This model was trained with illustrations by Joseph Christian Leyendecker, with an emphasis on his more "cartoonish" ones. A full list of works used in training, as well as captions, can be found in the training data.
Description
Trained on a small subset of his works (x29?).
Resolutions (counts): 832x1152 (7), 896x1088 (6), 896x1152 (4), 832x1088 (3), 832x1216 (2), 640x1344 (1), 640x1600 (1), 896x896 (1), 960x1024 (1), 960x1088 (1), 1024x896 (1), 1152x832 (1)
Not perfect. I want to train a better one, but I haven't got around to it yet.
FAQ
Comments (5)
knowsnose, this is really impressive!
which base model do you think is best to use to train painterly style models for?
Thanks! I'm not in the know as much these days, but Flux was distilled which may cause problems with learning; good detail and capacity to learn though.
I prefer Illustrious because of Booru tagging and that it can learn better than Pony. Only down side to Illustrious is that it's SD XL based, which is a bit out of date now.
@knowsnose I see thanks. Do you have issues with illustrious making things too specifically 2D or anime?
@olivereads38255 yeah, kind of; it's not perfect. It's possible to overfit, which at least fits the style closer. Overfitting generates deformities though; regularization might help, but I haven't tried. And it can be biased toward anime and a single average face.
@knowsnose oh good to know! thanks I'll experiment with this more!
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.