Alibaba’s AI Unit Qwen Releases Image-Editing Model That Separates Photos into Layers

The AI division of Alibaba, Qwen, has launched a new image-editing model called Qwen-Image-Layered. This model decomposes photographs into separate, editable components. It works by splitting images into multiple independent layers with transparent backgrounds (RGBA layers), allowing users to modify each layer individually without affecting the rest of the image.

The model supports basic editing operations like resizing, repositioning, and recoloring specific elements. Users can change backgrounds, replace people, alter text, as well as remove, move, and enlarge objects. Images can be separated into either 3 or 8 layers, and this process is repeatable—each layer can be further broken down into additional sub-layers as needed. The Qwen team describes this approach as bridging the gap between standard images and structured, editable representations.

The Qwen team has published the code on GitHub, and the models are available for download on the Hugging Face and ModelScope platforms.