The company says Craft is available globally now to all Canva users in Dream Lab and to developers through its public API.
Craft generates images up to 2K resolution in more than 30 aspect ratios. Canva says prompts can be written in any language and that the model can generate text in 55 languages. Its system separates composition planning from rendering: a planning model specifies elements, layout, scale, hierarchy, and text before a rendering model produces the image.
Canva says the model was trained with emphasis on millions of human-made design templates from its content library, then fine-tuned using designer feedback. It also says Craft can support Magic Layers, which turn generated images into editable design elements.
The announcement includes Canva's own DesignArena rankings and performance claims, but does not provide independent evaluations, API pricing, rate limits, model documentation, or a release date for the planned faster variants.