AI BriefingPickQwen-Image-2.1: Transparent background generation and 10-image reference editing with 7B parameters
Qwen/Qwen-Image-2.1
About the project
This model integrates two capabilities: generating images from text and editing existing photos. The visual generation component features 7B parameters, balancing computational efficiency and generation quality through 32 Single-Stream DiT layers.
It natively supports transparent background (RGBA) generation, making it ideal for creating stickers and icons. It accepts up to 10 reference images to maintain the identity of people or products during editing, and allows precise modifications to specific areas using circular or mask-based constraints.
It supports various aspect ratios from 1:1 to 16:9, with improved text rendering and human lighting processing for realistic textures. It can be easily integrated via the Diffusers library, and memory usage can be reduced by utilizing CPU offloading.
Qwen/Qwen-Image-2.1
The original page has no description.
text-to-image
This introduction was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report errors, attribution issues, or removal requests via Contact.