The training materials for this version of the v3 model are four times larger than those of v2, and they can better reproduce the texture of the League of Legends illustrations.

After testing, the League of Legends style is very good: strong lighting and shadow effects create an atmosphere, and detailed hand-drawn texture processing.

Sampler recommendation: Any will do, and it has good applicability.

Recommended number of iterations: around 30

It is recommended to enable high-definition restoration: The zooming algorithm is R-ESRGAN 4x + Anime6B, with a zoom factor of 2 and a re-drawing amplitude of 0.3.

ADetailer is activated. The model recommended is: face_yolov8n.pt

The CFG value is: 5.5

CLIP recommendation: 2

VAE recommendation: It can be enabled or disabled. The model comes with VAE encoding. You can also enable sdxl_vae, and the results are all good.

The generalization is very good. I hope everyone can visit more often and give more likes. Thank you!

This model is only for communication and learning purposes and cannot be used commercially. It is available for download on all platforms. If you wish to integrate it, please indicate the original model to avoid any unnecessary issues.

这一版v3模型的训练素材比v2的素材集多4倍，能更好还原英雄联盟插画质感

经过测试英雄联盟风格非常不错：强烈光影塑造氛围 和 细腻的手绘质感处理

采样器推荐：任意都可以，适用性还不错

迭代步数推荐：30左右

推荐开启高清修复：放大算法为R-ESRGAN 4x+ Anime6B，放大2倍，重绘幅度为0.3

ADetailer开启，AD模型推荐：face_yolov8n.pt，

CFG为：5.5

CLIP推荐：2

VAE推荐：可开启也可不开，模型自带vae编码，也可以开启sdxl_vae,效果都不错。

泛化性很好希望大家多返图多点赞，谢谢

此模型仅用作交流和学习，不可商用，在各个平台开放下载，如果融合请标注原模型以免不必要的麻烦。