Open Invoice OCR to Excel and drop in as many invoices as you like — PDFs, JPGs or PNGs. When you press Process, the tool works through them one at a time: for a born-digital PDF it reads the embedded text layer with pdfjs (fast, no OCR), and for a scanned page or an image it renders and recognises it with tesseract.js. A shared rule-based engine then pulls the standard fields from each invoice. Every invoice becomes a single row in an editable Invoices grid — File, Invoice No, Invoice Date, Due Date, Vendor, Bill To, PO, Currency, Subtotal, Tax, Total and Terms — and any detected line items are collected on a second Line Items sheet, each tagged with its invoice number. You can correct any cell, see a running total of the invoice amounts, and then export a real two-sheet .xlsx workbook (built locally with exceljs), CSV files for each sheet, or copy the table straight into Excel or Google Sheets.

The whole pipeline runs client-side: pdfjs reads each PDF’s text layer, tesseract.js OCRs scanned pages and images, and exceljs builds the workbook — all from the app’s own bundle, with no server, no API, no AI and no CDN call for your documents, and it works offline once cached. The files you upload never leave the page; nothing is sent, logged or stored. That is the key difference from cloud invoice-to-Excel or “AI” extraction services, which upload your documents to their servers and sometimes use them to train models. Because your invoices stay on the device, the tool is safe for confidential or client invoices, and only the Excel or CSV you choose to export ever leaves the page.

The compiled spreadsheet uses the columns bookkeepers expect when invoices become a table. The first sheet (Invoices) has one row per invoice with the file name, invoice number, invoice date, due date, vendor, bill-to, PO number, currency, subtotal, tax, total and payment terms. The second sheet (Line Items) lists every detected line item across all invoices, each row carrying its invoice number alongside the description, quantity, unit price and amount, so you can do line-item spend analysis or import detail into accounting. In the .xlsx the amount columns are written as real numbers (so they sum and pivot correctly) with a styled, frozen header row; everything is editable before export, and a running total of the invoice totals is shown.

Accuracy varies per invoice: clean, born-digital PDFs extract very well from the real text layer, while low-quality scans or photos rely on OCR (which misreads characters) and unusual or multi-column layouts can defeat the rule-based matching — line-item tables in particular are notoriously hard. The tool does not truly understand the invoices; it matches labels and patterns, so it will sometimes miss a field or pick the wrong amount. Speed is the other honest caveat: because each invoice is OCR’d locally and the batch is processed one after another, many image invoices can take a while, and a very large batch may strain memory — process in smaller groups if needed. Every cell is editable for exactly this reason: use it to do the bulk of the data entry and to keep confidential invoices on your device, but review and correct the spreadsheet before relying on it for bookkeeping, tax or payment. For a single invoice with confidence scores and a raw-text view, use the companion Invoice Field Extractor.