.jpeg)
Valorant Lora Style
Valorant Style flux:
Trained using 76 artworks. Captions generated using Qwen2-VL-7B and reviewed manually.
Valorant Style SDXL:
Trained using 91 artworks using CyberRealistic as base model. I tested it with RealVis40 lightning and looks good.
Sometimes backgrounds get blurry, just add "blur, blurry background" on negative prompt, anyway I fixed the dataset so this version should work fine without negatives.