Canva’s new AI tool will break your design process (in a good way)

The newly launched Magic Layers feature dissects flat images into native objects, putting an end to the frustrating era of pixel-locked assets.

Canva’s new AI tool, launching today, is going to save time, money, and headaches for so many people. Called Magic Layers, it turns any flat bitmap image into a fully editable Canva project, extracting text, objects, and components into individual layers.

This tool marks a fundamental shift in how we handle digital assets. Until now, a rendered image was basically a locked vault of pixels. If you wanted to change a typo or swap a background, you had four options: 1) Hunt down the original project file, 2) painstakingly change it in Photoshop, 3) accept a generative AI patch job, or 4) close the laptop and escape to live a real life somewhere by a nice beach.

Magic Layers shatters the vault. By reverse-engineering a flat picture into its constituent parts, Canva cofounder and Chief Product Officer Cameron Adams tells me, Magic Layers empowers users to resurrect and tweak any image they have on their hard drive.

Canva uses many models from OpenAI, Anthropic, and other developers, but the secret sauce behind this new layering capability is its proprietary AI design model, which the company unveiled last October. Think of it not just as a random design and image generator, but as a model that understands the elements of design.

It looks at a picture and sees its skeletal structure—distinguishing the foreground subjects from the background scenery, and recognizing typography as actual text rather than just colored shapes. When you feed it an image, whether it was spat out by an AI prompt or dragged from an old folder, it dissects those elements perfectly. The new Canva multilayer tool is the implementation of those abilities.

“Most AI outputs are fixed, really flat things, and they’re not easy to edit. You either have to, like, live with an 80% solution or you have to spend time reprompting, trying to get that little bit of the image that you wanted to get fixed,” Adams says. But now, he adds, “the model identifies everything in the frame and converts it into native Canva objects.”

So text isn’t just a cutout anymore. It becomes a live, editable text box. You can correct spelling errors, swap the font, adjust the size, or even translate the copy for international markets. The same goes for visual objects. Once separated, elements like a product bottle or a butterfly become completely independent actors on the canvas. You can move them, resize them, change their color, or banish them from the composition entirely without leaving a gaping hole behind, Adams explains.

Claire's went from tween mall icon to bankrupt — twice?

© Fast Company