Good luck, I've never found AI art to be consistent enough for my tastes.
PonyXL is.. pretty consistent. If you don't confuse it with prompts, and match the prompt to the image size, provide enough steps, it'll get it right 95% of the time. It's several times slower than SD 1 though.
If masks were used to generate images, it'd work flawlessly though.
Biggest issue is stuff like posing, multi-character images and all that, though I think people who are more than amateurs could get even that done.
Ofc, FC's style of image generation would be pointless, as last time I played the game, MC's state was a one-dimensional affair..