[REVISION HISTORY]

Canva launch of Craft AI model

Updated 1 time since CLSTR started tracking revisions of this situation.

What changed

2026-10-26 07:41 UTC → 2026-10-27 06:00 UTC · added ~~removed~~

Canva has introduced Craft, its first proprietary image generation foundation model developed by Canva Research. Designed specifically for design tasks like creating posters, infographics, and presentations, the model differs from competitors by focusing on design principles rather than purely photographic realism. The model employs a two-stage architecture: an internal planning model that interprets prompts to create a compositional plan for object placement and visual hierarchy, followed by a rendering engine that produces a final image up to 2K resolution. Because it was trained on millions of human-designed templates, the model supports high precision in layout, typography, and legible text across 55 languages. It also supports more than 30 aspect ratios. Craft is available via the Dream Lab interface and a public API, with integration planned for third-party chatbots. The technology allows for targeted edits on specific elements through multi-layered designs. To optimize cost-efficiency, the company is also developing lighter versions named Craft-fast and Craft-ultrafast. Early benchmarks on DesignArena placed the model fourth overall, ranking highly in categories such as portraits and landscapes.

Versions

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.