The dataset uses images from the character's 3D model, a small amount uses official illustrated images of the game.
Trained in Illustrios v0.1
🌺 STEPS: 25 ~ 32
🌺 CFG: 5 ~7
🌺 SAMPLER: Euler A
🌺 RESOLUTION: 1MP
The sample images are of the training and the rest using my own checkpoint with a Lora style.