汉服-碧水孔雀

模型介绍：

这是一个基于MJv6生成素材训练的Lora！主要效果为水下摄影加彩色汉服的结合！！

具体食用方法请看例图！！

推荐参数：

大模型：麦橘写实v7

VAE：真实感840000

CLIP跳过：1

推荐正面tag：无

推荐负面tag：logo,text,badhandv4,EasyNegative,ng_deepnegative_v1_75t,bad_prompt Negative Embedding_v2.0,坏图修复veryBadImageNegative_veryBadImageNegative_v1.3,negative_hand-neg,mutated hands and fingers,poorly drawn face,extra limb,missing limb,disconnected limbs,malformed hands,ugly,

Lora权重：0.6-0.9

迭代步数：20-30    

采样推荐：DPM++2M Karras

分辨率：512*1024（重要！！）

最后：

一般直出，3-5张就有好好看的，当然也考验你的提示词，感觉面部效果不佳的话可打开ADetailer。

祝大家玩的开心，欢迎来测试和返图！！

v2.0版本会改进眼睛的问题！！

Model Introduction:

This is a Lora trained on MJv6 generated materials! The main effect is the combination of underwater photography and colored Hanfu!!

Please refer to the example picture for specific consumption methods!!

Recommended parameters:

Big Model: Maiju Realistic V7

VAE: Realistic 840000

CLIP skip: 1

Recommended positive tag: None

Recommended negative tags: logo, text, badhandv4, EasyNegative, ng_deepnegative_v1_75t, bad_prompt Negative Embedding_v2.0, bad image repair veryBadImageNegative_veryBadImageNegative_v1.3, negative hand neg, mutated hands and fingers, poor draw face, extra limb, missing limb, disconnected limbs, malformed hands, ugly,

Lora weight: 0.6-0.9

Iteration steps: 20-30

Sampling recommendation: DPM++2M Karras

Resolution: 512 * 1024 (Important!!)

Finally:

Usually straight out, 3-5 photos will have a good look, and of course, it also tests your prompt words. If you feel that the facial effect is not good, you can open ADetailer.

Wishing everyone a happy time, welcome to test and return pictures!!

V2.0 version will improve eye problems!!