this LORA on the Illustrious-XLv0.1 training standard, it is recommended to keep the weights between -2 and +2 (usually). As shown in the figure, the function of this LoRA is to deepen/lighten the contrast in the image. Different effects can be achieved by using positive or negative weights.

For different versions, please read About this version .

After testing, I discovered some weird features (?), for models trained on the Illustrious-XLv0.1, their stability and performance are much better. but for the Illustrious-XLv0.1 model itself, its stability is slightly worse... I started having some doubts about my work, but I still uploaded it.

Warning:

In some complex scenes or scenes with monotonous tones, this model may produce elements that differ significantly from the original image.