Canva Craft: An Image Generator Based on Design Layouts

Canva Craft is trained on millions of design templates and takes composition, hierarchy, and typography into account. The model creates images up to 2K resolution and allows individual elements to be edited.

Canva Craft: An Image Generator Trained on Design Layouts

Canva has introduced Canva Craft, its first image-generation model. Unlike systems trained primarily on photographs, Craft was developed using design templates from the company’s library.

The Canva platform, which competes with Adobe and is used by 95% of Fortune 500 companies, developed the model with its own research team. According to TechCrunch on October 22, 2026, Craft is intended to compete with image generators from Google, OpenAI, ByteDance, and Meta.

The model creates images at resolutions of up to 2K and supports more than 30 aspect ratios. It can be used to create social media and marketing materials, as well as realistic scenes, digital illustrations, and environmental images.

Training on Templates Instead of Photographs

Canva emphasizes that Craft was trained not on photographs, but on millions of human-created layouts from the service’s library. These include professionally designed posters, social media posts, presentations, infographics, menus, and text-heavy compositions.

According to the company, finished flat images do not give a model the same understanding of how design works. Templates, by contrast, show how element placement, visual hierarchy, and typography relate to one another. This helps Craft learn not only to reproduce an image’s appearance, but also to account for the principles of effective composition.

How Craft Builds an Image

Image generation takes place in two stages. First, the model interprets the prompt and plans the composition: it determines where objects will be placed, how large they will be, how to establish visual emphasis, and where to position text. Craft then creates an image at up to 2K resolution, choosing from more than 30 aspect ratios for different platforms.

Canva links this approach to more precise detailing and improved text readability. The company also says users will be able to edit individual parts of an image without changing the other elements.

Thus, Canva sees Craft’s key feature as not only image generation, but also training on materials created by designers: the model is meant to take a layout’s structure and design into account from the moment it begins creating the result.