dcy
/

AsiaFacemix / README.md
wizardt's picture
Added English translation. Correct errors.更新了中文,修正语病,增加了英文。
f6ccc34
|
raw
history blame
2.52 kB
metadata
license: openrail
datasets:
  - Gustavosta/Stable-Diffusion-Prompts

Model Description

该模型基于basil mix,dreamlike,ProtoGen等优秀模型微调,融合而来。 用于解决上述模型在绘制亚洲、中国元素内容时,只能绘制丑陋的刻板印象脸的问题。 同时也能改善和减少绘制亚洲、中国元素内容时,得到更接近tags的绘制内容。 This model based on basil mix,dreamlike,ProtoGen,etc. After finetune and merging, it solved the big problem that the other model can only draw ugly stereotyped woman faces from hundreds years ago When drawing Asian and Chinese elements. This model can also improve the drawing content of Asian and Chinese elements to get closer to tags.

基于dreamlike微调与AsiaFacemix效果图

Based on dreamlike finetune example: 1.png 2.png 37.png

基于Image to Image效果图

Based on Image to Image example: s1.png S2.png

添加国风汉服lora模型

Added Chinese Hanfu LORA model lora-hanfugirl-v1 V1模型基于真实的汉服照片训练,相对于v1-5,有更细腻美丽的脸部。 The V1 model is trained on real Hanfu photos and has more delicate and beautiful faces than v1-5. lora-1-05.png lora-1-0502.jpg

lora-hanfugirl-v1-5 V1.5模型同样基于真实的汉服照片训练,相对于v1,对不同的多个模型和不同分辨下的图片兼容性更好。 V1.5 model is also trained on real Hanfu photos. Compared with v1, it has better compatibility for different multiple faces, scene and pictures under different resolution. lora0501.png lora0502.png