The new model operates in a two-stage process. Initially, it translates user prompts into a structured plan that maps spatial relationships, visual hierarchy, and object scale. This foundation allows the system to output 2K resolution images across 30 distinct aspect ratios. Because the underlying data is derived from Canva’s professional design catalog, the generated assets are compatible with the platform’s Magic tool, enabling users to convert static images into fully editable, layered designs.
Canva challenges AI giants with proprietary image generation model
Design platform Canva has entered the crowded generative AI field with its own model, Canva Craft, aiming to surpass existing tools from Google and OpenAI. By training the system on its vast library of human-designed templates rather than raw photographs, the company hopes to offer superior control over typography and layout composition.
Accessibility remains a core priority for the rollout. Users can currently find the model within the Dream Lab tab, with broader integration planned via Canva’s API and the Model Context Protocol for third-party chatbots like ChatGPT and Gemini. The project incorporates expertise from Leonardo.ai, a media generation platform acquired by Canva in 2024. As the company navigates a reported valuation shift from $42 billion to $35 billion, this in-house technology serves a dual purpose: improving creative output while curbing the long-term API costs associated with third-party providers. Future iterations, dubbed Craftfast and Craft-ultrafast, are already in development to further optimize speed and efficiency.
Comments (0)
No comments yet. Be the first!