Illustrious version : LatestIllustrious base merged with a bunch of sdxl models in mbw train difference. Very different from the previous version. Excels in dynamic images, sfw and nsfw, colors and detailed textures. I prefer it with low cfg like 3.5, or with extensions such as automatic cfg which relax the cfg a bit, otherwise you get very strong colors. I find i like my model's outputs with automatic cfg, even better than with skimmed cfg in comfyui. Try different settings. I prefer euler ancestral as a sampler or dpmpp_sde. Preferred schedulers : sgm uniform or align your steps from nvidia. AYS requires less steps generally. When in normal use is use 35 steps, with AYS i use 15 steps base + 12 steps a second step with the same model as a refiner. Here you will find images with both methods.Older version V1+ V2 : This is a very versatile sdxl model with good anatomy and a artistic creative style. It has DPO (Direct Preference Optimization) baked into it, which i found improved a lot the composition, prompt compression and anatomy. Give it a test. I use 2M karras sampler and 35 steps but feel free to choose. I suggest using the suggested hands and feet lora if you have trouble with anatomy (suggested ressources). They work very well with my model. I did not bake them into the model because the hands one changes composition a bit at full weight, but again, please experiment !Here is the prompt structure i use :(Style) (in artstyle by artist X), (other type of style for example surrealism if you wish). BREAK(Composition) (A full body shot of a X girl for example). BREAK(Character) description of the character, BREAK(background) (description of background or just aesthetic and simple backgournd). BREAK(detail) (masterpiece, high quality),

ReyMix_XL

A fusion between digital art and anime while maintaining good anatomy.

Files with SHA256: 6b6b7cf8df62b648288e568a4e72e4087cfa38d207538e0996258c6c84fcf8b5