Why the early access? BECAUSE I NEED BUZZ!! (without getting too deep) I underwent emergency surgery in the beginning of the new year. I'm just NOW (In August) nearly fully recovered from this ordeal. I've been home healing, and also.. not working. So money is tight all my access to Grok, and other services has since been cut off. So CIVIT is my primary platform right now. ANY AND ALL HELP YOU CAN PROVIDE WILL BE APPRECIATED <3
This model was designed to be a bridge between western comic style, anime, and fantasy art all rolled into a tight yummy ILL burrito. It can do flat / line art to semi-realistic when pushed through prompts with ease. It's highly variable, detailed, and it includes so many characters all ready inside the model itself it's scary. I tested this model on about 100 random characters from comics, video games, card games, anime, and fandom and out of the 100, it was able to accurately produce all of them with the exception of like 5-6 (not counting all clothing detail) without any Loras to help aid the character results. I spent almost 3 months creating and testing this model for the absolute best results. I generated close to 15GB of test images in the creation of this model. It can do anything between flat color, line-art, semi-realistic, and 3D with prompts. It's versatile and handles loras very well but also powerful and inclusive enough that no lora is needed for you prompt snipers.

Please use the model and upload your images. I want to see what you can make and generate with it!

In the meantime ENJOY THIS MODEL! THERE IS ALSO A 2ND MODEL ON THE HORIZON THAT IS BASED OFF THIS & MORE REALISTIC / SIMILAR BUT ONE THING AT A TIME. ;)

THIS MODEL HAS NO VAE -->RECCOMENDED VAE: illustriousXLV20_v10.safetensors

RECCOMENDED CFG: 3-7 (4.5-6.5 sweet spot)

RECCOMENDED STEPS: 23-35 (25-30 sweet spot)

RECOMMENDED HIGHRES PASS: Double your base render steps, 0.35-0.42,

RECCOMENDED CLIP SKIP: 2