man murmur
Documentation
Getting Started
Murmur is a voice dictation gadget for Linux. Hold a hotkey, speak, and text appears at your cursor in any application.
1. Install
2. First run
On first launch, Murmur will download the Whisper Tiny model (~75MB). This is a one-time download.
3. Dictate
Press Ctrl+Shift+Space and hold. Speak naturally. Release to transcribe. Text appears at your cursor.
Settings
Click the gear icon on the Comm Badge to open settings. Settings are stored at ~/.local/share/murmur/settings.json.
Global keyboard shortcut for push-to-talk. Default: Ctrl+Shift+Space
Whisper model size. Tiny (~3s), Base (~8s), Small (~20s). Auto-downloads on first use.
Customise the Comm Badge glow colour via hue rotation.
Select your widget style. Currently: Gemini v1. More coming.
Hotkey Reference
| Action | Default |
|---|---|
| Push-to-talk (hold) | Ctrl+Shift+Space |
| Open settings | Click gear icon |
FAQ
Where are models stored?
~/.local/share/murmur/models/. Models auto-download from Hugging Face on first use.
Can I use my own Whisper model?
Not yet. Custom model paths are planned for a future release.
Does it work without internet?
Yes, 100%. After the initial model download, Murmur never makes network requests.
Why xdotool?
xdotool uses the XTEST extension to simulate keystrokes, which works in any X11 application. Wayland support via ydotool is planned.
How do I report a bug?
Open an issue at github.com/murmurlinux/murmur/issues with your distro, desktop environment, and steps to reproduce.