How to make a spot the difference puzzle

Any photo can become a puzzle. The hard part is not editing the image — it is choosing changes that are findable but not obvious. Here is the method, plus the shortcuts worth knowing.

What makes a difference fair

Play a few puzzles and you notice the same handful of moves. They work because they hide in busy parts of the image, where the eye already has too much to track.

• Remove something small. A mug from a desk, one apple from a bowl, a bird from a fence. The gap is invisible until you look for it.
• Change a colour. A red cushion becomes blue. This one is easy to spot on a plain wall and hard on a patterned rug.
• Shift an object. Move a chair two inches left, or slide a book along a shelf. Small displacements are brutal on cluttered backgrounds.
• Add something that belongs. A second cloud, an extra flower, a stone on a path. The trick is that it must look natural — a floating object gives itself away.
• Change a shape. Bend a branch, widen a doorway, shorten a shadow. Good for architecture and landscapes where colour is uniform.

The changes that ruin a puzzle are the ones that break the scene: a lamp that no longer stands on the table, a person with three arms, a shadow pointing the wrong way. Players stop hunting and start pointing at the error.

How many differences should a puzzle have?

Five is the standard for a reason. Three feels thin, ten turns the game into a chore, and anything above that forces you to make changes so tiny that players argue about whether they count.

Adjust by audience instead of by count:

• Young children (4–7): five differences, all large, all colour-based. Nothing hidden in texture.
• Adults, casual play: five differences, mixed difficulty, at least two in busy areas.
• Puzzle enthusiasts: keep it at five but make three of them subtle — a shifted object, a removed detail, a slightly rotated sign.

Making one by hand

This is the traditional route and it still produces the cleanest results if you have the time and the software.

1. Pick a photo with detail. Gardens, kitchens, bookshelves, street scenes, crowded beaches. Avoid plain skies, blank walls, and heavy blur — there is nowhere to hide a change.
2. Duplicate the layer. In Photoshop, Affinity Photo, GIMP, or Photopea, copy the background layer so you always have the original intact.
3. Make five edits on the copy. Use clone stamp or content-aware fill to remove things, hue/saturation to recolour, and free transform to shift objects. Work at 100% zoom, not fit-to-screen.
4. Write down each change as you make it. You will forget the fourth one. Every time.
5. Export both images at the same size. Side by side, same dimensions, same crop. Any difference in framing makes the puzzle unsolvable.
6. Build the answer key. Draw numbered circles on the modified copy and list the five changes underneath. This is the part most people skip and later regret.

Realistically this takes 30–60 minutes per puzzle once you are practised, and most of that is deciding what to change rather than doing it.

Making one with AI

The faster route is to let a model propose the changes and then verify them. That is what this site does: upload a photo, and it returns two versions plus a numbered answer key.

1. Upload a photo — phone shots work fine. JPG, PNG, and HEIC are all accepted.
2. Wait 20–40 seconds while the image is analysed and five changes are generated.
3. Compare the two versions and check that every change looks natural. If one looks wrong, generate again from a different photo — cluttered scenes give the best results.
4. Download the pair and, if you are sharing or printing, the answer sheet with the differences circled and described.

Five mistakes that make a puzzle unplayable

• Differences clustered in one corner. Spread them across the frame so the eye has to travel.
• Changes that alter the composition. If a difference moves the horizon, everything else looks wrong and players get lost.
• No answer key. Half of all arguments about these puzzles are caused by a missing key.
• Different crops. The two images must be pixel-identical except for the five changes.
• Too many differences. Five is plenty. Ten is a punishment.