search

m-bain / whisperX

WhisperX: Automatic Speech Recognition with Word-level Timestamps (& Diarization)
BSD 2-Clause "Simplified" License
12.61k stars 1.33k forks source link

Fail to generate segment #920

Closed leinace1001 closed 1 week ago

leinace1001 commented 1 week ago

When I call model.transcribe, it seems that some chunks are failed to segment, or just like after text normalization. Instead of outputting sentences, it just outputs a sequence of words. Like:

um yes just in terms of manifestation and believing in myself and you know daring to want the things that I genuinely want instead of settling for what is realistic

Anyway to fix it?

leinace1001 commented 1 week ago

Saved by providing an initial prompt with sentences. i.e. Um, yes. Just in terms of manifestation and believing in myself, and you know, daring to want the things that I genuinely want instead of settling for what is realistic.