Brief
Gemini 3.8 Live models expose speech-to-speech via WebSocket endpoint
Google released two new speech-to-speech models, Gemini 3.8 Live and 3.8 Live Extended Thinking, and a developer built a browser UI to try them. The useful integration detail is a WebSocket endpoint for bidirectional audio.
Google released Gemini 3.8 Live and 3.8 Live Extended Thinking today, two speech-to-speech models similar in shape to OpenAI's GPT-Live models. A developer used GPT-6 Astra Extra High to build a web UI for trying the models, selecting a model and voice preset, entering an optional system prompt, and starting a voice conversation through the browser, with the ability to interrupt the model while it is talking.
The implementation uses no libraries. It connects to a WebSocket endpoint at wss://generativelanguage.googleapis.com/ws/google.ai.generativelanguage.v1alpha.GenerativeService.BidiGenerateContent?key=... and uses a Web Audio API AudioContext for both capture and playback. For builders, this shows a direct integration path for adding real-time voice interaction to a workflow step without third-party libraries.
Our reading
Our reading is that the WebSocket endpoint and no-library approach offer a concrete integration pattern for builders who want to add interruptible voice conversations to a workflow.
Source details and supporting facts
Each line is stated by the page named above it.
Stated by simonwillison.net
- Google released Gemini 3.8 Live and 3.8 Live Extended Thinking today.
- The two models are speech-to-speech models that are a similar shape to OpenAI's GPT-Live models.
- The web UI implementation uses no libraries.
- It connects to the wss://generativelanguage.googleapis.com/ws/google.ai.generativelanguage.v1alpha.GenerativeService.BidiGenerateContent?key=... WebSocket endpoint.
- It uses a Web Audio API AudioContext for both capture and playback.
- The UI allows selecting a model and voice preset, entering an optional system prompt, and starting a voice conversation through the browser, including the ability to interrupt the model while it is talking.
Sources
- Simon Willison's WeblogText stored 16 September 2026
How this story was checked. Written from the 1 page listed above, stored 16 September 2026; claims checked against that stored text on 16 September 2026.
What that means
- 6 of 6 reported statements were confirmed against the page that carries them; the rest were removed rather than published.
- Figures in the text were required to appear in the stored source text: yes. Identifiers: yes.
- The check reads stored text only: no claim rests on a fresh look that did not happen.
- Where the reporting was silent, the text says so instead of filling the gap.