MODEL OVERVIEW
Qwen Image 2.1
Qwen Image 2.1 is a 7B image model for text-to-image creation and image editing. Its published workflow highlights typography, native transparent RGBA output, and multi-image references.
One workflow for creation and editing
The model is designed to generate new scenes and revise existing images. A creation prompt describes the subject, setting, framing, and finish. An editing prompt names the change and protects the parts that should remain. This makes the written brief useful for both a blank canvas and an existing image.
Typography and transparent output
The release focuses on more legible text and layout in images, along with native transparent output for supported workflows. For a poster or package, put exact copy in quotation marks and describe its placement and hierarchy. For an isolated asset, request transparency and inspect the delivered file.
References for multi-subject scenes
Multiple reference images can help define subjects, wardrobe, objects, and settings. Give each image a specific role and describe how the final scene should combine them. The official repository and deployment documentation provide the current setup details and examples.