🙌yes you are right, i kept a cleaner dataset to train only a part of the look. i prefer to add blur or stronger scanlines later. the first experiment which had a "dirtier" dataset didn't turn out so well, so I thought I'd try this

I'd say it's authentic for a new tape. They all looked good/clean on the first play... And after that it's a coin toss on whether the tape gets eaten or not, while the VCR's quality also adds a half-life factor into it. Media consumption was so much more literal back then 😄

You can probably emulate the effect with consecutive vid2vid passes, maybe some acceleration, being a little stingy with the step counts, and I'm sure it will fall apart in no time—or your money back!