Canva releases Craft, its first image model, to all users

Canva released Canva Craft, its first image generation model, to compete with image models from other large firms. All users can open it in the Dream Lab tab, and the Canva team made it. Canva says Craft makes accurate objects and text that a person can read, but it gave no test results to show this.

How it works

Craft works in 2 steps. First it reads the user text and makes a plan for the position and dimension of each element, and for the text. Then it draws the image at a maximum of 2K resolution, in 30 shapes of image (aspect ratios).

The proof

Canva gave no test results. Canva says the text is accurate and that a change is only in one part. A Canva tool, Canva Magic, can turn a Craft image into a file with layers.

What is not known

Canva says the model gives more control of quality and speed. It can also lower the cost of models from other firms, but Canva gave no numbers. Next come 2 models with a lower cost, Craftfast and Craft-ultrafast.

Insight: Canva makes its own model as growth slows

Canva used some other image models before Craft, and it cut its 2026 growth forecast. Canva plans to put Craft in chatbots such as ChatGPT through MCP (a standard that connects AI chats to tools). This reads as a move to control the model, lower the cost and keep Canva in the chatbots that compete with it. The next step to monitor is if Craftfast lowers the cost.