Trained on just 29 (13, 21) images of Billie Eilish.
Tried a new training method with 10 epochs, 2 batch size, network 32/32, (10,15) repeats and no regularization images.
Trained on base sd 1.5, strength of 0.8 should work on most models.
Trained with WD14 captions in under 25 minutes. Smaller file size as well.
Let me know if you prefer this model over the original larger ones.