For the last two or three weeks I’ve been having problems with the Whisper API. Some transcripts are generated fine, but sometimes the operation is very slow or doesn’t work at all. Now it’s random. The same file may work very well, or not at all. Do you also experience this problem?
You can test with this file for example:
Do you have any idea what might be causing this problem? Is the API temporarily slow?
did a biggish transcription Job and experienced the same. Eventually I used RunPods serverless faster Whisper to get the Job done because of the deadline I had…
Whisper via the API seems to have issues with longer audio clips and can give you results like you are experiencing. You might have better success if you split up the audio into multiple audio clips and then combine after.
Or if you have the hardware, run whisper locally with GPU acceleration.
I use this API for an audio file hosting site, so I can’t ask each user to install Whisper on their own computer ^^. But I’d at least like to know if I’m the only one experiencing this problem or not. Knowing that it’s random, but that billing for “transcribed” minutes is obviously systematic.