How to Read a Scanned or Phone-Photo Invoice Automatically
Paper invoices and phone photos are the part of bookkeeping that never gets easier. A clean PDF usually carries its text inside the file, so software can read it directly. A photo or a scan is only an image of a page, and the software has to work out what it is looking at before it can read anything. Angles, shadows, folds, faded print and a thumb over one corner all change what comes out the other side.
Why photos and scans are harder than PDFs
A PDF generated by a billing system keeps its text as text. A phone photo of a receipt does not: the letters are pixels, the page may be tilted, and a crumpled thermal receipt can lose the lines at the bottom where the total sits. Small vendor logos, tax IDs printed in a footer and handwritten notes on a purchase order add more places where a reading can go wrong.
The same invoice can look perfectly clear to a person, who fills gaps from context, and still be hard for a machine. That is why the useful question is not only "did it read the file?" but "which fields did it read with confidence, and which ones should a person check first?"
What to check after the data is extracted
Automatic reading is a starting point, not a finished entry. Cifra extracts the vendor, tax ID, invoice number, date, subtotal, tax, total, currency and category from each document. Whatever the tool, these are the checks worth making on every document:
- The vendor name and tax ID belong to the company that issued the invoice.
- The date and invoice number match what is printed on the page.
- Subtotal plus tax equals the total. Cifra compares these figures and warns before export if they do not match.
- The currency and category fit how your books are organized.
Each extracted field comes with a confidence score. The least confident one is flagged in amber, so the reviewer looks at it first. A person reviews and fixes anything with one click, and nothing is exported without that review. This is general guidance and does not replace the judgment of a qualified accountant or tax professional.
How Cifra handles PDFs, photos and scans
Cifra accepts a vendor invoice, expense receipt or purchase order as a PDF, a phone photo or a scan, up to 12 MB per file. A document can have any number of pages, and it counts as one document regardless of length. The review screen is the same for all three input types, so paper and digital files do not need separate handling.
Results can be exported as CSV, JSON or Excel. Exports are file downloads. Uploaded documents are not stored. Every account, including the free one, gets an API key for sending documents without opening a browser. The Business plan adds a forwarding email address: invoices sent there arrive already read and waiting for review.
Plans are priced in US dollars per month. The Free plan costs $0 and includes 10 documents a month, with no card required. The Professional plan costs $29 and includes 250 documents. The Business plan costs $89 and includes 1,000 documents.
Try it free
If you handle a steady stream of paper and phone-photo invoices, the simplest test is your own pile. Run a batch of real documents through the free plan and review the amber fields and any total mismatch warnings yourself before you decide anything else.