Before you post a group photo, a street scene or a picture from a school event, there is one question worth asking: does everyone in this photo want their face on the internet? Children, bystanders, strangers in the background and friends who never agreed to be posted all have a reason to say no — and once a photo is shared, it is effectively shared forever. Our free face blur tool fixes that in minutes: it auto-detects faces in any photo and blurs, pixelates or blacks them out before sharing, free and unlimited, with your photo never leaving your device. This guide explains how to use it, which blur style fits which photo, what auto-detect misses, and the checks to run before you press share.
Step one: add your photo. Drop a photo onto the tool, click to browse, or paste it straight from your clipboard with Ctrl+V — one photo at a time. Large images are downscaled for smooth editing, and your download is exported at up to 4000 pixels, full quality for sharing and printing. Step two: find the faces. Tap Auto-detect faces to find faces automatically, or drag boxes over faces yourself. Click any box to select it, then restyle it or press Delete to remove it. Each box can be blur, pixelate or a black bar, so one photo can mix styles — blur for a friend who asked not to be tagged, a black bar for a stranger who must be unmistakably gone. Step three: download the result. Export a clean PNG with EXIF location data stripped. Your original photo is never uploaded or stored anywhere, and the blurred regions are permanently baked into the downloaded pixels — that permanence is exactly what makes it safe to share, and exactly why you should keep your original somewhere safe too.
Blur, in the gaussian style, is the friendliest option — faces melt softly into the background and the photo still looks like a normal photo, which is what you want when the point is privacy rather than a statement. Pixelate screams censored and draws the eye, which is sometimes exactly the point: it signals loudly that a face has been hidden, and it reads clearly even at small sizes on a phone screen. Black bar is the strongest anonymisation but looks harsh, and it is the right choice when a face must be unmistakably gone. The rule of thumb from the tool page is simple: blur for privacy in otherwise normal photos, black bar when the face must be unmistakably gone. The same thinking extends beyond faces: draw boxes over number plates, house numbers, addresses, names on documents or anything else you would rather not publish. For screenshots full of text, the Screenshot Redactor tool is tuned for exactly that job, but for photos, a hand-drawn box here does the same work.
The tool page's own example is a street fair album: 24 crowd photos, 67 faces in total. Auto-detect finds 61 faces in about 40 seconds. Six misses — sunglasses, sharp profile angles — get manual boxes in roughly two minutes. All boxes are set to blur, and the batch is exported as PNGs with EXIF location data stripped. Total time: under five minutes for a share-safe album, with the originals untouched. That example is worth studying because it shows the workflow honestly: auto-detect does the bulk of the work, manual boxes finish it, and the whole job is measured in minutes, not an evening of editing. It also shows why the manual step matters. If those six missed faces had gone out unblurred, the album would not have been share-safe at all — the misses are the whole story, and the tool is designed around catching them rather than pretending they do not happen.
Auto-detect loads a small on-device face-detection model, face-api.js, the first time you tap the button, and detection itself runs entirely in your browser. But it misses some faces: sunglasses, face masks, extreme angles and very small or distant faces are the common gaps, because the model looks for the pattern of a visible face and those cases break the pattern. Detection also needs one internet connection to download the model; everything after that is local, and if auto-detect is unavailable or you are offline, manual mode keeps working fully — just drag boxes over the faces yourself. So before sharing, run this checklist on every photo, zoomed in: every face in the foreground, every face in the background, every reflection in a mirror or window, every child at the edge of the frame, and every number plate or address caught in shot. The blur is permanent in the export, so a missed face cannot be fixed after posting — review first, share second. And keep your original: once you download the PNG, the blurred regions cannot be un-blurred, by you or by anyone else, because there is no hidden original data left to recover.
Trusting auto-detect without reviewing. Sunglasses, masks, profiles and tiny distant faces are missed regularly — check every photo zoomed in before sharing. Sharing the original by accident. The export is the safe file; the original still has every face and its location data, so post the PNG you downloaded, not the file you started with. Forgetting location data. A photo's EXIF data can carry GPS location, device info and timestamps; the exported PNG is re-encoded from the canvas, which drops that data — but only if you share the export. Blurring the face but leaving the name tag, plate or address. Identity leaks through details, not just faces; box those too. Expecting it to work on video. This tool is photos only — for video, extract the frames you need or use a video editor with tracking blur. Deleting the original. The blur is permanent, so if you might want the unblurred photo later — for family, for print, for any reason — keep the original somewhere safe before you export.
The honest limits are five. Auto-detect misses some faces, as above, so review is part of the job, not an optional extra. Detection needs one internet connection to download the model; everything after that is local, and manual mode works fully offline. The blur is permanent in the export, so keep your original if you might want it later. The tool handles one photo at a time and photos only — no video, no batch upload. And while the exported PNG drops EXIF metadata including GPS location, device info and timestamps, that protection applies to the export only; the original file still carries all of it. Within those limits, the tool does one job completely: it takes a photo you cannot safely share and returns one you can, without an account, a watermark, a photo limit or an upload. The companion guides are our background removal guide, for when the whole backdrop has to go rather than a detail in it, our passport photo guide, for the opposite problem — a photo where the face must be fully visible and correctly sized — and our image upscaling guide, for sharpening a small photo before you blur and share it.
When the whole backdrop has to go rather than a detail in it — the AI cut-out route.
The opposite problem: a photo where the face must be fully visible and correctly sized.
Sharpen a small photo before you blur and share it — 2x and 4x, in your browser.