m-bain / whisperX

WhisperX: Automatic Speech Recognition with Word-level Timestamps (& Diarization)
BSD 2-Clause "Simplified" License
11.55k stars 1.22k forks source link

subtitles contain non-English text when task=translate,but faster-whisper subtitles is only english【es audio translate】 #593

Open pxEkin opened 10 months ago

pxEkin commented 10 months ago

audio: Video-es-24.zip

faster-whisper: image

whisperX with vad_onset = vad_offset = 0.0 【vad_onset=0.50 and vad_offset = 0.363 have similar results】 image

I think they deserve the same results. Can anyone know the reason?

KossaiSbai commented 9 months ago

Hey there I would be interested in taking charge of this issue