Completely rebuilt dataset by hand. ~130 painstakingly hand captioned images. Using short phrase captions.
I am building this for the entire community and need help building and improving the dataset. If you can help gather or caption images, please contact me. The more people we get involved, the better this will become. Share with your friends!
Goals:
Types: White, Printed, Pullups,...
Sizes: Small, Medium, Large, Huge,...
Any Art Style
Custom Designs
Alternative Models for Specific Use Cases
Tips:
Seems to work really well with Pandora's Box(see resources). I usually run Pandora's Box as the main model and use Better Than Words(see resources) as the refiner for extra detail. Double it with img2img and finally double it again with your favorite 4x upscaler.
v0.1 info:
~380 mostly autotaged image dataset
White Female Bias
Biases to specific poses: shots and legs spread specifically
No distinction between types or size
More difficult to work with than I would like

Completely rebuilt dataset by hand. ~130 painstakingly hand captioned images. Using short phrase captions.
1.転載モデルは学習・共有目的のみで使用し、著作権は原作者に帰属します
2.モデルの認証は公式チャンネルでご連絡ください。クリエイターの権利保護に努めています クリックして認証