How to Review Automatic Face Redaction Before Sharing

Use local face detection as a starting point, then check every mask, surrounding detail, and exported PNG before you share an image.
Sep 6, 2026

Last reviewed: September 6, 2026

Automatic redaction is the first pass, not the final decision

Automatic face detection can save time when a photo contains several people. It is still only a suggestion: a face can be small, turned away, partly hidden, near an image edge, or confused with a non-face object. A safe workflow always includes a human review before the image leaves your device.

PrivaCanvas finds likely face regions locally in your browser. It does not identify people, compare faces to a database, or scan the rest of the image for text, credentials, addresses, or other private details.

A five-step review before you share

  1. Choose an image in the PrivaCanvas editor, then select Detect faces. Suggested masks are created on your device.
  2. Check the whole image first, then zoom in on every suggested mask. Make sure it covers the complete face, including the outline and nearby pixels.
  3. Look for faces that may have been missed: small people in the background, side profiles, reflected faces, faces hidden by hair or an object, and people close to the edge of the frame. Draw a manual mask wherever needed.
  4. Check everything that is not a face. Names, address labels, screens, ID badges, vehicle plates, maps, messages, and account information require a manual mask because face detection does not look for them.
  5. Download the new PNG, open that downloaded file, and inspect it one more time. Share the exported copy, not the original image.

Choose the right mask for the risk

For a face in an ordinary group photo, Blur or Pixelate can preserve enough visual context while reducing recognition. For exact secrets such as a password, API key, account number, address, or ID number, use Blackout with some padding around the content.

The exported file is flattened into a new PNG, so the drawn masks are rendered into its pixels. This is still not a guarantee that an image is anonymous: review the result in the context where it will be posted or sent. For a detailed comparison, see blur, pixelate, and blackout.

Why the final exported file matters

Reviewing the image only on the editing canvas is not enough. The actual file you attach may be the original, a different version, or a screenshot of the editor. Reopening the downloaded PNG gives you one last chance to confirm that every mask is present and that you selected the correct file.

If you are sharing a screenshot, also use the screenshot redaction checklist. It covers browser tabs, notifications, file paths, API keys, and other details that automatic face detection cannot find.

Questions

Does face detection tell me who a person is?

No. It only suggests likely face regions for manual review. PrivaCanvas does not perform identity matching or age, emotion, or other trait inference.

Can automatic detection redact a screenshot's text?

No. Text and other private details need manual masks. Use blackout for exact secrets, and rotate a credential if it may already have been exposed.

Does the review upload my image?

No. Image decoding, face detection, editing, and export run in the browser. PrivaCanvas does not have an image-upload endpoint for this editor.

Review a photo before sharing