colorfulart

Trained on aesthetic images (modern art). The colorful model was used as the basis (colorful 3.1 has this lora with 0.3 weight), so this model probably is the best choice, but feel free to experiment. Best, but slow generations, results with DPM++ SDE Karas (DDIM, DPM++ 2M Karas are good to but with less details). Some blurs with euler family samplers. I use Hires fix for more details. Minimal affect on style at 0.3 (only noise/contrast will be added), balanced results around 0.8 and strong with 1.0. My goal was create something like perlin noise as base for more Artful in results