Canva announced the launch of Canva Craft, its first image generation model developed and trained from scratch by the internal Canva Research team. The model is available starting October 22, 2026, to all platform users in Dream Lab, and also via public API.
Unlike conventional models that generate pixels directly from the prompt, Craft uses a two-stage architecture. In the first stage, a planning model interprets the request and creates a detailed composition plan, including elements, spatial relationships, visual hierarchy, and text. In the second stage, a rendering model produces the final image at up to 2K resolution, in more than 30 aspect ratios.
Focus on design
Training emphasized millions of human-created templates in Canva's library rather than primarily photographs. This helps the model understand layout, typography, and hierarchy. According to the company, Craft ranked fourth overall on DesignArena and led categories such as people/portraits and landscapes/nature.
Outputs are particularly well-suited for conversion into editable layers via Canva's Magic Layers. The company plans faster and cheaper variants (Craft-fast and Craft-ultrafast) and future integration with assistants via MCP.
The model includes Canva Shield safety controls. With more than 250 million users generating hundreds of designs per second, Canva aims to reduce dependence on external models from competitors like OpenAI, Google, and others.
Sources
Transparency: This content was created, edited, or reviewed with the aid of artificial intelligence. Information was cross-checked with public posts on X and sources available on the internet. Consult the original sources for the full context.
By GeekikiBot
