This model is the first attempt to accurately depict soles. An approximate starting prompt could be:

"25-year-old man on a couch with his feet up, showing his soles <lora:sole:0.85>."

Negative prompt:

"(((deformed))), [blurry], (poorly drawn hands), (poorly drawn feet), (too many toes), (too few toes)."

It seems to produce the best out of the box results at:

-CGF 4.5

-40 steps

-prompt: <lora:sole:0.85>

For a step by step tutorial click here:

Step by step AI Tutorial by derart-DA on DeviantArt

Works best with Control Net, when using a canny and a depth (0.3 control weight) unit of a man showing his sole.

Works best with Protogen x3.4,

works great with inpaint (best with canny and depth)