Restoring old class photos
Thirty faces at forty pixels each, none of them individually legible.
Why class photos come out this way
A class photograph spreads its resolution across a whole room of people. Each face gets a tiny share of the frame to begin with, and decades of fading and a phone capture of the print leave almost nothing per person.
What to run, and in what order
This is the strongest case for face restoration, because the tool processes every detected face separately instead of averaging one sharpening pass across the whole group. Repair first if the print is faded or creased, then Restore Faces.
The tool for this is Restore faces in a photo.
Practical notes
- Capture the print in one flat, straight-on shot instead of photographing sections and stitching them
- Expect variation: faces near the centre come back convincingly, faces at the edges less so
- Check any face you plan to identify by name before telling someone that is them
Questions
Why do some faces come back better than others?
Faces near the centre are larger and better focused. Faces at the edge are smaller, often partly turned away, and give the model less to work from.
Can I crop one child out and enlarge them?
Yes, and it works better than you would expect. Restore faces first, then crop, then upscale the crop.
Going deeper
How to unblur a face in a photo explains the method behind this, including what it cannot fix.