Canva has released Craft, its first in-house image generation model, built from scratch by its Canva Research lab and available to all users through Dream Lab and to developers via a public API.
Craft uses a two-stage architecture that separates compositional planning from image rendering, allowing for legible text and precise object placement, according to Canva.
Canva says it trained Craft on millions of human-made templates from its content library, including posters, social posts, presentations, and infographics, rather than primarily on photographs.
Craft ranked fourth on DesignArena's leaderboard behind three OpenAI models, making Canva Research the second-ranked AI lab for image generation on the platform, the company said.
Canva plans to release two cheaper variants of Craft called Craft-fast and Craft-ultrafast, and researchers from Leonardo.ai, which Canva acquired in 2024, contributed to the model.
