I played around with the AI function a little and it works mostly ok and sometimes very good.
AI transcribing depends heavily on the audio quality of the YouTube track. Sometimes that quality s#cks!
So I am now uploading my own YouTube Videos .
I create my own videos in a batch with my original FLACS (lossless compression) using this script:
ffmpeg -loop 1 -i image.jpg -i input.flac -c:v libx264 -acodec flac -vf “pad=ceil(iw/2)*2:ceil(ih/2)*2” -shortest output_image.mp4
You need ffmpeg and some image. I just use a small white jpg.
Of course you can also try your own MP3 or whatever you have - potentially your audio tracks are better than the rubbish found on YouTube.
Example: