license: mit | |
A concept model trained on a variety of varying styles but same concept under the class of "style". | |
For best results use "cyberware style" and the token "mechanical parts". For more guided generations replace "mechanical parts" with "mechanical [arm/leg/spine/etc]" | |
This model was trained on Waifu diffusion 1.3 so it has some difficulty generating realism but a 0.3 merge with SD has yielded good results. | |
Img2img capabilities are where this model really shines in my opinion. |