Speech to text software for writers & authors
Talk the draft out. Get your words, not an AI's.
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.
Open source · 80,000+ downloads
What you're choosing between
You said the line. It came back as somebody else's.
You want the draft down at the speed you talk. You do not want it back sounding like everyone else's.
The draft comes back in nobody's voice.
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.
Dictate a line of dialogue, get an answer back.
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.
You cannot see what got changed.
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.
An unfinished manuscript in someone else's account.
Choose local transcription when you want your manuscript handled on your own machine. It needs no upload.
What lands on the page
Faster drafts. Still your prose.
Local dictation is unlimited on every plan, runs on your own machine, and gives back the sentences you said.
It types what you said, at your cursor
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.
The raw transcript, kept beside the clean one
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.
Your character names, spelled your way
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.
Your recordings on our servers
None
audio is never stored, and on the local route it never leaves your device
Cost per 100,000 words
$0
local dictation is unlimited on every plan, including Free
Setup
Your pace
download, pick a model, choose a hotkey
Download to first paragraph
Talk it through once. Then edit, not rewrite.
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.
Download for free.
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.
Hold the hotkey and talk.
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.
Read it back, then write.
If a line reads wrong, open the history and check it against the raw transcript.
What it will and will not do
Ten reasons writers keep their own words.
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.
It transcribes, it does not draft.
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.
The AI only composes when you ask it to.
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.
Questions stay questions.
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.
Deliberate bad grammar survives, if you want it to.
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.
You can turn cleanup off.
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.
Your invented vocabulary goes to the transcriber.
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.
Unlimited local dictation, on every plan.
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.
It works offline.
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.
100+ languages, and you can switch mid-draft.
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.
It writes into whatever you already use.
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.
Before you install it
The questions writers ask first.
Straight answers on voice, privacy and cost.
Will it rewrite my prose?
How do I check what it changed?
Will it get my characters' and places' names right?
Is my unpublished manuscript private?
Is it really free for a whole book?
Does it work in Scrivener, Ulysses or Word?
What people tell us
Local models. No WiFi. Better than typing.
From our Discord, GitHub and inboxes. They write code, not chapters, so read them for the tool rather than the trade.
Raw transcript
Available to check
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.
Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.
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.
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.
Pricing
Free forever.
Upgrade when you want more.
FreePopular $0No card details needed | Pro $6.67/user/mo$80/user billed annually Start with Pro | Business $13.33/user/mo$160/user billed annually Start with Business | |
|---|---|---|---|
| Best for | Writers drafting a book who want unlimited, offline dictation for free | Writers who also want cloud transcription without managing a model | Editorial teams and agencies who want unlimited recordings and priority support |
| Main reason to choose | Unlimited local dictation at no cost. Upgrade for unlimited Cloud transcription and workflow features. | Unlimited cloud transcription when you would rather not run a model | Everything in Pro plus unlimited meeting recordings, agent mode and priority support |
| Offline speech-to-textOffline speech-to-text runs entirely on your own device, it works without internet connection, and it's unlimited and free. | Included | Included | Included |
| Cloud speech-to-textCloud speech-to-text sends audio to OpenWhispr Cloud for the highest accuracy and speed, with no local models to download. | 2,000 words/week via OpenWhispr Cloud, or unlimited using your own API keys | Unlimited OpenWhispr Cloud speech-to-text | Unlimited OpenWhispr Cloud speech-to-text |
| Meeting recordings | 5 hours/month | 20 hours/month | Unlimited |
| Privacy & data handling | Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored. | Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored. | Offline audio stays on your device. Cloud audio is sent only to transcribe it, then never stored. |
| Setup experience | Hands-on: install app, choose local models, optionally add API keys | Easy: install app, upgrade, use unlimited cloud speech-to-text | More: unlimited meeting recordings, Agent mode and chat over your data |
| Advanced features |
|
|
|
| Team management | None | None | Basic team / workflow use |
| Support | Community support | Email support | Priority support |
Need SSO, compliance features, admin controls, or dedicated support for a larger team?
Contact us about Enterprise.
Talk to us
Why we built it
The line you got right out loud, first time. We built for that.
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.
Free · open source · local-first
Nothing on the page that you did not say.
Download OpenWhispr, take the local route at setup, and your next chapter gets talked out on your own machine, in your own sentences.
Loading...Free to download on Mac, Windows and Linux.