6 hours ago
2
Canva unveiled a slew of new AI products and features at its fall product launch on Thursday, but this technical, kind of boring new drop might actually hold the most promise. Canva has developed a new foundational AI model for design. And that model promises to solve one of my biggest frustrations with AI images.
Foundational models are the digital frameworks running behind the scenes to process your AI requests. Traditional diffusion models can create "flat" images and powered many initial AI image generators. Over the past few years, those models have been upgraded to become omni LLMs -- basically, they are multimodal (can handle text, image and other inputs), became more context-aware, and they are capable of handling more complex tasks. For image generators, that meant it could process reference images, as one example.
Canva's new model works a little differently. It generates images in layers while maintaining its contextual awareness. What all that means for you is that you will be able to select certain elements in an AI-generated image and use all of Canva's editing tools to adjust them. You won't have to entirely regenerate an image to fix one small error, a massive quality of life upgrade.
Canva has become a $65 billion company on the promise of making design accessible to everyone, especially those of us who aren't Photoshop experts. AI image generators, although not without their controversies and concerns, aim to achieve the same goal. I've used many different AI image programs, and one thing that consistently disappoints me is their post-generating editing tools. Many programs, if they have any at all, are bare-bones. Sometimes editing makes the errors worse. So you can see why the idea of editing AI images with Canva's many one-click tools is not only appealing but should put other AI image companies on notice.
"With Canva's foundational model, you can go from prompt to a fully editable, layered design," Robert Kawalsky, global head of product at Canva, said in an interview. "You can use the richness of Canva's editor to literally click any element, any ingredient, and manipulate it, change it in the way that you used to with a design that you manually created, from the ground up. This is just a really fundamental shift."
While Canva is the first major design company to release layer-based AI image editing, it won't be the last. Adobe's vice president of generative AI Alexandru Costin told me in an interview at Adobe Max this week that the company plans on releasing similar capabilities in the near future. Editing abilities overall remain a top concern for designers and illustrators who use creative AI tools.
Don't miss any of our unbiased tech content and lab-based reviews. Add CNET as a preferred Google source.
There's another plus here: Canva's in-house AI image generator is great for beginners, but it wasn't the most impressive of the bunch. This new model should bring some more firepower.
There are a bunch of other updates, including the ability to make forms for data capture, a new Premiere Pro-looking video timeline and HTML-compliant email templates. You'll also be able to copy the art style of an asset and paste it onto another object. But the only other major news is actually about an acquisition of Canva's from last year, Affinity.
Canva acquired the professional design program Affinity in 2024, and now Canva users will be able to use Affinity for free, no subscription required, ever. It's a surprising choice and makes Affinity undeniably the most affordable choice for professional editing software, lapping Adobe's increasingly expensive Creative Cloud subscriptions.
The last major batch of updates from Canva was AI-focused, too, leaning into helping its users with STEM-oriented tasks like coding.
For more, check out how to write the best AI image prompt and the best AI image generators.
