whatever I say its market as IVR activity. why so?...
# ask-ai
s
whatever I say its market as IVR activity. why so?
IVRActivity: silence detected; sending notification
✅ 1
t
Thanks for reaching out, Gurkirat. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
Thanks — I understand you're seeing every utterance marked as IVR activity with the log:
IVRActivity: silence detected; sending notification
. I want to help get this fixed. To diagnose why speech is being labeled as IVR activity I need a few details so I can reproduce and pinpoint the cause: • Which LiveKit product/SDK and exact version are you using (browser JS SDK, native iOS/Android, server-side component, or a third-party client like Limechat)? • Exact steps to reproduce and the moment the log appears (e.g., right after join, after speaking, when silence begins). • The full, contiguous logs (timestamps) around the event — ~30 seconds before and after the
IVRActivity
message. • Does this happen for every participant and every call, or only specific users/rooms/devices? • Any custom VAD/IVR settings you changed (thresholds, silence timeout, language/model selection)? • Are any automated prompts, prerecorded audio, or DTMF tones present during the call? • Brief description or short sample of the audio where it misclassifies speech as silence (background noise level, short utterances, overlapping speech, codec used). • Room/connection metadata: region, protocol (WebRTC), browser and version or device OS, and codecs in use. • A reproducible test case (room name and approximate timestamps or participant IDs) I can use to correlate logs. If you prefer, you can redact sensitive fields in logs or audio — just let me know what you remove. Next steps once you provide those items: • I’ll look for patterns in the logs and audio to determine whether VAD thresholds, codec issues, or a timing/timeout are causing early silence detection. • If it’s a configuration issue I’ll give exact settings to change. If it’s a bug, I’ll explain a short-term workaround and the next steps I’ll take to escalate to engineering. Which SDK/version are you on and can you attach the logs or a short audio snippet? , with interactive elements