Not just designs: Canva enters the big battle of artificial intelligence

Canva announced the launch of Canva Craft, its first image generation model, competing with models from Google, OpenAI, and Meta. The model was trained on unique design templates and enables the creation of high-resolution images with readable text.

The popular design platform Canva announced the launch of its first in-house image generation model, named Canva Craft. This move positions the company in direct competition with tech giants like Google, OpenAI, and Meta, which already offer similar models in the generative AI market. The new model operates using a unique two-stage technology: in the first stage, it interprets user instructions and creates a detailed plan, and in the second stage, it generates the image itself at a high resolution of up to 2K. Unusually, the model was trained on design templates from Canva's library, rather than on photo databases, allowing it to better understand composition and typography. According to the company, this approach enables the creation of images with precise details and clear, readable text, along with advanced editing capabilities that allow for targeted changes without affecting other parts of the image. The model is now available to all platform users and is expected to also be available via APIs for developers. In addition to upgrading capabilities, the independent model is expected to save the company significant costs previously associated with using external models. Canva, which acquired the Leonardo.ai platform in 2024, noted that researchers from the acquired company contributed to the model's development. In the near future, the company plans to release two faster and more cost-effective versions of the model.