SD1-5 百变二次元梦幻灵眸 LoRA ，这个模型在人像特写和简单的场景中表现良好,具有非常丰富和快捷的换装效果，现代和复古风格都可以通过少数单词轻易表现。如果不用任何提示词，直接就是复古油画风格，用一两个简单的单词即可触发惊人的表达效果。有图片演示，用户可以自行探索更多意外的惊喜。

在我上传的这些图片生成过程中用启用了颜色控制插件CUTOFF，yellow,white,green,blue，权重调的是0.66。这次上传的图片，其中封面和前面的几张白头发的画只用了关键词：bule eys;   绿眼睛的那几张，只用了sparkling green eyes；黄头发的只用了：blue eyes ,yellow hair;复古油画风格的图片，是没有用任何关键词的，我用底膜是：2.5D_XXMixUnreal.safetensors， 一秒出图，基本上不需要调整，图片合格率基本百分之九十以上。以后会继续关注和完善，非常高兴能分享给大家。

SD1-5 Variable quadratic dream eye LoRA, this model performs well in portrait close-up and simple scenes, has a very rich and fast dressing effect, modern and retro styles can be easily expressed by a few words. If you do not use any prompt words, it is directly retro oil painting style, with one or two simple words can trigger amazing expression effects. There are picture demonstrations, and users can explore more unexpected surprises on their own. The color control plug-in CUTOFF, yellow,white,green,blue was enabled during the generation of these pictures that I uploaded, and the weight was set to 0.66. In the picture uploaded this time, the front cover and several paintings of white hair are only used keywords: bule eys; The ones with green eyes, just used sparkling green eyes; blue eyes,yellow hair; Vintage oil painting style pictures, is not used any keywords, I use the bottom film is: 2.5D_XXMixUnreal.safetensors, a second out of the picture, basically do not need to adjust, the picture pass rate is basically more than 90%. It will continue to be improved in the future, and I am happy to share it with you.

推荐参数

Recommended Parameters:

Sampler: Euler  a

Steps: 20~40  >38

The recommended base model： 2.5D_XXMixUnreal.safetensors or similar models

CFG scale:  3.5

clip skip 2

VAE:   vae-ft-mse-840000-ema-pruned.ckpt

Color control plug-in CUTOFF, yellow, white, green, blue, the weight is 0.66