本模型有着轻微硅胶质感的LORA模型头发细节都是动漫偏写实质感不影响真实背景
This model has a slight silicone texture of the LORA model hair details are animation realistic texture does not affect the real background
模型通过最初的30几张训练集增加到54张 训练使用了之前的Kigurumi的素材 本模型灵感也是来源于他 我像训练一个介于现实世界中的动漫人物 但是效果可能没那么理想 反倒有点像动漫硅胶娃娃 但是训练集里没有任何硅胶元素 这个很神奇 我只是将少量的Kigurumi素材和PVC模型的素材混合训练就达到了这效果
The model was expanded from 30 training sets to 54 training sets using the previous Kigurumi material from which the model was inspired. I trained an anime character in the real world, but the effect may not be so ideal, it is a bit like an anime silicone doll, but there is no silicone element in the training set. This is amazing I achieved this effect by mixing a small amount of Kigurumi footage with PVC model footage
切记模型的提示词不要加特别多的风格词 质量词 即可出效果 建议使用少量的效果词加上以很少的触发词和模型触发词效果最佳
Remember that the model prompt words do not add too many style words quality words can be effective. It is recommended to use a small number of effect words with few trigger words and model trigger words for best effect
模型触发词:(Model trigger word): mk_jq
推荐模型权重(Recommended model weight): 0.6~0.8
推荐SDXL底模（Recommended SDXL base mode）：
Fantastic hacker_Style_bailing
DreamShaperXL ,
SDXL Unstable Diffusers,
StarlightXLAnimated,
推荐参数（Recommended parameters）：
dpm++ 3m sde exponential 45
dpm++ 3m sde karras 45
高清修复（hires.fix）: 4x-ultrasharp , 高
分迭代步数 10~15
重绘幅度 0.3~ 0.5
clip skip: 2
CFG：3.5~6
等倍高请修复人像人物分辨率： 768x1280 ，1024x1024
模型反馈请站内私信
本模型中不包含任何真人素材图像源和训练代码来自互联网，模型仅用于科研兴趣交流
This model does not contain any real human material image source and training code from the Internet, the model is only used for scientific research interest communication