About

Sheet from PDF turns the tables inside PDFs into clean spreadsheet data — bank statements, invoices, receipts, reports. Your browser does the work, so the file never leaves your device.

Why it exists

A bank statement is one of the most sensitive documents most people ever handle. It carries your name, your address, your account number, your salary, and a map of everywhere you have spent money. An invoice carries who you deal with and what you pay them. Neither belongs on somebody else's server.

Almost every mainstream converter asks you to upload it to a server first — often capping the free tier at a fraction of a real document, and holding the file for a while afterwards. That is a strange thing to ask of people who just want their transactions in a spreadsheet.

The technology to avoid it has existed for years. So this tool reads the PDF locally, inside your browser, and never sends it anywhere. If you want to check that claim, open the page and then turn off your internet connection — the table is still read and the rows are still built, because none of that ever needed a connection.

What makes it different

Generic converters extract text and hand you whatever they find. Statements are not generic: the debit and credit columns are each empty on most rows, long merchant names wrap onto a second line, and the date format differs by country. Getting a usable table out means understanding the document, not just reading it.

This tool also adds the transactions up and compares them against the running balance column on every row. If a line does not reconcile, you are told which line. That check is deliberately unglamorous, and it is the reason the output can be trusted rather than merely exported.

How it works

• The PDF is opened locally and its text is read with its position on the page.
• Lines are regrouped into rows, and column boundaries are found from the layout.
• Headers are matched to date, description, debit, credit and balance.
• Every row is checked against the running balance.
• You correct anything wrong in the editable grid, then export CSV.

Being honest about limits

Scanned or photographed PDFs are not supported yet, because there is no text layer to read — that needs optical character recognition, which is on the roadmap. The tool tells you when it thinks it is looking at a scan rather than guessing.

Support for individual banks is being expanded. Where a bank's layout is not yet well-supported, the column mapper lets you correct it, and the tool says so rather than pretending otherwise.

Who runs it

Sheet from PDF is a small, independent project. It is paid for a conversion at a time, or by a monthly or yearly pass. There is no free tier and no advertising. Nothing about your documents is used to train anything, sold, or shared — because nothing about your documents reaches us in the first place.

Get in touch

Feature requests, banks that do not convert well, and corrections are all welcome at hello@example.com. Please see the contact page before sending an example, and never send us a real statement.