This checkpoint is not good at all. Avoid it.

It'a reactive to prompts like an anime, and this is a positive point.

But the big problem is the quality of the output, where images details are always "stressed out", like when you have two LORAs in conflict on the same image, just to give the idea. Color contrast is exaggerated, making the result even worse.
I suspect is trained in a wrong way, adding incompatible LORAs and pretending this would be a new "style"...
I've generated dozen of images and, honestly, I dropped all...