What OCR is, and why it matters where it happens

What OCR does

OCR, optical character recognition, finds the lines and letters in a picture of a page and turns them into text. That text is what lets you search a scan for a word, copy a reference number out of it, or have it read aloud.

It is a careful guess, not a copy. A crease, a smudge, handwriting or a dim room can make it read a letter wrongly, so check an amount or an account number against the page itself before you rely on it.

On the phone, or on a server

Many scanning apps send the photo to a server to be read, which means the page passes through someone else's computer before you can search it. A recent iPhone can do the reading itself: Apple's text recognition runs on the phone, so a page does not have to leave it to become searchable.

It also works with no connection at all: in airplane mode, abroad, or in a cellar with the old files.

Languages and accents

Recognition reads better when it knows which languages to expect, and a household's paperwork is often in more than one. Accents matter twice: once when the page is read, and again when you search and leave them out. Why searching “ucet” should still find “účet” explains the second.

How Pocket Aurora Papers does it

Papers reads every page on your iPhone with Apple's on-device text recognition, in English, Czech, German, French and Spanish, and search works with no connection. A page does not need to be uploaded to be read.

If you sign in, the page and its text are also kept in our cloud in the EU, encrypted while stored with keys we manage, so your other devices can search it and read it aloud. If you turn on suggestions, which are off until you do, Papers sends the text, never the pages, to an AI model in the EU. The privacy page has the details.