LiveOptions tuning - Spanish diarization issues in a Google Meet

[ivandda]

I'm having issues with Deepgram's real-time transcription in a Google Meet using Python, specifically with speaker diarization.

The diarization sometimes adds non-existent speakers and has inconsistent speaker recognition. I think that it might be caused by the LiveOptions parameters.

Are there some recommended parameters for this use case? Specially for endpointing and utterance_end_ms.

Current setup:
options = LiveOptions(
model="nova-2",
language="es",
smart_format=True,
punctuate=True,
diarize=True,
channels=1,
interim_results=True,
utterance_end_ms="1100",
vad_events=True,
endpointing=500,
encoding="linear16"
)

Thanks!

[deepgram-community[bot]]

Thanks for asking your question. Please be sure to reply with as much detail as possible so the community can assist you efficiently.
_{Consider joining our Discord community for more opportunity to engage with your fellow Deepgram users. You can earn points which can be redeemed for cool stuff by being active in our communities!}

[deepgram-community[bot]]

It looks like we're missing some important information to help debug your issue. Would you mind providing us with the following details in a reply?

• A request ID that triggered your error or issue.

[ivandda]

6460f705-8374-4ef4-9441-5be33a5b7f56

[deepgram-community[bot]]

Combining Deepgram's Multichannel and Diarization features can provide very specific, useful information about the people speaking in multiple audio channels. This might help your situation, check our docs on this topic. Hope this helps!

https://developers.deepgram.com/docs/multichannel-vs-diarization

This message was sent by John Vajda from Deepgram, via our community automation.

[ivandda]

Hi, my audio is not multichannel. So I can't use of that feature.

Do you have any example of code that shows how to use Live Audio + diarization in the PythonSDK? So that I can see how to implement the on_message function to properly save the speaker diarization. I don't see any example like that on the docs or repo.
Ty