Auto-Anonymize
Automatically detect and blur visible frontal faces in an image.
Read the guide and worked example ↓
Recent Results
The download appears directly below its progress bar when processing finishes.
How to use Auto-Anonymize
Apply blur to frontal faces detected in a still image. This is an assistance tool for image review, not a guarantee of complete anonymization. Check the output carefully before sharing it, especially when the scene contains profiles, small faces, reflections, or identifying text.
Prepare, run, and review
- Upload a supported still image and choose a blur strength. Start with a setting that makes the detected face region difficult to identify without obscuring more of the surrounding scene than necessary.
- Run the tool. The current implementation uses a frontal-face cascade detector and applies Gaussian blur to detected rectangles.
- Download anonymized.png and inspect every person at full image resolution. Compare against the original so missed regions are easy to identify.
- Review other identifying information separately, including license plates, visible names, badges, screens, reflections, and file metadata. Decide whether additional editing is required for your intended use.
A successful job can still need more review
Input scene:
• one large frontal face
• one small side-profile face
• one readable badge
Review all three details in the output.The frontal face may be detected while the profile and badge remain unchanged. Increasing blur strength affects detected regions; it does not teach the detector to find missed faces or text. If the output appears unchanged, inspect whether any faces were detected rather than assuming the image was already anonymous.
Limits and common mistakes
- The detector can miss faces that are small, rotated, occluded, poorly lit, or unlike its training examples. False detections are also possible.
- This tool does not automatically detect license plates, remove identifying text, anonymize audio, or process an entire video.
- Blur strength is not a formal privacy guarantee. Review the actual output at the resolution and context in which it will be shared.
- Keep access to sensitive originals controlled. Producing a blurred copy does not change who can access the original file elsewhere.
Will stronger blur fix a missed side-profile face?
No. Blur strength applies to regions the detector found. A missed face requires additional detection or manual editing and review.
Continue learning
ToolsInVision guide · Updated September 26, 2026 · Editorial approach · Report an issue