AI Face Blur
Blur, pixelate or black out faces in a photo. A small detector boxes the faces it finds, small ones in a group included, and you box whatever it misses.
The photo
Photo
Drop the photo here, paste one, or choose a file
JPEG, PNG, WebP — HEIC in Safari · faces are found on arrival
—
—
Boxes · 0
Zoom
Effect
Fill paints the box one solid colour — the one for number plates, badges and screens, where a blur or a mosaic can still be read back.
A quarter of a box's shorter side at 100%, so the blur follows each box — the same in the preview and the file.
Save as
PNG keeps every pixel; JPEG at 92 is about a tenth of its size for a photo. Either way the file is drawn fresh, so EXIF and GPS are left behind.
Download
Add a photo and the detector runs on arrival — no press. Download is held until at least one box exists, because with none the file would be the photo again.
How to use
- Add the picture under Photo, or press Take photo if you are standing at the event with a phone. Detection starts as soon as the image loads.
- Read the box list and then correct it. Drag a box onto a face that was skipped, resize one over a plate or a badge, and remove a false hit with its ×.
- Choose Blur, Pixelate or Fill, switch to As saved at 1:1 to check nobody is recognisable, then press Download photo.
About this tool
A school concert, a busy street, a screenshot with a colleague's face in the corner. Blurring faces in a photo before it goes public is a small courtesy, and here it takes about thirty seconds. Add the picture and a face detector runs over it straight away, dropping a box on each face it recognises. Every box is padded 15% outward from the detection, so you get the chin and the ears rather than a rectangle sitting neatly on the nose.
The detector is YuNet, a small open model trained on faces down to about ten pixels across. It reads the photo whole, then again in overlapping tiles, so the back row of a group shot gets a close look instead of being skipped. On a twenty-person group photo it boxed all but a face or two, and those were half hidden behind a neighbour. That is the honest limit: turned-away, covered and very distant faces can still slip through, so count them yourself against the list.
Everything after that is yours to correct. Add box drops a square in the middle at a fifth of the photo's shorter side. Drag a box where it belongs or pull a corner to resize it; with a keyboard, the arrows move the box you are on and Alt with the arrows resizes it. On a phone, tap the photo and the selected box jumps there, and Shrink and Grow change it by a tenth. Remove a box with its own ×, the × on its row, or the Delete key.
Pick Blur, Pixelate or Fill and set the Intensity slider, which starts at 50%. The blur is a quarter of each box's shorter side at full strength, with no ceiling, and the line under the slider says what that means on your largest box. Pixelate wound up to 100% collapses every box to three blocks across. As saved hides the outlines and numbers and shows the file exactly as it will be written, and 1:1 lets you check it pixel for pixel. Save as PNG to keep every pixel, or JPEG for a file about a tenth the size.
Before you post it
- Faces turned away from the camera, heads behind a hand or a mask, and faces only a few pixels tall are the ones the detector can leave alone, and now and then it boxes a pattern that only looks like a face.
- Reflections give people away. A mirror, a shop window, a laptop screen, even a pair of glasses can hold a face that sits outside every box you drew.
- A gentle blur or a fine mosaic hides less than it seems to. Software trained for the job still recognises many blurred and pixelated faces, so push the slider up and judge the file at 1:1 rather than at a glance.
- One photo at a time, and loading a new one clears the boxes. There is no batch mode, so a folder of twenty pictures is twenty passes.
- The largest photo it takes is 67.1 megapixels, 8192 × 8192 or the same area in any shape. That is as much as Safari on an iPhone or iPad can draw into one picture, and the saved copy is always full size, so shrink anything bigger with the Image Resizer first.
Frequently asked questions
How do I blur a face in a photo?
Add the photo, wait a beat for the detector to box the faces it finds, then pick Blur under Effect and raise the intensity. Press Download photo and you get a fresh PNG or JPEG with the effect baked in. Anything the detector missed you cover yourself with Add box, and anything it boxed wrongly you remove with its ×.
Blur or pixelate: which one hides a face better?
At full strength they are about equal, and both destroy detail rather than covering it. Pixelate replaces the box with a coarse mosaic, three blocks across at 100%, and reads as a deliberate censor bar. Blur smears the area into a soft patch and looks more natural in a photo you still want to look like a photo. Pick on appearance and push the slider high either way.
Can someone reverse the blur and see the original face?
There is nothing to undo, since the file holds only the blurred pixels. A face can still be recognised, though. In one study, software trained for the job named blurred faces 57% of the time and saw through mosaics more than 95% of the time, in a group of forty people. Gentle settings are the weak spot. Raise the slider until the line under it reports a large blur or only a few blocks, and check the result at 1:1.
A face wasn't detected. How do I cover it myself?
Press Add box. A square appears in the middle of the picture at a fifth of the shorter side; drag it over the face and pull a corner until it fits, or tap the face and let the box jump there. Boxes can overlap freely and there is no cap on how many you draw, so a group shot with a dozen strangers is fine.
Can I use this on a number plate or an ID badge?
Use Fill for anything with letters or numbers on it. Blurred and pixelated text can often be read back, which is why researchers call both unsafe for redacting it. Fill paints the box solid black, which leaves nothing to recover. Draw a box over the plate, badge, screen or envelope and pick Fill for that photo.
Does the saved file still carry the photo's GPS location?
No. The download is drawn onto a new canvas and written as PNG or JPEG, and that process keeps only pixels. Camera model, timestamp, and the GPS coordinates that phones write into a photo are all left behind with the original file, which stays untouched on your disk.
Missing something in AI Face Blur? Suggest a feature →
Tell someone who needs this.