Download
1 variant available
Update 2023-10-03: v1.1b: Expanded dataset to 41 images.
Trying out low resource settings. Trained on 10 images collected from X. The hood is quite consistent. Others, not so much.
Known issues:
Incorrect back view and lower body
Single images in the preview are cherry-picked from random prompts. Image grids are unfiltered (seed=0-9).