Canva Craft, the company’s first image-generation model, is available globally in Dream Lab and through its public API. Canva announced the model on October 22.
Developed by Canva Research, Craft creates images at up to 2K resolution, supports more than 30 aspect ratios and can generate text in 55 languages. Users can also specify a custom aspect ratio.
Canva says the model first plans an image’s composition, including text, object placement and relative size. A separate rendering stage produces the picture from that plan.
The company reports that Craft debuted fourth among image models on DesignArena, which ranks models through community comparisons.
Magic Layers handles the editable elements
Craft’s planning structure supports Canva’s existing Magic Layers tool, which converts flat pictures into designs with individually editable text and objects.
Canva’s Magic Layers explanation describes how it identifies components in a flat image and reconstructs them in the editor. It can recognize text and rebuild it as editable text boxes, alongside separate objects and backgrounds.
This gives a designer a way to change a heading’s wording or font while working with the surrounding layout. A flat JPEG or PNG normally stores that heading as pixels, with no separate text box to select.
The company’s Magic Layers teaching resource covers updating text, colors and elements after conversion.
Canva lists Craft’s MCP connection for AI assistants as coming soon. Faster Craft-fast and Craft-ultrafast variants are also planned, without a release date in the announcement.
Source: Canva
Start the conversation by posting the first comment