Trained on NAI.

V2 has fewer lower bodies (and even you can not see legs and crotch) .

V2的下半身更少（甚至看不到腿和裆部）。

V1 sometimes looks like the girl is standing in the pool in the middle of the flower.

V1有时候像是女孩站在花朵中心的池子里。

===============

She does not have lower half of human body.

她没有人类的下半身。

===============

V1：

Some Negative Embeddings（for examplpe EasyNegative）and ckpt models focus on drawing normal human body, which might influence the output.

某些负面的Embedding（比如EasyNegative）或者模型专注画正常的人类身体，这可能会影响结果。

It seems that AnythingV4.5 can bring excellent result?!

似乎 AnythingV4.5的效果很好？！