Alibaba's AI unit Qwen has launched an image editing model called Qwen-Image-Layered. This model allows users to separate images into individual layers with transparent backgrounds (RGBA layers), enabling detailed edits on each layer without affecting others. Users can perform tasks like resizing, repositioning, and recoloring specific elements, swap backgrounds, replace people, modify text, and manipulate objects by deleting, moving, or enlarging them. The model can split images into either three or eight layers, and the process can be repeated—layers can be further subdivided into additional layers for more granular editing. This feature positions Qwen’s approach as a practical tool for creating structured, editable image representations, bridging the gap between traditional images and advanced editing capabilities. Qwen has made the code publicly available on GitHub, with models accessible on platforms like Hugging Face and ModelScope. For those interested in testing the model, demos are offered on the same platforms, providing a hands-on experience with this new image editing technology.