Call ID: 3fd6e07d-3f4d-46d8-a882-d73072fa8d62 Time: between 1:41 and 1:53 User talks with no interruption from Assistant but in the transcriptio it appears overlapped.

@giorgior27 I have reviewed your call recording and examined the logs. It appears that there is an endpointing issue that requires attention. Specifically, I recommend increasing the duration for which the assistant waits for the user to finish speaking. This adjustment will allow for a more seamless transcription flow. Currently, the transcripts are appearing out of order due to the system detecting that the user has stopped speaking prematurely. I suggest configuring the start speaking plan with custom endpointing rules to ensure that the system waits for the user to complete their speech, utilizing REGEX as the endpointing parameter. You can learn more about custom endpointing from the provided resources. Please let me know if this information is helpful, and I would be happy to incorporate it into the documentation.  https://gist.github.com/bajajcodes/cd664c81344ef9f27a4d304de43bd313

Thank you very much for that, I will have a look and see if I get better results. The problem is that I tried a bit playing with these values but sometimes it waits too much after the user has stopped and some other times it talks over. Maybe is a problem about latency? That is not fixed.

Can you send a new call_id so that we can take a look into it?

Sure but I still didn't do what you suggested. I played with these values from dashboard before switching to api. So I was asking if latency could impact that. I will do what suggested above and run few more tests. Then I will give you new id calls

Yes, it will impact the latency a little bit but it will solve your issue.