Blur part of an image

1. Upload your image. Drag it onto the box above, or click to browse. It loads straight into your browser tab — nothing leaves your device.
2. Pick an effect. Choose Blur, Pixelate, or Black box from the segmented control. This is the effect that new regions will use.
3. Drag a rectangle over anything sensitive. Click (or tap) and drag on the image to draw a box around a face, plate, signature, or line of text. Release to apply the effect.
4. Repeat for every region. Draw as many rectangles as you need. Made a mistake? Use Undo last, delete a single region from the list, or Clear all to start over.
5. Download. Choose PNG or JPG and click Download redacted image. The export is re-rendered at your image's original resolution, so it matches the source file size exactly.

Every redaction region remembers exactly where it sits on the full-resolution source image, not just the smaller preview you see on screen — so what you drag on screen is what actually gets covered in the downloaded file, pixel for pixel.

All three redaction styles hide the region underneath, but they are not equally safe for every situation. Here is how to choose:

For faces or scenery, blur or pixelate both work well. For anything you truly need gone — a password on a sticky note, an account number, a name — use black box, since it is the only effect that discards the original pixels entirely rather than obscuring them.

A blur or a low-resolution pixelation is a lossy transformation, but it is not random noise — it is a predictable mathematical operation applied to real pixel data. For photos of faces or complex scenes, that rarely matters: there are too many plausible underlying images for an attacker to recover the original with any confidence. Short, structured text is a different story.

Text has a tiny alphabet (a handful of letters, digits, and symbols) and a predictable layout (fixed font, fixed spacing, a short known-ish string like a password or account number). Researchers have repeatedly demonstrated that this makes pixelated or blurred text de-anonymizable: by rendering candidate strings in the same font and comparing how they degrade under the same blur or pixelation kernel, an attacker can narrow down — and sometimes fully recover — short strings like passwords, IP addresses, or usernames. The lower the pixelation block count or blur radius relative to the text size, the more the redaction leaks about the original shapes.

The practical takeaway: use blur or pixelate for faces, backgrounds, and general scenes — they are visually convincing and adequate for casual privacy. But whenever you are redacting text in a screenshot — a password, an email address, an account number, a name in a chat log — use the black box effect instead. It permanently discards the pixel data rather than degrading it, so there is nothing left for any reconstruction technique to work from.

Almost every screenshot shared outside a small trusted circle contains something that shouldn't be public. The most common categories people run through a redaction tool before posting, filing, or emailing a screenshot are:

• Faces in photos shared publicly, group chats, or support tickets where consent wasn't given to show a bystander.
• License plates in dashcam footage, real estate photos, or street-view screenshots.
• Names and usernames in chat logs, forum screenshots, or bug reports shared publicly.
• Account numbers, order IDs, and reference numbers in screenshots of invoices, bank statements, or support tickets.
• Email addresses and phone numbers visible in inboxes, contact lists, or CRM screenshots.
• API keys, tokens, and passwords accidentally visible in terminal or settings screenshots shared for troubleshooting.
• Home addresses shown on package labels, maps, or delivery app screenshots.
• Signatures on scanned contracts, delivery confirmations, or shared PDFs converted to images.
• Children's faces in family photos posted publicly, per many parents' and schools' privacy guidelines.

Both the blur and pixelate effects expose a strength slider, and the right setting depends on how large the sensitive region is relative to the rest of the image and how much risk you are willing to accept.

• Blur strength (px): a low value (4–10px) softens details but can still leave recognizable outlines, especially on high-contrast edges like text or plate characters. For a genuinely unrecognizable result, push it well past the point where you personally can still make out the shape underneath — as a rule of thumb, the blur radius should be at least a third of the region's shorter side.
• Pixelate block size (px): small blocks (4–8px) can still preserve enough structure for a trained eye — or an algorithm — to guess at shapes. Larger blocks (20px and up relative to the region size) destroy far more information and are safer for anything you actually need hidden, at the cost of a more visibly “blocky” look.
• When in doubt, go stronger — or use black box. There is no visual downside to over-redacting a region; the only cost is aesthetics. If the region contains text, numbers, or anything structured, skip the strength question entirely and use the black box effect, which removes the ambiguity altogether.

The right level of redaction depends heavily on where the image is going. A screenshot posted casually to a private group chat has different stakes than one attached to a public bug report, a job application portfolio, or a document submitted as part of a legal or compliance process.

For social media and casual sharing, blur or pixelate is usually sufficient — the goal is to avoid drawing attention to a face or detail, not to defend against a determined attacker. For public bug reports, open-source issues, or anything indexed by search engines, treat every redaction as permanent and worst-case: assume someone will eventually try to read what is underneath, and use black box for any text. For legal, compliance, or journalistic use where a redaction failure could have real consequences (protecting a source, meeting a data-privacy obligation, preparing a document for discovery), black box every sensitive region and, where possible, double-check the exported file at full resolution before it leaves your hands — a redaction that looks fine in a browser tab but was drawn slightly off-target is a common and entirely avoidable mistake.

Yes. Everything happens locally in your browser using the HTML canvas API — your image is decoded, drawn, redacted, and re-encoded without ever being uploaded to a server. There is no server-side processing step at all, so the tool works offline once the page has loaded, and nothing about your image is logged or stored anywhere. That makes it safe to use for genuinely sensitive material: leaked screenshots, ID scans, internal documents, or anything else you would not want to hand to a third-party service.

Because redaction is entirely client-side, the strength of your privacy is only as good as the effect you choose. Remember: for scenery and faces, blur or pixelate is fine; for text you must permanently hide, use black box.

Faces, license plates, and names in photos are increasingly treated as personal data under privacy laws like the EU's GDPR and various US state privacy statutes. If you handle screenshots or photos that contain identifiable people as part of a business process — customer support tickets, HR records, incident reports, marketing photography releases — redacting anything not strictly needed before the image is stored, forwarded, or published is good practice even where it isn't strictly mandated. It reduces your exposure if the image is later leaked, shared beyond its intended audience, or retained longer than planned.

This tool is a good fit for that kind of one-off, ad hoc redaction: a support agent blurring a customer's face out of a screenshot before attaching it to an internal ticket, or a property manager covering a house number before posting a listing photo. For high-volume, automated redaction across thousands of images — for example, blurring every face in a dataset before training a machine-learning model — you'll want a dedicated pipeline with automatic face and plate detection rather than manual dragging, but the same blur-versus-black-box logic still applies: detected faces are usually fine with blur or pixelate, while any detected text region should be blacked out rather than merely obscured.

• Support and QA teams blur customer PII out of bug-report screenshots before filing tickets or sharing them in Slack.
• Journalists and researchers redact sources' faces or identifying details in leaked documents and photos.
• Real estate and rental listers blur license plates and house numbers in exterior photos before publishing listings.
• Developers black out API keys and tokens in terminal screenshots shared in documentation or Stack Overflow questions.
• HR and finance teams redact employee names or account numbers from screenshots used in internal training materials.
• Parents and social media users blur children's faces or home addresses before posting photos publicly.

• Draw your rectangle slightly larger than the sensitive area — a tight box that clips the edge of text or a face defeats the purpose.
• Use black box, not blur, for any short text string (passwords, account numbers, single names) — see above for why.
• Increase pixelate's block size for small regions; a block size close to the region's own size gives the strongest redaction.
• Always export and re-check the downloaded file at full size before sharing — a region that looks covered in a shrunk preview should look covered at 100% too.
• Zoom out mentally, not just visually: check whether a face, a plate, or a name is visible in a reflection, a mirror, or a second copy elsewhere in the same image before you consider it done.
• If the image has metadata (GPS location, device info) you also want to remove, run it through the EXIF remover after redacting.

The tool loads your image into an off-screen HTMLImageElement and keeps track of two coordinate spaces: the full-resolution source image, and a smaller, display-scaled canvas sized to fit comfortably in the page. Every rectangle you drag is captured with pointer events — the same API handles mouse, touch, and stylus input, so the tool works the same way on a laptop trackpad or a phone screen — and is immediately converted from on-screen pixels back into source-image pixels before it's stored.

That means the live preview you see is always an accurate, scaled-down rendering of what the final export will look like. When you click download, the tool builds a brand-new canvas at your original image's exact width and height, redraws the source image onto it, and re-applies every region's effect at full scale — using ctx.filter for blur, a downscale-then-upscale trick with image smoothing disabled for pixelation, and a simple fill for the black box. The result is exported as PNG or JPG and handed to your browser as a normal download, with no intermediate upload step anywhere in the pipeline.