Black-and-white glitch + Bright red accents = Distortion
Carefully captioned 55 image dataset with greatly varied subjects: does nice portraits, landscapes, cityscapes, architecture, wilderness, and more.
I was listening to this ↗️ while making the dataset and waiting for it to train.
Usage guide
Most subjects don't need anything special, but sometimes you'll find a subject the model is unfamiliar with. In this situation, I turn up the LoRA tag weight to 1.5 at least and/or start the prompt with "glitch art of" before the subject.
No comments yet