PDF to Excel
Detect the tables on a PDF's pages, check and correct the grid on screen, then export only the pages you keep as an .xlsx workbook or CSV.
🔒 Processed in your browser — never uploaded. Tables are rebuilt from text position, so the result is text-faithful, not pixel-perfect.
Drop PDFs here or click to choose
Several at once is fine.
Turn PDF tables into an editable spreadsheet
Drop a PDF and this tool reads its text layer, clusters the words into rows and columns by where they sit on the page, and shows you the grid it found — click through the pages one at a time to check them — then writes the pages you keep to a real .xlsx workbook that opens in Excel, Google Sheets, LibreOffice or Numbers, with one worksheet per kept page or every page stacked into one sheet when a single table runs across several. The same corrected grid downloads as a CSV file too, and a whole folder of PDFs can go in at once — every page of every one is detected in the same pass, and each sheet is then named for the file it came from. Pages that read as ordinary prose rather than a table start unticked, so a report full of paragraphs does not turn into a workbook full of one-column sheets. It is built for the everyday job of getting a table out of a report, invoice or bank statement without retyping it, so you can sort, total and chart the numbers straight away.
Runs in your browser — nothing is uploaded
Text extraction and the table reconstruction both happen locally with pdf.js, so your PDF never leaves your device — safe for financial and internal documents. Because a PDF stores no table structure, the columns are inferred from geometry: clean, left-aligned ruled tables come across accurately, while complex layouts, merged cells or multi-column pages may misalign. Because that guess can be wrong, you can correct it before you download: a slider sets how wide a horizontal gap has to be before it starts a new column, so a cell that was split down the middle or two columns that were run together are fixed in one drag; adjacent columns can be joined by hand; and any cell can be typed over. Moving the slider re-reads the page, so make those hand corrections last. The result is text-faithful rather than pixel-perfect, and a scanned PDF with no selectable text is offered on-device OCR here, after which detection runs again on what was read.
Frequently asked questions
How well does it detect tables?
It works best on clean, ruled tables where each column is left-aligned. Rows are found by matching text baselines and columns by the gaps between text, so a tidy table comes across accurately. Complex layouts — merged cells, nested tables or multi-column pages — can misalign. You do not have to find that out in Excel: the detected grid for every page is shown before you download, a mis-split or over-merged column is fixed with the column-gap slider or the join button, and any cell can be corrected on the spot.
Does it work on a scanned PDF?
Yes. A scanned PDF is only images of pages, so when no selectable text is found the tool offers to read the words on your device first — the recogniser is downloaded once and nothing is uploaded — and then detects the tables in what it read, without leaving the page, reading it as English or as Korean + English, whichever you pick. That offer is made for one document at a time; in a run of several, a file with no text on any page is named, so you can drop it on its own.
Is my PDF uploaded anywhere?
No. pdf.js reads where each piece of text sits on the page, in this tab, and the .xlsx is written by SheetJS, which this site sends only when you press download. Your file never leaves your device.