Docs tower · floor

OCR: turning a scan into text you can use

ocr — optical character recognition — is the step that turns a photograph of a page into something a computer can read. It is the hidden requirement behind half the complaints in this tower: the converter that produced gibberish, the PDF that will not search, the form that could not be filled.

It is also, quietly, one of the great free wins of the last few years. The recognition built into an ordinary phone is now good enough for most everyday jobs, costs nothing, and never sends the page anywhere.

What recognition adds to a scan A scanned page is a picture: nothing in it can be selected, searched or copied. Recognition reads the shapes and writes a layer of real text underneath, which is what makes the page searchable. Recognition Before After A picture of a page Nothing can be selected. Nothing can be searched. Copying gives you an image. A screen reader finds nothing. The same picture, plus text The image is unchanged. Underneath it sits a layer of real characters — searchable, selectable, and sometimes wrong.
Recognition does not redraw the page. It adds a transcript beneath it, and that transcript is a guess — which is why the last step of this job is always reading it back.

The words people use track how much they already know. Image to text is what somebody types the first time, usually holding a phone. Ocr pdf and pdf ocr come from people who have met a scanned document and understood that it is a picture. Pdf to text converter means they want a plain file rather than a document, and searchable pdf is the most precise of all — that person knows exactly what is missing. Ocr software is a purchase decision, and text recognition is the name of the field itself.

The floors below cover what is already on your devices, how to make a scanned PDF properly searchable, the specific things recognition still gets wrong, and the cases — handwriting, bad light, unusual alphabets — where expectations need adjusting.

Where to start

Five ways in. The first is free and takes about four seconds.

“I need the words out of this photo.”
Start at already on your devices
“My scanned PDF will not let me search it.”
That is making a pdf searchable
“The recognised text is full of mistakes.”
Go to when it gets it wrong
“It is handwritten.”
The honest answer is in the hard cases
“I have two hundred pages to do.”
That case is in making a pdf searchable

Already on your devices

Free, offline in most cases, and documented by the companies that build it. For a single photograph this wing is the entire answer.

On an iPhone or iPadLive Text: point at the words in a photo and copy them. Apple documents it for iPhone XS, XR and later on iOS 15 or newer.Being built
On a MacApple: “In Photos, you can use Live Text to copy and use the text that appears in a photo” — hover, Control-click, Copy Text.Being built
On AndroidThe recognition built into the camera and the photo app, and where it differs by manufacturer.Being built
With Google DriveGoogle converts PDFs and photos up to 2 MB, and warns that “lists, tables, columns, footnotes, and endnotes are not likely to be detected”.Being built
Image to text, in secondsThe fastest route from a picture to words you can paste, on each device.Being built

Making a PDF searchable

A different job from copying a paragraph: the recognised text has to be written back into the file, underneath the picture, so the document itself becomes searchable.

What searchable pdf meansThe invisible text layer, how to tell whether a file has one, and why it changes everything downstream.Being built
Ocr pdf on each systemThe free and paid routes that add a text layer rather than just showing you the words.Being built
Pdf to text converterGoing from scanned pdf to text in a plain file rather than a document, and the tools that do only that.Being built
Many pages at onceRecognising a stack of scans without doing them one at a time.Being built
Ocr software worth havingThe point at which a dedicated program earns its price, which is later than most people think.Being built

When it gets it wrong

Recognition is a guess expressed with total confidence. These floors are about knowing where the guesses go wrong so you can check the right places.

The characters that get confusedOnes and lower-case L, zeros and O, commas and full stops — and why numbers are the worst case.Being built
Layout that defeats itColumns, tables and footnotes, which Google itself says are not likely to be detected.Being built
Checking the resultA short routine that catches most errors without reading every word.Being built
When accuracy really mattersInvoices, prescriptions, legal documents — and why automatic recognition is a draft, not a record.Being built

The hard cases

Where the technology is still genuinely limited. This wing exists so nobody spends an afternoon fighting something that was never going to work.

Handwriting to textWhat works, what does not, and why your own notes are harder than a printed page.Being built
Bad photographsAngle, shadow and focus, and why Google asks for text “at least 10 pixels high”.Being built
Other alphabetsWhere recognition is excellent, where it is patchy, and how to tell before you start.Being built
Old and damaged documentsFaded print, typewriter pages and photocopies of photocopies.Being built

What this tower will not do

It will not tell you recognition is accurate. It is a transcription made by a machine that cannot read, and on numbers it is wrong often enough to matter. Every floor here ends by telling you what to check.

It will not rank the ocr software yet. That verdict needs the same set of pages — clean, awkward and terrible — run through each program with the results published, which is a test.

And it will not send a private document to a website when the phone in your pocket does the job offline. On this floor more than any other, the free built-in answer is also the private one. What holds instead is simple: the recognition on your own phone is free, offline, and enough for most of this floor.

Where this page got its facts

  1. Apple on using Live Text to copy text from a photo in Photos on Mac — support.apple.com, read 21 August 2026.
  2. Apple on copying and translating text from photos on iPhone and iPad, and the devices required — support.apple.com, read 21 August 2026.
  3. Google Drive on converting PDFs and photos to text — the size limit, the tips, and what is not detected — support.google.com, read 21 August 2026.

Written by Alberto Gulotta

Founder and editor of AI Tools Primer, writing from Palermo, Italy. Thirty-five years of taking computers apart, starting with a Commodore 64 — the long version is on the about page.

Something wrong on this page? Write to aitoolsprimer@gmail.com and it gets fixed.

Independence and limits

No affiliate links and no paid placements anywhere on this site. Nobody pays to appear here, and no company has seen this page before you did.

This is general information, not professional advice. Where a page touches money, health, safety or the law, it names its source and the date it was read — and your situation may still differ. See the privacy page and the cookie policy.