This is to kind of emulate the style of women and posing in Adriano's scenes.

V2:

Trained on more specific data from Adriano scenes -- including each sub-channel as much as possible. The new training is a bit more sensitive and also tends to create video-like artifacts from time to time. It's better at lower settings, like 0.3 to 0.6.

I definitely recommend using a face fix, too.