Photo spot the difference: how real photographs become puzzles

Choosing a source image

Not every photograph makes a puzzle. The ones that do have distinct objects with clear edges, an even spread of interest across the frame, and no single dominant subject that pulls all attention to the centre. A market stall, a cluttered desk, a living room, a beach with people and objects scattered across it: all good. A tight portrait against a plain wall: hopeless, because there is nowhere for a change to hide.

Resolution and lighting matter for a practical reason. Soft, low-light images make it hard to distinguish a genuine edit from photographic noise, and noise is exactly what an automatic difference measurement mistakes for a change.

Which edits work, and which do not

The edits that produce satisfying puzzles are structural: an object removed entirely, an object added, an object swapped for a different one, something moved to a clearly different position, something resized noticeably.

The edits that produce complaints are cosmetic: a slight tint shift, a small brightness change, a texture tweak. They register as differences in the data and are close to invisible to a player. We now explicitly steer away from those, because a puzzle where the answer is "that shirt was a slightly different blue" is not a puzzle, it is a trap.

Measuring the differences, rather than trusting them

Once the edited copy exists, the two images are compared pixel by pixel. Regions that differ beyond a contrast threshold are grouped into zones, and those zones become the tap targets. Nothing is taken on faith: a change only counts if it is measurably present in the image.

This is also why a photo puzzle sometimes has a different number of differences than the edit intended. If an edit was too faint, the measurement drops it, and the puzzle publishes with fewer differences rather than with one nobody can find.

The repair pass

Raw measurement rarely produces a clean puzzle on the first attempt. Zones come out too small to tap, or two nearby edits merge into a single blob, or a zone straddles the frame edge. So the measurement is followed by an automatic repair pass: zones grow to a tappable minimum, nudge inside the frame, and separate from neighbours that sit too close.

Only after repair do the fairness gates run. If the result still fails — fewer than three usable differences, or zones that cannot be separated — the puzzle is not published.

Why spacing is adjustable

The right minimum distance between two differences depends on the photograph. In a wide landscape, changes ten percent of the frame apart feel comfortably separate. In a close interior, the same distance puts two changes on the same shelf. So separation is a setting rather than a constant, adjustable per puzzle when a scene needs more breathing room.

How photo puzzles feel different to play

Photographs carry far more texture than drawn scenes, which changes the strategy. Counting works less often because real scenes repeat objects imperfectly, and silhouette comparison works better because real objects have complex outlines that are hard to fake.

Most players find photo puzzles slower and more absorbing. The scene rewards looking at it, rather than just scanning it, which is a different pleasure to the sharp rhythm of a drawn puzzle.

Rights and originality

Every photograph used on the site is either generated for us, licensed for this use, or owned by us outright. If you believe an image on FreeInteract is used without permission, email us with the puzzle link and it will be taken down while we check.