欢迎加入的qq交流群:817686141
请看使用说明
推荐使用默认的sd原版的大模型运行会有更好的细节,
这个lora你可以理解为一个q版转化器,可以让角色人物有更多的可爱感,也就是我们说的变得更加q。
触发词是chibi,权重我推荐给0.5-0.85,随着权重的增加和减少会增加这个q版的程度。
这是我的一些好的在线出图参数,推荐使用这些参数运行可以得到更好的效果
采样:euler a 迭代步数:30-50 放大倍数:1.5-2
高分修复:Latent (bicubic antialiased) 高分迭代:20 重绘幅度:0.5
Welcome to join the qq exchange group: 817686141
Please see the instructions for use
It is recommended to use the default sd original version of the large model to run with better details,
You can understand this lora as a q-version converter, which can make the characters more cute, that is, what we call becoming more q.
The trigger word is chibi, and the weight I recommend is 0.5-0.85. As the weight increases and decreases, the degree of this q version will increase.
These are some of my good parameters for online plotting. It is recommended to run with these parameters to get better results.
Sampling: euler a Iteration steps: 30-50 Magnification: 1.5-2
High score fix: Latent (bicubic antialiased) High score iteration: 20 Redraw range: 0.5