Gemini 2.5 Flash Native Audio — Telephony Compatibility & Output Sample Rate Support

[smeetagrawal]

Hi,
I've been integrating the Gemini 2.5 Flash Native Audio model with telephony vendors (Plivo, Exotel). The model accepts input at 16 kHz and produces output at 24 kHz. While 16 kHz input works seamlessly with telephony, I'm observing noticeable latency on the output side at 24 kHz.
The model works well over browser-based WebRTC, but the telephony experience is impacted. I have two questions:

Is the Gemini 2.5 Flash Native Audio model intended to support telephony use cases, or is it currently optimized primarily for browser/WebRTC?

Are there plans to support lower output sample rates (16 kHz / 8 kHz)? I understand this would reduce audio quality, but it would significantly improve latency for telephony and unlock a wide range of voice-based use cases.

[Varun-S10]

Hi @smeetagrawal, To help us check this latency issue, could you please tell us if you are using a Gemini API key or Vertex AI? Could you please also share your google-genai version and a minimal reproducible code snippet?

[github-actions]

This issue has been marked as stale because it has been open for 7 days with no activity. It will be closed in 2 days if no further activity occurs.