I am using the realtime api (no beta header) and around 8:20AM EST it stopped working. No audio transcription, and only a single input audio in traces. Yesterday everything was working great. Did something get updated, or is it under maintenance? Anyone else experiencing no voice input/output?
Using gpt-realtime model, via websocket, no beta header.
Confirmed, simply using the “audio” playground in the dashboard, which uses websockets:
text input: OK
audio input: no response, failures, extreme lag.
It seems the issue and behavior is worse with stateful conversation history. The first may or may not succeed, then after, mostly an API spinning its wheels and not triggering or generating, or extreme waits that start running older turns after more attempts at interruption. Same with other models.
Couldn’t confirm the issue right away.
Can you share the session ID?
trace_sess_CDEX8wnd9H01MiPK6WUko
I am correctly updating the session, getting session updated response from OpenAI. Requesting a completion, but nothing. Doesn’t respond to my speech, but seems to be getting my audio input according to the traces.
And of course it was my account balance. After logging events I realized that the failures were balance related.
Disregard this entire post.
@_j : do you still have this issue on the platform?
