If you ♥ it, I will update it in the future.

0.8~1.0 can achieve relatively good results

The character model is highly reproducible, but not very flexible, and seems to be a bit overfitted; it uses about 140 images (including some fan-created artwork) for training. The training set contains some images that are high contrast, or contain watermarks. So to avoid this in the build, you guys need to use the

((text)), ((watermark)), ((logo)), ((chinese text)), ((twitter username)), ((artist name)),twintails, high contrast, severed hair, tail animation, animated, anime screencap.

I know it's a pain in the ass to deal with hair ....... Because she has a short blade tied to the end of her braid, I'm having a hard time getting the ai to be flexible enough to learn it, you could add dagger tied to hair, but it wouldn't work very well, sorry.

Here are the appearance types:

1. normal: meihuashisan,1girl,very long hair,bodysuit,dress,green eyes,braided ponytail,

2. mask: plus mask can be

3. Vests: remove dress and bodysuit and add sundress

4. long hair: remove braided ponytail

5. As a child: meihuashisan (baby), child, others plus normal and remove braided ponytail.

Also use a cape, or red eyes to show she's fighting.

There are still a lot of niggling issues that I will deal with if I have time.

Please note that this is still in beta!!!!