Search

Qwen December 2025: Qwen-Image-Layered revolutionizes image generation

Qwen December 2025: Qwen-Image-Layered revolutionizes image generation

Qwen-Image-Layered: Image generation moves to the next level

December 19, 2025 — Alibaba Cloud unveils Qwen-Image-Layered, an image generation model with native decomposition into layers, completely open-source.

This model marks a significant advance in the field of AI image generation, offering a feature previously reserved for professional software like Photoshop: the management of independent RGBA layers.


Key Features

FeatureDescription
Isolated RGBA LayersEach layer is physically separated with native transparency
Prompt ControlExplicit specification of 3 to 10 layers
Flexible StructureFrom coarse layouts to fine details
Native EditabilityIndividual modification of each layer

Technical Architecture

Qwen-Image-Layered relies on the VLD-MMDiT (Variable Layers Decomposition MMDiT) architecture and uses an RGBA-VAE encoder to produce layers with a native alpha channel.

The model allows for infinite decomposition of images, offering creators granular control over each visual element.


Available Integrations

The ecosystem reacted quickly with integrations available from launch day:

PlatformIntegrationDate
fal.aiAPI availableDec 19
ComfyUIOfficial pluginDec 19
PrunaAISpeed optimizationDec 20

Qwen Image Layered is live on fal! — @Alibaba_Qwen on X


Use Cases

  • Graphic Design: Creation of compositions with editable elements
  • Game Art: Asset generation with background/character separation
  • Marketing: Modular visuals for multi-channel campaigns
  • Prototyping: Quick mockups with repositionable elements

Availability

ResourceLink
Source CodeGitHub QwenLM
ModelHugging Face Qwen
fal.ai Demofal.ai

The model is published under an open-source license, allowing for commercial and academic use.


What this means

With Qwen-Image-Layered, Alibaba positions Qwen as a major player in image generation, offering professional capabilities for free that were previously inaccessible. This open-source approach accelerates adoption and innovation in the creative ecosystem.


Sources