Sudoku Scanner

Sudoku Scanner: scan a puzzle from a photo, screenshot, or camera

Our free Sudoku scanner turns a printed Sudoku, magazine puzzle, newspaper grid, PDF, screenshot or camera photo into an editable digital board. Instead of typing 81 cells by hand, you load an image, line up the four corners of the puzzle, scan the grid, correct any uncertain digits, and send the result to the Sudoku Solver or the Sudoku Helper.

It runs in your browser. It can take an uploaded image, a pasted screenshot, or your device camera. Once the image is loaded, corner handles let you mark the exact outside edge of the grid. The tool then straightens the image, splits the corrected grid into 81 cells, reads the visible givens, and shows the result in a digital grid you can edit before using it anywhere else.

A good scanner needs more than optimism about OCR. Sudoku photos vary: some are tilted, some have shadows, some use thin fonts, and screenshots come with grid lines and highlights. The workflow here is deliberately practical — scan quickly, flag the cells it is unsure about, make correction fast, and then hand the puzzle to the right tool.

How the scanner reads a grid

The four corner markers tell the scanner which quadrilateral in your image should become a square 9x9 grid. A photo rarely captures a puzzle perfectly square: it may be taken at an angle, the paper may curve, or the puzzle may fill only part of the frame. Dragging the handles to the outside corners of the Sudoku is what makes perspective correction possible.

After that the image is straightened and divided into 81 cells, and each cell is measured to decide whether it holds a printed digit or nothing. Recognition works by template matching: the scanner renders the digits 1 to 9 in thirteen different fonts, and compares each cell against every template, scoring how closely they correlate. The best-scoring digit wins, and the score becomes a confidence value.

That confidence is what drives the review step. Any cell scoring below 70% is highlighted in the grid, and hovering it shows the actual figure. The status line after a scan also counts them — "3 cell(s) flagged as low confidence" — so you know how much checking is waiting before you look.

Why the review step is not optional

Even good template matching misreads a digit sometimes. A 1 can look like a 7 in some fonts. A 5 can look like a 6 when the photo is soft. A handwritten 9 can look like a 4. One wrong given is enough to make a solver report a contradiction, so review is part of the workflow rather than an optional extra.

The obvious defence is to hunt for repeated digits, and it is worth doing. But it is not enough on its own, and this is the case that catches people out:

That is why the real check is comparison, not inspection. Read the scanned grid against the original row by row and confirm the same givens sit in the same places. Do not check only the highlighted cells: a highlight means the scanner was unsure, and the dangerous errors are the ones where it was confident and wrong. Empty cells should stay empty.

How to use the Sudoku scanner

1. Load an image. Use Upload Image to choose a file or drag one in, paste a screenshot, or switch to Use Camera and press Capture.
2. Place the corner markers. Drag the four handles onto the outside corners of the Sudoku grid, not the page border.
3. Press Scan Puzzle. The tool straightens the grid, reads the cells, and shows a digital version.
4. Review the result. Check the highlighted cells first, then compare every row against the original.
5. Correct the grid. Click a cell and use the number pad; Erase clears one, Clear empties the board.
6. Send it onward. Send to Solver for the full answer, Send to Helper for hints, or Export the puzzle as text.

Corner placement is the part that matters most. Handles slightly inside the grid will crop the outer digits; handles outside it will pull in page text or borders. Zoom in if you need to and put each handle exactly where the thick outside grid lines meet.

Taking a photo that scans well

Use good light, keep the page flat, and avoid glare from glossy paper. The puzzle does not have to fill the frame, but each digit should be clearly legible. A slightly angled photo is fine because perspective is corrected, though extreme angles make digits harder to read, and a shadow falling across the grid blurs the contrast between printed digits and empty cells.

On a phone, hold steady and tap the puzzle to focus before capturing — a crisp photo from slightly further away beats a blurry close-up. When scanning from a book, press the page flat enough that the grid lines stay straight; if the middle curves into the binding, back off a little and use the corner markers to crop instead.

Screenshots and digital puzzles

Screenshots usually scan better than camera photos: the lines are sharp and there is no perspective to correct. Paste one directly or upload the file. You do not need to crop first, because the corner markers isolate the puzzle from the rest of the image, though a screenshot with nothing overlapping the grid will still do better.

Digital puzzles often carry coloured highlights, selected cells, pencil marks or adverts near the grid. If the scanner reads a pencil mark as a given, correct it before sending the puzzle on. The scanner is meant to capture the starting givens, not every note or interface highlight in the source image.

What to do after scanning

Once the grid is right, choose the tool that matches what you want. Send to Solver gives the completed answer, a validity check, and confirmation that the puzzle has exactly one solution. Send to Helper is the better choice for learning, because it works interactively and gives hints without revealing the whole grid.

You can also Export the puzzle as an 81-character string, read from the top-left cell to the bottom-right with 0 or a dot for blanks. That is the format to keep if you want to save the puzzle, share it, report a problem, or move the same grid between tools — and Import takes the same string back, which is a quick way to reload a puzzle you scanned earlier without scanning it again.

Privacy and local processing

The scanner works entirely in your browser, and that is stronger than it sounds: there is no upload step anywhere in it. The page reads your file locally, or takes camera frames locally, does the straightening and the recognition on a canvas in the tab, and never sends the image anywhere. Nothing about your photo reaches a server, because there is no code that could send it.

That matters because Sudoku photos come from personal notebooks, puzzle books, screenshots and documents you may not want to hand over. As with any browser tool, still apply common sense with sensitive images: crop to the puzzle when you can, keep unrelated personal information out of the frame, and close the page when you are done on a shared device.

Troubleshooting a poor scan

If the result looks wrong, check the corner markers first. Most failed scans come from inaccurate corners rather than from digit recognition. Make sure each handle sits on the outside Sudoku border and that the four are in the right order. Then look at the image: if the grid is blurry, dark or half in shadow, a better photo will do more than any amount of correcting.

If only a few digits are wrong, fix them by hand — that is faster than rescanning. If many are wrong, try a cleaner image or a screenshot. If the scanner fills cells that should be blank, the source probably has pencil marks or background texture in them; clear those before solving, because the solver can only work with the givens you hand it.

The accuracy checklist

Before exporting or solving, spend a minute on this. Count the givens in each row and compare with the source. Check that the first and last rows have not been cropped by the corner markers. Look for repeated digits in each 3x3 box — two of the same digit means one of them is wrong. Confirm that blank cells are genuinely blank, especially if the printed puzzle has pencil marks or small notes.

Then do the one thing that catches what the rest cannot: read each row against the original and check the givens match position for position. As the second diagram above shows, a misread can survive every duplicate check ever devised and still leave a puzzle with no answer. Scan first, verify second, solve third — in that order the tool is dependable.

Scanner or manual entry?

Manual entry is reliable but slow. Typing a whole puzzle can take longer than solving the opening moves, and it is easy to drop a digit into the wrong row. The scanner gives you a first draft in seconds and moves your attention from typing to checking, which is the more useful place for it.

Manual entry still wins sometimes. If a photo is poor, or the puzzle is handwritten in an unusual hand, typing is quicker than correcting. The practical rule: scan when the image is clear, correct when only a few cells are uncertain, and type when the source is too messy for recognition to help. It is useful for casual players moving a newspaper puzzle online, for teachers turning printed worksheets into checkable grids, and for anyone who finds a large high-contrast digital grid easier to read than a small printed one.

Limits: handwriting, variants, and what it does not read

The templates are printed digits in ordinary fonts, which is exactly why printed puzzles scan well and handwriting does not. Neat handwriting may work, but handwritten digits vary far more than any font does, and the scanner has no examples of them to compare against. Expect to correct more cells. Thick pencil marks, crossed-out numbers and tiny candidate notes can all be mistaken for givens.

The scanner reads a standard 9x9 layout and the digits in it, and nothing else. A 9x9 puzzle with extra colours or highlights can usually be scanned and then corrected, but the extra rules of a variant are not captured at all: jigsaw regions, killer cages, diagonal constraints and larger grids such as 16x16 all need their own solvers. If your puzzle has extra rules, keep them separately and treat the scan as the digit layout only.