Canva launches Craft image model with planning stage for text and layout control
The AMW Read
Craft meaningfully updates Canva's role in generative media by combining an in-house, design-trained model with platform distribution, though workflow gains remain unproven.
Canva launches Craft image model with planning stage for text and layout control
Canva has launched Craft, an in-house image generation model that plans a design before rendering it. Canva says the two-stage architecture improves legible text and precise object placement in complex compositions. Trained on millions of human-made templates from its content library, Craft supports output up to 2K resolution and more than 30 aspect ratios. It is available globally through Dream Lab. Canva also plans Model Context Protocol access for agents and faster Craft-fast and Craft-ultrafast variants; those remain planned offerings.
The market significance is the connection between a creative platform's distribution and its own design-trained model. Within generative media, Canva is competing on the requirements of finished marketing assets: text, composition, and layout control. Its template library gives it a source of design-specific training material, although the announcement does not establish a durable performance advantage. The Verge reports that Craft ranks fourth on Design Arena's crowdsourced image-generation leaderboard, a result that offers a preference signal rather than proof of production reliability.
For builders and investors, the concrete test is whether the planning stage reduces corrections between an initial prompt and a usable asset. Buyers evaluating Craft should compare text accuracy, object placement, and revision effort on their own social posts, menus, and presentations. If those tasks require fewer manual fixes, Canva could capture more of the creative workflow inside its platform. The launch alone does not establish that productivity gain or its effect on customer spending.
