qwen-image-edit

qwen-image-edit is an open-source AI image editing foundation model (20B parameters) created by Alibaba’s Qwen team. Built on top of the Qwen-Image generative model, it extends generation into practical editing workflows, letting you modify an existing image with natural-language instructions while keeping important identity cues intact (subject, layout, and overall semantics). It supports both high-level semantic edits (changing what’s in the scene) and appearance edits (changing how it looks), making it useful for everything from quick marketing iterations to more controlled creative production.

A standout capability is its text editing and rendering: Qwen-Image-Edit can add, delete, or replace text directly inside images, and it is designed to handle both Chinese and English with strong fidelity. This is particularly valuable for posters, product banners, UI mockups, and localization tasks where maintaining a believable font style, placement, and visual integration is critical.

You can run Qwen-Image-Edit locally with Python by loading the official pipeline and passing an input image plus an editing prompt to produce an updated output image. For interactive usage, the project also provides a Gradio-based demo experience, and it can be deployed with multi-GPU setups for heavier workloads. If you prefer minimal setup, it’s commonly available through web demos (such as Hugging Face Spaces or Qwen’s chat-style interfaces). For production needs, it can also be integrated via API-based workflows (e.g., REST endpoints) to support batch processing and commercial deployments.

Qwen-Image-Edit is released under the Apache 2.0 license, making it friendly for commercial and product integration. It also integrates well with popular ecosystem tools (including Hugging Face and ComfyUI), which helps teams plug it into existing creative pipelines.

Source code and updates are available on GitHub at: https://github.com/QwenLM/Qwen-Image