Voice dictation for macOS that types into whatever app you already have focused, on your own OpenAI key
Hold a hotkey, say the thing you were going to type, and the transcript lands at your cursor.
brew install --cask mblode/tap/commandmentOr download the latest release. Requires macOS 15 Sequoia or later and an OpenAI API key.
Press Option + Shift + D to open Settings and paste your OpenAI key. It goes into the macOS Keychain.
Then hold Option + D anywhere, say a sentence, and let go. The text appears where you were typing.
- Streaming transcription: audio is captured at 24 kHz mono and streamed to OpenAI's Realtime API, transcribed with
gpt-4o-mini-transcribeby default. - Menu bar only: Commandment runs as a menu bar agent, so there is no dock icon and no window.
- Your key, your bill: the key sits in the macOS Keychain and requests go straight to OpenAI, with nothing in between.
- Accessibility is optional: grant the permission and transcripts type themselves into the focused app. Skip it and they land on your clipboard instead.
| Key | Action |
|---|---|
Option + D |
Start or stop recording |
Option + Shift + D |
Open Settings |
Both are rebindable in Settings.
The repository is a Turborepo with the macOS app in apps/macos, its tests in
apps/macos-tests, and the Next.js site in apps/web.
npm install
npm run web:devRun npm run build and npm test for the whole workspace, or use
npm run macos:build, npm run macos:test, and npm run web:build for one app.
- Commandment updates itself through Sparkle, on EdDSA-signed appcasts, or on demand from Check for Updates... in the menu bar or Settings.
- If the menu bar icon is hidden behind an overflow app,
Option + Shift + Dstill opens Settings.
MIT
Crafted by 