Private, on-device voice assistant built with local AI models. Runs entirely offline after initial setup—no cloud, no data sharing.
- Real-time speech-to-text: Fast dictation using distilled Whisper.
- Local text generation: Contextual responses powered by Qwen2.5-0.5B-Instruct.
- Voice-controlled automation: Launch apps and execute workflows via spoken commands.
- Text summarization: Instantly summarize selected text with one hotkey.
- Seamless desktop integration: Auto-types output directly into active windows using xdotool.
Built for speed and privacy on Linux desktops. Lightweight models run efficiently on consumer hardware (CPU/GPU).
Experience fluid, Mac-like voice interaction—completely local and under your control.
pip install torch transformers sounddevice scipy pynput sudo apt install xdotool xclip
Run with python voice_assistant.py. Models auto-download on first launch (internet required once).
Hotkeys:
- Alt+1: Dictation mode
- Alt+2: Generate response from speech
- Alt+4: Launch apps by voice
- Alt+5: Summarize selected text
Designed and built by a GenAI engineer focused on practical, private Linux tools.