Type | |
Stats | 32,502 |
Reviews | (289) |
Published | Feb 27, 2023 |
Base Model | |
Hash | AutoV2 759D6FDF53 |
This LoRA tries to improve the outlines performance, most of the popular anime models (e.g. Anything V4.5, Counter Feit, AbyssOrangeMix, etc.) have some sort of pastel-like feeling and rather thin outlines.
This LoRA is trained on stylized images and then merged with several LoRA models I trained previously (most of them are already published on Civitai, you can check my homepage).
The example images are generated with Anything V4.5 + orange mix VAE, but this LoRA is capable of multiple models (also including NSFW), along with different character LoRA.
For strength, I'd say 1.0 is good enough.
I've also attached a checkpoint version, merged with Anything V4.5. It's for more sophisticated use (e.g. using another style LoRA).
中文:
本LoRA模型是为解决多数动漫风格模型过于柔和导致描线较弱的问题而训练的。基于风格化的图片训练,并与我此前训练的多个LoRA模型进行了融合而成。
示例图片为Anything V4.5 + orange mix VAE生成,但本LoRA并不限定基础模型的种类(包括NSFW类),并可兼容多数人物LoRA。
推荐强度为1.0。
附带的Checkpoint版本是和Anything V4.5融合的版本,用于一些更复杂的场景(如使用其他的画风LoRA)。