Nonogram
Play nonogram (picross) puzzles in your browser. Every board is generated fresh and checked to be solvable by pure logic — no guessing, no sign-up, three sizes.
How a nonogram works
Every row and column carries a list of numbers. Those numbers are the lengths of the filled runs in that line, in order, with at least one empty square between consecutive runs. A row labelled 4 2 contains a block of four, then a gap of one or more, then a block of two — and nothing else.
The clues never say where the blocks go. Your job is to find the only arrangement that satisfies every row and every column at once. There is no counting trick and no hidden information: everything you need is on the edges of the grid.
Marking squares you have ruled out is not optional decoration — it is half the technique. An ✕ carries the same information as a filled square, and lines become solvable as often from the crosses as from the fills.
The techniques, in the order you need them
1. Overlap
The workhorse. Push a block as far left as it goes, then as far right, and anything covered both times is filled no matter what. In a 10-wide row with a clue of 8, the block spans columns 1–8 or 3–10, so columns 3–8 are certain. The rule of thumb: any run longer than half the line yields overlap immediately.
2. Edge forcing
A filled square touching the wall anchors the first clue exactly. If column 1 is filled and the clue starts with 3, then columns 1–3 are filled and column 4 is empty. The same logic runs backwards from the right edge.
3. Completion and closing
Once the filled squares in a line add up to the sum of its clues, everything else in that line is empty — cross it all out. Equally, a run that has reached its full clue length gets crosses at both ends.
4. Gap elimination
Crosses chop a line into segments. Any segment shorter than the smallest remaining clue cannot hold anything — cross the whole thing out. This is usually what breaks a stalled puzzle open.
5. Cross-referencing
Nothing you deduce is local. Every square you fill or cross is new information for the perpendicular line, so after resolving a row, immediately re-read the columns it crosses. Solving is a loop between the two directions, not a sweep through one and then the other.
Why these puzzles never need a guess
A random grid produces valid clues, but not necessarily a good puzzle — the clues might permit several different pictures, or require trial and error to narrow down. Both make for a frustrating solve.
So each board here is generated and then tested by a line solver: for every row and column it enumerates every arrangement still consistent with what is known, and keeps only the squares that all of them agree on. That is exactly the overlap-and-elimination reasoning a human uses, applied mechanically until nothing more can be deduced.
If the solver finishes the grid, the puzzle is solvable by logic alone — and, because the deduction never branches, that also proves the solution is unique. Boards that fail the test are thrown away and a new one is generated. The line under the grid tells you which kind you got.
If you are stuck
Look for the longest clue in the puzzle, not the line you were last working on. Long runs generate overlap, and overlap generates crosses in the perpendicular lines. Lines whose numbers nearly fill their length — 5 4 in a 10-wide row leaves only one free square — are almost always the cheapest place to restart.
Where nonograms come from
The puzzle was invented twice, independently, in Japan in 1987. Non Ishida won a window-art competition in Tokyo with a design made by switching building lights on and off in a grid, and turned the idea into a puzzle. In the same year Tetsuya Nishio published a near-identical format. The two designs converged so closely that both are credited.
James Dalgety, a British puzzle collector, coined the English name nonogram after Non Ishida when The Sunday Telegraph began running them in 1990. Nintendo licensed the format as Picross in 1995, which is why the two names are used interchangeably today. They also appear as griddlers, hanjie, and paint-by-numbers.
Deciding whether a given set of clues has a solution is NP-complete in the general case — the same complexity class as sudoku. Single lines, though, can be resolved in polynomial time, which is why the line-by-line approach both humans and the generator use is practical at these sizes.
Controls
- Click or drag: fill squares. Dragging paints whatever the first square became, so a stroke either fills or clears consistently.
- Right-click, or switch to ✕ mode: mark a square as ruled out. On a phone, use the ✕ button — it makes dragging out whole empty regions fast.
- Fading numbers: a clue greys out when that row or column matches it exactly. Grey does not guarantee the line is correct in the final picture, only that it fits its own clue right now.
- Undo: steps back one stroke, not one square.
- Sizes: 5×5 takes a minute, 10×10 five to fifteen, and 15×15 is a proper sit-down puzzle.