Recommended Prompts
(((masterpiece,best quality,official art,extremely detailed CG unity 8k wallpaper))),illustration,landscape,solo,1girl,bishojo,beautiful detailed eyes,Student uniforms
Recommended Negative Prompts
no prompt
| samplers | DPM++ 2M Karras | |
| steps | 20 | |
| cfg | 10 | |
| resolution | 512×512 | |
Recommended Hires (High-Resolution) Parameters
| upscaler | Latent (nearest-exact), Latent | |
| upscale | 2 | |
| steps | 20 | |
| denoising strength | 0.6 | |
Tips
- When using veryBadImageNegative_v1.2, you can reduce some weights to obtain better flexibility. For example: (veryBadImageNegative_v1.2-6400:0.9)
- The number of training steps (1600 ~ 6400) represents the number of steps in training. In theory, the more steps, the better the effect.
v1.3
veryBadImageNegative_v1.3使用了新的训练图集，在AOM3和viewer-mix_v1.7中使用时表现较好。
veryBadImageNegative_v1.3 uses a new training atlas and performs well when used in AOM3 and viewer-mix_v1.7.
v1.2
veryBadImageNegative是使用viewer-mix_v1.3生成的特殊图集训练而来的负面嵌入。
所以veryBadImageNegative是viewer-mix_v1.3的专用负面嵌入。
或许在其他的扩散模型中也有不错的效果，但缺乏验证。
veryBadImageNegative is a negative embedding trained from the special atlas generated by viewer-mix_v1.3.
So veryBadImageNegative is the dedicated negative embedding of viewer-mix_v1.3.
It may also have a good effect in other diffusion models, but it lacks verification.
关于v1.2：
增强图像的质量，削弱了风格。
在使用v1.2版本时，可以降低一些权重以获得更好的灵活性。
例如：(veryBadImageNegative_v1.2-6400:0.9)
About v1.2:
Enhance image quality and weaken style.
When using v1.2, you can reduce some weights to obtain better flexibility.
For example: (veryBadImageNegative_v1.2-6400:0.9)
关于v1.0与v1.1的差异：
几乎没有差别，仅用于训练的图集在生成参数有一些微妙的差别。
关于1600~6400的差异：
数字代表了训练步数，理论上步数越多效果越好。
About the difference between v1.0 and v1.1:
There is almost no difference. The atlas used for training only has some subtle differences in the generation parameters.
About the difference between 1600 ~ 6400:
The number represents the number of training steps. In theory, the more steps, the better the effect.
