It works differently now, more like version 2.5-04-17 or 2.5-05-20. I feel very uncomfortable with this.
Hello,
To help me understand and address your concerns more effectively, could you please elaborate on what specifically feels different or uncomfortable to you?
Knowing the details will allow me to provide a more accurate and helpful response.
Best Regards
Lalit Kumar
Emotionally flat/unresponsive “Modified by moderator”
Moreover, the topP and topK parameters have no observable effect on this model whatsoever.
Im having the same issue with the live api gemini-2.0-flash. it’s straight up not running inference properly on the video feed and the voice feels flat.
Yes just a little but easy to adjust to
I feel this model has come back today, it’s great ! 
1 Like
I hope your issue has been resolved with this.