I realize I’m late to this thread, but if you’re serious about archiving a VHS in the best manner possible, you have to go the RF capture route: https://github.com/oyvindln/vhs-decode
This method effectively captures the “raw” signal stored on the tape, allowing you to convert it after you’ve captured it however you see fit. You don’t have to worry about cheap digitizers/capture cards/etc distorting the signal.
just a guess, but in order for an LLM to generate or draw anything it needs source material in the form of training data. For copyrighted characters this would mean OpenAI would be willingly feeding their LLM copyrighted images which would likely open them up to legal action.