Canva is entering the AI image generator race with a model of its own. Called Craft, it was developed by the company's in-house research team and is available now to all users worldwide, both inside Canva and through a public API for developers.
What sets it apart from most generators is its architecture. Instead of turning text directly into pixels, Craft works in two stages. First, it parses the prompt and produces a plan describing spatial relationships between elements, relative scale, visual hierarchy and the text to insert. Then it generates the image from that plan. According to Canva, this separation helps produce more precise results on complex compositions, such as scenes with legible text or objects that must sit in exact positions.
Image: MSN — https://www.msn.com/en-in/money/news/canva-s-new-creative-os-brings-ai-design-smart-branding-and-video-editing-upgrades/ar-AA1PxzMX
The model outputs images up to 2K resolution in more than 30 aspect ratios. Beyond social posts and marketing material, Canva says Craft can create realistic scenes, digital art and environmental imagery. Training, the company says, used "millions of human-made templates" from the platform's content library, including social posts, presentations, infographics and menus.
Canva ranks its model fourth on DesignArena, a crowdsourced benchmark for AI image generators. Three OpenAI GPT-Image variants sit ahead of it. The claim comes from the company itself, which calls its research lab the second most advanced in the world for image generation.
The launch fits Canva's broader AI push. In April the company introduced editing through text descriptions, and competition is intense: Krea 3 focuses on speed and on changing a single area of an image, while OpenAI split its image API into two tiers, Flare and Sunburst.
Image: canva.in — https://www.canva.in/
Canva also says Craft includes input and output moderation, privacy controls and reporting tools. Faster and lighter versions, Craft-fast and Craft-ultrafast, are planned, and AI agents will be able to connect to the model through the open Model Context Protocol.
To try it, open the three-dot menu in Canva's left sidebar and select Dream Lab.
No comments yet. Be the first!