I see only two requests , is this a persistent issue? how frequently have you encountered this ?
I will test more, now.
What seems to be a problem is that when the assistant goes off to search the internet, he completes the search, has the result already, but keeps silent so that the user would have to ask him again if he is ready to answer the request he already received before and has completed the search. He does not reply back on his own… for example.
For Czech, the speech sounds ok but he mixes word sequences.
What is a big problem is that the woman 's voice speaks as he would be a man in Czech, which is completely wrong.
It’s Friday night, the API seems more stable today thaen last 2 times I tested it.
This is from a demonstration to a customer yestedrday, the Gemini 2.5 flash dropped dead mid demo call again. (Disconnected, started fresh.) It is NOT reliable at all.
do you still see this issue ?
Hello,
I am not using Gemini, till you put it on open GA, like fully on Vortex. For the delay I went to Deepgram Live Audio Agents. I did a Flash 2.0 based AI call center operator presentation to my clients in February - it was a success, then 2.0 messed up in March and April. Then you skip Voice Live Audio for 2.0 - never to production, replaced by the 2.5 flash preview. Not in production now. Will you ever release that? I asked Google rep in Vertex to allow that GA in my API, never happened. Be using it when in Vertex, my testing is over. You dropped the ball on this. This is my feedback as of today. Thanks for asking.
Matej