Canva releases Canva Craft, its first in-house image generation model
The model is trained on Canva's human-designed templates rather than photographs, is available to all users from the Dream Lab tab, and is coming to the Canva API.
Canva released its first image generation model, Canva Craft, on October 22, TechCrunch reports. The model plans an image first, working out layout, scale and text, then "generates an image of up to 2K resolution in a possible 30 different aspect ratios." Rather than photographs, Canva trained it on designs: "human-designed templates from Canva's library," including posters, social posts, presentations and menus. Images it produces can be turned into layered, editable designs.
Where it is available
Canva Craft is "available on the company's platform to all users, who can access the model from the Dream Lab tab," and "will also be available through the Canva API." Canva also plans to support it in third-party chatbots such as ChatGPT and Gemini through its MCP server, and says two cheaper versions, Craftfast and Craft-ultrafast, are coming. Canva has relied on several outside image models, and TechCrunch notes the in-house model is likely to save the company API costs; Canva itself cites quality and speed control as well as cost.
Why it matters
Software vendors building their own models to cut what they pay outside providers is a pattern worth tracking: it changes the vendor's margin, not necessarily your price. If generative features are part of why you pay for a design tool, check whether the plan you are on includes them, and how usage is counted, before the cheaper Craft versions arrive.