About the Image Redaction Tool
This tool permanently hides sensitive parts of an image — faces, names, account numbers, license plates, addresses, signatures — directly in your browser. Drag a box over anything you want gone and pick how to cover it: a solid black or white box, a pixelated mosaic, or a Gaussian blur. The key difference from a screenshot annotation app is that the redaction is destructive: when you export, the pixels under each box are genuinely replaced in the output file, not hidden behind a layer someone could peel back. The original content cannot be recovered from the downloaded image. Everything happens locally using the HTML5 Canvas API — the image is never uploaded, which is exactly what you want when the whole point is to remove sensitive information.
How to Redact an Image
- Upload your image — Drag and drop a PNG, JPG, WebP, BMP, or GIF, or click to browse. It stays on your device.
- Pick a redaction style — Black box for maximum certainty, white box to blend into documents, pixelate, or blur. You can mix styles across regions.
- Cover the sensitive areas — Drag a box over each thing you want hidden. Add as many regions as you need; drag an existing box to reposition it.
- Tune the intensity — Adjust blur strength and mosaic block size until the content is unreadable.
- Apply and export — Choose PNG, JPG, or WebP and click Apply & Export. The pixels are flattened at full resolution and the file downloads with its metadata stripped.
Why "Destructive" Redaction Matters
A surprising number of leaked documents were "redacted" with a black rectangle that sat on top of selectable text or a separate image layer — and the hidden content was trivially recovered. The same mistake happens with images when a tool only overlays a shape. This tool avoids that class of failure by rasterizing every region into the pixels of a freshly encoded image: there is no separate layer, no hidden data, and no metadata carried over. For black-box and white-box modes the covered pixels are overwritten with a flat color; for pixelate and blur they are rebuilt from a downsampled or blurred version of the area, so the fine detail that made the content readable is gone. Blur and pixelate look softer, but for truly sensitive data — IDs, financial figures, anything you would not want reconstructed — a solid box is the safest choice.
Common Use Cases
- Hide faces and license plates before posting photos publicly
- Black out account numbers, balances, and addresses on financial screenshots
- Redact names and emails in bug reports, support tickets, and documentation
- Cover API keys, tokens, and credentials in terminal or dashboard screenshots
- Remove personal details from ID photos, invoices, and contracts
- Blur bystanders or sensitive backgrounds in product and listing photos
What actually happens to the pixels
When you export, the tool draws your original image onto an off-screen canvas at full resolution, then paints each region directly into that pixel grid before encoding a brand-new file. There is no overlay and no second layer:
- Black box / white box overwrite the covered pixels with a single flat color. The original values are gone — not hidden under an opaque shape, replaced.
- Pixelate downsamples the region to a handful of blocks and scales it back up, so a block of many pixels collapses to one averaged color repeated across the area.
- Blur rebuilds the region from a Gaussian blur sampled from the surrounding pixels.
Because the result is a freshly rasterized, single-layer image, it also carries none of the original file’s EXIF metadata. The output is exactly what you see and nothing more — no hidden text layer, no recoverable original, no GPS tag. If you want to inspect what metadata an image holds before sharing it, the EXIF Viewer shows the full record.
Choosing a mode
| Mode | Reversible? | Best for |
|---|
| Black box | No — pixels overwritten | IDs, account numbers, signatures, anything guessable |
| White box | No — pixels overwritten | The same, on white documents where black looks heavy |
| Pixelate | Sometimes, for simple content | Faces, logos, backgrounds where a hint is acceptable |
| Blur | Sometimes, at low strength | Softening backgrounds, bystanders, general clutter |
The rule of thumb: if an attacker could guess and check the hidden value, only a flat box is safe. Blur and pixelate protect appearance, not low-entropy secrets.
Redaction failures worth learning from
The reason “destructive” is the headline feature is that the most famous redaction leaks were not destructive at all. Black bars have been laid over selectable text in PDFs and lifted with a copy-paste. Mosaicked numbers have been reconstructed by re-pixelating every candidate until one matched. Swirl-obscured faces in old photos were unswirled because the transform was invertible. Every one of these shared a root cause: the original data was still present, merely disguised.
This tool is built to avoid that category of mistake — the covered pixels are replaced in the exported file — but the mode you pick still matters. When in doubt, a black box over the sensitive area, then a quick check in your browser’s Network tab to confirm nothing was uploaded, is the safe workflow. For removing an unwanted region entirely rather than covering it, crop it out instead.
