Can't ML just do that directly? #3
Description
I saw this before, and it got me thinking some of the key criteria for NotebookLM spam https://github.com/ajayyy/SponsorBlock/wiki/Dataset-Uses
- The voice of the audio is based on the default settings, rather than being "custom" like uberduck-esque comedy videos
- The length of the video is suspicious, the velocity of speech is all weird relative to the content
- The content is full of filler since it has no nuance/surprise, and often the dialogue roles get mixed up in podcasts
- The background of the video is either AI-generated (not even prompted with care), clip art, or sometimes "gameplay"
How hard would it be to automate this?