Why blur faces before sharing a photo
Posting a photo of a protest, a school event, a street scene or a customer review often shows people who never agreed to be online. Blurring their faces respects their privacy, keeps children safe, and in many places is expected or required: privacy laws such as the EU’s GDPR treat a recognisable face as personal data. Journalists, teachers, landlords, real-estate photographers and anyone posting dashcam clips blur faces and number plates routinely.
The irony of most online face blurring tools is that you have to upload the very photo you want to protect. This one runs entirely in your browser. The photo is never uploaded; only a tiny face detection model is downloaded.
How automatic face detection works
Click Find faces automatically and a face detector called YuNet looks at the photo. YuNet is a compact neural network from the OpenCV project (232 KB, MIT licence) designed to find faces quickly on ordinary devices. It runs in your browser through WebAssembly.
- The photo is scaled into a 640 × 640 square and the network scores every position for “this looks like a face”, on three grids, 8, 16 and 32 pixels apart, which catch small, medium and large faces.
- Positions that score above the sensitivity threshold become boxes, and overlapping boxes for the same face are merged.
- For large photos (over about 1,280 pixels on the long side), the image is also scanned in overlapping tiles, so a small face at the back of a group photo is still big enough to detect.
- Each box is grown a little, more at the top than the bottom, so hair, ears and chin are covered too.
| Sensitivity | Threshold | Use when |
|---|---|---|
| Low | 0.8 | You only want clear, front-facing faces |
| Normal | 0.6 | Most photos |
| High | 0.4 | Crowds, small or turned faces (may add false boxes, which you can untick) |
Always check the result. Detection is very good at clear faces but can miss faces in profile, in deep shadow, behind sunglasses or partly hidden. Draw a box over anything it missed before you share.
Blur, pixelate or solid?
| Effect | What it does | Strength of protection |
|---|---|---|
| Blur | Averages each pixel with its neighbours (three passes of a box blur, close to a Gaussian blur) | Good at high strength; a light blur can leave a face recognisable |
| Pixelate | Replaces the area with flat squares of the average colour | Good with small blocks at high strength; very large blocks can sometimes be matched |
| Solid | Covers the area with one colour | Complete: nothing of the face remains |
The strength scales with the size of each box, so a big face in the foreground and a small one in the background get the same look:
For a 200-pixel-wide face at 60% strength, that is a blur about 30 pixels wide, enough to make features unrecognisable. Researchers have shown that weak blurs and coarse pixelation can sometimes be reversed or matched by software, so for sensitive cases (witnesses, children in care, people at risk) use Solid.
Blurring things that are not faces
Drag on the photo to draw a box anywhere. Typical uses:
- Number plates in car sale photos or dashcam stills.
- House numbers, street signs and door numbers in property photos.
- Name badges, school logos and uniforms that reveal where someone is.
- Screens and documents in the background of office or desk photos: emails, passwords on sticky notes, ID cards.
Each box can be moved by dragging it and resized with its corner dot; with the keyboard, select a box and use the arrow keys to move it or Delete to remove it. The oval shape suits faces; the rectangle suits plates and text.
Worked example: a classroom photo
A teacher wants to post a photo of 24 pupils working on a project, but three families have not given permission.
- Drop the photo in and click Find faces automatically with Normal sensitivity: 22 faces are found (two pupils were looking down).
- Draw boxes over the two missed faces, so every face is now covered.
- Untick the 21 faces of pupils whose families agreed, leaving the three who must stay hidden, plus the teacher’s choice to hide any badge with the school’s name.
- Choose Pixelate at 80%, download, and post.
After blurring
- Photos from phones often contain the GPS location where they were taken. Remove it with EXIF Remover before posting.
- Shrink the file for social media with Compress Image.
- Add your name or a copyright line with Watermark Image.
- Trim away unneeded parts with Crop Image: what is not in the photo cannot be identified.
Privacy
The photo is decoded, analysed and blurred inside your browser. The full-size image is processed when you download, not the small preview, so the saved file keeps its original resolution. There is no upload, no account and no limit on how many photos you blur.