Free OCR Tools for Windows That Actually Read Receipts Corre

PDF-XChange Editor is the free OCR tool we actually trust on receipts, and it has nothing to do with the OCR itself. It is the only one on our shortlist that keeps the whole job on the PC, opens a scanned thermal receipt in under a second, and does not demand an account before it will look at the file.

We run a small company and do our own bookkeeping. Every month there is a shoebox of fuel slips, courier receipts, hardware invoices and the odd till roll that has faded to a pale grey smear. Typing them into the accounts package by hand is about the worst hour of the month, so we went looking for something free that would pull the text out reliably enough to check rather than retype. Here is what we found, including the parts that annoyed us.

Why receipts are a harder test than documents

Most OCR demos use a clean A4 page, black text on white, scanned flat. A receipt is the opposite. It is small, curled, printed on shiny thermal paper that fades, often photographed at an angle on a phone, and full of tiny numbers that all look alike. If an OCR engine gets the shop name and the total right, that is already a win. The VAT line is where it usually falls over.

So the test we settled on was simple. Take last month’s pile of real receipts, run them through each tool, and see how many totals came out correct without us touching them. No synthetic samples, no borrowed test PDFs. Real till rolls with coffee stains.

The free options worth trying

PDF-XChange Editor. This is a Windows-only desktop program and the free tier is unusually generous. Text editing, annotations, form filling, page management and OCR are all free with no watermark on the core features. Some advanced bits do stamp a small watermark on the free version, but OCR is not one of them. In testing it was also the quickest PDF viewer we tried, opening large files noticeably faster than Adobe Reader. For receipts that arrive as phone photos, you can drop the image in, run OCR, and get a searchable layer you can copy from. It is our default.

Sejda. This one surprised us. The free tier handles documents up to 200 pages and 50 MB with no watermark and no sign-up, which is far more room than a month of receipts needs. It does genuine text editing, not just annotation. There is a web version and a desktop app. The desktop app processes files locally, which matters if the receipts carry card numbers or client names. We keep it as a second opinion when PDF-XChange misreads a fuzzy total.

A browser-based converter as a fallback. When a receipt is too smudged for either of the above, we upload it to a free web OCR utility, let it produce a text file, and copy the numbers across. This is the option of last resort because it means the image leaves our machine. For a fuel slip that is fine. For anything with a card number on it, we do not.

What the OCR reliably gets and what it does not

After a few months of running receipts through these tools, the pattern is clear enough to plan around.

Shop name and date: almost always correct, even on faded paper.
Line items: usually right, occasionally mangled where two columns sit close together.
Totals: correct more often than not, but worth a glance every time.
VAT breakdown: the weak spot. Small fonts, tight spacing, and our local VAT rate printed as a single digit. We check this field by hand on every receipt, no exceptions.
Handwritten notes: forget it. If a driver scribbled a job number on the slip, you are typing that in yourself.

The honest position is that no free OCR tool will file your receipts for you. What a good one does is turn an hour of typing into ten minutes of checking. On a shoebox of thirty receipts that is a real saving. On two receipts it is not worth opening the program.

Scanning properly makes more difference than the software

We learned this the slow way. The same receipt that OCR mangles on a bad phone photo comes out clean when scanned properly, and you do not need expensive hardware for that.

Scan flat, not held at an angle in your hand.
Use greyscale or colour, not pure black and white. Thresholding destroys faint thermal print.
Use the highest resolution the scanner or phone offers, then let the OCR tool deal with the size. Downsampling first loses the small digits.
If the paper has curled, flatten it under a book for a minute before scanning. Curl is what causes those smeared lines at the bottom of a receipt.

None of this is clever. It is just the difference between feeding the engine something it can read and feeding it a guess.

The free tools we would not use for this

Some free PDF editors cap the number of documents per day on the free tier, which rules them out for a monthly batch. A two-document daily limit sounds workable until you have thirty receipts and a deadline. Others watermark the output, and a watermarked scan is no use as a bookkeeping record. And several require an account before you can run OCR at all, which for a two-person shop is one more login to keep track of.

We also tried the OCR built into a couple of free office suites. It exists, and it works on clean documents, but receipts are not its strength. For a typed invoice it is fine. For a curled till roll it is a coin toss.

How we actually use it

Our routine, for what it is worth. Receipts go into a single folder as scans or photos, named by date and vendor. Once a week we run the folder through PDF-XChange, check the totals and VAT lines against the paper, and copy the figures into the accounts package. Anything that failed OCR gets re-scanned at higher resolution and tried again. About one in ten still beats us, and those we type by hand. That is a fair trade for a free tool.

If you only take one thing from this, take the folder habit and the manual VAT check. The software is the easy part. The checking is what keeps the books straight, and no free OCR engine is going to do that for you.