Seriously, could anyone shed a little more light on techniques to get "cute" stuff instead of generic AI styles? Making images that look like
>>4364787 is dead easy, but I've yet to figure out how to reliably make anything that looks like
>>4365243,
>>4365601, or even
>>4367606.
I mainly use Animagine, which recognizes quite a few artists with enough representation on danbooru, so that works for aping a very specific style (pic rel). But on the other hand, danbooru tags are generally terrible for describing the aesthetic qualities of an image like artstyle, complexity, detail, shading, etc, so there's no way to prompt a more nuanced style.
I've found LoRAs to mostly be worthless, but it also seems like there aren't many anime models for SDXL with positive reviews.
What am I missing? Do I just need a more thorough knowledge of artists with cute styles?
