PDF to Excel

How PDF to Excel works

The file is read with a FileReader, rewritten in memory and offered as a download. This is why large PDFs are limited by your device's RAM rather than by an upload size limit. A 200 MB scan never crosses the network, so the slowest part is your own disk.

That is the whole architecture: static files, one script, no upload endpoint to leak or log. Close the tab and the work is gone.

Why a PDF has no table

In a spreadsheet, a cell knows its row and column. In a PDF there is no table object — only glyphs at coordinates, plus thin grey lines that happen to look like a grid. Converting to Excel therefore means inferring structure, not translating a format.

This tool infers two things: rows from text baselines that sit within a few points of each other, and columns from distinct x positions across the page. Everything else — merged cells, wrapped text, a header spanning three columns — is a judgement the reader has to finish in Excel.

Getting the tolerance right

The tolerance slider is the whole conversion in one control. Too tight and one visual row becomes five spreadsheet rows, because the PDF places every word at a slightly different y. Too loose and two separate lines merge into one row of nonsense.

Start at 6 points. Dense tables with a small font need 3–4; a report with generous spacing can take 10. The preview under the button shows the first rows exactly as they will land, so you can adjust before saving.

Where the numbers come from

pdf.js exposes each text run with a transform matrix. The fourth value is x and the fifth is y measured from the bottom of the page, so the baseline is flipped into screen space before grouping. Nothing is guessed from visual gaps alone: the coordinates are the ground truth.

The workbook is written by SheetJS as a real .xlsx — not a CSV renamed — so Excel opens it without a warning and keeps the sheet names.

How it compares with a converter that uploads

Every web converter solves the same problem the same way: the file goes to a server, a library extracts the text, a spreadsheet is written, and the result comes back as a download. That works, and it also means a bank statement or a payroll sheet sits on a machine you do not control for at least a few minutes.

This page does the extraction in the tab with pdf.js and writes the workbook with SheetJS. The practical differences are the ones in the table, and they are all consequences of that one decision.

When the output needs a second pass in Excel

Two mistakes are the tool's, not yours, and both are fixable in seconds. If each visual row arrives as two spreadsheet rows, the tolerance is too tight for that font — raise it by two points and convert again. If separate columns collapse into one, the gap between them is smaller than the clustering threshold, so lower it instead.

The other three are properties of the PDF itself. Merged cells become repeated text, because a PDF has no merge: the cell is simply drawn once and spans visually. Wrapped text splits into two rows when the second line sits a full line-height below the first. A header row is just text like any other — nothing marks it as a header, so turn on filtering in Excel and set it yourself.

What to check before trusting a big conversion

Numbers arrive as the strings they were in the PDF. A cell reading 1,234.50 is text, and Excel will not sum a column of text, so convert the column once with Text to Columns rather than retyping values. Currency symbols and thousands separators survive as characters.

For anything longer than a few pages, spot-check the last page: that is where a rotated table or a footer that overlaps the last row shows up. Rotated text (a landscape table inside a portrait page) is the one case this extraction cannot repair, because the x/y coordinates no longer line up with the visible grid.

Good to know

• Works offline once loaded, including on flights and in locked-down networks.
• No file size limit beyond your device's own memory.
• Free to use, with no account and no usage counter.
• Results are produced locally, so they are not logged or shared with anyone.

Questions