How to Blur Faces in Photos Before Posting

When you need to blur a face — and why it matters

There are a hundred situations where you have a photo worth sharing but one problem: someone's face is in it and they didn't consent to being published online. A protest photo. A street scene from your trip. Kids at a public event. A screenshot from a video call. A photo of a crime scene or accident you're reporting.

In most of these cases the solution seems obvious — blur the face. But the tools to do it without uploading to a stranger's server are surprisingly hard to find. Most "blur face" tools online require you to hand over your photo to a third party, which defeats the purpose if the photo is sensitive.

Why uploading your photo to blur it is a bad idea

If you're blurring a face because the photo is sensitive — a protest, a confidential meeting, a person who doesn't want to be identified — then the last thing you should do is upload it to some random website to process it. You've just given a copy of the unblurred original to an unknown server.

Client-side processing is the only approach that makes sense here. The blurring happens in your browser using your device's CPU. The photo never travels anywhere. Close the tab and it's gone.

How face detection works in the browser

Filesmith's face blur tool uses BlazeFace — a lightweight face detection model originally developed by Google, which runs entirely in JavaScript via TensorFlow.js. It loads directly into your browser tab from a CDN, analyzes the image locally, and returns bounding boxes around any faces it finds.

The model is fast and fairly accurate on front-facing portraits, but it can miss faces that are: turned sideways, very small in the frame, partially obscured, or in low-contrast lighting. That's why the manual brush exists — you're always in control.

Step by step: blurring faces with Filesmith

1. 1
   Drop your photo Drag and drop any JPG, PNG, or WEBP onto the tool. Nothing is uploaded.
2. 2
   Auto-detection runs immediately The BlazeFace model loads and detects faces automatically. Detected faces are blurred right away.
3. 3
   Adjust blur intensity Use the slider to make the blur heavier or lighter. It re-renders live so you can see the effect instantly.
4. 4
   Fix anything the model missed Switch to "Paint blur" and brush over any faces the auto-detection missed. The brush has soft edges so it blends naturally.
5. 5
   Erase any over-blurred areas Switch to "Erase blur" to remove blur from parts of the image you didn't mean to cover.
6. 6
   Save the photo Hit "Save photo" — downloads as JPG (if original was JPG) or PNG. The original file is untouched.

When auto-detection misses faces

BlazeFace is optimized for frontal, well-lit faces. It will reliably miss faces that are:

• Turned more than ~45° to the side
• Very small relative to the image (crowds in the distance)
• Obscured by hats, sunglasses, or masks
• In dim or backlit conditions

For any of these, just use the paint brush. It's a 30-second fix — set the brush size to cover the face and paint once or twice. The soft radial gradient on the brush means it blends naturally rather than leaving a hard rectangular box.

Does blurring actually protect someone's identity?

A strong enough blur does. The key is using enough blur strength that the face can't be reconstructed — even by someone running AI image enhancement. A blur radius of 15+ in this tool is sufficient for most purposes. If you need strong anonymization, go higher.

What doesn't work: light blurring, pixelation at low levels, or black bars applied with low opacity. These can sometimes be partially reversed. Heavy Gaussian blur — which is what this tool applies — cannot be reversed.