How to use Blur faces & private details
- Choose a photo or screenshot. The first time, a small face model (about 1 MB) downloads once.
- Wait a moment while faces are found on your device; each one gets a box.
- Check every box: switch off any you want to keep, and drag on the photo to cover number plates, names, addresses or screens. Drag a box to move it, or a corner to resize it.
- Pick blur, pixelate or a solid box and its strength, then download the photo.
Share the picture, not the people in it
A street scene for a blog, a school event for a newsletter, a car park photo for an insurance claim: often the picture is fine to share but the faces, number plates or names in it are not yours to publish. This tool hides them before you post. It finds faces automatically and gives each one a box, and you can drag on the photo to add boxes over anything else, such as a licence plate, a name badge or an address on a parcel.
Everything happens in your browser. The photo is read, searched and blurred on your device, so pictures of children, colleagues and private places never pass through anyone else's computer.
How faces are found, and when they are missed
Detection uses BlazeFace, a face detection model from Google's MediaPipe project, published under the Apache 2.0 licence and trained on photos Google collected with consent. It runs with ONNX Runtime Web inside a background worker in this tab. The model downloads once, about 1 MB, and your browser keeps it. Small faces are easy to miss when a whole photo is shrunk to the model's size, so the tool also checks overlapping close-ups and merges what it finds.
No detector is perfect. A face turned to the side, tilted, very small, out of focus or partly covered can be missed, and now and then a pattern gets a box it does not need. Check every box before you share: switch off any you want to keep, move or resize one that is slightly off, and draw your own over a face it missed.
What the download contains
Choose blur, pixelate or a solid box. Strength is measured against each box, so a face looks equally hidden in a screenshot or a large camera photo, and the preview matches the download. The effect is applied to a fresh copy of the full-size photo. With blur and pixelate, the pixels inside each box are first averaged into blocks, which throws the detail away, and blur then smooths those blocks; a solid box paints over them. The original pixels are not kept in a layer that could be peeled back.
The result is saved as a new JPG or PNG, so the EXIF data, GPS location and camera details of the original are not copied, and a sideways phone photo is saved upright.
Questions
- Is my photo uploaded?
- No. The photo is decoded, searched for faces and blurred in your browser. The only download is the face model itself, from WeaveAI's own engine server; your photo never leaves your device.
- Will it find every face?
- Not always. Faces turned to the side, tilted, very small, blurred or partly covered can be missed, and now and then something that is not a face gets a box. Look over the whole photo before you share it, and draw a box over anything it missed.
- Can the blur be undone?
- The download is a new image: inside each box the original pixels are averaged into blocks and then smoothed, or painted over, so the detail is gone rather than hidden in a layer. For the strongest result, use a solid box.
- Does the download keep the photo's location and camera details?
- No. The result is drawn fresh from the pixels and saved by your browser, so EXIF data, GPS location and the camera details of the original are not copied. A photo stored sideways by a phone is saved upright.