v3 is a bit more subtle than v2, more focused on achieving the downblouse pose/concept than on ensuring nipples pop out. It was achieved by training a LoRA on a subset of the best downblouse images, then adjusting weights and merging with v2. It is less sensitive to the nipslip concept.
v3 is a bit more subtle than v2, more focused on achieving the downblouse pose/concept than on ensuring nipples pop out. It was achieved by training a LoRA on a subset of the best downblouse images, then adjusting weights and merging with v2. It is less sensitive to the nipslip concept.