Answer in brief
Identify whether the problem is a raster image, a PDF or a locked Canva element. Grab Text and Magic Layers rebuild structure from an image; neither restores the original layers that existed before flattening.
What Canva means by “flattened” text — and why it has more than one cause
This guide separates three different technical causes why text in Canva stops behaving like text, showing exactly what file structure actually survives in each case and what is lost for good — instead of one generic “click here and it comes back” instruction.
Canva’s official help center introduces the term “flattened text” — in the company’s own wording, text “that’s been converted into an image and can’t be edited” (editorial translation). The short answer by file type: from a plain image, editable text can only be rebuilt from scratch — either with Grab Text, which swaps text on the image, or with Magic Layers, which splits the whole image into layers — never restored to the original; with a PDF, everything depends on whether the file was scanned; and if a design is already open in Canva and the text simply will not move, the first thing to check is an element lock (Lock), not flattening. Behind this general picture are at least three distinct technical situations that are fixed differently: what follows goes through each one separately — what actually happened to the file’s structure, which tools Canva offers for that specific case, and where the conclusions rely on documentation alone rather than on testing a real file.
Case one: a JPG or PNG that contains text
If the source is an ordinary raster image — a photo, a screenshot, an exported banner — the text on it technically never existed as a separate editable object: the letters are just pixels, part of one flat image. Getting editable text back here is not restoration; it is creating a new text layer on top of, or next to, the old pixels. Canva offers two tools for this, with different mechanics.
Grab Text — swapping text on an image for editable blocks
The Grab Text tool is opened from Edit → Tools → See all → Grab Text and lets you select a specific word or phrase on an image, then turn just that piece into a separate editable text block. The old letters are removed from the image as this happens: the official help page warns that “Grab Text may sometimes leave small traces of text” on the image (editorial translation), which the Magic Eraser tool can clean up. The result can be undone two ways — a regular Undo, or pressing Grab Text again, which returns the image “to its original state” (editorial translation). Per the feature page’s wording, the process itself “doesn’t overwrite the image or reduce its resolution” (editorial translation) — that is about image quality, not about old letters staying visible under the new text. The tool has documented limits: it does not support non-Latin-script languages, does not work with rotated or curved text, and will not accept files longer than one page. On patterns or lines it can “invent” text that was never there, and such fragments have to be removed manually. The help page does not name the exact plan tier Grab Text requires, saying only that it is “available on eligible plans” (editorial translation; Canva’s own phrase); availability is worth checking against your own account rather than trusting outside reviews.
Magic Layers — an attempt to rebuild the whole image at once
When an image has many text and graphic elements and a spot fix will not do, Canva offers Magic Layers — a feature officially announced on March 11, 2026. It tries to split an uploaded image (JPEG and PNG are the recommended formats) into separate layers — text, images, background. Canva’s official Design School guide describes the result this way: “the material is essentially rebuilt from scratch as a Canva design” (editorial translation) — that is, it is a layout rebuilt based on the image, not a recovery of the original layers that existed before the design was flattened. Cameron Adams, Canva’s co-founder and Chief Product Officer, described the problem the feature solves in a quote carried by the trade outlet PetaPixel: “you’d end up with a finished image that couldn’t be edited, iterated on, or made your own” (editorial translation). As of the September 15, 2026 check date, the feature remained a public beta — the button on the official page is literally labeled “Try Magic Layers Beta”; at its March 2026 announcement, access was limited to the U.S., UK, Canada and Australia, and Canva itself wrote that global access “would follow” (editorial translation) — no more recent confirmation of an expanded country list was found by September.
Case two: a PDF whose text will not edit
PDFs are more complicated than a plain image, because the same file format behaves differently depending on how that PDF was created.
A digital PDF with a live text layer
If a PDF was exported directly from software — Canva, Word, InDesign, Figma and similar — it may retain a real text layer. Canva’s official PDF-import help page puts this carefully: “depending on the file, text and images may be separated into editable elements” (editorial translation) — separation is possible but not guaranteed for every such file; the outcome depends on how the source program wrote the text into the PDF.
A scanned PDF — import always produces an image
For scanned documents the rule is different, and here Canva makes no exception for the import step itself: per the help page, “scanned PDFs are imported as a single flattened image, and their text and design elements can’t be edited” (editorial translation). Re-uploading the same file changes nothing: a scan is, by definition, a photograph of a paper page, and it never had a text layer. But further options depend on language and page count rather than being closed off entirely. Canva’s official Grab Text help page directly advises for this case: “scan your document and upload it to Canva” (editorial translation) — meaning that after import, a one-page Latin-script scan can be tried as an ordinary image through Grab Text, with the same tool limits that apply to a photo. For a Cyrillic-script document there is no documented AI route: Grab Text supports Latin script only, and Canva publishes nothing about Magic Layers’ language limits — before relying on it for a Cyrillic scan, that should be checked separately rather than assumed to work or not work. Canva’s official PDF-import help page separately offers its own path for non-editable text — “replace or recreate the content using Canva’s text elements” (editorial translation).
When a flat PDF was made on purpose, not by accident
A separate cause of a non-editable PDF is not an import error but a deliberate choice at export time. When downloading a finished design via Share → Download → PDF Standard or PDF Print, Canva has a “Flatten PDF” checkbox: per the feature’s wording, flattening merges all of the document’s layers into one, after which the content looks like a single non-editable image (editorial translation). This is done on purpose — for an already-finished, finalized file — so a recipient cannot accidentally or deliberately change text, a logo or the placement of elements when printing or forwarding it; Canva separately notes that for documents still under review or awaiting approval, compression (Compress PDF) rather than flattening is the better fit, since it keeps the file editable. The official feature page does not directly describe what happens to the original design in the author’s account after a flat file is downloaded — no separate confirmation was found; interface logic (Share → Download creates a separate file for download without replacing the design in the editor) points that way, but it has not been verified in practice and remains an open question. If someone else sent a flattened PDF, it is more sensible to first ask the author for the original file or a link to the Canva design than to try to re-recognize text on the exported image. It is also worth remembering the technical limits of the import itself: Canva caps an uploaded PDF at 500 pages, 300 megabytes and 1,400 elements and images per file — exceeding the limit can lead to an incomplete import regardless of the document’s origin.
Case three: the design is open in Canva, and the text still will not move
In practice, the third case is easy to confuse with the first two, though the cause here is fundamentally different: the file is not flattened and the text remains text — a specific element, or the whole page, is simply protected from changes.
Lock versus Lock position — the difference
Canva’s official help page describes two lock modes: full Lock prevents moving, resizing, restyling and editing the content of an element, while “Lock position only” fixes not just the position but also the text’s formatting — font, size, color, alignment — leaving only the text itself editable. Images and frames work on similar logic: effects and filters are fixed, while the picture inside the frame can still be replaced, cropped or moved. Anyone with edit permission on the design can apply and remove an ordinary lock — on a personal account this is usually one click on the lock icon on the element itself. One separate thing to note: if the whole page is locked, an individual element on it cannot be unlocked until the page lock is removed.
Why the lock sometimes cannot be removed on an education account
The exception is education accounts on the CEDU, CEDD, CEDS and CEMS bundles: if a teacher locked an element, a student cannot remove that lock on their own, even with permission to edit the design itself. For every other user and brand, locking works the same way: anyone with edit permission on the design can remove it. A narrower exception is not the element lock itself but Brand Templates (a team’s branded templates): applying and changing locks there is not open to any team member, only to admins and brand designers.
Canva’s AI quota: how Magic Layers spends the limit, and whether it is on Free
Magic Layers is officially classified as a Premium AI tool and draws on the account’s shared monthly AI-usage quota — it is not an unlimited feature. Canva’s official AI-access help page lists the plans eligible for Premium AI tools: Pro, Teams, Business, Enterprise, Nonprofit. The phrase “paid plans only” is imprecise here: Canva for Nonprofits is a free program for nonprofits, not a paid plan, and Canva Education is not mentioned in that list at all — even though further down the same page names a separate limit for it, up to 200 Premium AI tool uses. Later on the same page, it states plainly that Free and Pro Lite get up to 20 uses per month of Standard AI tools or Premium AI tools — meaning Free also has access to the category Magic Layers belongs to, just with a lower limit. Canva’s official pricing page confirms this separately: the Free card lists the same limit, and “Try Magic Layers” appears as a listed new feature. Bottom line: despite the five-plan list on the help page, two other official sources show a Free account does have access to Magic Layers — up to 20 uses a month, shared with any other Premium and Standard AI tools. Magic Layers’ own beta region and status are a separate limit, unrelated to plan tier.
Format limits of Grab Text and Magic Layers worth knowing in advance
Neither tool returns the original — both return a rebuild — and both have format limits worth checking before trying, not after. Grab Text officially does not work in China, in Canva Docs, in Canva Video, or in the standalone Photo Editor outside the main editor; supported document types include Social Media, Document, Presentation, Print products, Websites and Whiteboards. Magic Layers is built for single-page JPEG or PNG images; at its March 2026 announcement, Canva specified the beta works with single-page PNG or JPG files and is built on the company’s proprietary Canva Design Model — multi-page files and complex PDF layouts are not part of its intended use. The documentation found publishes no percentage accuracy figure for either Grab Text or Magic Layers — the result has to be judged on the specific file, not on a general service promise.
A route: where to start on a specific file
It makes more sense to structure the diagnosis by file type, not by favorite tool. If it is a plain image, the question is which script the text uses: for Latin script, Grab Text officially works both for individual words and — per Canva’s own recommendation — for all the text on an image at once (editorial translation: “for the best results, select all the text in the image”), so the “Grab Text is for spot fixes, Magic Layers is for the whole thing” contrast is not quite accurate: it is more precise to talk about different mechanics — Grab Text swaps specific words, Magic Layers rebuilds the whole image into a layered layout. For Cyrillic text, Grab Text is documented as unsupported; whether Magic Layers handles Cyrillic, Canva does not say anywhere — there is neither confirmation nor an exception in the documentation found, and that is the first thing worth checking on a real file rather than assuming in advance. If it is a PDF, first find out whether the document was physically scanned: for a scan, there is exactly one documented AI route — Grab Text after import, and Latin script only; for a digital PDF, there is a chance part of the text imports as editable without extra tools, but that is not guaranteed. If the design is already open in Canva and edits simply will not go through, before hunting for a complicated cause it is worth checking the lock icon on the element itself and on the page as a whole: that is not about flattening but about a lock, and it is fixed differently — by removing the lock, not by searching for text on an image.
Practical checklist
- Start with the source type — a photo or image, a PDF, or an already-open Canva design: the next steps differ for each.
- For a PDF, work out its origin: a scan of a paper document, or an export from digital software.
- For a design already open in Canva, check the lock icon on the element first, before looking for a more complex cause.
- For an image with non-Latin-script text, remember that Grab Text will not recognize it.
- Before using Magic Layers, check your account’s AI-usage counter — the quota is shared across all Premium AI tools.
- Keep the original editable file separate from flattened exports: a file downloaded with the “Flatten PDF” box checked is a ready copy for printing or sending, not a replacement for the editable design in your Canva account.
Questions and answers
How do I undo Grab Text and restore the image to its original state?
Use Undo (or its keyboard shortcut), or press Grab Text again — per Canva, this returns the image “to its original state” (editorial translation).
Where can I check how many Premium AI uses are left on my account?
In Settings → Billing, under AI usage: it shows the remaining monthly limit.
I unlocked the element, but it still will not edit — why?
Check whether the whole page is locked: while a page is locked, an individual element on it cannot be unlocked until the page lock is removed first.
After importing a PDF into Canva, part of the text changed font — is that a bug?
No: if the source font is not in Canva’s library, the help page describes this as expected behavior — the font is automatically swapped for a similar one.