Ultraskin 0.9 - SD 2.1 768 finetuned on images of ultra-detailed human skin

This model will add a LOT of skin detail compared to SD 2.1. Sometimes this makes images look more realistic, sometimes less realistic!

Prompting for this model is the same as 2.1 and can be tricky compared to 1.5.

Try these tags in your prompts: "teen","twenties" or "twenty year old", "thirties" or "thirty year old", continued all the way to "eighties" or "eighty year old", "elderly", "moles", "freckles", "scar","skin detail"

For stronger results, use the activation word ultraskin in your prompt. I recommend using this token toward the end of your prompt as it can lead to exaggerated detail and plastic looking skin.

Recommended negative prompts (if you want photographic results): rendered, 3d, blender, octane, unreal engine, video game character, cat

This model is based on the 768 model and was trained on 768px x 768px images, so you will have better results creating images at least this size.

Sample outputs:

Samples Samples Samples

🧨 Diffusers

This model can be used just like any other Stable Diffusion 2.1 model. For more information, please have a look at the Stable Diffusion Pipeline.

If using this with a web UI that requires a YAML file, download the YAML file and place it next to the model.

Downloads last month
14
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Spaces using VegaKH/Ultraskin 48