Bank statement OCR

Bank statement OCR for scanned statements and phone photos

A scanned statement has no text to copy: it is a picture of a page. OCR reads the dates, descriptions and amounts from that picture and turns them into rows you check beside the scan, then download as Excel, CSV or JSON.

Page 1 free, no sign-up to see it; just an email to download. The whole document is one payment: $3 up to 5 pages, $9 up to 25, priced by page count above that. No subscription.

Several statements? Drop up to 100 at once: one payment for all of them, priced by their total pages.

Output
Structured output
Speed
Fast preview
Exports
Excel, CSV, JSON
Sample result Ready

Source documentBANK_STATEMENT_APRIL_2026.pdf

Bank Statement
Statement identity
Bank name Account number Statement period Currency
Transaction rows
Transaction date Description Debit Credit
Balances
Opening balance Closing balance Running balance Totals

OutputJSON

{
  "document_type": "bank_statement",
  "bank_name": "Northstar Bank",
  "account_last4": "4821",
  "statement_period": "2026-04-01 to 2026-04-30",
  "opening_balance": 18420.25,
  "closing_balance": 21378.44,
  "transaction_count": 87
}

What gets extracted

What OCR reads from a scanned statement

The account details, every transaction row and the opening and closing balances, kept in the statement's own columns so the running balance can be followed from row to row.

Statement identity

Bank name Account number Statement period Currency

Transaction rows

Transaction date Description Debit Credit

Balances

Opening balance Closing balance Running balance Totals

From scanned page to checked rows

  1. Upload the scan, photo or image PDF

  2. OCR reads the page into rows while you wait

  3. Compare each row with the scan beside it and fix any misread figure

  4. Download Excel, CSV or JSON

Scans that copy and paste can't handle

  • Statements printed, stamped and scanned back in

  • Phone photos taken at an angle or with a shadow across the page

  • Photocopied or faxed pages with faint digits

  • Handwritten notes in the margin beside the transactions

FAQ

Common questions

What is bank statement OCR?
Optical character recognition: software reads the characters in an image of a statement, such as a scan or a photo, and turns them into text and numbers. pdf2text then puts them into transaction rows.
How accurate is it on scanned statements?
On our accuracy benchmark, whose pages include scans and phone photos, pdf2text got 98% of numbers correct across 82 hand-checked real-world pages. Faint or skewed pages are where mistakes happen, so check those rows against the scan.
Do I need OCR for a digital PDF statement?
No, a digital PDF already has text. pdf2text reads both kinds the same way, so you don't need to know which one you have.

Read a scanned statement into rows you can check

A scanned statement has no text to copy: it is a picture of a page. OCR reads the dates, descriptions and amounts from that picture and turns them into rows you check beside the scan, then download as Excel, CSV or JSON.