使用的是wainsfwillustriousv140训练，所以理论上用这个模型效果最好

底模本来是自带naga_u的画风和兽耳酱的角色的，但是不太还原，就做了这个模型

这个模型既是画风也可以是角色。画风用naga_u，角色用kemomimi-chan_(naga_u)

角色可能会需要这两个tag的帮助：sleeves past wrists, sleeves past fingers

经过测试，这个模型有小概率会把角色皮肤变成小麦色的，这时候负面写一个black skin就行了（也许过段时间会去找原因，应该是打标出问题了）

ps：只能说不愧是兽耳酱，他的图里不管出现什么怪东西都很合理，有效解决了sd乱出东西的问题

Since it was trained on Wainsfwillustrious v1.40, this model should theoretically give the best results when used with that base.

The base model already comes with the naga_u art style and the kemomimi-chan character, but the reproduction wasn’t accurate enough, so this model was trained to improve that.

This model works both as a style and as a character:

use naga_u for the style, and kemomimi-chan_(naga_u) for the character.

The character may also benefit from these two tags:

sleeves past wrists, sleeves past fingers.

In testing, the model has a small chance of turning the character’s skin into a tan / wheat-colored tone. When that happens, just add “black skin” to the negative prompt.

(I’ll probably investigate the cause later — it’s most likely a tagging issue in the training data.)

P.S. Honestly, that’s exactly what you’d expect from kemomimi-chan — no matter how weird something looks in the image, it somehow still feels “correct”, which helps reduce Stable Diffusion’s tendency to generate random nonsense.