UnreaLibrary-Mix / README.md
Fre2C's picture
Update README.md
c7f0313
|
raw
history blame
6.38 kB
---
license: creativeml-openrail-m
---
所有的预览图没有使用embedding,lora
我的融合模型:
DreaMirror:https://civitai.com/models/30294 / https://huggingface.co/Fre2C/DreaMirror-Mix
**这个模型的方向是尽可能忠于提示词(在2D模型中这好像有点难),保留2D模型的创造性(所以我并没有融合3D/2.5D模型。所以大部分时间都在与手部进行搏斗),合适的明暗对比。**
**你可以用它尝试任何东西!如果你觉得这个模型不错,可以评价一下以及上传你的图片。**
**从以下地方学习了很多,十分感谢。**
https://huggingface.co/WarriorMama777/OrangeMixs
https://civitai.com/models/9409/or-anything-v5
https://economylife.net/u-net-marge-webui1111/
https://docs.qq.com/doc/DTkRodlJ1c1VzcFBr?u=e7c714671e694797a04f1d58aff5c8b0
https://docs.qq.com/doc/DQ1Vzd3VCTllFaXBv?_t=1685979317852&u=e7c714671e694797a04f1d58aff5c8b0
https://www.figma.com/file/1JYEljsTwm6qRwR665yI7w/Merging-lab%E3%80%8CHosioka-Fork%E3%80%8D?type=design&node-id=1-69
**使用建议:**
脸部出现崩坏的情况,以及想**提升面部质量**,使用**局部重绘** **重绘区域**使用**仅蒙版**(效果最好)获得**更好的面部**,或使用Hires. fix改善,使用**其他随机种**或者**工具**也是不错的办法。
**较高**的分辨率(比512 * 512高一点)再加上**Hires. fix**,图片质量**会更好**(如果显存不够你可以尝试**低倍率**的Hires. fix或者**其他放大方法**)。
**用于画面质量的正面提示词(像 best quality)是不必要的,会减少画面的可能性,还会使画面趋于一种风格。**
**将你原本用在正面质量提示词上的权重,用在负面质量提示词上,那是更好的选择。**
如果觉得画面内容**不够丰富**,你可以尝试**细致地描述**,使画面更加**贴近你的想象**
*如果你想用很少的提示词抽奖的话,最好把雨伞(umbrella)加进负面提示词(至少在V1是这样的)。*
我一般在效果不符合预期时使用clip2。
**随你喜好使用lora!**
All preview images do not use embedding,lora.
**The direction of this model is to be as faithful as possible to the prompt words(This seems a bit difficult in a 2D model), preserve the creativity of 2D models(So I did not merge the 3D/2.5D models. So most of the time is fighting with the hands), appropriate light and dark contrast.**
**You can try anything with it!**
**I have learned a lot from the following places, thank you very much.**
https://huggingface.co/WarriorMama777/OrangeMixs
https://civitai.com/models/9409/or-anything-v5
https://economylife.net/u-net-marge-webui1111/
https://rentry.org/Merge_Block_Weight_-china-_v1_Beta#1-introduction(This is the translated version)
https://docs.qq.com/doc/DQ1Vzd3VCTllFaXBv?_t=1685979317852&u=e7c714671e694797a04f1d58aff5c8b0
https://www.figma.com/file/1JYEljsTwm6qRwR665yI7w/Merging-lab%E3%80%8CHosioka-Fork%E3%80%8D?type=design&node-id=1-69
**Suggestions for use:**
If the face appears to be falling apart, and you want to **improve the quality of the face**, use **Inpaint** and **Inpaint area** use **only Masked** (Best results) to get a **better face**, or use **Hires. fix** to improve, use **other seed** or **tools** is also a good way.
**Higher** resolution (a little higher than 512 * 512) plus **Hires. fix**, picture quality will **be better** (if the gpu memory is not enough you can try a **Low magnification** of Hires. fix or **other upscale tools**).
**Positive prompt for image quality (like best quality) are unnecessary and reduce the possibilities of the picture, also make the picture tend to be in a style.**
**It's better to Use the weight you would have used for positive quality prompt on negative quality prompt.**
If you feel that the content of the picture is **not rich enough**, You can try to **describe in detail** to make the picture more **closely to your imagination.**
*If you want to sweepstakes with few prompts, it is better to add umbrella to the negative prompt (at least in V1).*
I usually use clip2 when the results don't meet expectations.
**Use lora as you like!**
我使用这两个VAE/I use these two VAEs:
https://huggingface.co/hakurei/waifu-diffusion-v1-4/blob/main/vae/kl-f8-anime2.ckpt
https://civitai.com/models/22354/clearvae
**V1**
![image](https://raw.githubusercontent.com/Fre2C/UnreaLibrary/master/Preview/V1/3124573465.png)
![image](https://raw.githubusercontent.com/Fre2C/UnreaLibrary/master/Preview/V1/1876586030.png)
![image](https://raw.githubusercontent.com/Fre2C/UnreaLibrary/master/Preview/V1/1210285340.png)
![image](https://raw.githubusercontent.com/Fre2C/UnreaLibrary/master/Preview/V1/1450654027.png)
![image](https://raw.githubusercontent.com/Fre2C/UnreaLibrary/master/Preview/V1/3519324115.png)
![image](https://raw.githubusercontent.com/Fre2C/UnreaLibrary/master/Preview/V1/2842647251.png)
**使用的模型/Models used**
kawaiimixNijiV5Cute_v10【58f37f4736】
Counterfeit-V3.0_fp32【17277FBE68】
pikasNewGeneration_v20【6C509880A5】
breakdomainanime_A0440【1870FA10C3】
plagion_v10【0C42B21C09】
AnythingV5V3_v5PrtRE【7f96a1a9ca】
tComicV35_v35【25750140EA】
**配方/Recipe**
**use:https://github.com/hako-mikan/sd-webui-supermerger/**
kawaiimixNijiV5Cute_v10 x (1-alpha) + Counterfeit-V3.0_fp32 x alpha)x(1-beta)+ pikasNewGeneration_v20 x beta
alpha:0.7,1.0,0.9,0.8,0.7,0.6,0.6,0.7,0.8,0.9,0.7,0.5,0.7,0.7,0.85,0.75,0.65,0.75,0.85,0.75,0.65,0.75,0.85,0.9,0.8,0.8
beta:0.75,0.35,0.45,0.55,0.65,0.75,0.85,0.75,0.85,0.75,0.6,0.6,0.6,0.5,0.35,0.45,0.55,0.6,0.65,0.55,0.6,0.5,0.35,0.4,0.5,0.4
**named as step1**
breakdomainanime_A0440 x (1-alpha) + plagion_v10 x alpha)x(1-beta)+ step1 x beta
alpha:0.25,0.35,0.45,0.55,0.65,0.55,0.45,0.55,0.4,0.6,0.7,0.75,0.8,0.4,0.4,0.5,0.6,0.7,0.8,0.6,0.5,0.4,0.5,0.4,0.7,0.7
beta:0.7,0.85,0.75,0.65,0.55,0.7,0.6,0.5,0.4,0.5,0.6,0.5,0.4,0.6,0.8,0.7,0.6,0.8,0.7,0.6,0.5,0.4,0.5,0.6,0.5,0.4
**named as step2**
AnythingV5V3_v5PrtRE x (1-alpha) + tComicV35_v35 x alpha)x(1-beta)+ step2 x beta
alpha:0.65,0.75,0.65,0.75,0.65,0.75,0.65,0.75,0.85,1.0,0.85,0.75,0.85,0.4,0.65,0.75,0.65,0.45,0.3,0.15,0.3,0.45,0.65,0.75,0.8,0.8
beta:0.75,0.25,0.35,0.45,0.55,0.75,0.85,0.75,0.85,0.75,0.85,1.0,1.0,0.7,0.35,0.45,0.55,0.75,0.65,0.75,0.65,0.55,0.45,0.35,0.75,0.85
**prune and get final fp16 version**