macOS · v1.0.0-beta.1

Press to speak.
It types for you.

A quiet menu-bar app that turns your voice into text anywhere on your Mac. Hold the right Option key, speak, release — the transcription lands in whatever you were typing in. Works with your own API keys. No accounts, no lock-in.

Universal build · Apple-notarized · Auto-updates · Apple Silicon + Intel

Three keys. That's the whole app.

No launcher, no UI gymnastics. Live in the menu bar, summoned by a hotkey.

1

Hold

Press and hold the right Option key from anywhere — Slack, a Doc, your terminal, an email reply.

2

Speak

Say what you want to write. A discreet overlay shows you it's recording. Release when done.

3

Pasted

Your words appear at the cursor, transcribed and optionally cleaned up by an LLM. That's it.

Built for people who type all day

Fast

Whisper-large-v3-turbo on Groq gives real-time transcription. A two-sentence reply is ready before you've reached for the keyboard.

Smart post-processing

Optional LLM pass cleans up filler words, adds punctuation, and formats lists. Toggle it per-session.

History, always

Your last 500 transcripts, searchable, stored locally. Nothing synced, nothing uploaded.

Stays out of the way

No dock icon, no windows you have to close. A tiny overlay while recording, then silence.

Works anywhere

Any app that accepts text: browsers, editors, chat, Figma, the terminal. It pastes into the focused field.

Notarized & auto-updating

Signed by Apple, shipped as a universal DMG. New versions install silently on the next relaunch.

Bring your own key

Your keys. Your provider. Your bill.

Voice never touches a server we operate. There's no sign-up, no subscription, no telemetry, no "free tier" meant to trap you. Paste your own API key in Settings and the audio goes directly from your Mac to the provider you chose.

No vendor lock-in

Switch providers in Settings and restart. Nothing migrates, because nothing was ever stored with us.

No accounts

There's no backend. You download the app, point it at a provider, and it works. That's the whole onboarding.

No data retention

Audio is sent to your chosen provider and discarded. Transcripts live locally on your disk, where you can inspect them.

Open options

Groq, Sarvam, OpenAI — pick whichever fits your latency, language, and budget. Add more as they're released.

🇮🇳 Made for India

Hindi, Tamil, Marathi, Telugu, Kannada, Bengali, Gujarati — and your accent.

Most dictation tools are trained on American English. When you speak Hinglish in a Bangalore café or dictate an email in Tamil, they choke. Voice ships with first-class support for Sarvam AI — built in India, trained on Indian speech — so you can dictate in the language you actually think in.

  • Saarika v2.5 & Saaras v3 — Sarvam's best STT models, selectable in Settings.
  • 12+ Indian languages including Hindi, Bengali, Tamil, Telugu, Marathi, Gujarati, Kannada, Malayalam, Punjabi, Odia.
  • Code-switching — natural Hinglish, Tanglish, and Kanglish without you toggling anything.
  • Indian accents in English work too. No more the-app-thinks-you-said-"vibe" when you said "five".
Try it with your Sarvam key
Sarvam · Saarika v2.5

Spoken:

"Yaar, meeting ko thoda reschedule kar do. Client ka call aa gaya hai at five thirty."

Transcribed:

Yaar, meeting ko thoda reschedule kar do. Client ka call aa gaya hai at 5:30.

Privacy that's not a marketing slide

What stays on your Mac

  • Transcripts (last 500, in ~/Library/Application Support/voice/)
  • Your API keys (in the same local store, no keychain roundtrip by default)
  • Settings (language, hotkey, provider choices)

What leaves your Mac

  • Audio — sent to the STT provider you configured. That's it.
  • (Optional) Transcript — sent to an LLM for cleanup, if you enable post-processing.
  • Nothing else. No analytics, no crash reports, no "anonymous usage" payload.

Download Voice

Universal macOS build, notarized by Apple. Works on Apple Silicon and Intel.

Download voice-1.0.0-beta.1.dmg

After installing, grant Accessibility + Microphone when prompted. First paste will ask for Automation access — click Allow.

1 Open the DMG, drag Voice into Applications.
2 Launch from Spotlight. Paste your Groq or Sarvam API key in Settings.
3 Hold right Option anywhere. Speak. You're done.