The Visual Privacy tab lets you examine the actual picture and select information that should be permanently concealed. This is separate from metadata sanitization because the information you are protecting is part of the image pixels.
Manual redaction is available in all PhotoSanitize editions. Pro and Enterprise add automatic detection to help identify common types of potentially sensitive visual content.
Examining the image
After adding and analyzing a photograph, highlight it in the Photo Queue and open Visual Privacy. If a compatible decoder is available, PhotoSanitize displays the picture in its preview area.
Use Fit when you want to see the complete image within the available space, and 100% when you need to examine details more closely. Zooming is particularly useful for documents, labels, screens, or faces that occupy only a small part of the photograph.
For supported multiframe images, you can navigate between frames using the available controls. This matters because an animation may contain information that is not visible in its first frame.
If the preview cannot display the picture, PhotoSanitize may still be able to inspect some of its metadata. Visual redaction and automatic detection, however, require access to decoded image pixels.
Automatic detection
Pro and Enterprise can automatically identify potential faces, license plates, visible text, QR codes, and barcodes. You can select which detector categories are enabled in Options and then run the scan using Detect in Visual Privacy.
The application displays the recognized regions in the image preview and findings list. Each automatic finding includes a type and confidence information.
For example, a photograph of an office reception area might produce findings for several faces, the text on a visitor badge, and a QR code displayed near the entrance. You can inspect these detections and decide which ones should be concealed.
Detection is performed locally on your computer. The image does not need to be uploaded to an external AI service.
Understanding confidence values
A detector’s confidence value describes its estimated confidence that a region matches the category being searched for. It does not indicate how private the information is.
A high-confidence face detection may involve someone who is allowed to appear in the photograph, while a lower-confidence text region may contain a confidential account number.
This is why the detection list should be treated as an aid to review rather than a final decision about what must be redacted.
The visible content and intended recipient determine which areas require concealment.
Detection limitations
Automatic detection depends on the visual characteristics of the source image. Low resolution, poor lighting, unusual angles, blur, reflections, partial obstruction, and small objects can all affect the results.
A face viewed from the side may not be detected. A partially obscured license plate may be missed, and a small name printed on a document may not produce a text finding.
Automatic detection may also identify ordinary content that does not need to be concealed.
For these reasons, examine the entire picture and add manual redactions for sensitive areas that were not detected.
What happens after detection
Detected regions appear as potential redactions. You can review their placement, adjust supported region boundaries, and select which findings should be included in the sanitized output.
When a profile requires visual review, use Mark Reviewed after completing this examination. The review applies to the current image and selected findings, not automatically to every photograph in the queue.
During sanitization, the selected regions are permanently concealed in the output. They are not simply covered by removable annotations.
For professional and organizational work, review the resulting file at a suitable zoom level before approving it for delivery or publication.