V2

This time the LORA was trained on WAI (v12.0) as target model. I cleaned the dataset a bit to have better training data, which resulted in a style more close to the OG videos. To avoid the unwanted brains showing randomly in the generated images, the trigger word is now "kanaria_brain_style".

V1

Style based on Kanaria's Brain MV. I extracted equidistant frames from the videos at the bottom.

The trigger word is "brain-style". Because of the "brain" in the trigger, this lora tries to add brains to the image; I plan on training a V2 with a different trigger (when I save enough buzz). In the meantime, try adding "brain" and "brain exposed", as negative prompt. Also, I trained on vanilla Illustrious; I works fin with JANKU, and I plan to train V2 with WAI-NSFW as I usually do.