We often experience that the model stops speaking in the middle of its response. The log shows “Turn detected” although VAD is disabled. Any ideas? Is there a better way to force the model to finish it’s response or is the cascading model the more reliable approach?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| 2.5 flash audio native - output broken in DE | 7 | 279 | August 25, 2025 | |
| Gemini 2.5 Native Dialog audio problems | 19 | 872 | September 1, 2025 | |
| Is it possible to manually set end of turns for the live api when inputing audio? | 1 | 72 | June 5, 2025 | |
| Gemini 2.5 flash audio - not reliable, the official demo agent dropping dead 50% of time within minutes | 5 | 140 | August 25, 2025 | |
| Error: "Gemini Live session closed: Precondition check failed." | 1 | 141 | July 22, 2025 |