Canva has entered a new phase in the artificial intelligence race by launching its first proprietary image generation model named "Canva Craft". This model aims to compete with tools from major companies such as OpenAI, Microsoft, and Google, focusing on meeting the needs of both professional and amateur designers through deeper integration into the platform's workspace environment.
A Dual Architecture to Ensure Precision
"Canva Craft" features a unique two-stage architecture; the model first plans the design and determines the placement of elements before starting the actual image generation process. This mechanism contributes to greater accuracy and better consistency, especially in complex tasks that require legible text or precise distribution of objects within a scene—a problem that traditional generation models have long faced.
The model supports image generation at resolutions up to 2K and provides more than 30 aspect ratios. Its use is not limited to social media posts and marketing materials, but extends to creating realistic scenes, digital art, and integrated environmental images, making it a comprehensive tool for creative production.
Specialized Training and Competitive Performance
The company explained that it trained the model on millions of human-designed templates found in the platform's content library, including presentations, infographics, and menus. This specialized training gives the model a deeper understanding of practical design contexts and how to aesthetically and functionally arrange elements.
Currently, "Canva Craft" ranks fourth on the "Design Arena" list of the best AI image generation models, a benchmark based on public evaluations. The model comes directly behind OpenAI's GPT-Image releases, reflecting the technical performance strength that Canva has achieved in its first independent endeavors.
Security and Global Availability
The model includes strict input and output control tools to ensure privacy and security, along with tools for reporting inappropriate content. Canva also plans to allow AI agents to connect to the model via the Model Context Protocol, with promises to release faster versions (Craft-fast) and ultra-fast versions in the future to meet immediate work needs.
The model is now available to all users globally via the "Dream Lab" option located in the platform's sidebar, allowing millions to immediately begin experiencing the new planning and generation capabilities.
What Actually Changes for Designers?
"Canva Craft" represents a strategic shift from merely integrating external technologies to owning proprietary AI infrastructure. The true value lies in the model's ability to "understand" the design structure before drawing it, solving the problem of distorted text or overlapping elements. For the technical user, this means reducing the number of prompts needed to reach a professional result ready for direct use in work, while ensuring that the text embedded in the images will be correctly spelled and visually clear.