Sign In

戏装 Chinese Opera Costumes

871
3.0k
16
Verified:
SafeTensor
Type
LoRA
Stats
2,098
Reviews
Published
Apr 6, 2023
Base Model
SD 1.5
Trigger Words
xiaosheng
huadan
Hash
AutoV2
1B031ABDF9

戏装 Chinese Opera Costumes

[Eng version below]

607323927751959196

一个用于添加人物装扮的Lora,目前仅有女旦和小生角色,后续其他装扮陆续添加中,慢慢更新吧……

得益于小人书的多人场景能力,两者混用的画会带来奇特的故事场景能力。但是由于戏曲类手势身段多有变形,手指问题基本不可救药了……

但同时,这个LORA可以作为低权重的人物装束补充,效果比单用好很多,请见例图1、2、3,使用了芳菲作为主LORA

推荐分辨率:512*768(hires required)

推荐模型:AOM3\ChilloutMIX (其实理论上其他的也行,没空测了……)

推荐设置:CFG 5 | 权重 0.75 | DDIM step25 | hires SwinIR 4x step 10 denoising 0.4

例图中多人场景是搭配小人书2.0的效果,权重比为0.75:0.5(小人书0.5)

A Lora for adding character costumes, currently only the role of female HuaDan and male XiaoSheng, followed by other costumes are added one after another, slowly update it ......

Thanks to the multiplayer scene ability of xiaorenshu, the painting of the two mixed will bring a strange story scene ability. But because of the opera class gesture body more deformation, finger problems basically irredeemable ......

But at the same time, this LORA can be used as a low-weight character attire supplement, the effect is much better than the single use, please see examples of Figure 1, 2, 3, the main LORA is flowergirl.

Recommended resolution: 512*768 (hires required)

Recommended models: AOM3\ChilloutMIX (in fact, theoretically the others are fine, no time to test ......)

Recommended settings: CFG 5 | weights 0.75 | DDIM step25 | hires SwinIR 4x step 10 denoising 0.4

The multiplayer scene in the example is with the effect of xiaorenshu 2.0, the weight ratio of 0.75:0.5 (xiaorenshu=0.5)