This is my first attempt at training a LoRA model for Z Image Turbo. I trained the model using the AI Toolkit on my RTX 5060 Ti with 16GB of VRAM. After some testing, I settled on a 2000-step model. It seems quite flexible, but time will tell. The dataset was small, only 14 images. I think that's why the variety of horns is limited. I might experiment with this model and its dataset further.
Triggers:
t1efling
red skin
horns
glowing eyes
black whites