Convert a bank statement PDF to CSV in Python.

01 / INSTALL AND GET A KEY

One dependency. Your first 100 pages are free, no card, at pennyocr.com/dashboard.

pip install pennyocr
export PENNYOCR_API_KEY=pk_live_...

02 / READ THE DOCUMENT

from pennyocr import PennyOCR

result = PennyOCR().ocr("statement.pdf")  # every page, markdown tables
text = result.text

03 / WHAT YOU GET BACK

Each page's transactions as a markdown table. Multi-line narrations are joined with <br>:

**Statement of account** 01-Sep-2026 to 30-Sep-2026

| Date | Narration | Chq/Ref No | Withdrawal | Deposit | Balance |
|---|---|---|---:|---:|---:|
| 01/09/26 | OPENING BALANCE |  |  |  | 12,500.00 |
| 03/09/26 | UPI/ACME STORES | 4521 | 1,250.00 |  | 11,250.00 |
| 05/09/26 | NEFT SALARY<br>SEPT |  |  | 45,000.00 | 56,250.00 |
| 09/09/26 | ATM WDL | 8812 | 5,000.00 |  | 51,250.00 |

04 / PARSE THE TRANSACTION TABLES

Collect the rows of every table that has a Balance column, across all pages:

import re

def money(s: str) -> float:
    return float(re.sub(r"[^\d.]", "", s) or 0)

def transactions(md: str):
    """Rows of every markdown table that has a Balance column."""
    rows, header = [], None
    for line in md.splitlines():
        line = line.strip()
        if not line.startswith("|"):
            header = None
            continue
        cells = [c.strip().replace("<br>", " ") for c in line.strip("|").split("|")]
        if set("".join(cells)) <= set("-: "):
            continue  # the |---|---| separator row
        if header is None:
            header = cells
        elif "Balance" in header:
            rows.append(dict(zip(header, cells)))
    return rows

txns = transactions(text)
print(len(txns), "rows")

05 / CHECK THE RUNNING BALANCE

This catches a misread amount, a dropped row or a column shift. Flag the statement for review if anything fails.

def check_running_balance(txns):
    """Every row must equal the previous balance - withdrawal + deposit."""
    bad = []
    for prev, row in zip(txns, txns[1:]):
        expected = money(prev["Balance"]) - money(row["Withdrawal"]) + money(row["Deposit"])
        if abs(expected - money(row["Balance"])) > 0.005:
            bad.append((row["Date"], expected, row["Balance"]))
    return bad

problems = check_running_balance(txns)
print("balance OK" if not problems else f"misread rows: {problems}")

06 / WRITE THE CSV

Only after the balance check passes:

import csv

with open("statement.csv", "w", newline="") as f:
    w = csv.DictWriter(f, fieldnames=txns[0].keys())
    w.writeheader()
    w.writerows(txns)
print("wrote statement.csv")