Story
Oct 23, 2026
Canva Thinks Planning Is the Fix for AI Images That Miss the Brief
Canva’s bet is that image AI needs to reason about a design before rendering it: a strategy it says will make generated visuals more accurate, editable and economical than the hit-or-miss outputs users have learned to expect.
Canva’s push into generative AI began before Craft. In April, the design platform added prompt-based editing, allowing users to create or alter designs through written instructions, and introduced layers for AI-generated images so one element could be changed without disturbing the rest.1
Now it is taking a more consequential step: building its own image model. Canva Craft, released to all users through the Dream Lab tab, is the company’s first in-house image generator and is also slated for the Canva API and third-party chatbots through its MCP integration.2
The company’s central argument is that image generation should not leap straight from prompt to picture. Craft works in two stages: first, it interprets a request into a design plan covering spatial relationships, scale, visual hierarchy and text; only then does it generate the image.2 Canva says that sequence improves two stubborn weak points for image AI — legible writing and precise object placement — while enabling edits that leave the rest of a composition intact.1
Craft can produce images up to 2K resolution across more than 30 aspect ratios, spanning social posts and marketing assets as well as realistic scenes, digital art and environmental imagery.1 Rather than training chiefly on photographs, Canva says it used millions of human-made templates — posters, presentations, infographics, menus and typography-heavy layouts — from its own library. That could also allow outputs to be turned into layered designs through Canva Magic.2
The release is as much about control as creative ambition. Canva has previously relied on several outside image models; an internal option gives it more say over quality and speed, and could lower API costs.2 Leonardo.ai researchers, brought in through Canva’s 2024 acquisition of the media-generation platform, contributed to Craft. Canva says cheaper Craftfast and Craft-ultrafast versions are next.2
Craft arrives in a tough field: it ranks fourth on Design Arena’s crowdsourced image-model leaderboard, behind three OpenAI GPT-Image-2 variants.1 But Canva is wagering that, for designers, following the brief may matter more than simply making a striking picture.