How to Blur Sensitive Information in Screenshots

Screenshots leak data constantly. You crop a bug report and the browser tab beside it still names an internal system. You share a dashboard and a customer email sits in the corner. The screenshot looks fine to you because you already know what to ignore.

This guide covers how to hide that information properly — and, just as important, where the common methods fail.

Start by cropping, not blurring

The safest redaction is removing pixels entirely. If the sensitive content sits at an edge, crop the screenshot so it is gone from the file rather than obscured within it.

Cropping cannot be undone, cannot be reversed, and cannot be misjudged. Reach for it first, and only blur what you cannot crop away.

Blur and pixelate: the honest version

Both tools here apply their effect to the whole image, not to a region you select. That is a deliberate design choice, and it shapes how you should use them:

• Screenshot blur — softens everything uniformly.
• Screenshot pixelate — replaces detail with blocks.

Whole-image treatment is the right default when sensitive data is scattered across the frame, or when you want an obviously-redacted overview image. It is the wrong tool when you need one field hidden and the rest readable. For that, crop to the area you want to show, or draw an opaque shape over the field with the screenshot annotator.

Weak blur can be undone

This is the part most tutorials omit, and it is the part that matters.

Blur and pixelation are not encryption. They are lossy transformations, and at low strengths they preserve enough structure for the original to be recovered:

• Light blur is a reversible convolution in principle. Deblurring is a well-studied problem, and text rendered in a known font at a known size is close to the ideal case for it.
• Coarse pixelation of text is worse than it looks. Each block averages a small, predictable region. Because there are only so many characters a block pattern can correspond to, an attacker can render every candidate string, pixelate each one the same way, and match. This has been used to recover redacted text in real published documents.
• Animated or partial obscuring has repeatedly failed in public, where a single unblurred frame or a slightly misplaced overlay exposed everything.

Practical thresholds with these tools:

• Blur strength 40 or above smears characters into their neighbours far enough that recovery is not realistic.
• Blur under 15 may leave text legible with effort, or recoverable programmatically. Do not use it for secrets.
• Pixelation should be coarse enough that a block spans several characters, not one.

If the content must stay secret, cover it

For passwords, account numbers, API keys, medical details, or anything with real consequences, do not rely on a blur at all. Use the annotator to draw a solid, opaque rectangle over the region.

An opaque shape carries no information about what was beneath it. There is nothing to recover, because the original pixels are replaced rather than transformed. It is the only method on this list with that property.

Then remove the metadata

A redacted screenshot can still carry metadata, and screenshots of photos can inherit the original EXIF. Once the pixels are handled, pass the file through the EXIF remover so timestamps and device details do not travel with it. See the metadata guide for what is actually stored.

A workflow that holds up

1. Crop away everything the reader does not need. 2. Cover any high-stakes field with an opaque shape in the annotator. 3. Blur or pixelate at high strength for anything remaining that is lower-stakes but still private. 4. Strip metadata from the finished file. 5. Reopen the exported file and read it as a stranger would. Check the corners, the tab bar, the taskbar, notifications, and any reflection in a window.

Step 5 catches more mistakes than the rest combined.

Everything happens on your device

These tools run in your browser. A screenshot you are redacting is, by definition, one you do not want strangers to see — uploading it to a server to have it blurred would defeat the point. The pixels are processed locally and never transmitted.

Troubleshooting

• Text is still readable after blurring. Raise the strength. Below roughly 15 it is not a redaction.
• The whole screenshot is unusable. Expected: blur is uniform. Crop to the region you want to publish first, then blur, or cover the single field instead.
• A large screenshot processes slowly. The blur runs in a background worker so the page stays responsive. Downscale first with the screenshot resizer if you do not need full resolution.
• The exported file is larger than the original. Re-encoding can do that, especially from PNG. Run it through the screenshot compressor if size matters.