New Gemini Live API "Native audio output" models not supporting System Instructions

Gemini API
, ,
1

The newly Native audio output models in Gemini Live API not follows Systen Instructions, please fix it.

2

I am an audio output user too.. Can you please elaborate on what do you mean by not supporting system instructions?

4

So basically the model “Gemini 2.5 Flash Preview Native Audio Dialog” not support or not follow “System instructions” like:

config: {
      responseModalities: [Modality.AUDIO],
         speechConfig: {
             languageCode: 'en-US',
             voiceConfig: {
                  prebuiltVoiceConfig: {
                  voiceName: 'Puck',
             },
         },
},
systemInstruction: {
      parts: [{
          text: `You are an good AI.`,
          }]
      }
}

So in this case the model “Gemini 2.5 Flash Preview Native Audio Dialog” not follows “You are an good AI.”

6

I wanna echo this, when I specify in the instructions to speak Chinese or Japanese it just ignores and replies in English

8

Hi @Samuel_Hvidager,

Can you post how you are using, system instructions and error that you received so that we will try to reproduce the issue and escalate to the appropriate team.