Canva Just Built Its Own Image AI — And It Changes How Designers Work

According to TechCrunch, Canva has launched its first proprietary image generation model, named Canva Craft. Developed with input from researchers at Leonardo.ai — which Canva acquired in 2024 — the model aims to compete directly with offerings from Google, OpenAI, and Meta. Unlike traditional generators trained purely on photographs, Canva trained this model using human-designed templates like posters, presentations, and infographics.

How does the two-stage generation work?

TechCrunch reports that Canva Craft operates in two distinct steps. First, it interprets the user's prompt and builds a detailed layout plan covering spatial relationships, scale, visual hierarchy, and text placement. In the second stage, it renders the final image at up to 2K resolution across up to 30 different aspect ratios.

How to access and use Canva Craft

Users can access the new model directly within the Canva interface. Here are the steps to start generating:

• Open any design project on the Canva platform.
• Navigate to the Dream Lab tab in the sidebar.
• Enter your detailed prompt specifying your layout or graphic needs.
• Generate the image and use Canva Magic tools to convert elements into fully editable, layered designs.

Is it worth it for your workflow?

For creative professionals who rely heavily on Canva for rapid asset creation, this release eliminates the friction of jumping between third-party image generators and design boards. Because the model was trained on actual graphic layouts, text legibility and object positioning are significantly sharper than standard generic generators. Upcoming variants like Craftfast aim to speed up generation further, making it a strong asset for daily production tasks.