For a long time, converting a regular flat photo into an editable layer file has been a "must-have" for professional designers. According to AIbase, Alibaba's artificial intelligence department Qwen has recently launched a revolutionary image editing model - Qwen-Image-Layered. This model can directly decompose static photos into multiple independent RGBA layers with transparent backgrounds, giving AI image editing the structured operation capabilities similar to Photoshop.

image.png

Traditional AI image editing often involves redrawing the entire image, making it difficult to achieve precise control over specific elements. However, this new model from Qwen can intelligently segment the image into 3 or 8 layers. Users can independently scale, reposition, change colors, or delete a specific layer, without affecting other parts of the image. For example, users can easily change the background, replace a person, modify text, or even enlarge an object and move it to another location, all as if they were working on a layered PSD file.

More innovatively, this layering process is repeatable. AIbase noticed that users can further subdivide individual layers into more sub-layers according to their actual needs, achieving extremely fine image modifications. The Qwen team stated that this technology bridges the gap between standard images and structured, editable representations, greatly lowering the barrier to complex image processing.

Currently, Alibaba has officially open-sourced the code of this model. Developers and users can obtain the model and conduct practical tests on Hugging Face and ModelScope.

modelscope:https://modelscope.cn/models/Qwen/Qwen-Image-Layered

Key Points:

  • 📸 Automated Layering: It can intelligently split a single-layer photo into multiple independent RGBA layers with transparent channels, turning a regular image into an "editable PS draft" in an instant.

  • 🎨 Precise Non-Destructive Editing: Supports independent adjustment of the size, position, and color of specific layers, enabling replacement of a person or changing the background color without affecting other elements in the image.

  • 🔓 Full Stack Open Source Technology: The model code has been released on GitHub, along with an online demo, aiming to promote the transition of AI image editing towards structured and professional approaches.