local model.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"A person smiling with a laptop while relaxing on a sofa."}},"trustBar":{"label":"Record lectures from:","names":["Zoom","Microsoft Teams","Panopto","Echo360","Google Meet","YouTube"]},"pain":{"label":"The problem with lecture notes","title":"You cannot listen and write at the same time.","subtitle":"A lecture runs at the speed of speech. Writing notes pulls your attention away from following the argument.","cards":[{"title":"The notes you take are the lecture you missed.","body":"Every sentence you write down is a sentence you stop listening to. Let the recording take the words and keep your attention on the whiteboard instead."},{"title":"A two-hour recording takes two hours to replay.","body":"Audio is hard to revisit when you cannot skim it. OpenWhispr hands back text and structured notes you can scan before replaying anything."},{"title":"Pick the right mode for exam month.","body":"Use local transcription for unlimited on-device transcription at $0, or Cloud for top accuracy and speed without downloading a model. Free includes 2,000 words of Cloud transcription a week and 5 hours of meeting recordings a month; Pro adds unlimited Cloud transcription and 20 recording hours."},{"title":"Your lectures, sitting on a company's server.","body":"Most transcription tools upload the audio to process it. In local mode OpenWhispr keeps the recording on your laptop, and the code is public on GitHub so you can check that yourself."}]},"features":{"label":"What you get","title":"The lecture, written up before you get home.","subtitle":"Recording is the easy half. What makes a lecture useful in week nine is text you can search and notes you can read.","items":[{"image":{"label":"A person listening beside a laptop in a tiered lecture hall."},"title":"Records the room or the stream","body":"In the lecture hall it records through the mic. For a Zoom, Teams or Panopto lecture it captures the audio playing on your laptop. Mac, Windows and Linux."},{"image":{"label":"A person sitting beside a laptop in a campus lounge."},"title":"Notes, not a wall of transcript","body":"The transcript goes through a write-up step that returns headings, key points and action items in markdown. With a local AI model that step runs on your laptop too."},{"image":{"label":"A person speaking while working on a laptop at a library desk."},"title":"Dictation for everything after","body":"Press a hotkey anywhere and talk. Clean text lands in Google Docs, Word or Overleaf, so the essay that follows the lecture gets said out loud instead of typed."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Lecture audio sent to us","value":"None","footnote":"in local mode, recording and transcription run on your laptop"},{"label":"Cost per student","value":"$$0","footnote":"unlimited local transcription and write-ups, no card, no expiry"},{"label":"Setup takes","value":"5 min","footnote":"from download to your first recorded lecture"}]},"howItWorks":{"label":"How it works","title":"Three steps, and you leave with the write-up.","subtitle":"The whole setup is one download and one model choice. There is nothing to sign up for.","image":{"label":"A person wearing an earbud beside a laptop and handwritten notes."},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model during setup and the whole thing runs offline."},{"title":"Hit record when the lecture starts.","body":"It takes the room through your mic, or the stream through your laptop's audio. Then you just listen."},{"title":"Tap Generate Notes.","body":"The transcript comes back as headings, key points and action items. Search those in revision week instead of scrubbing audio."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons students choose OpenWhispr","subtitle":"On a student budget, cost comes first.","items":[{"title":"Free forever, no card details","body":"Unlimited local transcription and note write-ups at $0, with no trial or semester limit."},{"title":"No bot joins the call","body":"OpenWhispr captures audio directly from your system; it does not join meetings as a participant."},{"title":"Your own notes go in the same file","body":"Type a line during the lecture and the write-up folds it in with what was said."},{"title":"Works with no WiFi","body":"Local models run offline. A basement lecture theatre, the train home, a flight: it still records and still transcribes."},{"title":"Recordings stay on your laptop","body":"In local mode the audio is transcribed on your device, and the open-source code shows the local path never uploads it."},{"title":"Searchable in revision week","body":"Search a term in your write-ups instead of scrubbing audio."},{"title":"Use a custom dictionary in the write-up","body":"Add lecturer names, module codes and technical terms to the custom dictionary to give the write-up the spellings you use."},{"title":"100+ languages","body":"Sit a lecture in your second language and read the notes in it, or dictate your essay in your first."},{"title":"Upgrade for unlimited cloud transcription","body":"The free plan includes 2,000 words a week of OpenWhispr Cloud transcription. Pro is $8/month for unlimited Cloud transcription. Students get 40% off for 12 months; email support@openwhispr.com with proof of enrolment for a code."},{"title":"Open source, under the MIT license","body":"Read the code before you install it. No privacy policy to take on faith."}]},"faq":{"label":"Before you download","title":"The questions students ask first.","subtitle":"Straight answers on cost, permission and what your university will think. If yours is not here, ask us in Discord.","items":[{"question":"Is it actually free, or free for a month?","answer":"Actually free. Local transcription, local note write-ups, 100+ languages and all three desktop platforms cost $0 forever, with no card details. The free plan also includes 2,000 words a week of OpenWhispr Cloud transcription and 5 hours a month of cloud meeting recordings. Pro is $8/month if you want unlimited Cloud transcription. Students get 40% off for 12 months; email support@openwhispr.com with proof of enrolment for a code."},{"question":"Am I allowed to record a lecture?","answer":"That is your university's call, not ours, and it varies by institution and by country. Most have a policy on recording teaching, often tied to accessibility. Check yours, and ask the lecturer. OpenWhispr does not record anything until you press the button."},{"question":"Does it work for online lectures as well as in-person ones?","answer":"Yes. In a lecture hall it records through your mic. For a Zoom, Teams, Google Meet or Panopto lecture it captures the audio playing on your laptop, on Windows, Linux and macOS 14.2 or newer. Already have the file, from your phone or a lecture-capture download? Drop it in and it transcribes that too, with no size limit on a local model."},{"question":"Will it get my subject's terminology right?","answer":"It runs Whisper-class models trained on a wide range of accents and 100+ languages. For the words a general model has never seen, add them to the custom dictionary in Settings: lecturer names, module codes and technical terms. Those spellings carry into your dictation and into the write-up of your notes."},{"question":"Is using the AI write-up cheating?","answer":"Transcription is not: the words are the ones your lecturer said, and reading them back is no different from reading your own notes. The write-up step is different, because a model summarises the transcript for you, so check your university's rules on AI before any of it goes into work you hand in."},{"question":"Where does my recording actually go?","answer":"It depends on the mode you pick during setup. OpenWhispr Cloud is the default: audio is sent only to transcribe it, and is not stored afterwards. Choose a local model instead and the answer is nowhere. It is transcribed on your laptop, and the open-source code shows the local path has no upload step. You can switch in Settings any time."}]},"reviews":{"label":"What users say","title":"Built by people who take notes in bad WiFi too.","subtitle":"From our Discord, GitHub and inboxes.","score":{"value":"0","label":"Lecture recordings kept on our servers","body":"In local mode the recording never leaves your laptop. On cloud, audio is sent only to transcribe it and is not stored after that."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for students","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Students recording lectures on their own laptop, with unlimited local transcription and notes"}]},{"parts":[{"text":"Students who want cloud transcription that just works, with no local models to download"}]},{"parts":[{"text":"Research groups and societies recording at volume"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"No meter and no card on the local side, with the recording staying on your laptop"}]},{"parts":[{"text":"Unlimited cloud speech-to-text without managing local models or keys"}]},{"parts":[{"text":"Unlimited cloud meeting recordings plus speaker labels for group work"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"A lecture you can read shouldn't need a subscription.","paragraphs":["OpenWhispr is for students who want to turn lectures into readable notes without a subscription.","We build in public. The code and releases are on GitHub, where 5,000+ engineers have starred the project. When we say the recording stays on your laptop, that's not a promise in a privacy policy. It's a property of the code, and you do not have to take our word for it.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Turn your 9am into notes with one tap.","body":"Download OpenWhispr, hit record when the lecture starts, and read the notes on the way home. On your laptop, at $0, with no card details.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Free forever on Mac, Windows & Linux."}}},"useCaseSpanishVeterinary":{"breadcrumb":{"home":"Home","useCases":"Use Cases","spanishVeterinary":"For Spanish-speaking vets"},"metadata":{"title":"Free Spanish Veterinary Dictation Software - OpenWhispr","description":"OpenWhispr turns Spanish veterinary dictation into text for QVET, Provet Cloud, or any clinical record, with free local models available.","keywords":["spanish veterinary dictation software","veterinary dictation software spanish","spanish vet dictation","veterinary transcription spanish","dictado veterinario en español","software de dictado veterinario"]},"schema":{"name":"OpenWhispr","applicationCategory":"Spanish veterinary dictation software","operatingSystem":"macOS, Windows, Linux","description":"Free, open-source dictation software for veterinary teams who document in Spanish. Select Spanish, dictate, review the transcript, and paste it into the clinical record you already use.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Spanish veterinary dictation software","title":"Dictate in Spanish. Get text ready for your PIMS.","body":"Put the cursor in the historia clínica, press the hotkey, and dictate in Spanish. Review the text before you file it.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Veterinarian working at a clinic computer beside a dog"}},"trustBar":{"label":"Use alongside:","names":["QVET","Provet Cloud","GVET","WinVet","Wirevet","Microsoft Word"]},"pain":{"label":"The problem with Spanish clinical documentation","title":"The consult ends. The historia clínica is still blank.","subtitle":"You still owe the anamnesis, exploración, diagnóstico, tratamiento, informe de alta, and sometimes a prescription. Every line still needs your review.","cards":[{"title":"The Spanish note starts after the patient leaves.","body":"The findings are fresh, but the historia clínica is empty. Dictate the note while you still remember the sequence, then review it before filing."},{"title":"Drug names, doses, and accents need a final check.","body":"Select Spanish before you dictate. Store recurring patient names, drug names, and clinic abbreviations in the custom dictionary, then check every dose and clinical term before filing."},{"title":"Your PIMS does not need a new integration.","body":"OpenWhispr does not connect to QVET, Provet Cloud, GVET, WinVet, Wirevet, or Word. It produces text you can paste into a standard editable field, so you keep the workflow you already use."},{"title":"A prescription still needs a veterinarian.","body":"Dictation can type the details. It cannot examine the animal, choose treatment, validate required fields, or sign the prescription. You stay responsible for reviewing the record in the approved system."}]},"features":{"label":"What you get","title":"Pick Spanish. Keep your PIMS.","subtitle":"Select Spanish, press the hotkey, then review what lands in the field. No migration and no new record system.","items":[{"image":{"label":"Veterinarian examining a cat beside an open laptop"},"title":"Spanish is an actual speech setting","body":"Select Spanish during setup or in settings. The shipped app lists Spanish for local Whisper and Parakeet models as well as supported cloud transcription."},{"image":{"label":"Veterinarian using a laptop beside a cat in a carrier"},"title":"From the microphone to a PIMS-ready transcript","body":"Use the system-wide hotkey in a clinical history, discharge, referral, email, or Word document. Automatic paste is a setting and needs the required OS permission. The clipboard option keeps text available for manual paste."},{"image":{"label":"Man using a laptop in a stable tack room"},"title":"A local option when the connection drops","body":"Download a local model once, then Spanish dictation can run on the computer without internet. Local models are unlimited and included on every plan."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"New PIMS required","value":"None","footnote":"paste into the veterinary record system you already use"},{"label":"Cost to start","value":"$$0","footnote":"unlimited local models are included on the Free plan"},{"label":"Setup takes","value":"5 min","footnote":"from download to the first Spanish dictation"}]},"howItWorks":{"label":"How it works","title":"From spoken Spanish to the right clinical field.","subtitle":"OpenWhispr handles the dictation step. You keep the PIMS, the clinical judgment, and the final review.","image":{"label":"Veterinarian using a laptop while comforting a dog on an exam table"},"steps":[{"title":"Download for free.","body":"Install on Mac, Windows, or Linux. Select Spanish as a spoken language."},{"title":"Put the cursor in the record.","body":"Open the relevant field in QVET, Provet Cloud, GVET, WinVet, Wirevet, or Word, then press the hotkey and speak."},{"title":"Review every clinical term.","body":"Check patient names, drug names, doses, accents, and required fields before you save or sign anything."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons Spanish-speaking vets choose OpenWhispr","subtitle":"Spanish is a speech setting in the app. The hotkey, local models, dictionary, and automatic-paste or clipboard workflow stay the same across your desktop.","items":[{"title":"Spanish appears in the language picker","body":"Spanish is a selectable spoken language in the shipped onboarding and settings flow."},{"title":"Two local model families list Spanish","body":"The app's model registry includes Spanish for local Whisper and multilingual Parakeet transcription."},{"title":"No dedicated PIMS integration required","body":"If a PIMS field accepts pasted text, use automatic paste or the clipboard. No dedicated QVET, Provet Cloud, GVET, WinVet, or Wirevet integration is claimed."},{"title":"Use the sections your team already knows","body":"Dictate anamnesis, exploración, diagnóstico, tratamiento, evolución, or informe de alta without changing the record's structure."},{"title":"Keep recurring terms in one list","body":"The custom dictionary stores the patient names, drug names, procedures, and abbreviations you add. You still verify each result."},{"title":"Local Spanish dictation can work offline","body":"After the local model is downloaded, transcription runs on the computer without an internet connection."},{"title":"Unlimited local models cost $0","body":"The Free plan includes unlimited local models, Spanish support, and the custom dictionary."},{"title":"Mac, Windows, and Linux","body":"Use the same Spanish dictation workflow across all three supported desktop platforms."},{"title":"The cleanup instruction is set for Spanish","body":"The instruction asks for Spanish grammar, spelling, and punctuation while keeping naturally spoken English technical terms. Review the result before filing."},{"title":"It does not pretend to prescribe","body":"OpenWhispr types what you dictate. The veterinarian still examines the animal, chooses treatment, checks the required fields, and signs in the approved system."}]},"faq":{"label":"Before you download","title":"The questions Spanish-speaking vets ask first.","subtitle":"Straight answers on Spanish support, terminology, PIMS workflow, offline use, prescriptions, and cost.","items":[{"question":"Does OpenWhispr actually support Spanish dictation?","answer":"Yes. Spanish is a selectable speech language in the shipped app. The code lists it for local Whisper, local multilingual Parakeet, and supported cloud transcription. This page does not claim equal accuracy for every accent or vocabulary."},{"question":"Will it recognize every veterinary term and drug name?","answer":"No blanket accuracy claim. The closed product sources do not document veterinary-specific or Spanish-veterinary tuning for the included local models. Add recurring terms to the custom dictionary, then review patient names, drugs, doses, abbreviations, and diagnoses before filing."},{"question":"Does it work with QVET, Provet Cloud, GVET, WinVet, and Wirevet?","answer":"No dedicated integration is claimed. If the field accepts pasted text, press the system-wide hotkey and dictate. Automatic paste needs the relevant setting and OS permission; the clipboard option supports manual paste."},{"question":"Can Spanish veterinary dictation work without internet?","answer":"Yes, in local mode after you download a model. Local models run on the computer with no internet required. Cloud transcription still needs a connection."},{"question":"Can OpenWhispr create or submit a veterinary prescription?","answer":"No. It can type the details you dictate into the active field, but it does not prescribe, validate a jurisdiction's required fields, submit to a prescription platform, or sign on your behalf. The veterinarian reviews and completes that workflow."},{"question":"Is it actually free?","answer":"Yes. The Free plan is $0 and includes unlimited local models, Spanish support, the custom dictionary, and Mac, Windows, and Linux. OpenWhispr Cloud on Free is limited to 2,000 words a week; Pro is $8 a month and makes cloud transcription unlimited."}]},"reviews":{"label":"What users say","title":"Real users. Their words, unchanged.","subtitle":"These reviews are not presented as veterinary endorsements. They describe the dictation experience the page is built on.","score":{"value":"ES","label":"Selectable speech language","body":"Spanish appears in the shipped language registry for local and cloud transcription."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for Spanish-speaking veterinary teams","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Vets and techs who want unlimited local Spanish dictation and are fine with a quick model download"}]},{"parts":[{"text":"Vets who want unlimited OpenWhispr Cloud transcription in Spanish"}]},{"parts":[{"text":"Veterinary teams that want unlimited recordings and advanced workflows"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Spanish dictation with local models at $0"}]},{"parts":[{"text":"Unlimited cloud transcription without downloading a local model"}]},{"parts":[{"text":"Unlimited recordings plus Agent mode and chat over your data"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" stores terms you add"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" stores terms you add"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" stores terms you add"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Why we built it","title":"Spanish should be a setting, not a workaround.","paragraphs":["OpenWhispr gives veterinary teams a direct path from spoken Spanish to the clinical field already open on screen. The app handles dictation. The veterinarian keeps the terminology review, clinical judgment, and signature.","We build in public. Every release and every line of the app is on GitHub, where 5,000+ engineers have starred and forked the project. So when we say the dictation can stay on your machine, nobody in your practice has to take our word for it. They can read the code.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"One of OpenWhispr's founders"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Put the cursor in the historia clínica. Dictate in Spanish.","body":"Review the text while the case is still fresh, then file it in the system you already use.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, we'll fix it with you in Discord. Free forever on Mac, Windows & Linux."}}},"useCasePhysicians":{"breadcrumb":{"home":"Home","useCases":"Use Cases","physicians":"For Physicians"},"metadata":{"title":"Free Physician Dictation Software for EHR Notes - OpenWhispr","description":"OpenWhispr is free physician dictation software for HPI, exam and assessment text at your EHR cursor, with on-device transcription.","keywords":["physician dictation software","medical dictation for physicians","EHR dictation software","clinical note dictation","voice recognition for physicians","offline physician dictation"]},"schema":{"name":"OpenWhispr","applicationCategory":"Physician dictation software","operatingSystem":"macOS, Windows, Linux, iOS","description":"Open-source physician dictation software with optional on-device transcription. Put HPI, exam and assessment text at the active cursor with unlimited local dictation at no cost.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Physician dictation software for clinical notes","title":"Press the hotkey. Fill the chart field you're in.","body":"Dictate the HPI, exam, assessment and plan into the EHR field you already opened. Choose a local model so transcription runs on your computer. OpenWhispr is free and open source.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Physician dictating at a computer in an exam room"}},"trustBar":{"label":"Physicians chart in:","names":["Epic","Oracle Health","athenaOne","eClinicalWorks","MEDITECH","Microsoft Word"]},"pain":{"label":"The problem with charting by keyboard","title":"The visit ended. The chart did not.","subtitle":"HPI, exam, assessment, plan, inbox. Every empty field is another piece of the visit waiting to be typed.","cards":[{"title":"Open charts stack up between patients.","body":"The encounter is over, but the HPI and assessment are still blank. Dictate while the details are fresh."},{"title":"Every field breaks your train of thought.","body":"Click HPI, type. Click exam, type. Click assessment, type again. OpenWhispr leaves the EHR open and puts the transcript at the active cursor."},{"title":"One drug name becomes five corrections.","body":"Generic dictation trips over medication names, eponyms and local abbreviations. Add the terms you use to OpenWhispr's custom dictionary."},{"title":"A five-minute trial needs an IT ticket.","body":"You want to test dictation on one chart, not provision a practice. Local mode starts with a desktop download and no account."}]},"features":{"label":"What you get","title":"Less keyboard. Fewer open charts.","subtitle":"Physician dictation software should fit the note template already on screen. OpenWhispr turns speech into text at your cursor without making you rebuild the workflow around it.","items":[{"image":{"label":"Physician dictating into a desktop microphone at a workstation"},"title":"HPI and assessment at the active cursor","body":"Open the note, place the cursor and press the hotkey. Your dictation is transcribed and pasted into that field, ready for review."},{"image":{"label":"Physician holding a medication bottle beside a computer"},"title":"Your medication names in the dictionary","body":"Add the drugs, procedures, clinician names and abbreviations that keep coming back wrong. OpenWhispr passes those terms to transcription."},{"image":{"label":"Clinician dictating beside a laptop inside a van"},"title":"Keep dictating when ward Wi-Fi drops","body":"Download a local model once. Then transcribe on the workstation in clinic, at home or on rounds without an internet connection."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Audio uploaded in local mode","value":"None","footnote":"in local mode, transcription runs on the physician's device"},{"label":"Cost per physician","value":"$$0","footnote":"unlimited local dictation on every desktop plan"},{"label":"Setup takes","value":"5 min","footnote":"from download to your first dictated chart field"}]},"howItWorks":{"label":"How it works","title":"Finish the chart you already have open.","subtitle":"Keep the EHR workflow you already know. OpenWhispr adds a hotkey and local transcription, with no account or card required to start.","image":{"label":"Physician working on a laptop at night"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Choose a local model during setup."},{"title":"Put the cursor in the chart.","body":"Open the HPI, exam or assessment field, press the hotkey and dictate."},{"title":"Review and sign.","body":"Read the inserted text, correct what needs correcting, then finish the note in the EHR."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons physicians choose OpenWhispr","subtitle":"Try one workstation, one chart field and one local model. These are the ten things to check.","items":[{"title":"No copy-and-paste detour","body":"Dictate inside the open chart field. Review the inserted text there, then continue the note."},{"title":"One hotkey across desktop apps","body":"Use the same dictation control in the EHR, a portal reply, an email or a document."},{"title":"A dictionary for your clinical vocabulary","body":"Add drug names, procedures, eponyms and abbreviations as recognition hints for supported transcription and cleanup paths."},{"title":"Local mode stays on the workstation","body":"Choose a local model and dictation is transcribed on the device. Cloud fallback is a separate opt-in setting."},{"title":"Works fully offline","body":"Once the local model is downloaded, dictation keeps working without internet in clinic, at home or on a ward."},{"title":"Unlimited local dictation at $0","body":"Every desktop plan includes unlimited use of local speech-to-text models."},{"title":"Mac, Windows and Linux","body":"Use it on the desktop operating system your workstation already runs."},{"title":"Open source, inspectable by IT","body":"Your organization can review the code on GitHub instead of taking the local-mode claim on faith."},{"title":"100+ languages","body":"OpenWhispr supports 100+ languages. Availability varies by local model and cloud provider."},{"title":"Local or cloud dictation in Settings","body":"Choose Local or Cloud for dictation during setup, and change that dictation setting later when your workflow changes."}]},"faq":{"label":"Before you download","title":"What physicians ask before changing dictation software.","subtitle":"Straight answers on EHR fields, terminology, patient data, offline use and cost.","items":[{"question":"What is the best physician dictation software for EHR notes?","answer":"If you want a hotkey that puts the transcript at your active cursor, unlimited local dictation at $0, and code your IT team can inspect, try OpenWhispr on one workstation. It does not replace the physician's review or the EHR's signing workflow."},{"question":"Does OpenWhispr integrate with my EHR?","answer":"OpenWhispr is not an EHR integration. It transcribes your speech and pastes the text at the active cursor, so you can use it in editable chart fields and other desktop apps. Test it with your organization's EHR and virtual-desktop setup before a wider rollout."},{"question":"How does it handle medication names and clinical abbreviations?","answer":"OpenWhispr includes a custom dictionary. Add the medication names, procedures, eponyms and abbreviations you use, and those terms are supplied to transcription. You still review the note before signing."},{"question":"Can I dictate patient information?","answer":"Your organization decides which tools are approved for patient information. In local mode, transcription runs on the device without uploading the audio. Confirm your organization's policy before use."},{"question":"Will it work when the hospital network or Wi-Fi does not?","answer":"Yes, after you download a local model. Local transcription runs on Mac, Windows and Linux without internet. Cloud transcription remains a separate mode you can choose in Settings."},{"question":"Is it actually free?","answer":"Yes. Every desktop plan includes unlimited local dictation, a custom dictionary and 100+ languages at $0. Pro is $8 per month and adds unlimited cloud transcription and the iPhone and iPad app."}]},"reviews":{"label":"What users say","title":"What users notice when voice replaces the keyboard.","subtitle":"Verbatim notes from our Discord, GitHub and inboxes.","score":{"value":"0","label":"Per-physician licences for local dictation","body":"Unlimited local dictation is included at $0 on every desktop plan."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Start free with local dictation.","Upgrade for cloud or team features."],"caption":"Compare OpenWhispr plans for physicians","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Physicians who want free offline dictation at the active cursor and can set up a local model"}]},{"parts":[{"text":"Physicians who want unlimited cloud transcription and mobile dictation"}]},{"parts":[{"text":"Practices and teams documenting at volume"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited on-device dictation with no per-physician fee"}]},{"parts":[{"text":"Unlimited cloud dictation without downloading a local model"}]},{"parts":[{"text":"Unlimited recordings and priority support for a practice"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"The path from microphone to chart should be inspectable.","paragraphs":["Physicians should not have to guess what a dictation tool does between the microphone and the chart. OpenWhispr makes that path visible in public code.","Your organization can inspect the local-mode implementation on GitHub instead of relying only on a privacy page.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"OpenWhispr founder and community"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Finish the chart before you open the next one.","body":"Review the inserted text, sign the note and move on.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, ask us for help in Discord. Free forever on Mac, Windows & Linux."}}},"useCaseDoctors":{"breadcrumb":{"home":"Home","useCases":"Use Cases","doctors":"For Doctors"},"metadata":{"title":"Free Doctor Dictation Software, No Middleman - OpenWhispr","description":"Doctor dictation software with no middleman: pick a local model and OpenWhispr types your letters on your own machine in seconds. Free and open source.","keywords":["doctor dictation software","physician dictation software","dictation app for doctors","voice to text for doctors","referral letter dictation","offline dictation for physicians"]},"schema":{"name":"OpenWhispr","applicationCategory":"Physician dictation software","operatingSystem":"macOS, Windows, Linux, iOS","description":"Local-first, open-source dictation software for doctors. With a local model chosen, referral letters, reports and notes are transcribed on the doctor's own device and never leave it. Free for unlimited local dictation, and works fully offline.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Dictation software for doctors & physicians","title":"Speak the referral. Send it before the patient leaves.","body":"Choose a local model at setup and the referral, report or patient letter is typed and drafted on your own computer. No transcription service, no per-seat licence, no one else hearing the audio.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Doctor speaking into a desk microphone at a computer"}},"trustBar":{"label":"Dictate into:","names":["DrChrono","Practice Fusion","Tebra","AdvancedMD","Outlook","Microsoft Word"]},"pain":{"label":"The problem with dictation services","title":"You already dictate. Now choose where it runs.","subtitle":"Choose local processing for sensitive or offline dictation, or Cloud for top accuracy and speed without models to download.","cards":[{"title":"Keep sensitive dictation on your device.","body":"With a local model, OpenWhispr transcribes the recording on your machine."},{"title":"Get the text while the case is still fresh.","body":"OpenWhispr turns your dictation into text as you work, so you can review it before sending."},{"title":"Free local dictation for every doctor.","body":"OpenWhispr includes unlimited local AI models on every plan, free. Choose Cloud when you want top accuracy and speed without models to download."},{"title":"The code is the confidentiality agreement.","body":"Closed tools hand you a privacy policy and ask for faith. OpenWhispr's code is public on GitHub, so anyone in the practice can verify that in local mode, patient audio never leaves the machine."}]},"features":{"label":"What you get","title":"Instant turnaround. Nobody else in the loop.","subtitle":"Doctor dictation software used to mean a slow service or a per-seat licence. Keep the typed letter and drafted report without sending dictation through a service or buying another seat.","items":[{"image":{"label":"Doctor using a desktop computer in an office"},"title":"Typed on your machine, in seconds","body":"Dictate the referral, report or reply into your mic. The words land on your screen seconds after you stop talking, transcribed locally and never uploaded."},{"image":{"label":"Doctor writing in a notebook beside a computer"},"title":"Dictate the detail. Review the text.","body":"Speak the gist, then review the transcription, edit it and use it in your referral, report or patient instructions."},{"image":{"label":"Doctor typing at a clinical workstation"},"title":"One hotkey. Any window, any EHR.","body":"Press the hotkey and speak into whatever is open: the chart, a reply to a colleague, a report in Word. The text lands at your cursor, punctuated, processed locally."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Audio sent to a transcription service","value":"None","footnote":"in local mode, dictation is transcribed on your device and goes nowhere else"},{"label":"Cost per doctor","value":"$$0","footnote":"unlimited local dictation with no per-doctor licence"},{"label":"Local setup","value":"Choose your model","footnote":"install the app and choose local models"}]},"howItWorks":{"label":"How it works","title":"Turnaround time: the length of the dictation.","subtitle":"Old dictation goes out to a service and comes back tomorrow. Local mode runs on your machine, with no account or card required to start.","image":{"label":"Doctor opening a laptop at a desk"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Install the app and choose a local model."},{"title":"Dictate as you always have.","body":"Speak the referral, report or note. It's typed on your device seconds after you stop."},{"title":"Review, send, delete.","body":"Check the draft, paste it into the EHR or an email, and delete the audio. You keep control of all of it."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons doctors choose OpenWhispr","subtitle":"Your dictation carries a patient's history, so check the code and the settings, not just the brochure.","items":[{"title":"No transcription service in the loop","body":"The recording stays on your device instead of going out for overnight transcription."},{"title":"The audio has nowhere to go","body":"Local mode has no upload step. The recording is transcribed on your machine and stays there."},{"title":"Free for every doctor in the practice","body":"Unlimited local dictation at $0, every seat, forever. No per-doctor licence to renew each year."},{"title":"Works fully offline","body":"Home visits, a locum shift on a ward with no signal, a flight to a conference. Once the model is on your machine, dictation never stalls."},{"title":"Types into the chart, the inbox and Word","body":"One hotkey puts the text at your cursor: a referral template, a colleague's inbox, the record itself."},{"title":"Letters and reports, not just transcripts","body":"Run a local AI model, or connect your own key, to turn a dictated summary into a formatted referral, report or patient letter."},{"title":"Your letters are your files","body":"Recordings and drafts sit on your disk. Keep, encrypt or shred them on your schedule, not a vendor's retention policy."},{"title":"Auditable by whoever runs your IT","body":"Your IT team can inspect the public code instead of relying only on a vendor claim."},{"title":"Learns your drug names and eponyms","body":"A custom dictionary picks up the drugs, procedures and abbreviations your specialty uses, so they come out spelled right."},{"title":"100+ languages","body":"Dictate in the language you think in, transcribed locally. Useful when the note is in one language and the referral goes out in another."}]},"faq":{"label":"Before you download","title":"What doctors ask before switching dictation software.","subtitle":"Confidentiality, cost and accuracy, answered straight. Anything else, ask in Discord.","items":[{"question":"What's the best dictation software for doctors?","answer":"OpenWhispr gives doctors a choice: local dictation on their device for offline or sensitive work, or Cloud for top accuracy and speed without models to download. The free plan includes unlimited local models, 100+ languages and a custom dictionary."},{"question":"Who hears my dictation?","answer":"In local mode, no one: it is transcribed on your machine. In Cloud mode, audio is sent only to transcribe it and is never stored. Cloud is the fast, high-accuracy choice when you do not want to download models."},{"question":"Can I use it with patient information under HIPAA?","answer":"In local mode, patient audio is transcribed on your device and never reaches a third party, so there is no vendor to cover with a BAA. For cloud transcription of PHI, OpenWhispr signs a BAA on the Business plan and up, and holds a SOC 2 Type 2 report and ISO 27001 certification. Whether your workflow is compliant still depends on how you use it; no tool can grant that on its own."},{"question":"Will it get the drug names right?","answer":"It runs Whisper-class models that handle a wide range of accents, and the custom dictionary learns the drug names, procedures and abbreviations your specialty uses. Add a term once and it comes out spelled right from then on."},{"question":"Does it work on home visits, or on a ward with no signal?","answer":"Yes. Once the local model is on your machine, it works offline. When you are online, Cloud provides top accuracy and speed without a model download."},{"question":"Is it really free for the whole practice?","answer":"Yes. Unlimited local dictation on Mac, Windows and Linux, in 100+ languages, at $0, for every doctor in the practice. Pro at $8/month adds unlimited cloud transcription and the iPhone app when you want them."}]},"reviews":{"label":"What users say","title":"Nobody here mails their voice to a stranger either.","subtitle":"From our Discord, GitHub and inboxes.","score":{"value":"0","label":"Consults heard by a transcriber","body":"In local mode the audio never leaves your device. No server stores it, and no transcriber ever hears it."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for doctors","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Doctors who want free, private, offline dictation and are comfortable choosing local models during setup"}]},{"parts":[{"text":"Doctors who want easy, reliable cloud transcription and mobile dictation"}]},{"parts":[{"text":"Practices documenting at volume, or that need a signed BAA"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Free offline dictation, with patient audio staying on your device"}]},{"parts":[{"text":"Unlimited cloud transcription with no local models to manage"}]},{"parts":[{"text":"Unlimited recordings plus advanced workflow features for a practice"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"A patient's history is not training data.","paragraphs":["OpenWhispr started with a simple objection: turning speech into text shouldn't mean the audio lands on someone else's server, or in someone's training set. A patient's history least of all.","We build in public on GitHub. When we say local-mode dictation stays on your device, your practice can inspect the code instead of relying only on a privacy policy.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"Team / community photo"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Keep dictating. Just stop paying someone to type it.","body":"Your next referral goes from your voice to the recipient without passing through anyone else's hands. No service, no seat licence, no waiting.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, we'll fix it with you in Discord. Free forever on Mac, Windows & Linux."}}},"useCaseProductivity":{"breadcrumb":{"home":"Home","useCases":"Use Cases","productivity":"For Productivity"},"metadata":{"title":"Free Dictation App For Productivity - OpenWhispr","description":"OpenWhispr is a free, open-source dictation app for productivity. Press one hotkey to put tasks, notes and replies in the app already open.","keywords":["dictation app for productivity","productivity dictation app","voice typing for productivity","speech to text productivity app","dictate tasks and notes","offline dictation app"]},"schema":{"name":"OpenWhispr","applicationCategory":"Productivity dictation app","operatingSystem":"macOS, Windows, Linux, iOS","description":"Local-first, open-source dictation app for productivity. Press a system-wide hotkey and speak tasks, notes and replies into the app already open. With local models selected, transcription and clean-up run on your own computer and work offline. Unlimited local dictation is free on macOS, Windows and Linux.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Dictation app for productivity","title":"Say the task. Get it into the list.","body":"Capture the thought before it disappears, right in the app you already have open; no separate inbox to check later. Use OpenWhispr Cloud for fast, accurate transcription, or local models to keep the audio on your machine.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Woman speaking while working at a laptop beside a notebook and material samples"}},"trustBar":{"label":"Dictate into the app already open, including:","names":["Todoist","Notion","Obsidian","Gmail","Google Docs","Slack"]},"pain":{"label":"The capture problem","title":"The thought is quick. Typing it down is the slow part.","subtitle":"A task arrives while you are already doing something else. The longer capture takes, the more likely the thought stays in your head.","cards":[{"title":"The task arrives in the middle of another task.","body":"You are in Slack when you remember the invoice, or in a doc when you spot tomorrow's follow-up. Put the cursor in your task list, press the hotkey and say it before the thought disappears."},{"title":"The note is longer than the time you have.","body":"A useful idea becomes three thin bullet points because typing the whole thought feels like work. Talk it into Notion or Obsidian instead. OpenWhispr removes filler and false starts, then pastes the cleaned text at the cursor."},{"title":"Five two-minute replies become an hour of avoidance.","body":"The answer is already in your head. Press the same hotkey in Gmail or Slack, say the reply once and review the text before you send it. No separate recorder to empty later."},{"title":"Your system has names only your system understands.","body":"Project codenames, client names, acronyms and recurring places get mangled by generic dictation. Add them to the custom dictionary once so the transcription and clean-up get the spellings you use."}]},"features":{"label":"One hotkey wherever you type","title":"Keep the system you built. Replace the typing.","subtitle":"One system-wide hotkey works wherever the cursor already is. Choose local models and it keeps working without a connection.","items":[{"image":{"label":"Man speaking into a desk microphone beside a monitor and open notebook"},"title":"Your apps stay exactly where they are","body":"Todoist for tasks, Notion for projects, Obsidian for notes, Gmail for replies. OpenWhispr types into all four. It pastes the finished text into the active field, so there is no new inbox to process and no integration to maintain."},{"image":{"label":"Woman speaking beside a laptop and architectural model"},"title":"A false start does not become a messy task","body":"Say 'Thursday, no, Friday' and the clean-up keeps the correction. It removes filler, fixes punctuation and turns spoken dates into written ones while keeping your wording and intent."},{"image":{"label":"Man working on a laptop and notebook aboard a train"},"title":"The capture habit still works offline","body":"Download local speech-to-text and a local clean-up model once. Then tasks, notes and replies can be transcribed and tidied on your computer with no connection and no per-word bill."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"AI processing location","value":"On your device","footnote":"when local transcription and local clean-up models are selected"},{"label":"Cost to capture locally","value":"$$0","footnote":"unlimited local dictation on every desktop you use"},{"label":"Local setup","value":"A few steps","footnote":"download, choose models, set the hotkey and start speaking"}]},"howItWorks":{"label":"Download to first captured task","title":"Three steps from a thought to the right inbox.","subtitle":"Nothing to migrate and no productivity system to rebuild.","image":{"label":"Woman typing on a laptop backstage beside camera equipment"},"steps":[{"title":"Download it free.","body":"Install OpenWhispr on Mac, Windows or Linux. Choose OpenWhispr Cloud for speed, or download local models for offline use."},{"title":"Put the cursor where the thought belongs.","body":"Open the task field, note, email or message you already use. Press your hotkey and say the whole thought, including the context you would normally skip."},{"title":"Review it and move on.","body":"OpenWhispr cleans the wording and pastes it at the cursor. Check the text, save or send it, and get back to the work you were doing."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"Why it earns a place in the stack","title":"10 reasons people use OpenWhispr in the apps they already use","subtitle":"Each one removes friction from capture without asking you to replace the apps that already work.","items":[{"title":"One hotkey across your apps.","body":"Tasks in Todoist, project notes in Notion, linked thoughts in Obsidian, replies in Gmail and Slack. If automatic paste is unavailable, the text stays on the clipboard."},{"title":"No new inbox to empty.","body":"The words land in the tool where they belong. You do not record a voice memo now and remember to move it later."},{"title":"You approve the text before it becomes a task.","body":"The cleaned text lands in the field first. Fix a name, date or label before you save or send it."},{"title":"Dates look like dates.","body":"Spoken dates, times, numbers and currency are converted to standard written forms before the text reaches your task, note or reply."},{"title":"Your project vocabulary follows you.","body":"Add client names, codenames, acronyms and product terms to the custom dictionary. The same list helps across the apps you dictate into."},{"title":"Unlimited local dictation is free.","body":"No trial clock, no word cap and no card on every desktop plan when you run the speech-to-text model on your own machine."},{"title":"Offline means the capture habit does not break.","body":"Local transcription and a local clean-up model keep working on a plane, during an outage or anywhere the connection drops."},{"title":"100+ languages.","body":"Capture the thought in the language you actually think in. The same app and the same hotkey support more than 100 languages."},{"title":"The privacy claim is code you can inspect.","body":"OpenWhispr is open source. You can inspect the local transcription and paste paths yourself."},{"title":"Mac, Windows and Linux.","body":"Use the same free app on the desktop at work and the laptop at home, with a hotkey you choose for each machine."}]},"faq":{"label":"Before you add voice input to your productivity stack","title":"What people ask before they put OpenWhispr in their stack","subtitle":"What it does, what it does not do and what stays on your machine.","items":[{"question":"Does OpenWhispr create and schedule tasks for me?","answer":"OpenWhispr types finished text where your cursor is. It does not organise or schedule tasks. Todoist, Notion and other apps still apply their own date and label rules, and you review the text before saving."},{"question":"Do I need an integration for every productivity app?","answer":"No. The system-wide hotkey pastes text into the active app, so there is no Todoist plugin, Notion connection or Gmail extension to set up. On Mac, OpenWhispr needs accessibility permission to paste automatically. If automatic paste fails, the transcription remains available to paste from the clipboard."},{"question":"Will it clean up a rambling thought?","answer":"Yes, when clean-up is enabled. The clean-up prompt removes filler words, false starts, stutters and accidental repetition. It fixes grammar and punctuation, converts dates and times to written forms and is explicitly told to preserve your wording, formality and intent."},{"question":"Can OpenWhispr process my dictation locally?","answer":"Yes. Choose local transcription and local clean-up so audio and transcript processing stay on your computer. Once text is pasted, the destination app's own storage and privacy rules apply. OpenWhispr Cloud audio is transcribed and discarded, not stored."},{"question":"Is unlimited dictation really free?","answer":"Unlimited local dictation is $0 on every desktop plan and does not expire. The Free plan also includes 2,000 words a week through OpenWhispr Cloud. Pro is $8 per user each month for unlimited cloud transcription, and Business is $16 per user each month for unlimited cloud meeting recordings and agent mode."},{"question":"Will it spell project names and acronyms correctly?","answer":"Add the exact spellings to the custom dictionary. OpenWhispr passes that vocabulary into supported transcription paths and the clean-up step so client names, project codenames and internal acronyms have more context. You can edit the dictionary any time."}]},"reviews":{"label":"What people tell us","title":"Better than typing. Clean after a stumble. Local without WiFi.","subtitle":"Read them for the tool, not for a productivity method.","score":{"value":"0","label":"Productivity apps to replace","body":"OpenWhispr is the voice input across the stack you already use. Your tasks, notes, docs, email and chat stay where they are."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for productivity","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"People who want unlimited local dictation across their personal productivity stack for $0"}]},{"parts":[{"text":"People who want unlimited cloud dictation without managing local models"}]},{"parts":[{"text":"Teams that want agent mode and unlimited cloud meeting recordings alongside dictation"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Capture tasks, notes and replies locally with no word cap, trial clock or card"}]},{"parts":[{"text":"Unlimited OpenWhispr Cloud transcription with no local model to download"}]},{"parts":[{"text":"Unlimited cloud meeting recordings and agent mode for shared workflows"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Why we built it","title":"A productivity system is only useful if the thought gets into it.","paragraphs":["OpenWhispr turns speech into clean text where the thought belongs: your task list, notes or reply field.","OpenWhispr is open source, so you can inspect the local transcription and paste paths yourself. With local models selected, dictation can stay on your machine.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"One of OpenWhispr's founders"}},"finalCta":{"tagline":"$$0 · open source · your apps stay put","title":"Press the hotkey. Get the thought into the app already open.","body":"Download OpenWhispr, choose a model and dictate into your task list, notes or reply field. No card and no migration.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Unlimited local dictation is free on Mac, Windows and Linux."}}},"useCasePathology":{"breadcrumb":{"home":"Home","useCases":"Use Cases","pathology":"For Pathologists"},"metadata":{"title":"Pathology Dictation Software for Reports - OpenWhispr","description":"OpenWhispr is free pathology dictation software for macroscopic descriptions and microscopic findings, with auto-paste in supported report fields.","keywords":["pathology dictation software","pathologist dictation software","pathology report dictation","anatomic pathology dictation","histopathology dictation software","pathology speech recognition","offline pathology dictation"]},"schema":{"name":"OpenWhispr","applicationCategory":"Pathology dictation software","operatingSystem":"macOS, Windows, Linux","description":"Local-first, open-source pathology dictation software that transcribes spoken report text for automatic paste into supported fields. Free for unlimited local dictation.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Pathology dictation software","title":"Dictate the findings. Get report text at your cursor.","body":"Keep the specimen or slide in view. Dictate the macroscopic description, microscopic findings or final diagnosis. OpenWhispr types into supported fields. You review and sign out.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Laboratory professional reviewing a computer beside a microscope and slide tray"}},"trustBar":{"label":"Reporting environments to test with OpenWhispr:","names":["Epic Beaker","Oracle Health PathNet","Clinisys PowerPath","Clinisys WinPath","Sectra Digital Pathology","Microsoft Word"]},"pain":{"label":"The problem with typing pathology reports","title":"The slide is in focus. The keyboard pulls you away.","subtitle":"The finding is clear. The field is empty. Every trip to the keyboard interrupts the read.","cards":[{"title":"Your eyes leave the slide for the keyboard.","body":"You see the finding, then look away to type it. OpenWhispr lets you speak the report text while the morphology is still in view."},{"title":"The gross description waits on your hands.","body":"Measurements, margins, sections and the block key arrive in sequence at the cut-up bench. Dictate each observation while it is in front of you, then review the text before it enters the final report."},{"title":"Structured fields still need precise prose.","body":"Synoptic fields capture element-response pairs. Free-text microscopic descriptions, comments and addenda can still need your words. OpenWhispr transcribes those words without building or validating the protocol."},{"title":"One unfamiliar term becomes a correction loop.","body":"Specimen labels, eponyms and local abbreviations recur case after case. Add them to the custom dictionary as context hints, then check every occurrence before sign-out."}]},"features":{"label":"What you get","title":"Keep the case in view. Put the words in the report.","subtitle":"Pathology dictation software should fit the reporting screen already open. OpenWhispr turns speech into text at the active cursor. It does not replace your LIS, protocols or judgment.","items":[{"image":{"label":"Person typing beside magnified tissue images on two monitors"},"title":"Report text at the active cursor","body":"Open an editable report field, place the cursor and press the hotkey. OpenWhispr pastes the transcript when the field and session support automatic paste. Otherwise it copies the text for manual paste."},{"image":{"label":"Laboratory professional using a desktop microphone at a workstation"},"title":"Local transcription when you choose it","body":"Download a local speech model and select Local in Settings. With cloud fallback disabled, dictation stays on the workstation and continues without internet after the model is installed."},{"image":{"label":"Laboratory professional holding a specimen tube and smartphone beside a microscope"},"title":"Your recurring terms as context hints","body":"Add specimen names, eponyms, stains and local abbreviations to the custom dictionary. OpenWhispr feeds them to supported speech models as hints. It does not promise specialist terminology accuracy."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Direct LIS integrations included","value":"None","footnote":"OpenWhispr uses active-cursor paste in compatible fields instead"},{"label":"Cost per pathologist","value":"$$0","footnote":"unlimited local dictation on every desktop plan"},{"label":"Setup takes","value":"5 min","footnote":"from download to your first dictated report text"}]},"howItWorks":{"label":"How it works","title":"Findings in the report before you leave the case.","subtitle":"Speech becomes editable text in the report field. You check it against the case and complete sign-out in the reporting system.","image":{"label":"Healthcare worker in blue scrubs opening a laptop in a hospital corridor"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Pick a local model during setup."},{"title":"Dictate the findings.","body":"Put the cursor in an editable field, press the hotkey and speak the macroscopic description, microscopic findings, ancillary-study results, addendum or diagnosis text."},{"title":"Review, then sign out.","body":"Check the text against the specimen, slides, ancillary studies and required report elements. The final report remains your call."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons OpenWhispr fits pathology reporting","subtitle":"A pathology report carries exact observations and a final professional judgment. OpenWhispr stays in the narrower lane of turning your speech into editable text.","items":[{"title":"Automatic paste in supported fields","body":"Use the same hotkey in a compatible report field, an email or a document. If automatic paste is unavailable, OpenWhispr leaves the text on your clipboard."},{"title":"No direct LIS plug-in required","body":"Compatible editable fields use desktop dictation and automatic paste. OpenWhispr does not claim a direct integration with Epic Beaker, PathNet, PowerPath, WinPath or another LIS. Test your reporting system and virtual-desktop setup."},{"title":"A local transcription path","body":"Choose Local in Settings and keep cloud fallback disabled. The speech model runs on the workstation and the dictation audio is not uploaded for transcription on that path."},{"title":"Works offline after model setup","body":"Once a local model is downloaded, dictation continues without internet at the microscope, gross bench or home workstation."},{"title":"A dictionary for recurring terms","body":"Add specimen names, stains, eponyms and abbreviations as context hints. The pathologist still checks the transcript before sign-out."},{"title":"Unlimited local dictation at $0","body":"Every desktop plan includes unlimited use of local speech-to-text models."},{"title":"Mac, Windows and Linux","body":"Use OpenWhispr on the desktop operating system your reporting workstation already runs."},{"title":"Open source, inspectable by IT","body":"The desktop app code is public on GitHub, so your technical team can inspect the local and cloud transcription paths."},{"title":"100+ languages","body":"OpenWhispr supports 100+ languages. Availability varies by local model and cloud provider."},{"title":"Local or cloud in Settings","body":"Choose the transcription path during setup and change it later in Settings. Cloud audio is sent only to transcribe it and is never stored by OpenWhispr."}]},"faq":{"label":"Before you download","title":"What pathologists ask before changing dictation software.","subtitle":"Straight answers on reporting fields, terminology, diagnostic boundaries, local processing and cost.","items":[{"question":"Does OpenWhispr integrate with my LIS or digital pathology viewer?","answer":"OpenWhispr is not an LIS or viewer integration. It transcribes speech and pastes text at the active cursor, so you can use it in compatible editable fields and other desktop apps. Test it with your reporting system and virtual-desktop setup before wider use."},{"question":"Does it generate a diagnosis or complete a synoptic report?","answer":"No. OpenWhispr turns what you dictate into editable text. It does not inspect slides, interpret findings, choose a diagnosis, populate required protocol elements or sign out a case. The pathologist reviews every word in the reporting system."},{"question":"Will it recognize pathology terminology?","answer":"OpenWhispr includes a custom dictionary. Add recurring specimen names, stains, eponyms and abbreviations, and supported speech models receive them as context hints. This is not a guarantee of specialist terminology accuracy, so review every report before sign-out."},{"question":"Where is the dictation processed?","answer":"Choose Local and transcription runs on the workstation. Choose OpenWhispr Cloud and the audio is sent only for transcription, then never stored by OpenWhispr. Your laboratory decides which mode and workflow are approved for its data."},{"question":"Will it work without internet?","answer":"Yes, after you download a local speech model. Local transcription runs on Mac, Windows and Linux without internet. Cloud transcription remains a separate mode you can choose in Settings."},{"question":"Is it actually free?","answer":"Yes. Every desktop plan includes unlimited local dictation and a custom dictionary at $0. OpenWhispr supports 100+ languages across available models and providers; local-model availability varies. Pro is $8 per month and adds unlimited cloud transcription, more meeting recording hours, mobile and device sync."}]},"reviews":{"label":"What users say","title":"The dictation is fast. The judgment stays yours.","subtitle":"Existing OpenWhispr user feedback. These testimonials are about the tool, not pathology reporting.","score":{"value":"0","label":"Diagnostic decisions made by OpenWhispr","body":"OpenWhispr transcribes what you say. It does not interpret a specimen or slide, choose a diagnosis or sign out a report."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Free forever.","Upgrade when you want more."],"caption":"Compare OpenWhispr plans for pathologists","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Pathologists who want free offline dictation at the active report cursor"}]},{"parts":[{"text":"Pathologists who want unlimited OpenWhispr Cloud transcription"}]},{"parts":[{"text":"Laboratories and teams documenting at volume"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited local report dictation at $0"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models"}]},{"parts":[{"text":"Unlimited recordings and priority support for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users and OpenWhispr features"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" supplies names and terms as context hints"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" supplies names and terms as context hints"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" supplies names and terms as context hints"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" runs user-configured actions on transcripts"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic OpenWhispr team use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, SAML, SCIM, audit logs, retention controls, admin controls, or dedicated support for a larger team? These are enterprise security and data-governance options, not clinical or regulatory certification.","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"The report remains yours. OpenWhispr handles the typing.","paragraphs":["OpenWhispr handles the mechanical part: speech becomes editable text at your cursor. The interpretation, structured elements, final diagnosis and sign-out remain with the pathologist and the laboratory's reporting system.","OpenWhispr is open source under the MIT License, and its source is available on GitHub. When we say dictation can run locally, your laboratory can inspect the code.","And if anything breaks, you talk to the people who built it, in Discord."],"image":{"label":"OpenWhispr founder portrait"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Finish the report while the case is still on screen.","body":"Put the cursor in the editable field, dictate the findings and review every word before sign-out.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work with your workstation, ask us for help in Discord. Free forever on Mac, Windows & Linux."}}},"useCaseRadiology":{"breadcrumb":{"home":"Home","useCases":"Use Cases","radiology":"For Radiology"},"metadata":{"title":"Best Radiology Dictation Software, Free & Local - OpenWhispr","description":"OpenWhispr is free, open-source radiology dictation software that transcribes locally and pastes text at the active cursor for review.","keywords":["best radiology dictation software","radiology dictation software","radiologist speech recognition software","offline radiology dictation","local radiology transcription","radiology voice to text"]},"schema":{"name":"OpenWhispr","applicationCategory":"Radiology dictation software","operatingSystem":"macOS, Windows, Linux, iOS","description":"Local-first, open-source dictation software for radiologists. Choose local speech-to-text and the audio is transcribed on the workstation, then text is pasted at the active cursor. Cleanup is configured separately; there is no native radiology-system integration.","price":"0","priceCurrency":"USD"},"landing":{"hero":{"eyebrow":"Best radiology dictation software for imaging specialists","title":"Keep the images in view. Get dictated text at the cursor.","body":"Choose a local speech-to-text model, place the cursor in an editable report field, and speak. Audio is transcribed on the workstation; use local cleanup or turn cleanup off to keep the text there too.","cta":{"label":"Download for free","href":"/download"},"trust":"Open source · 80,000+ downloads","image":{"label":"Radiologist dictating beside chest scans at a reading station"}},"trustBar":{"label":"Common reporting tools, not integrations:","names":["PowerScribe One","Epic Radiant","Sectra IDS7","Visage 7","InteleViewer","Microsoft Word"]},"pain":{"label":"Where radiology dictation breaks","title":"The images need your eyes. The report still needs your words.","subtitle":"A radiology dictation tool has to respect the active field, the vocabulary and the final read. Miss one, and the keyboard comes back.","cards":[{"title":"Your eyes leave the study for the keyboard.","body":"Place the cursor once, then hold the hotkey and dictate the finding. OpenWhispr pastes the transcript into the active field so you can stay with the images."},{"title":"The same term comes back wrong every case.","body":"Add names, modalities, anatomy and local abbreviations to the custom dictionary. They are supplied as hints to supported transcription paths."},{"title":"One missing 'no' changes the report.","body":"Speech recognition can change clinical meaning. OpenWhispr supplies text, not clinical quality control, so the radiologist still proofreads before signing."},{"title":"A simple dictation test turns into an integration project.","body":"OpenWhispr is a desktop speech-to-text layer, not a PACS or RIS replacement. Try it in an editable field without changing the reporting stack."}]},"features":{"label":"What you get","title":"Cursor dictation, not another reporting platform.","subtitle":"OpenWhispr handles speech to text. Your worklist, images, templates, measurements and signing workflow stay in the systems that already own them.","items":[{"image":{"label":"Person working at two monitors beside a desktop computer"},"title":"Findings transcribed on the workstation","body":"Choose a local speech-to-text model and audio is processed on the device. Use local cleanup or turn cleanup off if report text must stay there too; after model download, speech-to-text works offline."},{"image":{"label":"Radiologist checking a reference card beside a medical scan"},"title":"A dictionary for the terms you actually dictate","body":"Add recurring anatomy, eponyms, protocol names and abbreviations. OpenWhispr supplies the list as recognition hints, and you still check the result."},{"image":{"label":"Radiologist typing beside monitors displaying medical scans"},"title":"One hotkey at the active report cursor","body":"Open the report field, place the cursor, hold the hotkey and speak. The transcript is pasted there for review, with no claim of a native PACS, RIS or EHR integration."}]},"stats":{"ariaLabel":"Key facts","items":[{"label":"Dictated audio uploaded in local mode","value":"None","footnote":"with local-to-cloud fallback off, audio stays on the radiologist's workstation"},{"label":"Cost per radiologist","value":"$$0","footnote":"unlimited local dictation for every radiologist"},{"label":"Setup takes","value":"5 min","footnote":"from download to a test dictation at the cursor"}]},"howItWorks":{"label":"How it works","title":"From spoken finding to text in the open report.","subtitle":"Keep the reporting workflow you already know. OpenWhispr adds a system-wide hotkey and a local transcription option.","image":{"label":"Radiologist working at a laptop beside brain scans"},"steps":[{"title":"Download for free.","body":"Mac, Windows or Linux. Choose a local model during setup."},{"title":"Put the cursor in the report.","body":"Open an editable findings or impression field, then press the hotkey and dictate."},{"title":"Proofread and sign.","body":"Check terminology, numbers, negation and laterality, then finish the report in your reporting system."}],"cta":{"label":"Download for free","href":"/download"}},"reasons":{"label":"No convincing needed","title":"10 reasons radiologists choose OpenWhispr","subtitle":"Use it for the narrow job it ships to do: turn speech into text at the active cursor. Keep the clinical and reporting workflow under your control.","items":[{"title":"Local transcription at the reading station","body":"Choose a local model and, with local-to-cloud fallback off, audio is transcribed on the device without an upload step."},{"title":"Text lands at the active cursor","body":"The desktop app pastes the transcript into the focused editable field. If automatic paste is unavailable, it can leave the text on the clipboard."},{"title":"No reporting-stack replacement","body":"Keep the worklist, viewer, report template and signing system your practice already uses."},{"title":"A dictionary for radiology vocabulary","body":"Add anatomy, eponyms, protocol names and abbreviations as hints for supported transcription and cleanup paths."},{"title":"Keeps working off the hospital network","body":"Once a local model is downloaded, dictation does not need the reading-room network or internet."},{"title":"$$0 local dictation per radiologist","body":"Every desktop plan includes unlimited use of local speech-to-text models."},{"title":"Three desktop operating systems","body":"Run the same desktop dictation app on the operating system your workstation uses."},{"title":"Your IT team can inspect the paste path","body":"Your organization can inspect the transcription and paste paths on GitHub."},{"title":"You read the images. OpenWhispr handles speech to text.","body":"It is not a diagnostic image reader and has no native patient, study or radiology-system context. Agent Mode can use optional screen context, but it does not diagnose or replace review."},{"title":"100+ spoken languages","body":"OpenWhispr supports 100+ languages. Availability and accuracy vary by local model and cloud provider."}]},"faq":{"label":"Before you download","title":"What radiologists ask before testing dictation software.","subtitle":"Straight answers on report fields, terminology, integrations, patient data, offline use and cost.","items":[{"question":"What is the best radiology dictation software?","answer":"It depends on the job. If you need a full reporting platform with patient and study context, templates, structured data, quality checks and enterprise interfaces, use a radiology reporting system built for that. If you want free, local-first speech-to-text that pastes at an active cursor, test OpenWhispr in your actual report field."},{"question":"Does OpenWhispr integrate with PACS, RIS or EHR software?","answer":"No native integration. OpenWhispr does not receive patient or study context from PACS, RIS, EHR, DICOM, HL7 or FHIR. It transcribes speech and pastes text at the active cursor. Optional Agent Mode can use screen context, which is not diagnostic integration. Test your reporting software and virtual desktop."},{"question":"Can it navigate templates or build a structured radiology report?","answer":"OpenWhispr ships no built-in radiology structured-reporting integration, report macros, voice-command field navigation or template logic. Generic Agent Mode can reformat text from your prompt, but it does not know the report fields. Use your existing system template and cursor."},{"question":"How does it handle terminology, numbers and laterality?","answer":"The custom dictionary lets you add recurring anatomy, eponyms, protocol names and abbreviations as recognition hints. It is not a radiology lexicon or clinical error checker. Proofread every report for terminology, measurements, negation and laterality before signing."},{"question":"Can I use it offline with patient information?","answer":"After a local model is downloaded, local speech-to-text runs offline and does not upload audio when the separate local-to-cloud fallback is off. Cleanup is configured separately. Your organization still decides which devices and applications are approved for patient information; cloud transcription is a separate global mode."},{"question":"Is local radiology dictation really free?","answer":"Yes. Every desktop plan includes unlimited local dictation, a custom dictionary and 100+ languages at $0. Pro is $8 per month and adds unlimited OpenWhispr Cloud transcription, more meeting hours, mobile and device sync."}]},"reviews":{"label":"What users say","title":"What users say about local dictation.","subtitle":"These quotes are verbatim. None is rewritten as a radiologist endorsement.","score":{"value":"1","label":"Job in the reporting stack","body":"OpenWhispr supplies speech-to-text at the active cursor. Your reporting system keeps the patient, study, template and signature workflow."},"items":[{"quote":"Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.","name":"0xSero","handle":"@0xSero"},{"quote":"I switched to voice input for my Claude Code chats. Even when I stumble and rephrase mid-sentence, the transcription comes out clean and polished.","name":"j3iiifn","handle":"@j3iiifn"},{"quote":"I needed a local model for when I'm travelling or have no WiFi, and this perfectly solves it. The UI and performance are great.","name":"Gabe","handle":"@gabe__perez"}]},"pricing":{"label":"Pricing","titleLines":["Start free with local dictation.","Upgrade for cloud or team features."],"caption":"Compare OpenWhispr plans for radiologists","tiers":[{"id":"free","name":"Free","badge":"Popular","price":"$$0","priceDetail":"No card details needed","cta":{"label":"Download for free"},"featured":true},{"id":"pro","name":"Pro","cta":{"label":"Start with Pro"}},{"id":"business","name":"Business","cta":{"label":"Start with Business"}}],"rows":[{"label":"Best for","cells":[{"featured":true,"parts":[{"text":"Radiologists who want unlimited local cursor dictation and can run a model on the workstation"}]},{"parts":[{"text":"Radiologists who want unlimited OpenWhispr Cloud transcription"}]},{"parts":[{"text":"Teams that want OpenWhispr Cloud, recordings and priority support"}]}]},{"label":"Main reason to choose","cells":[{"featured":true,"parts":[{"text":"Unlimited local speech-to-text at $0, with audio staying on the workstation"}]},{"parts":[{"text":"Unlimited cloud transcription without managing local models or keys"}]},{"parts":[{"text":"Unlimited recordings and priority support for a team"}]}]},{"label":"Offline speech-to-text","tooltip":"Offline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free.","cells":[{"featured":true,"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]},{"parts":[{"text":"Included","emphasis":"accent"}]}]},{"label":"Cloud speech-to-text","tooltip":"Cloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download.","cells":[{"featured":true,"parts":[{"text":"2,000 words/week","emphasis":"strong"},{"text":" via OpenWhispr Cloud, or unlimited using your own API keys"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"},{"text":" OpenWhispr Cloud speech-to-text"}]}]},{"label":"Meeting recordings","cells":[{"featured":true,"parts":[{"text":"5 hours/month","emphasis":"strong"}]},{"parts":[{"text":"20 hours/month","emphasis":"strong"}]},{"parts":[{"text":"Unlimited","emphasis":"strong"}]}]},{"label":"Privacy & data handling","cells":[{"featured":true,"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]},{"parts":[{"text":"Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored."}]}]},{"label":"Setup experience","cells":[{"featured":true,"parts":[{"text":"Hands-on: install app, choose local models, optionally add API keys"}]},{"parts":[{"text":"Easy: install app, upgrade, use unlimited cloud speech-to-text"}]},{"parts":[{"text":"Moderate: configure users, Agent mode, and data workflows"}]}]},{"label":"Advanced features","cells":[{"featured":true,"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"}]},{"bullets":[{"title":"Custom dictionary","body":" improves names/terms"},{"title":"100+ languages","body":" multilingual speech-to-text"},{"title":"Device sync","body":" use across devices"},{"title":"Mobile companion app","body":" dictate on mobile"},{"title":"API / MCP access","body":" connect tools/workflows"},{"title":"Agent mode","body":" automate actions"},{"title":"Chat over your data","body":" ask about files/transcripts"}]}]},{"label":"Team management","cells":[{"featured":true,"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"None","emphasis":"muted"}]},{"parts":[{"text":"Basic team / workflow use"}]}]},{"label":"Support","cells":[{"featured":true,"parts":[{"text":"Community support"}]},{"parts":[{"text":"Email support"}]},{"parts":[{"text":"Priority support"}]}]}],"enterprise":{"paragraphs":["Need SSO, compliance features, admin controls, or dedicated support for a larger team?","We also offer a custom Enterprise plan with managed onboarding and tailored features."],"link":{"label":"Talk to us","href":"/contact-sales"}}},"about":{"label":"Built in the open","title":"We built the speech-to-text layer, not a radiology stack.","paragraphs":["Radiologists already work across a worklist, image viewer and reporting system. OpenWhispr does one narrower job: turn speech into text and paste it at the active cursor.","We build in public. Your reading-room or IT team can inspect the local transcription and paste paths on GitHub instead of relying only on a product page.","And if a report field will not accept the paste, you talk to the people who built it, in Discord."],"image":{"label":"OpenWhispr founder and community"}},"finalCta":{"tagline":"Free · open source · local-first","title":"Test one dictated finding at the cursor.","body":"Start in a blank text field, then confirm cursor paste in an approved report field before clinical use.","cta":{"label":"Download for free","href":"/download"},"finePrint":"Takes 5 minutes to set up. If it doesn't work, ask us for help in Discord. Free forever on Mac, Windows & Linux."}}}},"now":"$undefined","timeZone":"UTC","children":"$L18"}] 19:I[89554,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"ThemeProvider"] 1a:I[50103,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"GitHubStarsProvider"] 1b:I[28473,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"ConsentProvider"] 1c:I[51627,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"default"] 1d:I[31494,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js","/_next/static/immutable/chunks/01pbao-w8op81.js"],"default"] 1e:I[16508,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"default"] 1f:I[95338,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"default"] 20:I[88667,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"AttributionCapture"] 21:I[24901,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"MetaPixel"] 22:I[21307,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"Rb2b"] 23:I[2355,["/_next/static/immutable/chunks/1p2ndgirh2-it.js","/_next/static/immutable/chunks/3zvxazlv9iyny.js","/_next/static/immutable/chunks/3sqnre1wqw5c3.js","/_next/static/immutable/chunks/3q341afyljkz1.js","/_next/static/immutable/chunks/2mda3sy1qsv_a.js","/_next/static/immutable/chunks/2w6h9kjzbvct2.js"],"Analytics"] 27:I[27201,["/_next/static/immutable/chunks/0r3sg558voqp8.js"],"IconMark"] 18:["$","$L19",null,{"attribute":"class","defaultTheme":"system","enableSystem":true,"disableTransitionOnChange":true,"children":["$","$L1a",null,{"children":[["$","$L1b",null,{"children":[["$","$L1c",null,{}],["$","main",null,{"className":"flex-1","children":["$","$L2",null,{"parallelRouterKey":"children","error":"$undefined","errorStyles":"$undefined","errorScripts"</article></body></html>