Qwen-Image-Edit

Precise text rendering inside edited pictures sets Qwen-Image-Edit apart: built on the 20 billion parameter Qwen-Image foundation, it can add, delete, or modify Chinese and English text in an image while preserving the surrounding font, size, and style. The Qwen team routes the input image simultaneously into Qwen2.5-VL for semantic control and a VAE encoder for appearance control, letting one model cover high-level edits like style transfer, object rotation, and character reinterpretation as well as low-level ones like adding, removing, or recoloring a single element without disturbing the rest of the frame. Dated iterations released as Qwen-Image-Edit-2509 and 2511 added multi-image composition and consistency improvements, keeping the family among the most downloaded open editing models on Hugging Face. Inference runs through the Diffusers QwenImageEditPipeline on a CUDA GPU, with community quantizations and ComfyUI support extending reach to consumer cards. Weights and code are Apache-2.0, so commercial use is unrestricted.