About Blur Sensitive Info
Bug reports, support tickets, demo screenshots — they always seem to include something you didn't mean to share. An email address. A real customer name. An API key in a console output. The Blur Sensitive Info tool is a focused redaction page: drop a screenshot, drag rectangles over the bits you want to hide, pick blur, mosaic or solid black per region, export. The blur engine is the same one used by the Screenshot Annotator, but the UI is stripped to the one job, so you spend less time clicking and more time confirming the redaction is complete.
- No uploads
- Browser-only
- Works offline
- 100% free
How it works
- 1
Drop a screenshot
JPG, PNG, or WebP. The image stays in your browser.
- 2
Drag to redact
Click-and-drag any number of rectangles over the regions you want to hide. Pick the redaction style per region — Blur (Gaussian), Mosaic (pixelate), or Solid (black bar).
- 3
Verify, then export
Zoom or pan the preview to confirm nothing's leaking. Export as PNG and the redactions are baked permanently into the pixel data.
Blur, mosaic, or black bar? A security ranking
The three redaction styles aren't interchangeable — they sit on a spectrum from 'looks nice' to 'provably gone'. Pick by what's underneath, not by aesthetics:
| Method | Reversible? | Use it for |
|---|---|---|
| Gaussian blur | Partially, if weak — deconvolution and AI sharpening recover low-radius blurs of known fonts | Faces, photos, anything where exact recovery has low value |
| Mosaic / pixelate | Yes at small block sizes — tools like Depix demonstrated recovering pixelated text | Mid-stakes content, with block size 24px or larger |
| Solid bar | No — the pixels are replaced, nothing remains | Credentials, API keys, account numbers, names: anything you'd call a secret |
Rule of thumb: if leaking it would cost you money or access, use Solid. Blur is cosmetic privacy; bars are real privacy.
How redactions actually fail in the wild
Most redaction disasters share one anatomy: the sensitive data was covered, not removed. Court filings have leaked because a black rectangle was drawn over text in a PDF — the text layer stayed underneath, selectable with Ctrl+A (the 2019 Manafort filing is the textbook case). Spreadsheets shared as screenshots have leaked adjacent rows through sloppy cropping. Photos have betrayed locations through GPS EXIF that survived the edit. And weakly pixelated passwords have been recovered with open-source tooling.
This tool's export is immune to the first class by construction: redactions are rendered into the raster pixel data and the output PNG carries no layers, no history, and no original underneath. The other classes remain your responsibility — which is what the checklist below is for.
The pre-share screenshot checklist
Before a screenshot leaves your machine for a bug report, support ticket or social post, sweep the pixels nobody means to share:
- Window title bars and browser tabs — they leak account names, document titles, and which internal tools you run.
- URLs — query strings carry session tokens, user IDs and search terms; redact from the ? onward unless the URL is the point.
- Browser chrome — bookmarks bar, extension icons, autofilled email in profile badges.
- OS edges — taskbar/dock, clock, Wi-Fi network name, notification toasts that pop mid-capture.
- Content periphery — sidebars with other customers' names, email previews, chat lists adjacent to the conversation you meant to share.
- Metadata — screenshots are usually EXIF-light, but photos of screens are not; strip EXIF when the source is a camera.
- After redacting, annotate — arrows and callouts belong on top of the redacted image, via the screenshot annotator.
Settings that make recovery unrealistic
If you stay with blur or mosaic for visual reasons, size them against the text you're hiding. Gaussian blur: the default 16px radius defeats casual recovery for normal UI text; go 24px+ when the text is small and dense, because small text blurred lightly is the easiest case for deblurring tools. Mosaic: the block size must be a multiple of the character size — 24px blocks reduce a 14px-tall password to roughly one block per character pair, which is where Depix-class attacks stop working. And for anything that grants access — tokens, keys, OTP codes, card numbers — skip the aesthetics and use Solid. There is no recovery argument to have about a region that contains no information.
Redacting beyond screenshots
Screenshots are one surface. For PDFs, drawing a bar in a viewer usually covers rather than removes — use the PDF editor's redaction bars, which are baked into the exported page image the same way this tool works. For documents you'll re-share as files (not images), remember the file format itself talks: author names, revision history and creation tools live in metadata — the PDF metadata editor clears those fields. The common thread across all of it: redaction is only real when the published artefact no longer contains the data, in any layer, at any zoom.
Related guides
All guidesPrivacy
How to remove metadata from photos before you share them
Every photo carries hidden data — including, often, the exact GPS location it was taken. Here's how to see it and strip it.
6 min read
Privacy
How to encrypt a file with a password (and share it safely)
Encrypting a file is easy. Getting the password to the other person without undoing the whole point is the part people get wrong.
7 min read
Frequently asked questions about Blur Sensitive Info
Is blurred text recoverable?
If the blur radius is too low, yes — there are public tools that can partially deblur weak Gaussian blurs. The default radius (16 px) is strong enough for normal-density text; for very small text or high stakes (auth tokens, financial data), use Mosaic with a 24 px+ block size or Solid black bars instead.
What's the difference between blur, mosaic and solid?
Blur smears the pixels — looks the smoothest but the least secure. Mosaic chunks the region into solid-colour blocks — irreversible if the block size is large enough. Solid fills the region with black — most secure, but visually obvious.
Can AI un-blur my screenshot?
AI sharpeners reconstruct plausible detail, and for weakly-blurred known fonts that reconstruction can be accurate enough to read. That's exactly why high-stakes content should get Solid bars: AI can't reconstruct information that's no longer in the image at all.
Is a black bar in an image as safe as PDF redaction?
Safer than most PDF redaction, actually. In this tool the bar replaces the pixels — the exported PNG contains nothing underneath. In PDFs, a drawn rectangle often sits on a layer above intact text, which is how famous court-filing leaks happened. If you redact PDFs, flatten or rasterise the result.
Can I redact a face?
Yes — drag a rectangle around the face and pick Blur or Mosaic. There's no auto-face-detection (privacy reasons — that would imply running an ML model on every uploaded image).
Will my redaction survive copy/paste / re-screenshotting?
Yes. The exported PNG has the redaction permanently rendered into the pixel data. There's no metadata or layer information that could be unredacted by a downstream tool.
Can I redact several screenshots in a row?
One image at a time today — drop the next file when you finish the first. The tool keeps your style settings between images, which makes ticket-queue triage reasonably quick.
Privacy, offline use, browser support, and pricing questions are answered on the site-wide FAQ.