One Flat Import Table, Not a Messy Sheet
The download is one import table: a single CSV of transactions. Source page and validation columns sit beside each row instead of living on extra Excel sheets.
STATEMENT FILE CONVERTER
Turn native or scanned bank statement PDFs into clean CSV rows via AI OCR. Built-in balance checks help you verify dates and entries before export.
Drop your bank statement here
PDF · JPG · PNG · large files process in the background
Choose FileThe download is one import table: a single CSV of transactions. Source page and validation columns sit beside each row instead of living on extra Excel sheets.
Text-layer statements convert from the existing characters. Scanned pages and JPG/PNG photos use local OCR. Blurry pages still need a human pass.
When the statement prints opening and closing totals, they are compared. Flagged rows are there so you can check them before you import the CSV.
Drag and drop a checking or savings statement PDF into this browser tab. No account is required until you hit the free page allowance. Password-protected files are rejected. Credit card PDFs have their own converter.
Native pages keep their text layer. Scans use local OCR. Remaining paid pages may use cloud AI repair when local recognition is unreliable. Keep the tab open on large files.
Check dates, descriptions, amounts, and validation flags, then download .csv. Confirm a few rows in your importer before you post.
Statements are designed for reading, not database feeds. Stacked payee details, multi-line descriptions, and repeating headers are where generic scrapers merge or drop rows.
Some bookkeeping tools reject XLSX and only map CSV headers. Use this page when the next step is a delimited file, not a workbook.
Paper prints, faxes, and smartphone photos have no reliable text layer. Financial OCR in this tab transcribes what it can. Unclear fee lines still need review.
Opening and closing checks catch math drift. They do not prove every description is the right payee. Fields are quoted UTF-8 with a byte-order mark; you still map columns in the destination app.
If the importer names CSV and the bank only gave a PDF, this page is the conversion step—not a finished set of books.
A reviewed CSV is easier to map into another system than a three-sheet workbook or a scanned PDF.
Local extraction runs first. Eligible paid sessions may send redacted page text for cloud AI repair, not statement photographs.
UTF-8 CSV with a byte-order mark opens cleanly in Excel and still works as a plain table in other tools.
The Excel converter is a separate page. This one writes the delimited file the next tool asked for.
If the bank portal only gives a PDF image, OCR plus review is the conversion step before you download CSV.
The download is a quoted UTF-8 CSV with a byte-order mark. You still map Date, Description, and Amount (or Debit/Credit) in Xero, Sage, QuickBooks, or another importer. This page is not a QBO Web Connect file and we do not claim every ERP accepts the headers unchanged.
Yes. Native PDFs use the text layer. Scanned pages and JPG/PNG photos use financial OCR in this browser. Blurry or angled pages need extra review. Free includes 20 pages per day. Keep this tab open on large files.
Date, Description, Debit, Credit, Amount, and Balance when the statement prints one, plus Source Page, Source, Sign Source, Confidence, Validation Status, Validation Notes, and Source Line. There is no separate check-number column.
When opening and closing totals are printed, they are compared to reconstructed amounts. A match is not proof that every payee is correct or that a fee line was not skipped. Review flagged rows against the PDF before you import.
Standard conversion reads the file in this browser tab. The original PDF is not uploaded to a conversion server. Closing the tab ends active processing; saved browser history and downloaded files may remain. Cloud AI repair uses remaining paid pages, not the 20 free pages. Account, card, IBAN, and government-ID numbers are masked except the last four digits. Statement photographs are not sent. That inference is not stored by BankExcel and is not used to train models.
Comma-separated, quoted fields, UTF-8 encoding with a byte-order mark so Excel on Windows still opens it as Unicode. Password-protected PDFs are not supported—unlock them in a reader first.
Read the PDF in this browser, review reconstructed rows, then download CSV. Confirm a few amounts in your importer before you post them.
Convert to CSV now