Does anything I dictate leave my machine?

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.

Does it work in a terminal, in Cursor and in VS Code?

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.

Can I write code by voice with it?

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.

Does dictation software for developers get library names and CLI flags right?

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.

Why not just wire up whisper.cpp myself?

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.

Is it actually free, and how does it compare to Wispr Flow?

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.

Dictation software for developers & engineers

Talk the change through. Get the prompt typed.

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.

Open source · 80,000+ downloads

Dictate into any app, including:

CursorVS CodeClaude CodeGitHubLinearSlack

What you're putting up with

So much of your typing isn't code.

Prompts, pull request descriptions, commit messages, review comments, standup updates. None of it is code. All of it is typing.

The prompt takes longer than the run.

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.

You already wired Whisper to a hotkey once.

Then the OS update broke the paste, the venv rotted, or you never got the clipboard back. OpenWhispr is that script, finished and maintained.

A privacy claim you can't check.

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.

Built-in dictation has never heard of kubectl.

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.

What the hotkey actually does

A paragraph of context. None of it typed.

Transcription runs on your machine. So can the cleanup. Neither costs anything.

Person speaking beside an open laptop at a shared desk.

Whisper runs on your machine, not ours

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.

Person gesturing while using a keyboard at a dimly lit desk.

One hotkey, any window

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.

Person speaking at a workstation in an office at dusk.

The system prompt is yours to edit

The prompt that cleans up your dictation ships as text you can open, rewrite, test against sample input and save.

Audio uploaded in local mode

None

the recording is transcribed on your own machine

Cost per developer

unlimited local dictation, every machine you own, forever

Setup takes

A few steps

download, pick a model, bind a hotkey, talk

What the install involves

Talking to your agent starts here.

No account, no card, and no telemetry unless you switch it on.

Person speaking at a home desk with a keyboard, laptop and monitor.

Download it free.

Mac, Windows or Linux. Setup offers OpenWhispr Cloud or local models. Pick local and it pulls them down once.

Hit the hotkey and talk.

Describe the change you want the way you would say it to a colleague at the next desk.

Read it, then send it.

The clean text is already at your cursor. Edit it, hit enter, and the recording is yours to keep or delete.

What you can check yourself

Ten reasons, and the repo is the citation for most of them.

Choosing dictation software as a developer comes down to what you'll take on trust. These are the parts you don't have to.

The audio is transcribed on your own disk

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.

Free on every machine you own

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.

MIT licensed, so fork it

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.

It types into a terminal

It pastes where your cursor is, including in a terminal. There's no editor plugin to install or maintain.

Your identifiers, spelled your way

Put library names, service names and the CLI tools you live in into the custom dictionary for sharper transcripts.

The cleanup prompt is editable

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.

It runs with the wifi off

A plane, a train, a customer's locked-down network. Local models need no signal once they're on disk.

A CLI, and reading your notes is free

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.

Telemetry is off unless you turn it on

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.

100+ languages

Standups, reviews and issue threads in the language your team actually speaks, transcribed on your own machine.

Before you install it

Six questions developers ask first.

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.

What people tell us

The people who found it before you did.

What people tell us.

Recordings kept on our servers

With a local transcription model the audio never leaves your machine. Cloud audio is sent only to be transcribed, then never stored.

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.

j3iiifn@j3iiifn

Started using OpenWhispr with my local models and hardware. I have to say, this is better than typing.

0xSero@0xSero

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.

Gabe@gabe__perez

Pricing

Free forever.
Upgrade when you want more.

Compare OpenWhispr plans for developers
	FreePopular $0No card details needed Loading...	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	Developers who want free, private, offline dictation and are fine with a local setup	Developers who want unlimited cloud transcription and MCP in their AI tools	Teams documenting and reviewing at volume
Main reason to choose	Unlimited local dictation, with the audio staying on your own machine	Unlimited cloud transcription plus API and MCP access, without managing local models	Unlimited recordings plus Agent mode and chat over your own notes
Offline speech-to-text	Included	Included	Included
Cloud speech-to-text	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	Moderate: configure users, Agent mode, and data workflows
Advanced features	Custom dictionary improves names/terms 100+ languages multilingual speech-to-text	Custom dictionary improves names/terms 100+ languages multilingual speech-to-text Device sync use across devices Mobile companion app dictate on mobile API / MCP access connect tools/workflows	Custom dictionary improves names/terms 100+ languages multilingual speech-to-text Device sync use across devices Mobile companion app dictate on mobile API / MCP access connect tools/workflows Agent mode automate actions Chat over your data ask about files/transcripts
Team management	None	None	For individuals and teams
Support	Community support	Email support	Priority support

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.

Talk to us

Why we built it

We wanted this for our own terminal first.

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.

$0 · open source · no extension to install

Say the next prompt. Don't type it.

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.

Free forever on Mac, Windows & Linux.