Open stasbel opened 3 months ago
Thanks for the report!
thanks what about predicting multiple languages and corresponding probs? why "no speech is detected"? audio example clearly has some :)
@ZachNagengast hi, is it on track to be completed anytime soon?
@stasbel This looks like it was actually a tiny oversight in the filter logic and not related to the sampling issues we just put out a fix for. Please check main
with commit https://github.com/argmaxinc/WhisperKit/commit/c93d613c8c6fc3ec5b9d1d9da5d1f4206183c5e4 for the fix.
@stasbel Checking on this, is your issue resolved with the latest change?
version:
0.7.2
snippet to reproduce:file: 0.wav.zip
tiny
version gives:which is not all languages and has negative prob
base
version gives:which is just plain error somewhere inside whisperkit