Gemini Live audio
Google’s new Gemini 3.8 Live models can be tested through a browser voice UI built directly on its WebSocket API.
Willison says Google released Gemini 3.8 Live and 3.8 Live Extended Thinking, comparing them in shape to OpenAI’s GPT-Live family. He used GPT-6 Astra Extra High to build a no-library web interface for selecting a model, choosing a voice, adding an optional system prompt, and holding an interruptible voice conversation. The implementation connects to Google’s bidirectional Gemini WebSocket endpoint and uses the Web Audio API for capture and playback. Source: Simon Willison's note
Willison says Google released Gemini 3.8 Live and 3.8 Live Extended Thinking, comparing them in shape to OpenAI’s GPT-Live family. He used GPT-6 Astra Extra High to build a no-library web interface for selecting a model, choosing a voice, adding an optional system prompt, and holding an interruptible voice conversation. The implementation connects to Google’s bidirectional Gemini WebSocket endpoint and uses the Web Audio API for capture and playback. Source: Simon Willison's note
score 8