OpenAI's new voice model doesn't stop talking to think

Full duplex means a voice model listens and speaks at the same time, over one audio stream that stays open in both directions for the whole call. It hears you while it's talking, and it decides on its own when to reply. Sign up for LiveKit Cloud: https://cloud.livekit.io/signup?utm_source=youtube&utm_medium=video&utm_campaign=devrel&utm_content=yj-K4kvGgsY This video is a voice agent built on OpenAI's GPT-Live-1, running on LiveKit. It's an airline rebooking agent with an ordinary Python function tool that takes a while to come back, so it shows what full duplex actually buys you. The agent keeps listening and talking while that lookup runs, and a request that arrives during the wait gets handled right then. Reasoning and tool calls go to a backend model you choose — OpenAI's GPT-6 Astra or GPT-6 Luna, or your own model and harness, including open source ones. Then we go through the code. You build a GPTLiveModel and pass it to AgentSession as llm, and the rest of the file is a standard LiveKit agent. LiveKit is an open source framework and cloud platform for building realtime voice agents, and we had support for GPT-Live-1 on day one. 📚 Resources 📚 LiveKit GPT-Live docs: https://docs.livekit.io/agents/models/realtime/plugins/gpt-live/?utm_source=youtube&utm_medium=video&utm_campaign=devrel&utm_content=yj-K4kvGgsY Agent docs: https://docs.livekit.io/agents/?utm_source=youtube&utm_medium=video&utm_campaign=devrel&utm_content=yj-K4kvGgsY 🤝 Join the Community: https://community.livekit.io/?utm_source=youtube&utm_medium=video&utm_campaign=devrel&utm_content=yj-K4kvGgsY #livekit #ai #voiceai