Canva Chases AI Domination With a New AI Image Model for Designers

Canva released Canva Craft, its first proprietary AI image model, globally to all Canva users, with the model built specifically for designers. Canva head of AI research Stefano Corazza told CNET that the company's acquisition of Leonardo.Ai and internal AI research ramp-up enabled the in-house model, which uses a two-stage architecture to plan a design's layout before generating it and is trained on Canva's library, licensed and public databases, and opted-in user data. Canva Craft ranks fourth overall on the DesignArena image-model leaderboard, trailing OpenAI's ChatGPT Images and ahead of Google's Nano Banana and Microsoft's MAI-Image, though it trails in the marketing subcategory.

Canva is going deeper down the AI rabbit hole with the release of its first proprietary AI image model, called Canva Craft. It’s now available globally to all Canva users and is made specifically for designers. This isn’t the first time Canva has added AI, to be sure. The beginner-friendly design program has been adding AI tools https://www.cnet.com/tech/services-and-software/canvas-new-ai-tools-want-you-to-embrace-your-stem-side/ and third-party AI models https://www.cnet.com/tech/services-and-software/canva-create-2026-news-ai-connectors-scheduling/ for years. Its first take on creative AI, Magic Media https://www.cnet.com/tech/services-and-software/canva-magic-media-review-ai-generated-images-for-amateur-creators/ , debuted in 2023. But Canva’s head of AI research, Stefano Corazza, told CNET that the company’s acquisition of Leonardo.Ai https://www.cnet.com/tech/services-and-software/leonardo-ai-review-a-versatile-image-generator-for-ai-creative-enthusiasts/ and internal ramp-up of AI research helped bolster its efforts to create its own AI image model, one made specifically for designers. Keeping all the AI tech in-house also helps the company keep cloud costs down and inference time quicker, Corazza said. Canva Craft is meant for designers, and a big part of that is how the model actually works. The model has a two-stage architecture, the company said in a blog post. That means the model, combined with Canva’s prompt-writing tools, can “reason” before generating a design. The model is trained on Canva’s library, licensed and public databases, and data from Canva users who opted in to training, Corazza said. The result is, ideally, more complex and comprehensive designs. More from CNET “ The AI can actually create a whole plan for how the image is going to look like, all the layout. If you think about designers, that’s exactly what they do right when they create a design,” said Corazza. Designers carefully place elements to create a story. “Our model does basically the exact same thing. So we are mimicking exactly the design process.” Early benchmarking results are positive. On DesignArena https://www.designarena.ai/leaderboard?tab=image§ion=image , Canva Craft is in fourth place overall for image models, trailing only OpenAI’s ChatGPT Images https://www.cnet.com/tech/services-and-software/chatgpt-images-2-release-openai-news/ and beating bigwigs like Google’s Nano Banana https://www.cnet.com/tech/services-and-software/google-nano-banana-pro-ai-image-generator-review/ and Microsoft’s MAI-Image https://www.cnet.com/tech/services-and-software/microsoft-new-ai-image-model-build-nano-banana-news/ models. Canva also ranks highly for generating portraits and product-focused imagery, though it trails in the marketing subcategory, which is where a lot of people use Canva. The ability to edit AI outputs is a priority for Canva, and it’s an important one for anyone who uses AI tools. Often, no matter how perfect your AI prompt is, an image or video generator will spit out a result that is almost, but not quite, right. Canva has been tackling this problem with tools like its Magic Layers https://www.cnet.com/tech/services-and-software/canva-ai-image-magic-layers-news/ , which let you break down an AI design into its most basic elements and adjust them, a technologically complex process that solves a big problem for AI enthusiasts. Canva Craft builds on that foundation and lets you edit results. Creative controversy Creative AI tools are controversial, to put it mildly. Photographers, illustrators and artists have been calling out concerns for years about how AI is allegedly trained on their work in violation of copyright protections https://www.cnet.com/tech/services-and-software/textbook-copyright-infringement-disney-and-universal-sue-ai-image-generator-midjourney/ , used to create abusive material https://www.cnet.com/tech/services-and-software/inside-take-it-down-act-ai-deepfakes-image-abuse-analysis/ and being offered up to executives as a replacement https://www.cnet.com/tech/services-and-software/amazon-cuts-14000-jobs-amid-ai-push/ for real artists. Creative companies from Adobe to Canva have maintained that AI is just a tool they want to offer their creators. In an apt analogy of the current state of generative media, Eric Wittman, CEO of the popular photo editing and sharing platform VSCO, replied to Canva CEO Melanie Perkins’ LinkedIn post https://www.linkedin.com/feed/update/urn:li:activity:7519145915362893825/ about Canva Craft and said that real photographers, not AI, will always rank No. 1 for portraits. While executives argue over the best way to incorporate AI, photographers have to deal with AI headaches on every design platform, including VSCO https://petapixel.com/2026/05/26/vsco-terms-of-use-explained-why-it-says-it-isnt-stealing-your-photos/ . As AI models improve, it’s also forcing real creators to consistently verify their work as human-made, not AI-generated. But digital verification methods are sporadic https://www.cnet.com/tech/mobile/apple-reference-images-explained-iphone-18-pro-ai-slop/ , creating an even more complex web for creators to navigate.