Can Blurred or Redacted Text Be Recovered?

Blurred text being recovered with a scan revealing the characters

Yes, blurred and pixelated text can often be recovered. Soft blur and pixelation are reversible transformations, and researchers have used them to read hidden passwords, names and account numbers. A solid black box that flattens into a flat image cannot be reversed, but a "redacted" PDF where the text still sits under the box very much can.

So the honest answer to how to uncover blacked out text is: it depends entirely on how the text was hidden. That distinction is the difference between a private screenshot and a leak.

This guide covers which redaction methods can be undone, how the recovery actually works, and how to hide sensitive text so it stays hidden for good.

Key takeaways

Pixelation is reversible: a Bishop Fox researcher built a tool that reconstructs the original text from a pixelated image
Soft blur is reversible too: academic teams trained models to read blurred text and rebuild blurred faces
"Redacted" PDFs leak constantly because a black box drawn over live text leaves the text in the layer underneath
A solid black mark that flattens into a raster image cannot be reversed, this is the only reliably safe method
The safest habit is to hide sensitive data before you capture it, so no recoverable original ever exists

The short answer: it depends how you hid it

Soft blur is reversible, a solid flattened mark is safe

Not all redaction is equal. Some methods are cosmetic and some are permanent, and they look almost identical at a glance.

Here's the quick verdict, riskiest to safest:

Soft blur (gaussian) — often reversible with machine learning. Risky.
Pixelation / mosaic — often reversible by brute force. Risky.
Swirl, warp or emoji covers — often reversible too, since a swirl is a math transform you can invert. Risky.
Black box over text in a PDF or doc — the box is cosmetic, the text stays in the layer underneath. Very risky.
Solid mark flattened into a raster image (PNG/JPG) — nothing left underneath. This is the only safe one.

The trap is that the risky methods feel the most thorough. You dragged a heavy blur across the screen, so it must be gone. It isn't. The one that actually works, a flat solid mark, looks the least fancy.

How pixelated text gets recovered

Pixelation looks like destruction. It's actually just averaging. Each block of pixels is replaced with the average color of that block, and that math runs in one direction, which means it can be run the other way.

If an attacker knows the font and layout, they can generate every candidate string, pixelate each one the same way, and compare. When a candidate pixelates to the same pattern as your image, they've found the original text. It's brute force, but for short strings like a name or an account number, it's fast.

This isn't theoretical. A researcher at the security firm Bishop Fox released an open-source tool that reconstructs text directly from pixelated images. Point it at a pixelated Social Security number and it reads it back.

Consider Daniel, a consultant who shared a case study screenshot with the client's account number pixelated. He thought he was being careful. A competitor with a weekend and that open-source tool read the number in an afternoon. The blur was there. It just wasn't doing anything.

Sharing screenshots with sensitive data? Our guide on how to blur a screenshot covers the methods that actually hold up.

How blurred faces get reconstructed

Faces are harder than text but not safe. Where text recovery is brute force, face recovery is machine learning.

Researchers at the University of Texas at Austin and Cornell Tech trained models that defeat standard blurring and pixelation, learning to identify faces and read content that was meant to be obscured, including the mosaic blur YouTube uses. Feed a model enough examples of blurred-then-clear images and it learns to invert the blur.

The practical takeaway: a light blur over a face in a photo, or over a license plate, is a speed bump, not a wall. If the stakes are real, a soft blur is not protection.

Why "redacted" PDFs still leak text

This is the most common and most embarrassing failure, and it has nothing to do with clever attackers.

When you draw a black rectangle over text in a PDF or a document, you're adding a shape on top. The original text is still there, in the text layer, underneath the box. Anyone can select the region, copy it, and paste the "hidden" text into a notepad. That's the whole trick behind searches for how to remove redaction from pdf, people undoing redactions that were never real.

Courts, law firms and companies leak this way over and over. The filing looks redacted on screen. Then a reporter copies the blacked-out paragraph and the names spill out. It happens often enough that the American Bar Association has warned lawyers about it, and courts can treat a copy-pasteable redaction as a failure of reasonable diligence, which risks sanctions.

When a legal team at a mid-size firm filed a "redacted" settlement, a journalist selected the black bars, hit copy, and pasted the counterparties' names into a story the same day. The redaction was a black rectangle. The text sat right under it.

True PDF redaction removes the underlying content, it doesn't just cover it. If your tool has a dedicated "redact" function that deletes the text, use it. If you only drew a black shape, you didn't redact anything.

How to redact so it can't be undone

Here's the reassuring part. Making text truly unrecoverable is simple once you know the rule.

The rule: the sensitive region has to become flat pixels with nothing underneath.

Flatten to a raster image. Export or screenshot the result as a PNG or JPG so there's no text layer, no vector shape, no separate blur to peel back.
Use a solid opaque mark, not a soft blur. A black or filled box with no transparency leaves nothing to reconstruct. Skip gaussian blur and pixelation for anything that matters.
Cover the whole thing, with margin. Extend the mark past the edges. A sliver of a visible character is enough to guess the rest.
Check the result at full zoom. Zoom into the saved copy. If you can make out anything, neither can you trust it.

Do that and there's no original left to recover. The security researchers who break pixelation say the same thing: the reliable way to hide text is a solid opaque bar, flattened, not a blur.

The safest habit: hide it before you capture

Every method above is damage control after the fact. The stronger move is to make sure the secret was never in the file to begin with.

DataBlur is a free browser extension for Chrome, Edge and Firefox that blurs any element on a live web page with one click, then you screenshot or share the already-hidden result. Because the sensitive region is covered before capture, the pixels you save are already gone, there's no unedited original to leak and no reversible blur to crack.

It runs entirely on your device, so nothing is uploaded, and it works live during Zoom, Google Meet and Teams. Auto-detection catches common leaks like emails, account numbers and card numbers before you reach for the screenshot key. For the full workflow, see our screen blur guide, and if you share your browser during demos, hide the URL bar too.

DataBlur blur sensitive data on a live page

Frequently asked questions

Can pixelated text be un-pixelated?

Often, yes. Pixelation averages blocks of pixels, and that can be reversed by generating candidate text, pixelating each option, and matching it to the image. A Bishop Fox researcher released a tool that reconstructs text from pixelation, so it's not safe for passwords, names or account numbers.

Can you recover a blurred face?

Sometimes. Academic teams have trained machine-learning models that defeat blur and pixelation, including reconstructing or identifying blurred faces. A light blur is a speed bump, not real protection, so use a solid cover for faces that matter.

Why do redacted PDFs still leak the text?

Because a black box drawn over text only sits on top of it. The original text stays in the layer underneath, so anyone can select, copy and paste it out. True redaction deletes the underlying content instead of covering it.

Is a black box safe to redact with?

Only if it flattens into a raster image with nothing underneath. A solid opaque mark on a PNG or JPG can't be reversed. A black shape over live text in a PDF or document is not safe, the text is still there.

How do I redact text so it can't be undone?

Use a solid opaque mark, cover the whole area with margin, and flatten the result to a raster image so there's no text layer or separate blur to recover. Then check the saved copy at full zoom to confirm nothing is readable.

Hide it once, hide it for good

So, can blurred or redacted text be recovered? If you used soft blur, pixelation, or a black box over live PDF text, assume yes. If you used a solid mark flattened into a flat image, no.

The safe method is boring on purpose: solid cover, full coverage, flattened to pixels, checked at zoom. Skip the pretty blur for anything that matters. And the strongest habit is to hide sensitive data before you capture it, so there's never a recoverable original in the first place.

Want redaction that can't be undone? Try DataBlur free and hide sensitive data on any page in one click, on your device, before you screenshot or share.