Hello,
I’m reporting a persistent audio-output issue with Gemini 3.1 Flash Live Preview in Google AI Studio Live.
Model:
gemini-3.1-flash-live-preview
Issue:
During Live conversations, the model’s voice output intermittently becomes heavily distorted and produces a high-pitched, glitchy sound resembling old dial-up modem noise. When this occurs, the speech becomes difficult or impossible to understand.
I have reproduced the issue under multiple conditions:
- Multiple preset voices, including Leda
- Arabic conversations
- English conversations
- Desktop/laptop
- Mobile
- Multiple separate Live sessions
Changing the selected voice does not resolve the issue.
The same behavior occurring across different devices, languages, voices, and sessions suggests that this may not be limited to a local audio-output configuration.
Date first observed:
31 August 2026
I have also captured a screen recording that clearly demonstrates the audio distortion and can provide the video if it would be useful for investigation.
I have already submitted the issue through the Google AI Studio “Send feedback” function.
Could you please confirm whether this is a known issue with gemini-3.1-flash-live-preview, and whether any recent backend/model updates could be related?
If there is a recommended workaround or alternative Live model to use temporarily, I would appreciate the guidance.
Thank you,
Hussain