Blur Faces in Photos & Videos — Free AI Privacy Tool
Automatically detect and blur faces for privacy.
Posting a crowd shot, a candid school photo, or a screenshot with bystanders means thinking about everyone's privacy in frame. Dropvert's face blur uses on-device face detection to find every face automatically and apply a pixelate or Gaussian blur — without sending the photo to a server.
How it works
3-step walkthrough
How it works
3-step walkthrough
- 1
Drop the photo
Drag any JPEG, PNG, or WebP. Dropvert runs MediaPipe's face detector locally and shows every detected face as an overlay on the preview.
- 2
Toggle individual faces, pick blur style and intensity
You can disable specific faces if you want some kept clear. Choose pixelate (visible, deliberate) or Gaussian (smooth, subtle). Adjust intensity and the region expansion (how far around the face to blur).
- 3
Download with everyone anonymized
Export as the same format. Batch mode runs the same settings across up to 20 photos at once.
Why use Dropvert
Local-first, free, no upload required
Why use Dropvert
Local-first, free, no upload required
- Automatic face detection via MediaPipe — no manual masking required.
- Toggle each detected face on or off independently.
- Two blur styles: pixelate for a clear "this is intentional" signal, Gaussian for subtle anonymization.
- Adjustable region expansion — blur a tight rectangle or extend out to cover hair and ears.
- Runs entirely in your browser; the photo and the model never leave your device.
Frequently asked questions
6 answered
Frequently asked questions
6 answered
- Does the face detector work on all kinds of photos?
- It works on most front-facing or 3/4-angle faces in reasonable lighting. Profile views, partial occlusion (sunglasses, masks), heavy motion blur, and very small faces (less than ~30 pixels wide) are harder. If a face isn't detected automatically, the tool unfortunately can't blur it — the manual paint mode is on the roadmap.
- Pixelate vs Gaussian blur — which should I use?
- Pixelate is visually obvious and signals "this was deliberate" — good for journalism, complaints, anything where the anonymization needs to be visible. Gaussian is subtle and blends better when you don't want the blur to be the focus of the image.
- Can the blur be reversed?
- Pixelate and Gaussian blur are both lossy and not reversible by any practical method. AI deblurring tools have caveats and can't reconstruct enough detail to identify a face from a properly applied blur.
- Why does it miss some faces?
- The detector's default confidence threshold favors precision over recall — better to miss an edge case than to blur something that isn't a face. If detection misses a face, increase the region expansion on the ones it found, or fall back to a manual editor for that photo.
- Does it work on videos?
- Yes — switch the page to Video mode and drop an MP4, MOV, or WEBM. Each frame runs through the same detector and blur, then re-encodes to MP4. v1 caps inputs at 30 seconds and 720p output to keep the in-browser pipeline responsive.
- Are the detected face boxes stored anywhere?
- No. Detection happens in your browser and the face boxes only live in memory while you're tuning the blur. Closing the tab discards them.
Related tools
5 suggestions
Related tools
5 suggestions