This model tries to replicate the style of the Ghost in the Shell movie from 2017. It's trained on 16:9 images and works for fl2va. I've not tested ref2va because I'm going to train a ref2va version anyway and see what happens. This is my first style lora, so it's meh, but at least it's also free ¯\_(ツ)_/¯

Trigger: gits2017

Strength: 0.9

The showcase videos were generated without turbo loras but with the Spectrum node, using the fp8 model.