I've been generating a lot of stuff (and seeing some stuff posted here by others) that is straying a little from SR7's SU style - mostly by being
too good! Too detailed, too well-shaded etc. This is
especially true in the backgrounds. To give SR7 his due, he
can do it to that standard if he wants, but he's aping Bruce Timm's DCAU which is mass-market animation so the level of detail is inherently low.
For some reason the usual SDXL SR7 LoRAs can allow too much of the subject LoRA through if it's not trained on a DCAU image source and don't simplify the art enough. I've tried using a DCAU-specific LoRA, but that has limited effect.
Fortunately, I've found that a
You must be registered to see the links
LoRA works pretty well.
Imagine that Lex decides he's "
Gotta catch 'em all!" and captures Misty from Pokémon. i.e. the LoRA is not trained on DCAU images.
No 'Detail Slider', or 'Detail Slider' set to 1:
View attachment 3694884
Very nice, but too much detail.
'Detail Slider' set to -2:
View attachment 3694888
Face, body & hair are not changed much, just a little more stylised in places. But the background is
much closer to SU.
Something to mull over, anyway.