Doing some tests with "faster-whisper" and the "large-v3" model on Google Colab, for transcribing the Korean episodes.
Seem to be more accurate than what I was using, or maybe it was just a lucky test.
If it ends up making better subtitles I'll update that S02 MEGA (>>149055500
) and leave a readme file in there.
So even if this thread dies it'll be there, no worries.