search

sanchit-gandhi / whisper-jax

JAX implementation of OpenAI's Whisper model for up to 70x speed-up on TPU.
Apache License 2.0
4.4k stars 375 forks source link

Transcription issues. #190

Open RezaTokhshid opened 6 months ago

RezaTokhshid commented 6 months ago

I've come across some issues while testing the repo. Have you seen any of these issues or do you have a solution for them?

Let me know if any of these ring a bell or if you need more info.

Jrcordal commented 1 month ago

Same, also sometimes a word repeates too much. I was thinking to use regular expression for the strings, did you arrived at a better solution?

I've come across some issues while testing the repo. Have you seen any of these issues or do you have a solution for them?

  • "thank you" or "alright" where it's not said and model stops transcribing after that point
  • Same words at the end of transcription when it's not said (on this one up to that word every thing is transcribed ok)
  • Audio not transcribe. I had a audio that got "So." as transcription. I tested the same with a diff model, your distill large, but it worked ok
  • Last 2-3 words just repeating

Let me know if any of these ring a bell or if you need more info.