//=time() ?>
Depth model + WD 1.4 merge test. If the ratio of the depth model is 0.7 or less, the probability of the original pose disappearing increases. Depth 0.8 WD 0.7 seems to be the most acceptable. #stablediffusion #AIイラスト #Depth
@fabianstelzer This is just creating an overtrained subject. Similar to "Mona Lisa". It does not work from SD 1.5 or higher trained using much more images.