Hi,
We have a working solution using gpt-realtime form open ai. Works great,
But we would like to move over to Gemini as the voices are better and prices a lot more attractive.
We have been testing with the “Gemini-2.5-flash-native-audio-preview-09-2025” model, when it works it sounds great but it has been very unreliable with random connection-timeout errors etc. We are using the api key from Google ai studio so not the Vertex one.
Anyone that has had any experience similar and found ways to solve it?