Missing corners are what's left when a photo has been handled too many times by too many people. The paper itself has gone — a triangular bite out of the top-left, a curved chunk out of the bottom, sometimes a whole strip. This is the damage type most likely to make a user ask "can you actually fix this?" and the one where honest answers matter most.

What causes missing corners on old photographs

Physical loss of the paper substrate happens in a handful of common ways. Album mounting is the most frequent: the Library of Congress photograph care leaflet explicitly warns that releasing photographs stuck to old album pages can cause tears, creases, and losses if attempted without patience or expertise (LoC care leaflet). Magnetic albums from the 1970s and 1980s are the worst offenders because the adhesive zones frequently outlast the photo's corners during removal attempts.

Other routes: corners crushed in a shoebox and snapped off, frames with sharp inner rebates that tore the print on extraction, water damage that turned the paper support brittle and made it flake, and children — who are statistically over-represented in the handling history of most family photographs.

The important distinction with this damage type: everything else on this list is about degrading information. Missing corners is about information that is physically not there anymore. There is nothing to recover. Anything that appears in the missing region is invented — educated, plausibly invented, but invented.

How AI handles missing corners — and where it can't

Large contiguous missing regions are a different inpainting problem from narrow scratches. A thin line can be spanned with local continuity; a whole corner has to be generated from learned priors about what typically sits in a corner of a photograph with that subject and composition. The IEEE research on regionwise generative inpainting for large missing areas explicitly calls this out as harder than small-region inpainting because the model has to maintain global scene consistency, not just local texture continuity (IEEE Xplore). Pathak et al.'s foundational work on semantic inpainting with deep generative models established the basic approach of using a learned prior to hallucinate plausible content in large gaps (arXiv 1607.07539).

PhotoFlip's pipeline for missing corners:

• Detects the absent region as an outside-the-frame gap rather than an inside-the-frame occlusion, which changes the boundary conditions the inpainter uses.
• Extends edges, horizons, walls, and backgrounds from the surviving portion into the missing region.
• Refuses, by default, to invent faces in a missing region — if a corner of the image originally contained a person and nothing of them survives, filling in a generated stranger is worse than leaving the region neutral. The pipeline biases toward extending background over manufacturing subjects.

Honest limit: if a missing corner cuts off half a face, the restore pass will not recreate the other half. The face-restore model is designed to enhance existing face data, not to fabricate facial anatomy that was never captured. A missing corner through a subject is always going to be a compromise between leaving the loss visible and filling it with plausible-but-not-real content. That's a call you make, not the model.

Example restorations

• Top-left corner missing from a landscape photo. Cropped sky and tree branches remained on the right and below. Easy case — the model extends both cleanly into the corner.
• Bottom strip torn off a family portrait. Grass and shoe-level detail gone. The pipeline generates plausible ground plane from surrounding image context.
• Corner crease + tear where a face used to be. Hard case. The model fills the corner background reasonably but leaves, or partly fills, the face region. Expect to crop rather than accept a fabricated face.

How to restore a photo with missing corners

1. Scan the whole damaged print, missing regions and all. Don't mask the damage yourself — the model needs to see the full boundary to place its inpainting correctly.
2. Upload at photoflipai.com/restore. For large missing regions, expect a longer processing pass.
3. Review the output carefully. This is the damage type where you should compare input and output closely and decide whether the fill is content you want or content you'd rather crop out.

Related damage: missing corners often travel with tape-marks (the tape took the corner with it) and creases near the tear line. For photos that are merely surface-scratched rather than physically incomplete, see scratch-removal or deep-scratches. Pricing: pricing. Details: how-it-works.