Canva has introduced Canva Craft, its first image-generation model, designed to create visuals with clearer details, readable text and more precise editing controls.
The model works in two steps: it first interprets a prompt and plans elements such as layout, scale and visual hierarchy, then generates an image at up to 2K resolution in 30 aspect ratios. Users can make targeted edits without altering other parts of a design.
Rather than training on photographs, Canva used human-designed templates from its library, including posters, presentations, menus and social posts. The company says this design-focused approach also means Craft outputs can be turned into layered projects with Canva Magic.
Craft is available through Canva's Dream Lab and will also be offered through the Canva API. Canva plans to enable image generation in third-party chatbots, including ChatGPT and Gemini, through its MCP integration. Researchers from Leonardo.ai, acquired by Canva in 2024, contributed to the model. Two faster, lower-cost versions--Craftfast and Craft-ultrafast--are in development.
By building image generation around design structure, Canva is aiming to connect visual creation more closely with editable design workflows, potentially making AI-generated content easier to refine and use across platforms.
