Canva has introduced Canva Craft, its first image generation model developed and trained internally, marking a new step in the company’s artificial intelligence strategy as competition intensifies among technology platforms offering generative AI tools.
Announced on October 22, the model was developed by Canva Research, the company’s in-house AI laboratory, and is designed to generate images with greater control over composition, typography and visual elements.
According to Canva, Craft can produce images at resolutions of up to 2K across more than 30 aspect ratios, supporting applications ranging from marketing materials and social media graphics to presentations, digital artwork and product visuals.
The launch places Canva in more direct competition with image generation technologies developed by OpenAI, Google, ByteDance and Meta.
How Canva Craft Works
Unlike conventional image generation systems that produce visuals directly from text prompts, Canva Craft uses a two-stage process.
The first stage interprets the user’s instructions and develops a composition plan, identifying the placement, size and relationships of visual elements, along with the text that should appear in the image.
A second model then generates the image based on that plan.
Canva says this architecture improves the accuracy of object placement, text rendering and overall design consistency.
The company trained Craft using millions of human-created design templates from its library, including presentations, posters, menus, infographics and social media materials.
Researchers from Leonardo.ai, the generative AI company Canva acquired in 2024, also contributed to the development, according to TechCrunch.
Canva Expands Control Over AI Technology
Developing its own image generation model represents an important step in Canva’s effort to expand its technology capabilities.
The platform has previously integrated models from external providers to support AI-powered image creation.
By developing proprietary technology, Canva gains greater control over model performance, product integration and the costs associated with delivering AI services.
The move could also reduce its reliance on external AI providers, although the company has not disclosed specific financial savings expected from Craft.
According to TechCrunch, Canva has been expanding its internal research capabilities while continuing to develop tools that combine artificial intelligence with editable design elements.
The company is also developing two additional versions, Craft-fast and Craft-ultrafast, intended to deliver faster image generation at lower computational costs.
Available to Users and Developers
Canva Craft is available globally through Dream Lab, the image generation feature within the Canva platform.
The company has also made the model accessible through its public API, allowing developers to integrate image generation into external applications and business workflows.
Canva plans to expand access through its Model Context Protocol (MCP) integration, enabling AI assistants and automated systems to generate visual content as part of broader tasks.
The company said the model debuted in fourth place on DesignArena’s crowdsourced image generation leaderboard, behind three OpenAI models.
Such rankings reflect performance within a particular evaluation system and do not necessarily establish superiority across all image generation tasks.
AI Investment Reshapes the Design Software Market
The launch comes as design software companies expand their use of artificial intelligence to automate content creation, editing and production workflows.
The technology also creates opportunities for the company to extend its services beyond its existing platform through developer integrations.
However, the financial impact of Craft remains uncertain. Canva has not disclosed the development cost of the model or specific revenue expectations associated with its launch.
Sources: Canva, official announcement, October 22, 2026; The Verge, reporting by Emma Roth, October 22, 2026; TechCrunch, reporting by Ivan Mehta, October 22, 2026.
