Anyone have a recent update here? Still experiencing this but seeing this thread is old so wondering if others found a fix or have just learned to live with it.
2 Likes
This bug has not been resolved yet. It is still reproducing even with the latest model, 2025-06-03.
Giving agent an instruction like pause at the end of sentence might be a possible workaround so far.
I am also still getting this issue on the latest model, I am using websockets. Has anyone found a reliable workaround?
I had an issue when using echo cancellation on a custom transport - it was interfering with VAD. Disabling it helped a lot with random cut-offs.
1 Like
Iām seeing the same behavior on the WebSocket connection. The audio stream is not generated for the final text response. Has anyone found a reliable fix for this issue?
Still having this issue - can see acknowledgements/small sentences at the end of utterances in our logged transcript, but a lot of the time the voice cuts off just before that final sentence as if itās structured like that. any danger of devs sorting this out lol itās been 16 months since the issue was raisedā¦and weirdly we did not experience it a single time with the old August model, and we used it daily for 4 months, but with the most recent december model it happens regularlyā¦
someone pls share a creative working solution
@OpenAI_Support Can you please fix this on realtime? The issue is still very much happening.
Did anyone find a working hack/fix?