Hi all.
I’m testing Live API with BidiGenerateContentRealtimeInput but, compared to OpenAI gpt-realtime, there is a HUGE different in speed. OpenAI reply in almost less than a second, always, Gemini, when it’s fast, is not less than 2-3 seconds and slow down a lot when the conversation length increase.
I’m using ti to test a phone agent and having 2-3 seconds of total silence between responses is unacceptable. Should I check something ? The websocket is running on a Google Cloud Run instance.