This Lora is aimed to give more consistent and flexible generation than my last edition.

First version is the legacy: Original style. Trained with only original version. I captioned the characters in 2 ways: FoxieAnna or Anna will aim to generate images of Anna but there will be less interference with the support checkpoint with the FoxieName trigger.

Further version aimed to keep the features of princesses with very flexible variation (style, outfit, actions, poses and so on...). I want same quality than my "cinematic disney princesses" but with more accuracy as it won't then be a merge but a trained version!