Blur people in a photo before you share it

Why you’d do this

Photos taken in public places and at gatherings often catch people who never agreed to be in them: passers-by, other parents’ children, a colleague in the background of a desk photo. Before you post such a picture, you may want to cover them. The same goes for private details: a name on a letter, a screen, a car number.

Dzzz’s Auto-Blur does this on your device. The photo is not uploaded, and the result is a flattened image saved straight from your browser. It can find people for you, but you stay in control: you can draw, move, resize or delete every box yourself.

How, with the named tool

1. Add the photo. Open Auto-Blur, then drop a photo on the dropzone or tap to choose one. The photo opens in the box editor.
2. Mark what to hide. Drag on the photo to draw a box over anything you want covered. Drag a box to move it, drag its corner to resize it, or tap × to remove it. The counter shows how many boxes are marked, and Clear all starts again.
3. Or let it find people. The first time, Download engine (~43 MB · one time) appears in place of the detect button. Once the engine is cached, tap Detect people (beta). The tool adds a box around each whole person it finds, and you can then move, resize or delete those boxes like any other.
4. Choose the style. Under Redaction style, pick Blur or Pixelate. Blur softens the region, and pixelate replaces it with coarse blocks. Pixelate is usually the harder one to see through, so prefer it for anything that must stay unreadable, such as text or a number plate.
5. Apply and save. Tap Apply redaction, check the result, then save the flattened PNG. The marked regions are permanently blurred or pixelated in that file and merged into the image. If something is wrong, Edit boxes again takes you back, and Redact another photo starts over.

Scanned documents

If the sensitive part is on paper, Document Scanner can flatten the page first. After you have captured or uploaded a page and corrected its corners, tap Redact something first. It opens Auto-Blur with that page, passed in memory rather than through the address bar. Note that it hands over one page, so for a multi-page document you repeat the process for each page you need to cover.

What to watch for

It detects people, not faces. The detector is a general-purpose object model, so its boxes cover a whole person, head to toe. If you need only a face covered, draw your own box over it. The detection step is labelled beta for a reason.

It only draws boxes. The tool never identifies or recognises anyone, never compares one person with another, and does not store any cropped image of a person anywhere. It places rectangles where it thinks a person is, and that is all.

Automatic detection will miss people. Someone small, partly hidden, turned away, cropped by the frame or in poor light may get no box at all. When nothing is found the tool says “No people detected” and asks you to add boxes manually. Always look over the whole photo before saving and add boxes for anyone it missed.

Detection also finds things you did not intend. A box may land on a statue, a poster or a mannequin. Delete what you do not need.

Only boxes you have marked are covered. Detection does not look for number plates, screens, documents or house numbers. Draw boxes for those yourself.

Check the saved file, not just the preview. Zoom in on the finished image before posting it. A light blur over a small face can still be recognisable to someone who knows the person, so make boxes generous and use pixelate when in doubt. The tool is a way to reduce what you reveal; it does not promise that nobody could work out who someone is.

Large photos are reduced. Anything over 4096 px on the long side is downscaled to 4096 px. If your device cannot do the blur effect, the tool uses pixelation instead and tells you.

Slow devices. If your device does not support WebGPU, detection runs in a slower compatibility mode, and the page warns you.

FAQ

Does Auto-Blur recognise who is in the photo?

No. It produces boxes only. It does not identify anyone, does not compute anything that could be used to match one person with another, and keeps no face or person image. The photo itself is processed on your device and not uploaded.

Should I use blur or pixelate?

Both hide the region in the saved file. Pixelate is the safer choice when the content must not be legible, such as text, number plates or small faces. Blur looks more natural for a background crowd. If in doubt, pixelate.

Can I do it without the 43 MB download?

Yes. The download is only needed for Detect people (beta). Drawing boxes by hand, choosing Blur or Pixelate and applying the redaction all work without it.