I started using Gemini Live API for my Livekit cloud agent. I appreciate the latency and low cost, but it seems to exhibit a “stutter” where it starts a word and is unable to finish it (see example).
The result is definitely uncanny and creepy. I was wondering what might contribute to this and whether there are approaches I can take to remedy the glitch.
Do you have LiveKit Agent Insights enabled? If so, can you download and provide the file? Did you see anythign in the agent log around the time that happened?
I believe that is Gemini sending that audio that way and LiveKit agent frame work is just playing what it received.
Have you checked in their forum for suggestions?
@CWilson I believe this is the case. I don’t see any issues in the logs for that session. And the audio wavform reflects that audio glitch in the insights. You can see the transcript getting cut off. I’ll see if the google forums pose any solution.
audio_glitch.zip (4.7 MB) Here are the insight logs for reference.
