Following the streaming implmentation here https://ai.google.dev/gemini-api/docs/speech-generation#stream
It seems that there is no proper implementation for streaming doesn't save much latency compared to non-streaming solution.

Could you please make an example or double check this?