For best outputs: Sampler: DPM++ SDE. Steps: 8, cfg scale 3, CLIP skip 2. VAE: sdxl_vae.safetensors Pos prompt: this is a versatile type of model, able from anime to photo-realism. Just play with pos and neg prompts pointing the opposite of the style in each. Neg prompt: can be left blank, or just put what you don't want to see. Also the neg can be used to guide the model between the different levels of realism or anime.