BoW - the best model SD2.1

Hello, I’m ready to present you the best model based on SD2.1-768px, below I’ll tell you why.

About

The model was created by merging many models. And as a result, it outperforms other models based on SD2.1 and SD2.0 in tests of all parameters.

Differences from others

• Negative Prompt: unlike the notorious Illuminati Diffusion, you do not need to download strange emdendigs and negative promt can in some cases not be filled out

• Faces and anatomy: previously it was believed that the faces and anatomy of Polomona were on SD2.1 and SD2.0, but this model uses more correct face shapes and more correct anatomy (but we still recommend using AfterDetailer)

• CFG Scale: previously, I had to strongly and unnaturally lower the CFG Scale, but now it works more intuitively and normally as on SD1.5

Merge details

To get this result, we mixed a lot of checkpoints and several loras.

• rMADA Merge

• Illuminati Diffusion

• Digital Diffusion

  and others models and loras...

Use

Optimal using in AUTOMATIC1111

Settings:

• Steps: 20-35

• Sampler: Euler a/DPM++ 2M Karras/DPM++ 2M

• CFG Scale: 4.5-7

License

Our license is CreativeML OpenRAIL-m

描述:

训练词语: