How to turn a scanned table into an Excel spreadsheet
Published September 26, 2026
Retyping a printed table is the least rewarding work a keyboard offers. If the table exists on paper — a price list, a bank statement, a lab report, a class roster — you can photograph it, scan it, and let OCR deliver the contents as spreadsheet rows. The result is rarely pixel-perfect, but it turns an hour of typing into a few minutes of checking.
How a picture becomes rows and columns
OCR first reads the words and, crucially, remembers where each word sits on the page. Table extraction then uses the geometry: words on the same horizontal band form a row, and larger gaps between word groups mark column boundaries. Scandut applies exactly this approach — each line of the page becomes a spreadsheet row, split into cells at the visible gaps, one worksheet per scanned page.
The workflow
Scan the page as usual — a straight, corrected, document-mode scan makes column gaps crisp and consistent. Run Extract text (OCR) on the results screen, then download the Excel (.xlsx) file. Open it and you will find the table roughly in shape: rows in order, labels in the first columns, numbers to the right.
Shoot the table as flat and squarely as you can. When a photo is heavily tilted, columns that align on paper drift diagonally in the image, and the gap detection starts merging neighboring cells. The perspective correction step fixes most of this — another reason to adjust the scan area corners carefully on table documents.
Five-minute cleanup in Excel
Expect three kinds of touch-ups. Merged cells: two columns separated by a narrow gap may land in one cell — split them with Excel’s Text to Columns. Misread digits: 0/O and 1/l/I confusions cluster in low-quality print, so scan the totals column against the image. Number formatting: OCR delivers text, so select numeric columns and convert them to numbers before summing. For most tables this is genuinely a five-minute pass.
What to expect — and when to retype after all
Clean printed tables with clear column spacing convert very well. Tables with vertical rules but almost no spacing, cells with wrapped multi-line text, and handwritten entries convert poorly — the geometry OCR relies on simply is not there. If a table is small (a dozen cells), retyping is honestly faster. The technique earns its keep on the long ones: transaction histories, inventory lists, anything with fifty-plus rows.