For best outputs: Sampler: Restart Steps: 50, cfg 7, CLIP skip 2 VAE: sdxl_vae.safetensors This model requires the LoRa "EnvyBetterHiresFixXL01" with a weight of 0.5 (included in the example runs) or else u get blurry images. Pos prompt: Keep it simple, for example -> "Masterpiece, best quality, high res, Christmas, arien monster, cute monster, cabin, Christmas setting", where "arien monster" and "cute monster" are the model's triggers to pop the cute monsters. Neg prompt: "oversaturated, blur, bad quality, worst quality, drawing, painting, crayon, sketch, graphite, impressionist, noisy, blur ry, soft, deformed, ugly, 3d render, low resolution, deformed, distorted, poorly drawn, bad anatomy, wrong anatomy, ugly, deformed, noisy, log, text"