Speak your coding prompts instead of typing them. Press Ctrl+Shift+V, say what you want, press again — your prompt lands in Claude Code, Copilot, or any AI assistant. Optionally enhance it with AI before sending.
Features
Voice-to-Text — Press Ctrl+Shift+V to start, speak, press again to stop. Text appears in your AI assistant's input box.
AI Enhancement — Click the Enhance button to improve your prompt with context-aware AI (knows your file, errors, git state, framework).
Multi-Provider — Use Gemini (free), OpenAI, Anthropic, or Groq for enhancement. Switch providers with one click in the status bar.
Multi-Target — Toggle between Claude Code and Copilot Chat right from the status bar.
Zero UI — No panels, no browser windows. Everything happens via the status bar.
Private — All voice processing runs locally (Moonshine ONNX). API keys stored in OS keychain. No backend, no tracking.
Quick Start
Install this extension from the Marketplace
Install Python dependency (one-time):
pip install useful-moonshine-onnx
The speech model (~400MB) downloads automatically on first use — no manual setup needed.
Set up enhancement (optional): Click the provider name in the status bar (e.g. Gemini) → pick a provider → paste your API key
Use it:
Ctrl+Shift+V → speak your prompt → Ctrl+Shift+V → text injected
Click Enhance in the status bar to improve it with AI
Status Bar
After installation, you'll see these items in the bottom-right status bar:
Item
What it does
Enhance
Click to enhance the last spoken prompt with AI
Gemini
Shows current LLM provider — click to switch
→ Claude Code
Shows injection target — click to toggle
Prother
Shows state (Idle/Listening/Processing) — click to start recording
