
Extract text from a phone photo of a receipt in your browser, keep the line breaks, and verify the numbers before they reach your books.
A shoebox of receipts is a bookkeeping problem. A camera roll of receipt photos is the same problem with better lighting. Typing every line by hand is slow and error-prone, but pasting unchecked machine output is worse, because a single misread digit lands in your records. Image to Text (OCR) reads the text out of a picture, and with a quick arithmetic check you can trust the result.
The clip below shows the drop, language and extract steps.
Take a photo of a till receipt from a made-up place, Cafe Luna. The printed lines are:
Drop the photo into the tool, leave the language on English and press Extract Text. Suppose the output comes back with the croissant price as 3.4S and the tax as O.91. Those two slips, S for 5 and the letter O for a zero, are classic misreads on small receipt fonts. Now run the maths yourself: 7.20 plus 3.45 is 10.65, and 8.5 percent of 10.65 is 0.905, which rounds to 0.91, giving a total of 11.56. The numbers prove the right values and expose the wrong characters. A receipt whose lines add up is almost certainly read correctly; one that does not add up needs a second look.
The reader is only as good as the image. A flat, evenly lit, in-focus photo gives far better text than a creased receipt in dim light. Before you shoot, flatten the paper on a dark table, stand square above it so the lines are not slanted, and let the camera focus. Thermal paper fades, so photograph old receipts promptly.
If the photo contains a lot of table around the receipt, crop it first with Resize & Crop so the receipt fills the frame. A sideways phone photo is usually fine: the tool applies the orientation stored in the photo, and if the first reading is weak it tries 90, 180 and 270 degree turns and keeps the best. A photo with no orientation data and very little text can still fail.
If the engine's overall confidence is under 70 percent, a yellow note shows the figure and suggests a clearer image or the right language. It is an average for the whole picture, so some lines may still be perfect, but treat the note as a request to check everything before relying on it. If it appears on a clean photo, you may have chosen the wrong language, which scrambles accented names and currency symbols.
OCR gives you text, not structure. To use the figures, copy the vendor, date, subtotal, tax and total into your expense list. Expense Tracker is a place to log them. Keep the original photo as the evidence, and keep the text file beside it, named for the date and vendor so both are easy to find.
Tax rules for what to keep, and for how long, depend on your country and your accountant. This tool reads text; it does not decide what is deductible, and extracted values are only as good as your check.
The reading happens in your browser and the image stays on your device. Only the language data comes from the network, and only on first use. There is no table export, and the tool will not split a receipt into columns, so you copy values yourself. Expect to review the result every time; a quick glance at the amounts is what turns a rough reading into a record you can stand behind. More utilities for documents and files are in the image tools category.
How do you keep receipts today: photos, a scanner app or a shoebox? Tell us which format you would like next on the contact page.