What’s a good min_endpointing_delay for non-telephony based systems that people like using?

Thanks for reaching out, Alexander. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

and what about min_silence_duration?

You're asking for a recommended min_silence_duration for non-telephony systems. Recommended value: 0.55 seconds (550 ms) — the Silero VAD default and a reliable baseline. Tuning guidance: • 0.3–0.5s — more responsive, higher risk of prematurely ending turns. • 0.7–1.2s — fewer false endpoints, feels slower. • In noisy or fast-speaker environments, try 0.4–0.5s. • For best results, keep

min_silence_duration

around 0.55s and combine it with the

livekit-plugins-turn-detector

(transformer-based) to reduce interruptions while allowing shorter silence thresholds. Would you like a suggested test plan or sample values to A/B test in your app? Sources: How does end-of-utterance detection work in conversations? | Silero VAD plugin | LiveKit docs | Agents Framework | livekit/livekit_composite | DeepWiki | Agents (Python) | livekit/livekit_composite | DeepWiki | Improving voice AI's turn detection with transformers