← Back to blog

How VideoPuzzle works: assembling images from live video

April 9, 2026 · 4 min read

Classic jigsaw puzzles are getting boring. A static image you scan in seconds and then play from memory. What if the image was moving? That's exactly the question that started VideoPuzzle.

Why video?

Solving puzzles activates your visual cortex — you hunt for patterns, edges, colors. With a static image, your brain memorizes the layout in a few seconds and play becomes mechanical. A moving video rewrites this process. Every piece is unpredictable: the puppy blinks, the panda raises a paw, steam rises from the cup. Your brain has to adapt in real time.

The result? A game that holds you longer even at the easiest difficulty, because you simply can't predict what a piece will look like a second from now.

Behind the scenes: canvas and Bezier curves

The implementation wasn't trivial. Classic puzzle pieces have characteristic rounded tabs and blanks. To make every piece look like a real jigsaw piece and have neighbors interlock, we generate the shape dynamically:

Drag & drop and scattered pieces

To make the game feel like a real jigsaw, pieces are scattered around the assembly area. A best-fit algorithm tries 80 positions per piece and picks the one with the largest distance from others. When you drop a piece within 32% of a cell from its correct position, it auto-snaps and locks in.

What's next?

We're planning more videos (now 11), more difficulties (up to 12×12), and maybe a co-op mode where multiple players assemble the same puzzle online. But that's future music — for now you can play the current version.

▶ Play VideoPuzzle