This is probably my most ambitious project to date. I combined over 260 images of original, lesser known and obscure superheroes to create this package.
No well-know superheroes were used for this LoRA, because the main focus is to create original concepts for heroes or villains.
The aim is to create your dream hero/villain with the focus of spandex, capes, crime-fighting or villainy.
You can simply just pop in the trigger word heromkr2023 and watch it make endless varieties of characters. Or you can get really specific and dictate the color palette of the costume, race, hair color, gender, amount of clothing...the sky is the limit. With such a large dataset, I believe that you can never run out of options.
For female characters you want to prompt the word 1girl as the LoRA tends to be male bias.
Weight is anywhere between 0.6 to 1. The higher the weight the more the image tends to look like superhero concept art. Works with all checkpoints.