Canva Releases Craft, Its First In-House Image Model

Design platform Canva has released Canva Craft, its first image generation model. The company's research team developed it to compete with comparable models from Google, OpenAI, ByteDance and Meta.

The model works in two stages. First it interprets the user's prompt and builds a plan that covers the spatial relationships of elements, their relative scale, the visual hierarchy and the details needed for text. Then it generates an image of up to 2K resolution, in any of 30 aspect ratios meant for different platforms.

Canva says this method produces accurate objects and legible text. Users can also make precise edits without changing the rest of the image.

Trained on designs, not photographs

The company did not train the model on photographs. It used human-designed templates from its own library, such as posters, social posts, presentations, infographics, menus and layouts heavy on typography. As a result, an image made with Craft can be converted into a layered design by Canva Magic, the company's image-to-design tool. Canva released a model that understands designs and layers last year.

Craft is open to all users on the platform through the Dream Lab tab. It will also come to the Canva API. Canva plans to bring image generation to third-party chatbots such as ChatGPT and Gemini through its MCP.

Researchers from Leonardo.ai, the media generation platform Canva acquired in 2024, contributed to the model. Two cost-efficient versions, Craftfast and Craft-ultrafast, are in development and are due soon.

Canva had been on track for $4 billion in revenue last year. It then cut its growth forecast for 2026, and outside analysts reportedly lowered its valuation from $42 billion to $35 billion.

Source: TechCrunch