nanowhisper 1.0.0

Local voice dictation: Whisper STT + Claude LLM post-processing

nanowhisper

Local voice dictation for Linux. Press a hotkey, speak, and your words appear as typed text in any application.

Pipeline: Microphone → Whisper (local STT) → Claude (optional cleanup) → typed into active window

Install

Requires uv.

uv tool install nanowhisper
nanowhisper setup

The setup command handles everything: system packages, display server detection (X11/Wayland), keyboard permissions, config file, API key, and starts nanowhisper as a background service.

After setup, log out and back in for keyboard access to take effect.

Configure

Edit ~/.config/nanowhisper/config.toml to change the hotkey, whisper model, output method, etc. The file is well-commented.

Usage

nanowhisper runs as a systemd service after setup. Manage it with:

systemctl --user status nanowhisper     # check status
journalctl --user -u nanowhisper -f     # view logs
systemctl --user restart nanowhisper    # restart after config changes
systemctl --user stop nanowhisper       # stop

Or run manually in the foreground:

nanowhisper --preload

Troubleshooting

Problem	Fix
No keyboard found	sudo usermod -aG input $USER, then log out/in
Text not appearing	Check [output] method matches your display server
First dictation slow	Use --preload or the systemd service (does this by default)
Mic permission denied	sudo usermod -aG audio $USER, then log out/in
LLM not working	Check ANTHROPIC_API_KEY in ~/.config/nanowhisper/.env

Uninstall

systemctl --user disable --now nanowhisper
uv tool uninstall nanowhisper
rm -rf ~/.config/nanowhisper ~/.config/systemd/user/nanowhisper.service