whisper-hotkey

Double-tap Control to dictate with Whisper on macOS. Transcribes your speech and pastes it into whatever text field is focused.

Uses faster-whisper (CTranslate2) with int8 quantization for fast local inference. No cloud APIs, everything runs on your machine.

Install

uvx whisper-hotkey install

This will:

1. Ask you to choose a model (tiny.en for speed, or distil-large-v3 for accuracy)
2. Download the model
3. Request microphone and accessibility permissions
4. Install a LaunchAgent so it starts automatically on login

Usage

After install, just double-tap the Control key to start recording. Double-tap again to stop — your speech is transcribed and pasted into the active text field.

That's it. It works across all apps, survives reboots, and runs in the background.

Commands

whisper-hotkey install    # Interactive setup (model, permissions, auto-start)
whisper-hotkey uninstall  # Remove auto-start
whisper-hotkey status     # Check if the service is running
whisper-hotkey             # Run the service directly (not needed after install)

Requirements

• macOS (Apple Silicon or Intel)
• Python 3.12+
• uv (recommended) or pip

Models

Permissions

The app needs two macOS permissions:

• Microphone — to record your voice
• Accessibility — to detect the keyboard shortcut and paste text

The install command will prompt for these. If the shortcut doesn't work, check System Settings > Privacy & Security and make sure both your terminal and Python have Accessibility access.

Logs

tail -f ~/Library/Logs/whisper-hotkey/stdout.log
tail -f ~/Library/Logs/whisper-hotkey/stderr.log

License

MIT