How fixed photos are labelled

What a fix actually is

When you fix a face, Open Eyes takes the eyes or the mouth from another of your shots, lines them up and blends them into your favorite. Every pixel was captured by your camera. On-device face detection only finds where the eyes and mouth are; nothing is drawn, invented or generated by AI.

That makes the result a composite capture: a photo combined from several real shots of the same scene. That's the term the photo industry's standard (IPTC) uses, and it's the one we use.

The three labels

Label	Where it lives	What it says
IPTC metadata	Inside the photo file (XMP)	Digital source type "composite capture" and a short description of how many eye and mouth regions came from other shots.
Content Credentials (C2PA)	A signed record inside the photo file, where your phone supports it	The same facts, signed so that changes to the file can be detected. Includes the app name and version.
Invisible watermark	In the pixels themselves	That the photo was edited by Open Eyes, and an edit number. It survives JPEG re-saving and cropping, even when metadata is stripped.

The labels never contain your name, your location, the other shots or anything about the people in the photo.

When a photo isn't labelled

If you save without changing any face, the copy is identical in content to your favorite shot, so it isn't labelled. The Saved screen tells you which labels were added.

Checking a photo

Content Credentials can be inspected with the free Content Credentials Verify tool, and the IPTC fields are shown by most photo managers that read XMP metadata.

Limits

• Many messaging apps and social networks remove metadata when you share, which removes the IPTC label and Content Credentials.
• The watermark doesn't survive resizing, rotation, heavy filters or deliberate removal. It's a helpful hint, not tamper-proof evidence.
• Content Credentials need a recent phone. Where they can't be added, the other two labels still are.