V2版本说明:
1.这个版本是用v1版本,调整LORA的层级数值来做出优化的。
2.这一版即使不用高清修复,效果依然还不错。
3.模型污染问题减弱,相对更受控。
4.减少油腻感。
Version 2 Release Notes:
This version is optimized by adjusting the layer values of LORA based on the V1 version.
Even without Hires Fix, the results are still good.
The model pollution issue has been reduced and is relatively more controlled.
Reduced the greasy feeling.
<lora:IcfgirlLora_v40:1> 0.6-1.1
这是一个网红脸模型,彩色的头发,半身肖像比较在行。
触发关键词 (xx)
This is an internet celebrity face model, with colorful hair, and is more skilled in half-body portraits.
keyword: (xx)
示例用到的名为model的基础模型在这里https://huggingface.co/a1079602570/animefull-final-pruned/tree/main
Description
FAQ
Comments (8)
请问一下,为什么这个模型看起来那么精美,但是它所需的空间却那么少
应该是训练图片质量比较高吧,和用的底模也有关系吧。我猜的
跟network_dim相关
Guess author knows a lot abt girls outfit看大佬的语言描述感觉太懂着装了
应该是我直接抄过来了,啥都超了0 0
为什么我生成的全身图片,始终没有作者的清晰,问题出在哪里?
我试过换VAE,采样器 都无法提高人物主体的清晰度。而换另外的lora就没这个问题。
Why is the full-body picture I generated not as clear as the author's? What is the problem?
I tried changing VAE, and the sampler can't improve the clarity of the main body of the characters. And changing to another lora will not have this problem.
这个lora有些问题,我这两天会更新一版,就会好很多了,我的图是放大过,修复过脸的。https://github.com/ototadana/sd-face-editor或者可以用这个插件修复脸。
@Zyx_xx 谢谢回复,期待你的下个版 ,很喜欢这个风格
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.
















