I originally wanted to create a more generalized model for a "Mrs. Claus" style costume dress, but not just in red. As I struggled to find pictures for what I actually intended, I broadened my scope to any dress that looked winter-y and with fur trim. And I'm quite pleased with the results, actually! This model was trained on 15 images of dresses as I just described. Usage <code>dswntr</code> is the bare minimum of what you need to see this model in action. I originally intended it to be used just with <code>dress</code> but it actually works on many types of clothes if you use <code>fur trim</code> instead. I tagged traits for dresses, as I normally do, such as <code>v-neckline</code> or <code>princess neckline</code> or <code>mini skirt</code>, but I think the real power of this model comes from utilizing <code>fur trim</code>, along with <code>fur hem</code>, <code>fur collar</code>, and <code>fur cuffs</code> along with other clothes to turn them closer to that "winter-y" style. Known Issues I recommend using ADetailer to fix faces in the generations. Animal ears: Because I used the keyword <code>fur</code> in my dataset, there is a tendency to get animal ears in the final image. It's not super common, so it can easily be fixed with inpainting or just run more generations. Brown hair: Through random happenstance, all the images I trained on are women with brown hair. If you don't specify a hair color, it will probably lean brown.

Files with SHA256: cb2e055e34b2d22d094244af7ac77806a4e7fe449c1d75aa5914f36728f6385a