Canva launches Canva Craft, its first in-house image generation model

Canva has released Canva Craft, its first in-house image generation model, available to all users through Dream Lab. The company said unlike traditional prompt-based generation, Craft first creates a plan that considers spatial relationships, relative scale, visual hierarchy, and text placement before producing images at up to 2K resolution. The model supports around 30 aspect ratios and allows users to edit specific elements without changing the rest of the image.

Rather than relying primarily on photographs for training, Canva Craft was trained on human-created templates from Canva's library, including posters, social media posts, presentations, infographics, menus, and typography-heavy layouts. Canva says this approach improves object accuracy, text legibility, and composition. Generated images can also be converted into editable, layered designs using Canva Magic's image-to-design technology.

Canva previously relied on external models for image generation, and developing its own model gives the company more control over costs, quality, and speed. Craft will also be available through the Canva API, with planned integrations for assistants such as ChatGPT and Gemini via the Model Context Protocol (MCP). Two more cost-efficient variants, Craftfast and Craft-ultrafast, are also in development.