Qwen Image Edit

Qwen Image Edit represents Alibaba's contribution to open-source AI image editing. Built on a 20B-parameter multimodal diffusion transformer (MMDiT) architecture, it excels at precise text editing, semantic modifications, and appearance adjustments, with industry-leading performance in Chinese text rendering. The model supports multi-image input and output for complex compositions. A related innovation, Qwen-Image-Layered, can split images into individual editable layers with transparent backgrounds (RGBA), similar to Photoshop's layer system. Users can accurately modify text in images, add or remove objects, adjust poses, transfer styles, and enhance details. Being fully open-sourced under the Apache 2.0 license, developers can freely access model weights through Hugging Face and ModelScope, making it ideal for custom implementations and commercial applications.