artium-sdxl

For best outputs:

• Sampler: DPM++ 2M Karras

• Steps: 22 ~ 14, depending on the simple of the output, cfg 7, CLIP skip 2.

• VAE: sdxl_vae.safetensors

• This is a Versatile type model, but focused and extremely well trained in Art. It knows a vast widely range of art styles, from RAW photography to anime, passing by manga, sketch, Rembrandt, fauvism, charcoal sketch, cubism, western comic and a very long list.

• In order to pop the art style you wish, just name it at the very beginning of the prompt, and keep the prompt as simple as possible, avoiding weights of any kind (model doesn't need them)

• Adetailer for the faces may be required for realism photo small faces (example, full body) at a weight of 0.45

• Couple of examples:

  • Charcoal Sketch -> "Charcoal sketch, a beautiful girl in victorian gown, intricate details, 8k, masterpiece"

  • Anime -> "Anime illustration, Kurisu Makise, brown jacket, red tie, ecchi, scifi, quantum entanglement, 8k, masterpiece, highly detailed"

  • Realistic photography -> "RAW Photography, a beautiful lady, short light brown hair, creative hairdo, beautiful hazel green eyes, imaginative use of floral patterns, she is very creative and a sensitive soul, a muse, vulnerable, fashion shoot, intricate details, 8k, masterpiece, dynamic play of light, deep darks, rich emotive colors, 35mm photograph, kodachrome, dust particles caught in the light" <-- adding the machine's name and photography tech details helps.

• The negs do not necessarily have to go paired with the art style. For example, you do not need to write "cartoon, anime" in the neg if u want to try to get realistic photo. The model knows what u want perfectly just by naming the art style. Take a look at my example runs to see examples of negs, they're quite similar to each other.