search

m-bain / whisperX

WhisperX: Automatic Speech Recognition with Word-level Timestamps (& Diarization)
BSD 2-Clause "Simplified" License
11.69k stars 1.24k forks source link

I need Filler words in transcription! #693

Open vladgrand2 opened 7 months ago

vladgrand2 commented 7 months ago

Hi! When I use key: --initial_prompt "эмм, вот, ага, ммм, угу, эээ..., ааа..." my transcription is done with filler words, but entire sentences are arbitrarily lost.

it is very close to this topic: https://github.com/openai/whisper/discussions/1594

maybe there is some other way to return the filler words back to the text?

realies commented 1 month ago

try disabling text normalisation

LaurinmyReha commented 3 weeks ago

checkout this whisper variant. Especially in English this should accurately detect fillers and give completely verbatim transcripts.

https://github.com/nyrahealth/CrisperWhisper

fznrs commented 2 weeks ago

checkout this whisper variant. Especially in English this should accurately detect fillers and give completely verbatim transcripts.

https://github.com/nyrahealth/CrisperWhisper

@LaurinmyReha do you know if there's any way to use the transcription from CrisperWhisper in WhisperX in order to get a diarized transcription?

vladgrand2 commented 2 weeks ago

try disabling text normalisation

Can you explain how I can do that?