The Australian digital design company Canva announced on Thursday the launch of its first AI image generation model, named Canva Craft, aiming to develop visual content creation tools and provide more accurate and editable images.
The tech platform TechCrunch reported that the new model was developed by the company's research team and relies on a two-stage mechanism; the first stage analyzes user instructions to determine the location, size, and relationships of elements, along with organizing text and visual sequence, while the second stage creates images with a resolution of up to 2K, supporting up to 30 different aspect ratios to suit various digital uses and platforms.
The company clarified that the model allows for creating images with precise details and clear text, as well as the ability to edit specific parts of an image without affecting other elements. The model was trained using designs created by people from the Canva library, including posters, social media posts, presentations, charts, and designs containing text, rather than limiting itself to photographs.
Images generated by the model can be converted into multi-layered designs using the Canva Magic tool, allowing users to continue editing and modifying elements independently. The company made the model available through the Dream Lab section on its platform and intends to provide it via an API, while working on making its capabilities available through external platforms, including ChatGPT and Gemini, using the Canva MCP platform.
Canva indicated that developing its own model could help reduce image generation costs and improve quality and production speed, noting that it is working on developing two lighter and faster versions, Craftfast and Craft-ultrafast, in preparation for their later launch. This development comes amid increasing competition among technology companies to develop generative AI models and improve their capabilities in creating and editing images, while reducing operational costs and expanding the range of uses available to designers and content creators.
Source: Syrian Arab News Agency (SANA)
Comments (0)