Canva releases its first AI image generation model, Canva Craft
Canva has introduced its first proprietary image generation model, Canva Craft, developed to compete against rivals like Google and OpenAI. Trained on human-designed templates, the model can generate images up to 2K resolution and convert them into layered designs using Magic Layers.
Design platform Canva has announced Canva Craft, its first image generation model developed to compete with models from tech giants such as Google, OpenAI, ByteDance, and Meta. Prepared by the company's own research team, with contributions from researchers at Leonardo.ai—which was acquired in 2024—the model has been made available to all users through the Dream Lab tab on the platform.
Canva Craft operates through a two-step process. In the first stage, it analyzes user prompts to plan details such as element relationships, scale, and text placement within the image. In the second stage, it generates images in 30 different aspect ratios and up to 2K resolution. This method ensures more accurate detailing of objects and keeps texts readable.
During the model's training process, human-designed templates from the Canva library, such as posters, presentations, and infographics, were used instead of photographs. As a result, the generated images can be transformed into editable, layered designs using the company's "Magic Layers" tool. Canva plans to release two new and faster model versions named "Craftfast" and "Craft-ultrafast" in the future.
Canva Craft is accessible via the platform's own interface as well as the Canva API. The company also aims to provide image generation support in third-party chatbots such as ChatGPT and Gemini through Canva's MCP architecture.
Sources
- techcrunch.com · Link