The Year in AI Papers
The year/Qwen

Paper 2512.15603

Qwen-Image-Layered: Towards Inherent Editability via Layer Decomposition

Shengming Yin, Zekai Zhang, Zecheng Tang, Kaiyuan Gao, Xiao Xu, Kun Yan, Jiahao Li, Yilei Chen, Yuxiang Chen, Heung-Yeung Shum, Lionel M. Ni, Jingren Zhou, et al.

Published
Dec 2025
Research lab
Qwen
Citations
24
GitHub
2.1K stars

01 In brief

Summary

Qwen-Image-Layered is an end-to-end diffusion model that decomposes a single RGB image into multiple semantically disentangled RGBA layers, enabling consistent image editing where each layer can be independently manipulated.

The model introduces three key components: an RGBA-VAE that unifies latent representations for RGB and RGBA images, a VLD-MMDiT architecture supporting variable-length layer decomposition, and a multi-stage training strategy that adapts a pretrained image generation model.

To address the lack of high-quality multilayer training data, the authors built a pipeline to extract and annotate layers from Photoshop (PSD) files.

Experiments on the Crello dataset show that Qwen-Image-Layered outperforms existing methods (e.g., LayerD) in decomposition quality, achieving lower RGB L1 and higher Alpha soft IoU.

The RGBA-VAE also surpasses prior VAEs in reconstruction metrics (PSNR, SSIM, rFID, LPIPS) on AIM-500.

Qualitative results demonstrate superior performance in image editing and multilayer synthesis compared to Qwen-Image-Edit-2509 and ART.

The model is released on GitHub.

02 From the paper

Abstract

Recent visual generative models often struggle with consistency during image editing due to the entangled nature of raster images, where all visual content is fused into a single canvas. In contrast, professional design tools employ layered representations, allowing isolated edits while preserving consistency. Motivated by this, we propose \textbf{Qwen-Image-Layered}, an end-to-end diffusion model that decomposes a single RGB image into multiple semantically disentangled RGBA layers, enabling \textbf{inherent editability}, where each RGBA layer can be independently manipulated without affecting other content. To support variable-length decomposition, we introduce three key components: (1) an RGBA-VAE to unify the latent representations of RGB and RGBA images; (2) a VLD-MMDiT (Variable Layers Decomposition MMDiT) architecture capable of decomposing a variable number of image layers; and (3) a Multi-stage Training strategy to adapt a pretrained image generation model into a multilayer image decomposer. Furthermore, to address the scarcity of high-quality multilayer training images, we build a pipeline to extract and annotate multilayer images from Photoshop documents (PSD). Experiments demonstrate that our method significantly surpasses existing approaches in decomposition quality and establishes a new paradigm for consistent image editing. Our code and models are released on \href{https://github.com/QwenLM/Qwen-Image-Layered}{https://github.com/QwenLM/Qwen-Image-Layered}