Persistent Error with no message
#1
by
MicahB
- opened
Whenever I'm trying to use this, the attempt simply errors with no explanation of what went wrong.
For your testing, here's the audio I'm using:
Fixed. Whisper model accept audios < 30 sec by default, now should be ok.
Additionally, I found that the audio you provided serves as a strong bad case for this model. It might be due to the high number of repeated words, which causes earlier pitch markers to affect subsequent recognition. Thank you for providing this example; it will be valuable for us as a reference for further improvement. Appreciate your help!
Thank you for your fast response! Incidentally, this data is from an American student learning Japanese, so it might not be the best use case of this model anyway.