2023-07-20, 10:25 PM
I've been looking into what you mentioned and man is it a rabbit hole. I'm a data engineer by trade, so AI upscaling is not so technically complex that I couldn't see myself figuring it out, but the time investment would be significant.
I suspect that innovative developers will consolidate/formalize these processes into a more user-friendly UI that will allow you to feed in some video, receive back 10 examples of how it might look, select the undesirable examples to eliminate, then keep repeating that process until an optimal model is made which can be saved and used on the media in question.
Shame it's not more user-friendly at this point, but I bet it's coming. I still have so much media to extract and transcode that delving into the world of upscaling would be a bit overwhelming right now.
I suspect that innovative developers will consolidate/formalize these processes into a more user-friendly UI that will allow you to feed in some video, receive back 10 examples of how it might look, select the undesirable examples to eliminate, then keep repeating that process until an optimal model is made which can be saved and used on the media in question.
Shame it's not more user-friendly at this point, but I bet it's coming. I still have so much media to extract and transcode that delving into the world of upscaling would be a bit overwhelming right now.
New to media handling? Check out my guide: From Disc to Drive: A Beginner's Guide to Preparing Your Media for Jellyfin