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

0.9~1.0 can achieve relatively good results

The character model is highly reductive, but does have flexibility. It was trained using about 50 images (including a few fan creations). The training set contains some high contrast, animated images. So this was avoided in the build.

I know hairstyles can be a pain to deal with, so you can add her thin ponytail when you need it, or leave it out if you don't need it to avoid the wrong hairstyle.

There are many more troublesome issues, which I will deal with again if I have time.

Note that this is still in the testing phase!!!