Canva Bets Its First Image Model Can Think Before It Draws
Canva Bets Its First Image Model Can Think Before It Draws
Canva’s push to bring generative AI deeper into its design platform began before Craft’s debut. In 2024, the company acquired media-generation platform Leonardo.ai; Canva says Leonardo researchers contributed to the new model.
By April, Canva had added prompt-based editing and the ability to preserve layers in AI-generated images, allowing users to alter one part of a design without remaking the rest. Those features foreshadowed the company’s broader argument: design work needs more than a striking single image.
Now Canva has launched Canva Craft, its first in-house image-generation model. Rather than immediately rendering a prompt, Craft works in two stages. First, it maps out spatial relationships, scale, visual hierarchy and text; only then does it generate the image. Canva says that planning layer is meant to deliver “accurate details of objects and legible text,” while retaining the ability to make precise edits without disturbing the rest of the composition.
The company trained the model on “millions of human-made templates” from its own library—posters, social posts, presentations, infographics, menus and other typography-heavy layouts—rather than photographs. That distinction is central to Canva’s pitch: Craft is not simply another image engine, but a system designed to turn generated imagery into editable, layered documents through Canva Magic.
Craft can produce images up to 2K resolution in more than 30 aspect ratios and is available through Dream Lab, with API access planned alongside integrations for third-party chatbots. Canva is also developing lower-cost Craftfast and Craft-ultrafast versions. The launch puts the company in more direct competition with Google, OpenAI, ByteDance and Meta—while potentially reducing its reliance on outside model APIs and giving it tighter control over speed, quality and cost.
Write a comment