Disable interruptions for audio streaming for multimodal live api

Gemini API
1

Hi,

I was wondering if there was a way to disable interruptions for audio streaming for multimodal live api.

Thanks

1 Like
3

Hi,

Actually I was looking for the opposite of this, meaning, I wanted interruption to work in the multimodal live api. However, by default this functionality is not working for me using the python genai library. (Posted about it here:
Interrupting Gemini 2 Flash Multimodal Live API seem not to work as expected)

Does the output audio stream from the model get interrupted for you by default when it detects an input? Also, does the ‘interrupted’ flag get set in the ‘server_content’ dict from the server? If so, could you please share your config?

Thanks!

5

Yes to both. However, I’m currently using the api through my own web socket connection, not through the genai library, similar to here:

7

Thanks for the info. Will try that out.

Regarding your requirement of uninterrupted audio streaming, guess you can try sending data
through BidiGenerateContentRealtimeInput instead of BidiGenerateContentClientContent, which according to this below link does not interrupt the model generation.

Cheers!

1 Like