Hi, We're using Vapi for voice assistant and noticing unexpected behaviour with transcript messages. Issue: We're receiving partial transcripts for the assistant role while the TTS is playing:

{type: 'transcript', role: 'assistant', transcriptType: 'partial', transcript: 'Hello.'}
{type: 'transcript', role: 'assistant', transcriptType: 'partial', transcript: 'Hello. This is...'}

Expected Behavior: Since the LLM generates the complete text before TTS starts speaking, we expected only final transcripts for the assistant role (not incremental partial transcripts). Questions: - Why are partial transcripts being sent for assistant speech? Is Vapi running speech-to-text on its own TTS output? - Is there a configuration option to disable partial transcripts for the assistant role? - Is this intended behavior for alignment/monitoring purposes? For user speech, partial transcripts make sense. Our Current Setup: Model: GPT-4o Voice Provider: 11labs (eleven_flash_v2_5) Transcriber: Deepgram Nova-3 Thanks for your help!

Hi Rahul Gupta, If you want to disable partial transcripts for the assistant role during TTS, you can configure your setup accordingly. For the ElevenLabs transcriber, there is an option to set

disablePartialTranscripts

to

true

. This will ensure that only the final transcripts are delivered. You can refer to the respective transcriber documentation to adjust other settings or modify the transcriber configuration as needed: - [ElevenLabs Transcriber Configuration](https://docs.vapi.ai/api-reference/squads/create) For any issues related to receiving partial transcripts unexpectedly, ensure that your configuration aligns with the expected behavior described in the documentation.