>>2866876https://civitai.com/models/339104?modelVersionId=427433They probably explain it better. But you put images into it and it fills up 150 style components with values that pertain to the images. It's very much like the block weights or it was unet of loras. Instead of being based on images it's on values those images would make. So it won't copy biases an artist draws (hair, mouths etc) but just the style aspect.
