Free PDF to Excel converter — tables to CSV in your browser

Every free PDF-to-Excel converter on the first page of results wants you to upload the file. Think about what that means for the file most people are actually converting: a bank statement, a supplier ledger, a payroll run. It goes to a machine you know nothing about, run by a company you have not heard of, and comes back watermarked.

None of that is necessary. A PDF that was made by software — exported from an accounting system, a bank’s portal, a supplier’s invoicing tool — already contains its text. Getting the table back out is geometry, and your browser can do geometry.

Drop a file above. It stays on your machine.

What it does

• Finds the rows. Two cells in the same row of an invoice are almost never at the same height — a smaller font, a superscript, and the baselines differ by a fraction of a point. Rows are clustered rather than matched exactly, so one row stays one row.
• Finds the columns. By looking for the vertical gutters that no text ever crosses, not by lining up where each cell starts. That distinction is the whole game on an amount column: 9.00 and 12,340.00 are right-aligned, so they begin four characters apart and end together.
• Rejoins split words. PDFs break text wherever the kerning changed, so Riyadh can arrive as Riy + adh. Left alone, that shifts every column after it — on that one row only, which is the kind of error you find three weeks later.
• Makes the figures numeric, if you ask. SAR 1,250.50 becomes 1250.50 so Excel will sum the column. Invoice numbers, dates and VAT numbers are deliberately left alone: rewriting 1,205 to 1205 does not tidy a reference, it corrupts it.

The one control worth touching

Column detection has exactly one number in it that depends on the document: how wide a gap has to be before it counts as a column break rather than a wide space. There is no value that is right for every layout, so instead of five named modes you get the number itself, as a slider, with the result redrawing as you drag.

Columns that should be separate came out merged? Drag left. One column split into two? Drag right. It re-grids from the text it already read — no re-upload, because there was no upload.

Bank statements, specifically

Statements are the most common thing put through a converter and the most likely to come out badly, for one structural reason: the description column contains free text of wildly varying length, and on some rows it runs far enough right to close the gutter that separates it from the next column.

Two things help. Convert one page at a time rather than all pages stacked, because a statement’s header and footer blocks have a different column structure to its body and they drag the detection around. And nudge the spacing slider a little right, since statement columns are usually generously spaced and the extra tolerance costs nothing.

What it cannot do

Scans. A PDF from a flatbed scanner or a phone is a photograph wrapped in a PDF container. There is no text in it to recover. This tool says so rather than handing back an empty grid that looks like an empty document — if that is your file, the document scanner will straighten and sharpen the image, and the document analyzer will read the figures off it with a model. That second step does send the document somewhere, and its page says so plainly.

Judge the numbers. It reports what the page said. If the layout was read wrongly the file will be wrong quietly, which is why there is a preview above the download button rather than a straight-to-download form. Look at it.

Remember anything. Close the tab and the file is gone. That is the same property as the privacy promise, seen from the other side.

Where the rows go next

Once you have columns, the spreadsheet opens them without an add-in. If what you converted was a quarter of sales or purchase invoices, the VAT return summary takes that CSV directly and gives you the ZATCA return box by box — including the rows whose VAT does not add up.