Canva can layer an image now — but only inside Canva

What Magic Layers actually does

Canva announced Magic Layers on 11 March 2026 and it does the hard part: upload a flat image and it separates the visual elements onto their own editable layers, restores text as real editable text rather than pixels, and keeps the layout intact. Anyone still telling you Canva cannot decompose an image is working from a pre-March-2026 picture of the product, and this page was one of them. If your work starts and ends inside Canva, and your file fits the constraints below, Magic Layers is very likely the right tool and you do not need us for it.

The limits worth checking before you rely on it

Canva is changing this feature quickly, so verify current limits in your own account rather than trusting an old launch article — or a competitor's page, including this one. We checked it directly on 8 September 2026: Magic Layers was available on a Canva account that still showed “Start your free trial”, and it split a 1024px image in about 12 seconds. Canva's own launch panel says it supports images, not documents or video. The output still needs review like any AI result: our test recovered the dog, rebuilt the covered background, and turned the logo into live Montserrat text, but a recognized font or reconstructed texture can be an approximation rather than the original.

The limit that is not a beta gap: the layers stay in Canva

The other four will probably loosen. This one will not, because it is architectural. Magic Layers produces layers inside a Canva design, and Canva's download list is JPG, PNG, SVG, PDF, GIF, MP4 and PPTX — there is no PSD export anywhere in the product, not in the editor, not in the Connect API, not through any Canva App. So the moment you export, the design flattens into an image and the layer structure you just generated stops existing outside Canva. That is fine if Canva is your final destination. It is a wall if the file has to go to a client in Photoshop, to a printer, to a rigging tool, or into any pipeline that expects a layered handoff — and it is why Canva's own sister product Affinity cannot open Canva files directly either.

The route that works on any plan, in any country

Separate the image outside Canva and bring the pieces in as elements. One: open your image in the Layersmith editor — free, no account to try, and the editor and file writers run in your browser so the image stays on your device for everything except the AI steps you start. Two: separate it, with one click per object using AI grab, or one button for Separate subject, or Split by color for flat graphics and logos. Three: sign in with a free account and export the Layer stack .zip, where every layer becomes a transparent PNG at the working canvas size, numbered in stacking order. Four: upload those PNGs to Canva and arrange them; each one behaves as an ordinary Canva element you can move, resize, rotate and reorder with the position controls. It works on Canva Free and spends none of your Canva AI credits.

Why transparent PNGs rather than a PSD

Canva's PSD import is plan-dependent and lossy, so a layered PSD is the wrong currency for getting things into Canva even though it is the right one for getting things out to a designer. Individual transparent PNGs are the lingua franca: they upload instantly, keep full quality with lossless alpha, and behave identically to native elements. The ZIP arrives numbered in stacking order, so rebuilding the arrangement is a matter of dropping them in bottom-first. And because the same document also writes a real layered PSD and an ORA, one separation serves both the Canva design and the version you hand to whoever needs a working file.

When to use which

Use Magic Layers when Canva accepts the image, its automatic result is good enough, and the design will live and die inside Canva. Use this when you need the layers as files somebody else can open, want to refine exactly what becomes a layer, need spot-colour separations, or do not want the image uploaded at all — the free paths here make no network call, which is a different kind of answer from a privacy policy. The overlap is real, but the end product is different: Magic Layers produces a Canva design; Layersmith produces PSD, ORA and transparent PNG layers you own.