Type | |
Stats | 114 1,708 |
Reviews | (12) |
Published | Sep 8, 2024 |
Base Model | |
Training | Epochs: 4 |
Usage Tips | Clip Skip: 1 |
Trigger Words | Proper Anatomy Motion Frames Multiple Panels 12 columns, 2 rows 12 columns, 3 rows 8 columns, 2 Rows 1person, 1 person 1woman, 1girl, 1 woman, woman 1man, 1boy, 1 man, man 2women, 2girls, 2 women, multiple people +3 more |
Hash | AutoV2 455F48D083 |
(Trigger word should be "Proper Anatomy")
(Given the majority of training data, negative tags of "sepia", "monochrome", and "frames" might be handy)
A first attempt to train a basic human anatomy Lora for SDXL architecture models on Public Domain artwork, especially progressive frame motion studies and posed statues, with a touch of historic photography, painting, drawing, and carved reliefs. Eventually I hope to refine the data set for training with the Common Canvas XL model and original artworks from CC applicable sources and pro-AI artists who've released their works for training. All 359 pictures used here are from online CC-0 and historic Public Domain artworks hosted online by the Met and National Gallery of Art.
(Note so far: Negative weights are really cool)