Problem with NovelAI (and models built upon it) is that everybody's seen it all. With all the shitty NAI generated images getting spammed out there, the brain just defaults it to boring and repetitive, even if it had work put into it, and it's not even an AI thing, plenty of human artists feel the same, it's just that humans are obviously (well, usually) more unique than a model that's built to draw the thing the way it was trained to.
I'd say something like ATitanStrawberry is a bit more robust when it comes to generating unique-ish-looking images (of course, "Uniqueness" lies in the eye of the beholder), and you don't need a lot of work or to be a prompt guru, just need dumb shit like (((Masterpiece, best quality))), [[realistic lighting]], [[[painting]]], [[vibrant]]. Here's some images from the first time i tried out SD i already managed something decent looking in less than 30 generated images and absolutely no editing, so very much beginner-level shit;
And even though i loved the results of the generation, by the 30ish generation i was already bored, because, again, i'd "Seen it all".
It's inevitable, everything looks boring if it's repeated long enough, but i guess if a game ran with a unique genned art it wouldn't really matter since it'd be different scenes and positions and it wouldn't go on long enough for repetition to be a problem.
(Not saying that you should redo this whole game's art, that would just be retarded lmao, just wrote this in the spirit of discussing AI shit cause why not)