It does do original things once you understand prompt language combinatin of positive & negative can do many things with the right steps & other tweaks. Then you train it from constant generations. Did you ever try to make half of a sky one thing then the other half another? Have you ever practiced precise detailing in one spot like an anime chick but there's a lot more intricate details & more picture going on in the eyes?

Have you ever tried to generate a reflection? Someone can be looking at something & the only way you can see it is from the reflection on their body?