Photograph a page of your handwritten khata / bahi khata and get a clean, editable Excel ledger back — Date, Particulars, Debit, Credit and Balance in proper columns.
PDF · JPG · PNG — free to try, no signup needed
Want higher volumes, table refinement and your full document library? Create a free account — 500 points included.
Lakhs of Indian shops, traders and societies still keep their daily accounts in a paper khata. The book works — until you need totals, a party-wise summary, or your CA asks for the year in Excel. Typing years of handwritten entries by hand takes weeks and invites errors.
DataExtract was built for exactly this document. Our handwriting pipeline combines Google Vision with a GPT-class model that understands ledger structure: it finds the entry rows even when lines are crooked, separates debit from credit amounts, keeps running balances, and reads mixed Hindi/Bengali/English entries. Every low-confidence word is highlighted so you know exactly what to double-check before you rely on the numbers.
Use your phone camera in good light. JPG, PNG and PDF up to 50 MB work.
The tool is preset for khata books — the AI knows to expect a handwritten ledger.
Get an .xlsx ledger with one row per entry, plus Word, PDF or JSON if you prefer.
Khata pages are automatically routed to our strongest handwriting model — smudged ink, ruled pages and cursive entries included.
Output is structured as Date | Particulars | Debit | Credit | Balance, not a wall of text. Opening and closing balances are detected too.
Mixed-script entries are read natively — party names in one script, amounts in another are handled on the same line.
Words the AI is unsure about are flagged with confidence scores, so a 5-minute review replaces a week of retyping.
Yes — khata pages are processed by a handwriting-specialised AI pipeline, not plain OCR. Accuracy depends on the handwriting and photo quality; every uncertain word is highlighted for review, so you always know what to verify.
Yes. The engine reads Devanagari, Bengali and Latin scripts, including pages that mix them — a common pattern where party names are in Hindi and amounts in English numerals.
A clean .xlsx with columns for Date, Particulars, Debit, Credit and Balance, one row per ledger entry. You can also download Word, PDF or structured JSON.
Your first pages are free — no card, no signup for a quick try. A free account adds 500 points (about 13 khata pages), and paid plans start at ₹399/month for regular digitisation.
Files are processed over encrypted connections (256-bit TLS) and are never shared or used to train models. You can delete documents from your library at any time.
Yes — scan the book to a multi-page PDF and upload it once; each page becomes rows in the same export. Bulk digitisation is exactly what paid plans are priced for.