Image Blur & Mosaic
Redact faces, plates or documents: drag one or more boxes, choose pixelate or Gaussian blur, set the strength and download as PNG or JPEG.
Blur and export happen locally in your browser. Your image never leaves your device.
Blur boxes on your image
Click to choose an image or drag a PNG, JPG or WebP here
How It Works
Every box you drag is recorded in the image's native pixel grid, and the preview canvas underneath the handles redraws the full-resolution file with the redactions applied — what you watch adjust on screen is the same math that produces the download.
- Two ways to hide
- Pixelate averages each block down to one color and paints it back with smoothing disabled, so detail is thrown away in big square bites — the classic mosaic and the stronger redaction of the two. Gaussian blur runs ctx.filter = blur(Npx) clipped to the box, over a source region padded by twice the radius so edge pixels mix in rather than bleeding transparent.
- Display space, native coordinates
- The stage scales the image to fit (up to 640 px across), but pointer positions are converted with the naturalWidth ÷ renderedWidth ratio — which also absorbs any browser zoom or DPI scaling — before boxes move or resize. Strength values are native pixels too: a 24 px block over a 4000 px photo behaves identically in the download and the preview.
- Deleting detail, honestly
- Blur is a mix, pixelation is a discard: light blur keeps enough structure for clever software to guess at, while large-block mosaics and big-radius blurs genuinely leave nothing to recover. The sliders go high on purpose (blocks to 64 px, blur to 40 px) so redactions can cross that safety line — and for legal-grade hiding, cropping the region out is the only absolute.
Frequently Asked Questions
Can a blurred or pixelated area be recovered?
Weakly hidden data can be partially reconstructed — light Gaussian blur is mathematically invertible to a surprising degree, and recent AI enhancers sharpen guesses about what was underneath. Strong settings are a different story: a mosaic with block sizes of 16 px or more relative to the original file, or a blur radius of 20 px or more, destroys the fine detail permanently; there is nothing left to invert. For genuinely sensitive content use the pixelate effect at high block size — or crop the region out entirely, which is the only guaranteed redaction.
I am redacting for GDPR-style sharing (faces, plates). What should I check?
Always open the downloaded output and inspect it before sending anything: zoom in on each redacted area and confirm the subject is not recognisable, including hair, clothing and background context that can identify people without the face. Keep your edits on the exported copy — the tool hands you a new file, but your original sits unchanged on your disk, and uploading the wrong file is the classic redaction failure.
Can I hide several faces or fields at once?
Yes — drag as many boxes as you need, move them freely and resize each with its four corner handles. The currently selected box is outlined in the accent color; press the Delete key (or the Remove selected box button) to drop it. Effects and strength apply to every box uniformly, which keeps the result consistent.
Does the export keep the original resolution?
It does. Boxes are stored in the image's native pixel coordinates, not the on-screen scaled preview, so the exported PNG (lossless) or JPEG (quality 0.92) is the same dimensions as the input with only the box regions rewritten. Block size and blur radius are measured in native pixels, so what you redact is exactly what you get, independent of how large the preview renders on your screen.
Is anything uploaded to a server?
No. The image is decoded into a canvas inside this tab, the pixelation and blur filters run there, and the output leaves as a local blob download. No upload endpoint exists for this page — disconnect the network and redaction still works.