Real-time voice AI agent that listens and responds while the user is speaking
Build a voice agent with bidirectional audio, streaming responses, interruption handling, and code-controlled actions such as playing, skipping, or pausing music.
From Google Cloud Tech — Build a real-time voice AI agent with Gemini Live API at 01:40
Problem: Text-to-speech systems can speak but cannot hear the user, while ordinary request-and-response interactions do not support continuous, interruptible conversation.
Examples
- Mira: a demonstrated voice agent that plays dream pop, skips a track when interrupted, discusses the music, and answers while the call remains live.