Canva launches in-house AI image generation model with two-stage planning

Canva announced the launch of Canva Craft, its first in-house AI image generation model, which uses a two-stage architecture to plan a design before generating an image. This approach aims to deliver "greater precision, consistency and control" compared to existing models, according to the company.

The model supports image generation up to 2K resolution and more than 30 aspect ratios, enabling use cases such as social media posts, marketing content, realistic scenes, digital art, and environmental imagery. Canva states that the two-stage system improves accuracy for complex images requiring legible text and precise object placement.

Canva trained Craft on "millions of human-made templates" from its content library, which includes social media posts, presentations, infographics, and menus. This training data is intended to align the model’s outputs with Canva’s design ecosystem.

Canva Craft is currently ranked fourth on Design Arena’s leaderboard for AI image generation models, trailing OpenAI’s GPT-Image-2.5 Sunburst, GPT-Image-2.5 Flare, and another unspecified model. The ranking reflects performance on a crowdsourced benchmark for AI-generated designs.

The company highlights "robust input and output moderation, privacy and safety controls, and reporting tools" as part of the model’s deployment. Canva also plans to enable AI agents to integrate with Canva Craft through the open Model Context Protocol, and it intends to release more efficient variants—Craft-fast and Craft-ultrafast—in the future.

Canva Craft is available globally to all users, accessible via the three-dot option on the left-hand menu under "Dream Lab." The launch follows Canva’s earlier AI initiatives, including prompt-based editing and layers for AI-generated images introduced in April.