PonyXL Version

Base prompt

1girl, jin, brown hair, twin braids, green eyes, green kimono

Second hairstyle

hair bun

Tips & problems

The goal of this model is to mimic the original art style as much as possible.

So the dataset includes only original show screencaps, and since 2/3 of it happens at night and everything is dark, the generated images created may appear faded. Unfortunately, adding night and dark to the negative prompt doesn't really help. Good at generating night scenes

We can fix it by adding more synthetic images that happens at least indoors and not at night.