Canva has officially launched “Craft,” its first AI-powered image generation model, developed by its in-house research team, Canva Research. Craft is now available to Canva users worldwide and can also be accessed by developers via a public API. The model is capable of generating images with resolutions up to 2K, supports over 30 aspect ratios, and can render text in 55 languages. According to Canva, Craft ranks fourth overall on DesignArena—a platform where users compare AI-generated designs—and holds the top spot in both the “People & Portraits” and “Landscapes & Nature” categories.
Features and Capabilities of Canva Craft
Canva reports that Craft ranks fourth overall on DesignArena, a crowdsourced platform for evaluating AI image generation models. Among 87 models, Craft ranks first in the “People & Portraits” category, and among 85 models, it leads in the “Landscapes & Nature” category. Additionally, the model ranks second in the “Abstract & Patterns” category and fourth in the “Products” category. Based on these results, Canva asserts that its research team is currently the second most advanced laboratory for AI-driven image generation globally.
Craft employs a two-stage process to generate images. First, it plans the composition based on the user’s prompt—determining the placement of objects and text—after which a second model renders the final image at a resolution of up to 2K. Users can choose from over 30 preset aspect ratios or define custom dimensions. This approach is designed to improve the rendering of text and the placement of objects within complex designs.
The model is trained on millions of templates from Canva’s content library, including posters, presentations, social media posts, infographics, and menus. This training enables Craft to handle layout and typography alongside image generation. It also integrates with Canva’s “Magic Layers” feature, which allows generated images to be converted into editable design elements. Canva has stated that it plans to introduce more advanced editing capabilities and reference image functionality in future updates.
Craft is capable of generating a wide range of content—from marketing materials and presentations to lifestyle photography, portraits, illustrations, digital art, landscapes, and architectural imagery. Users can input prompts in any language, and the model can generate text in 55 different languages.
Canva also noted that it has reduced image generation time for Craft through model refinement and a customized inference engine. The company is currently developing “Craft-fast” and “Craft-ultrafast” versions, which are expected to deliver faster image generation at a lower cost.
Availability of Canva Craft
Canva users can access Craft via Dream Lab, while developers can utilize its public API to integrate image generation capabilities into their own tools. Canva also plans to bring Craft to its MCP platform, which will enable AI assistants to generate images for designs as part of broader, automated tasks.
Craft also features Canva Shield, the company’s AI safety system. It includes content moderation, privacy and security controls, and tools for reporting potentially harmful or misleading content.