Just tested it with a hindi meeting recording that I had lying around. The model kept repeating the word 'Okay' the entire time. Is there a way to select which whisper model is used. Otherwise its only good for english transcripts, even then I doubt it will be able to do it as I believe the Whisper model being used is small.
Thats one of the limitations of small Whisper models, which often loop like that on non-English content.
How does this work? What does "Whisper-class" model mean? Does it have to download a full model?