interdiffusion-xl

For best outputs and usage:

• Sampler: DPM++ 2S a Karras.

• Steps: around 30, cfg 4 ~ 7, CLIP skip 1.

• VAE: sdxl_vae.safetensors

• This is a Versatile type of model, which means it can go from a quite decent anime to nearly photorealism, knowing many forms of art and artists.

• In order to switch from a level of realism to another:

  • Realism:

    • Pos prompt: wrap your text between "(full body professional photo)" at the beginning and "hyperrealistic art cinematic film still photography in the style of detailed hyperrealism photoshoot . 35mm photograph, film, bokeh, professional, 4k, highly detailed, Ultra-HD-details, RAW candid cinema, 16mm, color graded portra 400 film, remarkable color, ultra realistic, textured skin, remarkable detailed pupils, realistic dull skin noise, visible skin detail, skin fuzz, dry skin, shot with cinematic camera, <lora:add-detail-xl:1>" at the end, being the LoRa "add-detail-xl" included in the realistic example runs.

    • Neg prompt: "blur, blurry, unrealistic, cartoon, drawing, 2d, 3d, cg, render, assymetric face, imperfect human face, stains, tattoo, epiCNegative", being the embedding "epiCNegative" included in the realistic example runs.

  • Anime:

    • Pos prompt: start your pos with "anime, anime style, japanese animation style, 2d plain colors,"

    • Neg prompt: "realistic, realism, photo, photorealistic, volumetric"

  • 3D:

    • Pos prompt: start your pos with "3D, pixar style".

    • Neg prompt: "realism, photo, cartoon, 2d, semi realism, anime".