Streamer AI (Like Neuro-Sama)

Funny thing you can do just base.en (better than the large models) and make a basic ai for transcription quality if you have a slow computer

So the ai determines what is “best”

Whisper transcribing 96000 samples…

Whisper result type: <class ‘dict’>

Whisper dict result: ‘What would they do again?’

Transcription quality (avg_logprob): -0.70

Whisper transcribing 96000 samples…

Whisper result type: <class ‘dict’>

Whisper dict result: ‘I think I’m…’

Transcription quality (avg_logprob): -0.69

Whisper transcribing 319904 samples…

Whisper result type: <class ‘dict’>

Whisper dict result: ‘What was I doing again? I think I’m chill out now or something.’

Transcription quality (avg_logprob): -0.52

You: What was I doing again? I think I’m chill out now or something.

(Actually what i said: You: What was I doing again? I think I’ma chill out now or something.)

That’s what i did personally and it is better than just using the normal system as it is wrong a lot of the time

-bro77xp

(Also look ali the ai jumpscare - but it’s more like a signing off photo)