search

huggingface / distil-whisper

Distilled variant of Whisper for speech recognition. 6x faster, 50% smaller, within 1% word error rate.
MIT License
3.33k stars 238 forks source link

Sentence separation #31

Open Flossertoday opened 7 months ago

Flossertoday commented 7 months ago

Hey there,

I've noticed that distil-whisper tends to chop sentences in half. It'd be great if it could wrap sentences properly, especially after commas and periods. Switching to word mode and using Spacy sounds a bit like an overkill. Any suggestions on how to fix this?

sanchit-gandhi commented 7 months ago

Hey @Flossertoday! Do you have an example audio on which Distil-Whisper does this? Happy to take a look at doing some hyper-parameter tuning to see whether we can mitigate this

murdadesmaeeli commented 6 months ago

Hi @Flossertoday , any update the problem that you mentioned?