A semi-soft anime-styled SDXL Lightning checkpoint.

It gives you digital illustration styled images of high quality, with an improved diversity.

The model's strength is creativity and complex prompts. It's trained on very long prompts with a mix of wd14 and natural language captions.

It's very good with creativity, and handling unique/weird combinations of things, as well as making breathtaking landscapes.

It's weaker with faces and people. They turned out a bit too anime for my preference and could use some tweaks. But it is good with diversity. It listens really well to nationalities and body types.

It's trained on the same 2100 image dataset as https://civitai.com/models/651146/semi-soft-style-flux and the https://civitai.com/models/477136/daiversity-detailer-sdxl-photosemi-realistic.

Recommended settings

Steps: 8

CFG Scale: 1.8

Sampler: Euler a