LongCat Image

LongCat Image prioritizes instruction-based editing, allowing users to modify specific regions using plain text. Its primary strength is maintaining perspective and lighting consistency during transformations, such as changing a subject's outfit or inserting objects into a scene. This is achieved through a 6B-parameter diffusion model that automatically infers which image regions need adjustment.