transcribrai 0.1.0

Speech-to-text application with push-to-talk hotkey for Linux (X11/Wayland)

TranscribrAI

Speech-to-text application with push-to-talk hotkey for Linux (X11/Wayland).

Features

• Push-to-Talk Recording: Hold a hotkey to record, release to transcribe
• Local Transcription: Uses OpenAI Whisper (via faster-whisper) - no cloud required
• Wayland & X11 Support: Works on both display servers
• Configurable Hotkeys: Set your preferred key combination
• System Tray: Minimize to tray for quick access
• German & English: Full language support

Installation

pip install transcribrai

System Dependencies

Fedora/RHEL:

sudo dnf install portaudio-devel ydotool

Ubuntu/Debian:

sudo apt install portaudio19-dev ydotool

Arch Linux:

sudo pacman -S portaudio ydotool

Wayland Setup

For global hotkeys on Wayland, add your user to the input group:

sudo usermod -aG input $USER
# Log out and back in

Start the ydotool daemon for keyboard simulation:

systemctl --user enable --now ydotool

Usage

transcribrai

Or run directly:

python -m transcribrai

Default Hotkey

• Ctrl+Shift+Space: Hold to record, release to transcribe

Configure in Settings or edit ~/.config/transcribrai/config.json.

Configuration

Configuration file: ~/.config/transcribrai/config.json

{
  "hotkey": "ctrl+shift+space",
  "audio": {
    "device_index": null
  },
  "transcription": {
    "model_size": "base",
    "language": "de"
  }
}

Whisper Models

Model	Size	Speed	Accuracy
tiny	75 MB	Fastest	Lower
base	150 MB	Fast	Good
small	500 MB	Medium	Better
medium	1.5 GB	Slow	High
large-v3	3 GB	Slowest	Highest

Requirements

• Python 3.11+
• Linux (X11 or Wayland)
• PipeWire or PulseAudio
• PortAudio library

License

MIT License