Audio popping/glitching issues when generating long-form audio (over 2 minutes) in Google AI Studio

Google AI Studio
1

Title: Audio popping/glitch issue during long speech generation (>2 minutes) in Google AI Studio

Category: Audio Generation / API Quality / Web Console

Description: I am experiencing a consistent audio glitch and popping/clicking noise issue when generating long audio outputs (approximately 2–3 minutes) using the audio generation models in Google AI Studio.

  • Issue Details:

    1. Popping / Clicking Glitches: During long continuous generation, audio artifacts, popping sounds, and sudden jumps occur repeatedly throughout the playback.

    2. Reproducibility: This is not a client hardware or network performance issue, as the exact same glitches reproduce consistently across multiple different PCs and network environments.

    3. Timing: The issue becomes significantly more noticeable when the generated content exceeds 1.5 to 2 minutes.

  • Suspected Causes:

    • Phase mismatch or click artifacts between audio token chunks during autoregressive generation.

    • Audio buffer under-run or packet drop issues in the web streaming decoder when handling extended outputs.

  • Requested Improvements:

    • Implement smoother zero-crossing or cross-fade blending between audio chunks during streaming synthesis.

    • Increase buffer stability and timeout margins for long-form audio generation in the AI Studio web UI.

    • Stabilize attention/state accumulation for extended outputs (>2 minutes) to prevent audio degradation.

Thank you for your continuous support and updates.

2

Hello @taejin ,

We have just launched the new Gemini 3.8 Flash TTS (gemini-3.8-flash-tts) and Gemini 3.8 Flash-Lite TTS (gemini-3.8-flash-lite-tts) models: Google AI Studio

Could you please test with gemini-3.8-flash-tts and let us know if you are still experiencing the same audio issue?