Design platform Canva has released Canva Craft, its first image generation model, built by its research team to compete with similar models from Google, OpenAI, ByteDance and Meta. Researchers from Leonardo.ai, which Canva acquired in 2024, contributed to its development.
The model works in two stages. It first interprets a user’s prompt and plans details such as spatial relationships, relative scale, visual hierarchy and text, then generates images of up to 2K resolution across 30 aspect ratios. Canva said this approach produces accurate object details and legible text, while allowing precise edits without altering other parts of an image.
Rather than photographs, Canva trained the model on human-designed templates from its library, including posters, social posts, presentations, infographics and menus. As a result, generated images can be turned into layered designs using Canva Magic, its image-to-design tool.
Canva Craft is available to all users through the Dream Lab tab and the Canva API, with support planned for ChatGPT and Gemini via Canva’s MCP. The company said the model gives it better control over quality, speed and cost, and that two cheaper versions, Craftfast and Craft-ultrafast, are coming soon.
The release follows Canva’s reduced 2026 growth forecast, which analysts say cut its valuation from $42 billion to $35 billion.
