We are buiding a coversational application using gemini-live-2.5-flash-native-audio. Although there has always been lag significantly(!!) worse than the lovely marketing promises, it has become unusable over the last week - 10 days. No code changes have been made around our speech handling, but to remove all possible variables I went to Vertex AI studio in our cloud console to give it a run. I am in Australia, and I see lag of 8-10+ seconds between responses there too. In the last week, that has been up to 20-30 seconds.
I have seen many lag reports in here, but few answers. Clearly this service isn’t production ready, but it would be helpful to get some clarity around what is going on and what the plan is.
Thanks
Lidija