Canva presents the AI model Craft for generating designs

Canva has introduced its own image generation model, Craft, developed from scratch by the Canva Research team. Unlike standard generators, the system first designs the composition, determining object placement and visual hierarchy, before creating the final graphic.

The tool is designed for creating meaningful design layouts for ads, posters, and covers.

The model supports resolutions up to 2K and over 30 aspect ratios. It works with multilingual prompts and allows adding text in 55 languages.

A key feature is integration with Magic Layers, which converts the result into editable layers within the Canva editor. This distinguishes Craft from other solutions focused solely on photorealism.

Key features of the model:

• High-resolution image generation with flexible proportions
• Text support in multiple languages
• Ability to adjust individual composition elements
• Conversion of results into editable layers

Analogues on the market include Recraft and Ideogram, which also pay attention to typography and controllability. However, Canva aims to combine generation with a full-fledged layout editor.

According to DesignArena, the model ranked fourth overall, leading in specific categories. Access to the tool is available via Canva Dream Lab and API.

Local deployment is not currently supported. Developers plan to release accelerated versions: Craft-fast and Craft-ultrafast.