Wait what? 😆 🤣

[Its3rr0rsWRLD]

large-v3

1 00:00:00,000 --> 00:00:09,960 Images that have notified everyone who left the group in July at 06 hours in the morning and realize that June also marked the beginning of Spring Ultra schedule.

Lil.Peep.-.sixteen.lines.og.version.Official.Video.-.AudioTrimmer.com.mp4

🤣

[jongwook]

real GPT-1 vibe

[Its3rr0rsWRLD]

Fr

[ryanheise]

I unfortunately couldn't manage to get anything so elaborate using your file, just "Thank you."

@Its3rr0rsWRLD If you manage to be able to reproduce this in particular with the --word_timestamps True option enabled, would you be able to share the .json output file? I'd like to test a theory that might help to improve the detection and filtering of hallucinations.

[Its3rr0rsWRLD]

Hello!

For whisper, I used a program called WhisperGUI because I was just using whisper for fun and did not want to set it up fully and deal with python, as python just doesn't like my pc. WhisperGUI, from what I have seen, does not have a repository.

I am curious though, so I will set up whisper in a more controlled and monitored environment and test it there.

Thank you for taking the time to share your thoughts and suggestions.

3rr0r.

Edit:

I also forgot that the source code is all in the folder. 🤦‍♂️

I want to apologize for my indecisiveness and lack of assistance, I have the flu right now and I woke up to a GitHub notification lol.

Model: Large-v2
OL: English
Device: 4070 Ti

I'll just give the whole program. https://qiwi.gg/file/7HMW1695-WhisperGUI

config_user.json

[
  [
    "name",
    "user"
  ],
  [
    "modelBox",
    9
  ],
  [
    "translateBox",
    21
  ],
  [
    "deviceList",
    0
  ],
  [
    "translateToEng",
    0
  ],
  [
    "btnTxt",
    0
  ],
  [
    "btnVtt",
    0
  ],
  [
    "btnSrt",
    1
  ]
]

[ryanheise]

Ah, I didn't realise it was a different program, so even if I were to fix something in whisper, it would not fix anything in WhisperGUI since it's a different program.

Anyway, the output JSON I was talking about is obtained by running whisper --model large-v2 --word_timestamps True file.mp3