i seen ppl posting lora with styles thats already in anima model, this one isn't exception.

its good at drawing stars ✨ and leafs 🍁. and paper looking something.

settings is usual,
ER-SDE-Solver or euler/ddim
strength vary from 0.6 up to 1.5, depend how much paper u want.
BUT less strength corrupts text, keep it in mind (i used it at str = 1)

errors and cursed stuff 🤩 (rarely)
- may draw silhouette instead of char
- can make char tanned (lul)
- sometime dont want draw text 😡
- struggle with multiple chars in shot

lil note: it was one of the first loras i trained (month ago 😨), overtrain was intended feature as i tried to call style without trigger (and it didnt work).

with all downsides i really love that iteration of lora and dont want to lose it, will be safer on civitai servers (maybe?)