We received some error from the /v1/audio/transcriptions API. For example,
{"code":524,"message":"Gateway timeout.","param":null,"type":"cf_gateway_timeout"}
The time from sending the request to receiving the 524 response is about 100 seconds. Considering that the Whisper API supports audio files up to 25 MB, 100 seconds might not be enough to process such a large file.
I’m getting the same thing. It seems not to be about file size, but about the length of the audio.
It truly is random. Some long audio files will not trigger the cf_gateway_timeout, while other shorter ones will.
Even splitting audio files down to shorter length seems to have little effect.
We need some official statement from OpenAI fast, it’s ridiculous
Having this problem too. It happens randomly.
Seems like OpenAI have fixed the issue. All failed (cf_gateway_timeout) transcription requests now succeed.
Do wish we were not charged for these failed requests though…
I still have this issue for relatively big files, happens randomly. does not happen for smaller files.