Hey there,
From last few hours I am suddenly seeing lots and lots of timeout exceptions and retries when generating TTS via the OpenAI api.
Even for small sentences like “hey, are you there ?”, it’s taking 30-40 seconds, anyone experiencing something similar ?
Below is the model I’m using.
Model: “gpt-4o-mini-tts”
Welcome to the forum!
The first place to check for such is
Then also search the recent topics as often if this is a major problem there will be one topic that gets most of the Me Too replies and/or other topics reported.
Currently yours seems to be the first.
Thanks for reporting.
Note: I am not an OpenAI employee.
Hi,
I could reproduce the issue with gpt-4o-mini-tss.
I tested the older tts models like tts-1 returned quickly, around 1–2 seconds.
I also tested the pinned gpt-4o-mini-tts versions, including gpt-4o-mini-tts-2025-03-20 and gpt-4o-mini-tts-2025-12-15 and those behaved normally too.
The only one where I saw the same delay was the unversioned alias:
gpt-4o-mini-tts
So from my side, this looks less like a general tts/audio issue and more like it may be specific to the current unversioned gpt-4o-mini-tts alias or whatever backend/version it is routing to.
If you contact Support, I’d include timestamps, request IDs if available, model name, input length and retry/timeout details when reporting it to Support or if OpenAI staff asks for examples.
I’m facing the same issue. The difference in delay between gpt-4o-mini-tts and gpt-4o-mini-tts-2025-03-20 is very noticeable.
We faced the same issue - we had to switch to older models!