Type | |
Stats | 1,819 0 |
Reviews | (169) |
Published | Jul 23, 2023 |
Base Model | |
Hash | AutoV2 0B4E471FDC |
No special prompt words are required, but the more detailed the prompt words, the better
Description: A Lora that can significantly increase light and shadow contrast, hair texture, and background details
1. The weight is between 0.5 and 1. The higher the weight, the better the performance, but the relevance of the prompt words will be reduced
2. Works better on poor images
3. Can be used together with other Lora
4. Use most of the 2D models, if the effect is poor, you should reduce the weight appropriately
Train with the Lion optimizer and noise-offset
set the block weights to increase generalization
ps: A very confusing Lora,sometimes the effect is inexplicably good, and sometimes it is not obvious
~~o(≧ω≦)o
///////////////////////////////////////////////////
不需要专门的提示词,但图片提示词请越详细越好
描述:一个能显著增加光影对比,头发质感,背景细节的Lora
1.权重在0.5到1之间,权重越高表现越好,但对提示词的还原度会降低
2.对于较差的图像作用更好
3.与其他Lora可以一起使用
4.使用大部分2D模型,如果效果较差应该适当降低权重
使用Lion优化器(optimizer)和noise-offset训练
设置分层权重以增加泛化性
ps:一个相当玄学的Lora,效果有时候莫名其妙的好。
~~o(≧ω≦)o