<h1>v1.3</h1>veryBadImageNegative_v1.3使用了新的训练图集，在AOM3和<a target="_blank" rel="ugc" href="https://civarchive.com/models/7813/viewer-mixv13">viewer-mix_v1.7</a>中使用时表现较好。veryBadImageNegative_v1.3 uses a new training atlas and performs well when used in AOM3 and <a target="_blank" rel="ugc" href="https://civarchive.com/models/7813/viewer-mixv13">viewer-mix_v1.7</a>.<h1>v1.2</h1>veryBadImageNegative是使用<a target="_blank" rel="ugc" href="https://civarchive.com/models/7813/viewer-mixv13">viewer-mix_v1.3</a>生成的特殊图集训练而来的负面嵌入。所以veryBadImageNegative是<a target="_blank" rel="ugc" href="https://civarchive.com/models/7813/viewer-mixv13">viewer-mix_v1.3</a>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 <a target="_blank" rel="ugc" href="https://civarchive.com/models/7813/viewer-mixv13">viewer-mix_v1.3</a>.So veryBadImageNegative is the dedicated negative embedding of <a target="_blank" rel="ugc" href="https://civarchive.com/models/7813/viewer-mixv13">viewer-mix_v1.3</a>.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.

veryBadImageNegative

verybadimagenegative_v1.1-1600.pt

v1.1-1600

veryBadImageNegative_v1.3

v1.2-1600

v1.2-3200

v1.2-4800

v1.2-6400

v1.0-1600

v1.0-3200

v1.0-4800

v1.0-6400

v1.1-3200

v1.1-4800

v1.1-6400

Files with SHA256: e7ccefc0490f22225d22678d4f7d9e8dca2ba83af9cc7288f9f98de0ebdc4e0f