The new model can generate structured, high-quality visuals, including comics, layouts, and multilingual designs, all directly from user prompts.