OboScan extracts invoice data from PDFs, scanned documents, and photos, converting them into accurate, structured Excel-ready data—all processed securely on your own machine.
Every accounts desk loses the same hours to the same job: “someone has to type all these bills in.”
Open the invoice and the fields are already filled — number, date, vendor, GST numbers, every line row, the taxes and the total. What used to be five minutes of keying per bill becomes a glance and a click — and a month’s pile becomes one Excel report.
Seconds instead of minutesTally bills, Zoho templates, government Excel grids, export invoices printed in two currencies, freight notes from abroad — OboScan picks the right way to read each one by itself. And a layout that beats it can be taught in a minute.
PDFs, scans and photosQuantity × rate, sub-total + tax = grand total — every figure is cross-checked before you see it. A bill that doesn’t add up is flagged, not quietly passed on. OboScan would rather admit doubt than invent a number.
Honest about what it readDrop the whole stack into the window — or point the command line at a folder — and OboScan reads every file in turn and writes one Excel workbook. A bad file in the middle doesn’t stop the run; it becomes a row with the reason, so 49 good bills still land while number 50 is being chased.
Measured on a folder of 16 digital invoices: 161 line items into one workbook, in about two seconds.
Invoices — a row per bill with its totals, status and how it was read. Line items — a row per item, linked back to its file. Extra fields — the labelled values your vendors print that no fixed column covers.
The header row is frozen, filters are switched on and money columns are formatted as numbers — so you can sort by vendor or total the moment it opens, without cleaning anything up first.
Every row carries OK, CHECK FAILED or FAILED, plus the exact check that didn’t add up. You review the handful that need eyes instead of re-reading all fifty.
Some bills are printed in a way no reader can guess. Instead of waiting for a software update, you open the bill in Teach mode and click each value on the page. OboScan works out which printed label sits beside it and saves that as a rule for this vendor. From the next bill onwards, those fields are pinned.
You click the value; OboScan finds the printed label beside it and remembers “the number that follows Bill No”. So when the vendor shifts the layout or the margin next month, the field still lands.
A taught vendor is read from the page geometry alone — on a PC with no model installed, and on a scanned copy of the same bill too. Teach it here, and it still works on the counter machine.
If a marked field would make a bill that already added up stop adding up, OboScan rolls that value back to what it read itself. A rule is allowed to repair an invoice — never to spoil a good one.
There is no template to build before you start. Open the app and read a bill.
Drag a PDF, a scanned page or a phone photo into the window — one bill, or the whole month at once. Multi-page bills are fine, and you can zoom the preview while you check.
A digital PDF is read exactly from its own text in about a second. A scan goes through built-in OCR. Only a photo nothing else can read is handed to the AI.
Fields, line items and totals appear with a badge saying how they were read, and every arithmetic check is shown. Take one bill as JSON, or the whole batch as an Excel report.
Only if a layout defeats the reader: click its fields once in Teach mode and that vendor is solved for good — see Teach mode.
Smart document processing that uses AI only when needed, ensuring faster and more cost-effective PDF data extraction.
Everything runs on one ordinary Windows PC — no server, no graphics card, no monthly bill.
For a digital PDF the figures are copied from the file’s own text, not recognised from a picture — so a 7 can never become a 1.
A folder of bills becomes a single workbook — Invoices, Line items and Extra fields — with frozen headers, filters and formatted money columns.
An unreadable or 0-byte file is recorded as a row with its reason, so the other forty-nine bills still reach the report.
Description, quantity, rate, discount, tax and amount for each row — including rows that wrap over two lines.
Batch no., MRP, HSN, warranty — any extra column the vendor prints is kept under its own printed heading instead of being thrown away.
Quantity × rate per row, sub-total + tax = grand total. A printed grand total is never overwritten — only ever confirmed or questioned.
GSTIN for both sides, IGST/CGST/SGST, several tax rates on one bill, discounts, round-off and tax-inclusive pricing.
₹, $, RM and more — including export bills that print two currencies side by side without mixing up which total is which.
Mark the fields once on a sample bill and that vendor is pinned for good — and no model is needed to use the rule afterwards.
Scans and photos are handled out of the box. Nothing extra to install, and it only trusts an OCR result that passes every check.
Each result carries a badge — exact parse, OCR, your rule, or AI — so you know at a glance which numbers deserve a second look.
The bill sits beside the data at 150%, 200% or 300%, with a full-screen view — so you can check a figure without opening another app.
No cloud account, no upload, no telemetry. Pull the network cable out and OboScan carries on working exactly the same.
Unzip the folder and double-click. It lives in the system tray with its own window — nothing for IT to maintain.
An invoice carries your prices, your vendors and your margins. That is exactly the kind of paper that should never be posted to somebody else’s server — so OboScan simply doesn’t have the option.
OboScan was built and tested on a modest desktop — a 6th-generation i5 with 8 GB of RAM and no graphics card.
Send us a few of your real supplier bills, or let us walk through OboScan on a call. No account, no upload, nothing to install to find out.