Canva Launches Craft, Its First In-House Image Model, in Dream Lab and the Canva API

Canva released Craft on October 22, its first image generation model, built and trained from scratch by its in-house lab, Canva Research. Canva says Craft debuted as the fourth-ranked image model on DesignArena, a crowdsourced ranking of AI design models; that is Canva's own claim based on that ranking.

How it works

Craft works in two stages. The first reads the prompt and makes a plan: where objects go, their relative size, the visual hierarchy and any text. The second renders the image at up to 2K resolution in more than 30 aspect ratios for different platforms. Canva says this gives accurate details and legible text, and lets you edit one part of an image without changing the rest.

The training data is unusual. Canva says it trained Craft on human-designed templates from its own library, such as posters, social posts, presentations, infographics, menus and text-heavy layouts, rather than on photographs. Images from Craft can then be turned into layered, editable designs with Canva's Magic Layers. Prompts can be written in any language, and Craft can render text in 55 languages.

Availability

Craft is available now to all Canva users worldwide in the Dream Lab panel and to developers through Canva's public API. A connection through the Canva MCP server, which lets AI assistants create images as part of a larger task, is marked as coming soon. Two faster versions, Craft-fast and Craft-ultrafast, are on the way and, Canva says, will cost a fraction of leading frontier models. Canva has not published per-plan limits.

Who it matters to

Craft targets the work Canva users already do: marketing graphics, social posts and slides, where an image has to fit a layout and carry readable text. That is a different strength from generators built mainly for artistic or photographic images.

What this means if you make images for design work

If you already design in Canva, try Craft in Dream Lab on a poster or social post that includes text. That is where Canva says it is strongest, and the result goes straight into an editable design.

If you want stylized art or photographic images, Midjourney is still a common choice. If text inside the image matters most, compare it with a text-focused generator in Ideogram vs Midjourney.