Model had a very high starting loss at 0.4 and was trained down to a loss of .198

Lora will reduce color in most situations and can be used in the negative to increase color.