This version update adds more different artist styles, especially in the light and shadow art and supernatural art that I personally like very much, focusing on adding materials for strengthening training, so that you can get richer picture content and more exaggerated picture presentation. In the portraits, fashion design, product design, architectural design, science fiction, animation and other aspects that are mainly needed for daily large models, balanced proportions of materials are also added to make the model output more controllable.

Sampling method:

DPM++ 2M Karras

CFG: 7

Steps: 25~30

VAE: None

HD upscaling algorithm:

R-ESRGAN_4x+, LDSR, SwinIR_4x, Latent (nearest)