Canva Craft: Canva’s First AI Image Generation Model Is Here

• Canva Craft is the company’s first image generation model, built from scratch by its in-house Canva Research team.
• The model creates images at up to 2K resolution, supports text in 55 languages and offers more than 30 aspect ratios.
• Craft is available globally through Canva Dream Lab and the public API, with Magic Layers enabling editable design layers.

Canva Craft is now available as the design platform’s first image generation model developed and trained from scratch by Canva Research, its in-house AI lab. The model is built for design-focused image creation, with a system that plans the composition before rendering the final image.

Canva says Craft debuted as the fourth-ranked image generation model on DesignArena, making Canva Research the world’s second-ranked AI lab for image generation on the platform. It is available globally to Canva users through Dream Lab and can also be accessed by developers through Canva’s public API.

Canva Craft Image Generation Model: Key Details

Canva Craft is designed to create a range of visuals, including marketing material, social media graphics, posters, portraits, landscapes, illustrations and other creative compositions. Canva says the model can generate images at up to 2K resolution and supports more than 30 aspect ratios.

On DesignArena, Craft ranks first globally in People & Portraits, a category covering 87 models, and Landscapes & Nature, which covers 85 models. It is ranked second for Abstract & Patterns and fourth for Products. Overall, the model debuted in fourth place on the image generation leaderboard.

These rankings reflect DesignArena’s evaluations and categories rather than a guarantee that every generated image will meet a particular user’s needs. Results can still depend on the prompt, intended design and the changes made after generation.

How Canva Craft Generates Images

Craft uses a two-stage architecture. First, a planning model interprets the prompt and prepares a detailed composition plan. This plan covers the elements to include, their spatial relationships, relative scale, visual hierarchy and any text that should appear in the image.

A separate rendering model then turns that plan into the finished image at up to 2K resolution. By separating planning from rendering, Canva aims to improve object placement, composition and the accuracy of text within more complex designs.

The model can render legible text in 55 languages, while prompts can be written in any language. This is intended to help users create graphics with headings, labels and other text without having to write prompts in a single supported language.

‹›1 / 6

Craft’s training focused on design rather than relying mainly on photographs. Canva says its training data includes millions of human-made templates from its content library, such as posters, social posts, presentations, infographics, menus and typography-rich layouts.

After pre-training, Canva refined the model using feedback from its professional designers. The company says designers took part in more than 25,000 head-to-head image comparisons to assess outputs and improve instruction following, typography accuracy and overall visual quality.

Canva Craft Works with Magic Layers

Craft is designed to work with Canva’s Magic Layers feature, which can turn generated images into editable design elements. Canva describes the process as using JSON boundary boxing to identify and separate parts of an image into layers, making it easier to adjust individual elements instead of treating the whole output as one flat image.

This approach can be useful when a design needs further changes after generation, such as adjusting the placement of an object or refining the layout. The editable output is a key part of Craft’s design-focused approach.

Canva Craft Availability

Canva Craft is available now to users globally through Dream Lab in Canva. Developers can also access the model through Canva’s public API, allowing them to build image generation into their own tools and workflows.

Canva MCP support for AI assistants and automated workflows is coming soon. The company also says faster Craft-fast and Craft-ultrafast variants are on the way. These versions are intended to improve speed and efficiency while maintaining image quality.

Canva Shield Safety Measures

Canva says Craft is backed by Canva Shield, its AI safety system. This includes input and output moderation, privacy and safety controls, ongoing evaluations and reporting tools that allow users to flag potentially harmful or misleading content.

The company says its evaluations consider visual quality, brand safety, cultural context and potential harmful content as part of the development and deployment process.

Conclusion

Canva Craft marks Canva’s move into image generation with a model built around design composition, readable text and editable outputs. Its global availability through Dream Lab and the public API puts the model within reach of both everyday Canva users and developers, while MCP integration and faster variants are planned for a later stage.

Frequently Asked Questions