这是我基于 Z 底图训练的第一个 LoRA 模型。目前的版本看起来还有些欠拟合。
推荐设置：
推荐权重： 0.7 - 1.0
采样器 ： 推荐使用 Euler, Euler a 或 res-multistep
调度器 ： simple 或 normal
This is my first LoRA trained on the Z image base. It still appears to be slightly underfitted.
Recommended Settings:
Recommended Weight: 0.7 - 1.0
Sampler: Euler, Euler a, or res-multistep
Scheduler: simple or normal
Since this is an early version, your image posts and feedback are incredibly valuable to me. Please share your results to help me refine the next version!
Description
FAQ
Comments (1)
Its a interesting and fun lora
Its horrible at nudity
And its proportions and finger count are off some times, also it has a blending problem with clothing or the background.
But its still a decent Lora. It just needs more work then most to look decent. Would recommend.