Live mode — real-time voice conversation
Live mode opens a bidirectional audio channel with the agent: you speak, it answers out loud, with no dictate-then-read latency.
Using it
- In the chat, click the microphone 🎙️ (Live panel).
- Allow microphone access in the browser.
- Speak naturally — interrupt whenever you want, the agent stops.
- Close the panel to return to text chat; session context is preserved.
What live can do
- Answer your questions (same knowledge as the text agent).
- Understand French, English, Arabic… (automatic detection).
- Great for: hands-free brainstorming, hearing/interview preparation, assisted dictation, accessibility.
Technical & privacy
- Authenticated WebSocket transport (
/agent/v1/live) — real-time voice model (Gemini Live). - Audio is streamed only for the session's duration; no listening outside a session.
- Credit consumption based on duration/tokens, like the chat.
Tip: live shines on mobile — the platform runs in the phone's browser.