这是做毛绒绒风格的lora,训练图用了92张训练图,
底模用的是revAnimated_v122.safetensors,尺寸是512x768,描述词简单描述就好了,
例如无人类, 黄猫, 动物焦点, 铃铛, 动物, 简单背景, 张嘴, 项圈, 粉色背景, 颈部铃铛, 黑色眼睛, 看着观众, 铃铛, 猫, 红色项圈, 脸红, 微笑...
因为训练图大多都是用的是毛绒绒动物的图,所以出毛绒绒动物图效果比较好,不能直接出毛绒绒的人;还可以把之前的动物萌图拖进图生图加上这个lora+controlnet,就会出现毛绒绒效果;
生成图尺寸不宜太大,权重调看着调就好;
我尝试了dreamshaper大模型和disneyPixarCartoon大模型,效果都还可以,大家可以试试;
毛绒绒的动物是最治愈的,希望大家能喜欢~~~
The following is machine translated
This is for a plush style lora, 92 training maps were used for the training maps, and the base model is revAnimated_v122.safetensors, the size is 512x768.
The base model is revAnimated_v122.safetensors, the size is 512x768, the descriptor simply describe it.
For example no humans, yellow cat, animal focus, bell, animal, simple background, open mouth, collar, pink background, neck bell, black eyes, looking at audience, bell, cat, red collar, blush, smile...
Because most of the training charts are used plush animal charts, so out of the plush animal chart effect is better, can not be directly out of the plush people; you can also drag the previous animal moe charts into the chart to generate charts plus this lora + controlnet, there will be a plush effect;
Generate a map size should not be too large, weights tuned to look at the tune is good;
I tried dreamshaper big model and disneyPixarCartoon big model, the effect is okay, you can try;
Fluffy animals are the most healing, I hope you will like ~ ~ ~ ~