Canva introduced Craft, its first image generation model built from scratch by the in-house Canva Research lab. Unlike most generators that produce pixels directly from the prompt, Craft uses a two-stage architecture: first it plans the composition (elements, scale, visual hierarchy and text) and only then renders the image at up to 2K resolution across more than 30 aspect ratios.
Focus on design, not photos
Training prioritized millions of human-made templates from Canva’s own library — posters, posts, presentations, infographics and menus — rather than photographs. This helps the model understand layout, typography and hierarchy. According to the company, Craft ranked fourth overall on DesignArena and first in categories such as People & Portraits and Landscapes & Nature.
The tool is available to all Canva users via Dream Lab and also through the public API. It accepts prompts in any language and generates text in 55 languages. Canva states that the model undergoes moderation and safety controls under Canva Shield. Faster versions (Craft-fast and Craft-ultrafast) are planned.
Why it matters
For content creators, designers and marketing teams, the ability to generate images with legible text and controlled layout reduces correction work. The plan-then-render approach is a differentiator versus generic image models. The launch reinforces the trend of design platforms building their own models instead of relying exclusively on third-party APIs.
Sources
This text was produced with the assistance of artificial intelligence and editorially reviewed by GeekikiBot.
By GeekikiBot
