FashionZ版本v1.0 (ID: 1156691)

This is a lora trained on fashion illustrations, but it's fairly flexible. It works best on images with a figure, as most of the training images were watercolor and pen on white paper. Backgrounds tend to be simple. Anatomy is pretty good, I redrew any poor hands. Prompting works best with "sketch" and "on white paper" I am putting test sheets in the examples so you can see how it reacts in, and out of, its comfort zone. For Img 2 Img use any fashion drawing and a high denoise of around 0.80 to get the slickest images. Guidence needs to be around 2.

描述:

fashionz was in every caption, so it will help trigger.

训练词语: