Canva releases the Craft image generation model; Shanghai AI Lab open-sources the ShuSheng-A1.5 research agent
Canva released the Craft image generation model on October 22. It uses a two-stage process—composition planning first, then pixel rendering—supports 2K output, more than 30 aspect ratios, and in-image text in 55 languages, and is available through Canva Dream Lab and a public API.