currently we are using silero VAD to detect speech: https://github.com/livekit/agents/blob/main/livekit-plugins/livekit-plugins-silero/livekit/plugins/silero/va

currently we are using silero VAD to detect speech: https://github.com/livekit/agents/blob/main/livekit-plugins/livekit-plugins-silero/livekit/plugins/silero/vad.py https://github.com/livekit/agents/blob/main/livekit-plugins/... it works well for voice activity; though it doesn't always detect end-of-turn correctly (humans often pause mid-sentence to think). we are working on improving this behavior.