Lora trained in recreate the Redlight style.
Version 2: This model had several problems with eyes and fingers, so I pruned a lot of scketchy images from the dataset, keeping the most interestings features of this style.The resource tag was buggy and it didn't tag the model used for the images.
Sometimes still appears "speech bubbles" you can just add them to negative. So that helps to get more defined generations and better eyes.
Weight: .8-1
