Quick Answer
Receipt OCR reads a photo of a paper receipt and turns it into structured data: merchant, date, line items, subtotal, tax, and total. Drop in JPG, PNG, or WebP images and this free tool runs the Tesseract OCR engine inside your browser — the photos are never uploaded; only the engine and its English language file download once from a CDN. Every field stays editable, and you can export everything to CSV or JSON with no account and no scan limits.
How It Works
Add receipt photos — drop JPG, PNG, or WebP images onto the dropzone, browse for files, or load the synthetic sample receipt
Watch the queue: each image is preprocessed (grayscale plus contrast stretch) and OCR-read in your browser, with per-file status and an overall progress bar
Review each parsed receipt — merchant, date, total, tax, currency, and payment are editable, with OCR confidence and fallback warnings on every card
Fix the line items: edit names, quantities, and prices, add or remove rows, and open the raw OCR text to verify anything suspicious
Export the batch to CSV with a totals summary row, download single receipts as JSON, or send a total to the Expense Splitter
Key Facts
- Queue many receipts at once — each file shows queued, reading, done, or error status with live progress
- Images are preprocessed in-canvas before OCR: grayscale plus a 2%–98% percentile contrast stretch
- OCR runs on tesseract.js (English) inside your browser — receipt photos are never uploaded
- Extracts merchant, date, total, subtotal, tax, currency symbol and code, payment method, and line items
- Date parsing covers 2024-08-12, 08/12/2024, 12 Aug 2024, Aug 12 2024, and 8/12/24, with day-first detection when the first number is over 12
- Every parsed field is editable; line-item rows can be edited, added, or removed
- Export all receipts to CSV with a totals summary row, or download any single receipt as JSON
- A synthetic sample receipt runs the full OCR and parsing pipeline so you can test without a photo
- Free and unlimited — no account, no watermarks, no per-scan limits
Frequently Asked Questions
How does the parser turn OCR text into structured fields?
The OCR pass returns raw lines of text; the parser then applies transparent heuristics. The merchant is the first letter-rich line that is not boilerplate (phone, web, thank-you, receipt headers). Dates are matched against five common formats, and day-first versus month-first is decided by whether either number exceeds 12. The total is the last line labeled total, amount due, or balance due that carries a currency amount; when no label exists, the largest amount on the receipt is used and the card is flagged so you can verify it. Line items are rows that end in a price with at least two letters of text before it, with quantities read from "2 x" prefixes, while totals, tax, change, and card-number rows are excluded. Payment method comes from brand names (Visa, Mastercard, Amex) and masked card-number patterns.
Is this receipt scanner free?
Yes — free with no scan limits, no account, and no watermarks. Expense apps that OCR receipts typically gate the feature behind a subscription (for example, SplitMyExpenses charges for its AI features via its $3.99/month Premium tier); this tool runs the same kind of extraction at no cost because all the work happens in your browser.
Is my data private?
Your receipt photos never leave your device. Decoding, preprocessing, OCR, and parsing all run in this browser tab, and exports are generated locally. The only network traffic is the one-time download of the OCR engine itself and its English language file from a CDN — after that, scanning works even if you go offline in the same tab.
Which image formats can I scan?
JPG, PNG, and WebP work in every modern browser, and you can queue a whole batch at once. HEIC photos from iPhones decode only in browsers with native HEIC support (Safari); elsewhere the file is marked as an error with a hint to convert it to JPG or PNG, and the rest of the queue continues normally.
How accurate is the extraction?
OCR quality depends on the photo: flat paper, even light, and sharp focus give the best results, while crumpled receipts and dark café lighting degrade them. Every receipt card shows the OCR confidence score, and fields that came from a fallback (for example, a total inferred as the largest amount) carry a visible warning so you know exactly what to double-check. All fields and line items stay editable before you export.
What can I do with the results?
Export every receipt to a single CSV (file, merchant, date, total, tax, currency, payment, and a JSON column of line items, plus a totals summary row that counts receipts and sums totals), or download any individual receipt as JSON. There is also a one-click link to the Expense Splitter if you want to divide a receipt across a group.