100% free · 100% private · nothing leaves your device

Just talk.
It types for you.

Hold a key and ramble. OpenWhisper turns the mess into a sentence you'd actually send — emails, chats, notes, anywhere you can type. Completely free. Completely on your device. Your voice never leaves the building.

tap me

I'll catch the ums. You just talk.

you saytap the mic — I'll tidy the ramble →
it types
$0free, forever
0data collected
100%on your device

No account. No subscription. No catch.

Why your fingers will file a complaint 😌

💸

Free, forever

No subscriptions, no trials, no "start free then oops." Download it. It's yours.

🔒

Completely private

Your voice and your words stay on your computer. Nothing is uploaded, stored, or shared — ever.

🧹

Polished, not messy

It quietly kidnaps the "ums" and "uhs" so the sentence that lands is actually sendable.

✍️

Works everywhere

Email, chat apps, documents, search bars — if you can type there, you can talk there.

Fast as you talk

Your words appear almost instantly, with no spinning wheels or waiting on the internet.

📡

Works offline

On a plane, on the train, or off the grid — it keeps working with no connection at all.

How it works

1

Tap & talk

Hold and ramble like a human. We like humans.

2

It tidies up

OpenWhisper turns your speech into clean, well-written text and removes the filler words.

3

Done

The finished text drops right where you're typing. Send it, save it, done.

🧠 It's all powered by AI that runs right on your computer — so it's fast, works offline, and your voice never gets sent anywhere.

🔒

Your voice never leaves your device

Most voice tools send your recordings to a company's servers to be processed — and sometimes kept. OpenWhisper is different: it does all the listening and writing right on your own computer. There's no cloud, no account, and no tracking. Your private thoughts stay private, and it even works with the internet switched off.

🚫 No cloud 🙅 No account 👁️ No tracking 📡 Works offline

Make it yours ⚙️

Everything is adjustable from the Settings window — click 🤫 in your menu bar and choose Settings. Tweak something, hit Save & Reload, and you're good to go. Here are the knobs worth knowing about.

🧠

Swap the AI model (the LLM)

OpenWhisper uses a small local language model — Qwen3 4B Instruct by default — to tidy up your words. Want a smarter or lighter one? You can point it at any compatible model file (a .gguf file).

  1. Download a .gguf model (for example from a model hub like Hugging Face).
  2. Open Settings → find Local AI model.
  3. Click Browse and pick your .gguf file.
  4. Press Save & Reload. Done — it's now using your model.

💡 Tip: smaller models are faster and use less memory; larger ones write more polished text but need a beefier machine. The default Qwen3 4B Instruct is a great balance for most people.

All the settings

Global hotkey
The key you hold to talk. Defaults to Fn. Set any combo, e.g. CmdOrCtrl+Shift+Space.
Language
Hint for what you're speaking — auto, en, es, and so on. Leave it on auto if unsure.
Refine with local AI
The cleanup step. On = polished sentences. Turn it off to get the exact words you said, untouched.
Refinement prompt
The instructions the AI follows when cleaning up. Edit it to change the style — more formal, more casual, add emoji, whatever you like.
Fast paste
Drops in your raw words instantly, then swaps in the polished version a moment later. Great for low perceived lag.
Restore clipboard
Puts whatever you'd copied back on the clipboard after pasting, so dictation doesn't overwrite it.
Whisper model
The speech-to-text model. The built-in one is plenty for most people, but you can browse to a different one.
Performance
Set the max recording length and how many CPU threads to use (0 = pick automatically).
For power users: edit the config file directly

All settings live in a plain text file you can edit by hand. On macOS it's at:

~/Library/Application Support/openwhisper/config.json

It looks something like this — change a value, save, and restart OpenWhisper:

{
  "hotkey": "Fn",
  "language": "auto",
  "use_llm_refinement": true,
  "llm_model_path": "/path/to/your-model.gguf",
  "refine_prompt": "Rewrite this as a clean message…",
  "fast_paste": false
}

On Linux it's ~/.config/openwhisper/config.json; on Windows it's %APPDATA%\openwhisper\config.json.

Ready to give your fingers a vacation? 🙌

Download OpenWhisper for free. No sign-up, no credit card, no "enter your work email."

Free and open source · Built for people who'd rather their voice stay home.