Qwen-Image: Crafting with native text rendering

9 months ago

Qwen-Image, a 20B MMDiT image foundation model, is released with advanced text rendering and image editing capabilities.
Key features include superior text rendering (supporting multiple languages), consistent image editing, and strong cross-benchmark performance.
Demonstrates high-fidelity text rendering in Chinese and English, including complex layouts and small text details.
Excels in generating posters, PPTs, and diverse artistic styles, from photorealistic to minimalist designs.
Supports various image editing operations like style transfer, additions, deletions, and detail enhancement.
Aims to lower technical barriers in visual content creation and foster an open generative AI ecosystem.

Hasty Briefsbeta