This is a large-scale model created by training and fusion based on Flyingsnow 2.0Hill bottom pattern. During the training of the model, 1500 high-definition materials were used, including 500 each of water powder, watercolor, and flat illustration. In the end, it slightly leans towards water powder. As for what style it resembles, let's call it "what's style" for now. Anyway, hope you like it. Check out the sample images for specific visual effects.

• Sampling Method: It is recommended to use DPM++2MKarras, Eulera, or Eulera. They can quickly produce high-quality images;
• Sampling Steps: It is suggested to be between 20-30, not needing too many steps to avoid wasting time for image creation;
• CFG Scale: Recommended between 6 to 9;
• Clip Skip: It is recommended to be between 2 to 5, but sometimes trying 6 to 10 might bring some surprises;
• VAE: Already built-in;
• Hiresfix (High-definition repair): After achieving a satisfactory effect, open Hiresfix for high-definition repair. This can save a lot of time;
• Upscaler: It is recommended to use 4x-AnimeSharp, R-ESRGAN4x+, or R-ESRGAN4x+Anime6B;
• Sampling Steps: 20-30 is sufficient; Denoising Strength: Recommended between 0.3-0.4 (values too high will deviate from the original image);
• Output size: 768x512, 512x768, 512x512, or 512xN, Nx512 (N represents any number). That is, keeping one value at 512 for width or height will yield a better effect. Because the original materials for Stabledifussion1.5's bottom model training are all 512x512, keeping one value as 512 will produce better results (this applies to all 1.5 large models);

This model has been synchronized and released on C site, recognized by C site as a

Featured Models

Homepage Recommended