It's very likely that the results won't be as desired. Anime is nearly always a synthetic image which is intended to be clean and geometrically based (even if there are gradients and more real world additions; it's a synthetic).

The application domain for this includes any other sort of abstract logical synthesis, charts and maybe videogames (even ones that look realistic).

Real world content also has sharp boundaries between objects, and whatever part happens to do that work might be shared, but within objects fuzzier is probably better. IIRC someone was making an AI assisted upscaling of DS9 which would probably be closer to a generic algorithm for 'filmed' content.

Not that great:

Input: https://i.imgur.com/YLYnxx9.jpg

Output: https://i.imgur.com/5DLHoSi.jpg

Anime and cartoons have very specific qualities that allow for these types of techniques to be effective (as other reply explains).

An algorithm that applied upscaling to a picture of a person would eventually have to find a way to draw skin cells.