https://dashboard.vapi.ai/calls/019b143f-24d4-700e-8fc2-133914c5aed1 I’m trying to make the voice agent more conversational so that the user can interrupt the agent at any time. However, this doesn’t seem to work, the agent only recognizes user speech after it finishes speaking. I want the user to be able to speak and interrupt the agent at any moment. I’ve tried enabling the Stop Speaking Plan in the advanced settings, but it doesn’t seem to have any effect. Regardless of the configuration, the agent still ignores the user while it’s talking. Has anyone else encountered this issue or found a workaround? https://cdn.discordapp.com/attachments/1449150460291448852/1449150460731592715/image.png?ex=693dda09&is=693c8889&hm=e6f95972903ae3e34c96eaa6c8191cbe891b0b0a457fe7f3c1b57e2e7d0a26c0&

To enable real-time interruption, update your Stop Speaking Plan as follows (set

voiceSeconds

to 0.2 instead of 0):

json
"stopSpeakingPlan": {
  "numWords": 0,
  "voiceSeconds": 0.2,
  "backoffSeconds": 1
}

This configuration uses Voice Activity Detection for fast interruption and is recommended for most conversational use cases. Try this adjustment and test if the agent now responds to user interruptions as expected. Source: See the official documentation for more details on optimal settings and why

voiceSeconds: 0.2

is recommended for VAD-based interruption detection.

It looks like the assistant isn’t yielding mid-utterance, which usually happens when barge-in isn’t triggered at the audio level. The Stop Speaking Plan only works if the provider supports partial-audio interruption and the model is configured to listen while speaking. The next step is to review your TTS/STT combination, ensure barge-in is enabled at the provider layer, and verify the call logs for partial ASR events during output. Can you share which TTS/STT provider you are using for this agent so I can pinpoint where the interruption is being blocked? @DmytroSonOfVolodymyr

Hey! To help track down this issue, could you share: \- The call ID \- When exactly this happened (the timestamp) \- What response you expected to get \- What response you actually got instead This would really help us figure out what went wrong!