11/24 Update:

version for illustrious XL based model is available now.

Trigger Words:

alp

4/2 Update:

new Locon version available now.

Trained on animagineXLV3.1

weight around 0.8 ~ 1.0 works just fine

stabler and much smaller in size.

Trained on animagineXLV3

weight around 0.8 ~ 1.0 works just fine

Might increase mistakes in drawing hands, but I think it's acceptable.

Train with this free colab notebook:

kohya-trainer/kohya-LoRA-trainer-XL.ipynb at main · Linaqruf/kohya-trainer · GitHub

with almost conpletely default setting, in free tier

底模为 animagineXLV3

权重0.8~1.0左右都很好

使用之后有可能画手会更容易出错但是体感还好，所以凑合用吧（

使用的以下colab脚本：

kohya-trainer/kohya-LoRA-trainer-XL.ipynb at main · Linaqruf/kohya-trainer · GitHub

几乎全用的默认参数，免费用户也可以白嫖