French transcription is really poor

#1
by BenTouss - opened

I don't know if this is known, but for French, the persormances of Distil-Whisper is really far behind whisper.
Maybe the 'distillation' (I think it is removing some parts of decoder) removes the multilanguage capability of whisper.

Whisper Distillation org

Hi there. As stated in the README and the paper:

Note: Distil-Whisper is currently only available for English speech recognition. Multilingual support will be provided soon.

Hope that answers your question!

Oh yes, sorry I missed that.
Thanks a lot for the explanation

BenTouss changed discussion status to closed
Whisper Distillation org

Thanks @Xenova for the swift clarification! We recently released training code that should facilitate you to train a Distil-Whisper model in your choice of language: https://www.linkedin.com/feed/update/urn:li:activity:7131004471806980096/

Sign up or log in to comment