>>2792922I'm not sure what you mean. If you upload a couple usage examples to
https://litterbox.catbox.moe/ we can try to diagnose.
One thing off the top of my head is that the models themselves have different styles. Pic related is the same model applied to several different models.
Some loras just aren't gonna be compatible, say if it's something trained in a realistic style, an anime style model just won't have enough in common to produce anything sensible.
