We talked about how LiveKit’s open-source infrastructure became the audio transport layer for ChatGPT’s voice mode, and what it takes to build reliable, production-grade voice agents.

Topics include: – Voice vs. text pipelines (cascade vs audio-in/out) – Turn detection and latency problems – Ambient computing and full-duplex models – Why LiveKit open sourced its stack

(Would love feedback, *especially* from folks building real-time or AI UX systems.

Youtube: https://www.youtube.com/watch?v=QD7pgrk1egA Spotify: https://open.spotify.com/episode/6v7qP0GQjE9kHAxYIzZV8h?si=ce2b07e7bc4f4c4e