To encourage bolder and more adventurous color combinations, the training dataset consists of 110 images featuring not only colorful subjects but also colorful backgrounds, images where the color scheme itself is a central theme, and images with asymmetrical color schemes.

Applying this LoRA makes "colorful" use more aggressive and adventurous color combinations, although the effect may be weaker with certain FT models.