ONLY work well on 2:3 aspect ratio, better to download it on your local environment.
Suggest to use in 2:3 aspect ratio; Weighting as 0.8-1.1
A emtpy space would remain in between thicc, so you can print it out as a 2legs (Futamata 二股) dakimakura cover.
All version are trained on my M2 mac locally. If you appreciate with this or all those models, please buzz me.
Description
This version train with same images data set of XL v1.0, with updated tags, so named as v1.1.
And the base model is now using pony-based AutismMix SDXL confetti.