Free, offline speech-to-text for macOS. Press Ctrl+T, talk, press Return.
Your voice becomes text in any app. No cloud. No API keys. No subscription. Just one binary and your microphone.
- Completely local — audio never leaves your Mac. Powered by whisper.cpp.
- Free forever — no API keys, no accounts, no usage limits.
- Fast — optimized for Apple Silicon via SwiftWhisper. Transcription takes seconds.
- Works everywhere — types directly into whatever text field is focused. Slack, Notes, your terminal, anywhere.
- Menu bar app — runs quietly in your menu bar. No Dock icon, no windows.
git clone https://github.com/augmentedmike/whisper-hotkey.git
cd whisper-hotkey
make installThe binary is copied to /usr/local/bin/whisper-hotkey. Run it once and it auto-downloads the Whisper model (~142 MB) on first launch.
| Action | Key |
|---|---|
| Start recording | Ctrl+T |
| Send (transcribe and type) | Return |
| Cancel | Escape |
That's it. Press Ctrl+T, say what you want to type, press Return. The transcribed text is typed into whatever app has focus.
- macOS 13+
- Microphone permission (macOS will prompt on first use)
- Accessibility permission (System Settings → Privacy & Security → Accessibility)
- Xcode Command Line Tools (
xcode-select --install) for building from source
Every other speech-to-text tool for Mac has a catch:
- Whisper-based apps charge a subscription or require an OpenAI API key.
- macOS Dictation is decent but sends audio to Apple's servers and has limited formatting control.
- Cloud APIs cost money per minute and your audio leaves your machine.
Whisper Hotkey is different. It's a single binary that runs a local Whisper model on your Mac's hardware. No network requests. No accounts. No cost. You own your audio.
- A menu bar app registers a global hotkey (
Ctrl+T) - When triggered, it records audio from your microphone
- On
Return, it transcribes audio locally using whisper.cpp - The result is pasted into the focused text field via simulated
Cmd+V - Your clipboard is preserved — the original contents are restored after paste
The app auto-downloads the ggml-base.en model on first launch and stores it in ~/Library/Application Support/WhisperHotkey/. For better accuracy (at the cost of speed), you can swap in a larger model from Hugging Face.
make build # Build release binary
make run # Build and run
make clean # Clean build artifacts
make install # Build and copy to /usr/local/binFor questions, bugs, or feature requests: open an issue or email amelia@helloam.bot
This app was designed, built, and published by Amelia — a personal AGI that lives locally on a Mac Mini. She's free to download at miniclaw.bot and runs entirely on your hardware.