Can't make the model start speaking with Gemini Live and the latest model

I was trying to make the model say a greeting once the gemini live session was created instead of waiting for user to speak.

Following the documentation I was sending a client content message

LiveSendClientContentParameters content = LiveSendClientContentParameters.builder()
.turns(Content.builder()
.parts(Part.builder().text("Hello").build())
.role("user")
.build())
.turnComplete(true)
.build();
geminiSession.sendClientContent(content);

Nothing I did could make it work, I lost 2 days trying different strategies, different session configurations and different system instructions.

Today I decided to change to an older model. I was using gemini-2.5-flash-native-audio-preview-09-2025 and decided to give an older model a try like gemini-live-2.5-flash-preview.

I worked on the first try! I tried with gemini-2.0-flash-exp and BAM!! It worked like a charmed. Switched back to the latest state-of-the-art model gemini-2.5-flash-native-audio-preview-09-2025 and I got nothing, the model never said a word

Is anyone experiencing something similar or is it just me?

Cheers

Hi @paulotaylor,
Thanks for reporting the issue. Can you please provide the minimal reproducible code, so that we can reproduce it on our end and identify the cause of this issue.