a screenshot of edges2cats, an old image model trained to make a realistic textured cat from a line drawing, making patchwork fur-eyed abominations out of lineart of Felix the Cat, Garfield, Heathcliff, Stimpy, Sylvester the Cat, the Cheshire Cat, Snagglepuss, and Blaze the Cat.
The whole DLSS 5 thing (which is bad and insulting) reminded me of this cheeky edges2cats compilation I did ages ago.
Models only spit out results based on what their training data knows. If their model was only trained on realistic cats, for example, it can't understand your artistic intent.