How to Turn Scanned PDFs Into Readable Text

How to OCR Scanned PDFs Into Readable Text

Spot an image-only scan, run OCR yourself, clean up the recognition errors, and export TXT or EPUB for focused reading. About a 9 minute read.

You open the PDF and try to select a sentence. Nothing highlights. You search for a word you can see printed on the page; the reader reports zero matches. So you pinch, zoom, and pan across a fixed two-page scan spread, and by the bottom of the page you have absorbed approximately nothing.

This is a mechanics problem, not a character flaw. A scanned PDF is a photograph of a page. It contains pictures of text rather than text, which is why nothing can search it, reflow it, read it aloud, or display it at a comfortable size. The fix is optical character recognition, OCR: software that looks at the image and reconstructs the words.

This guide covers how to read scanned PDFs from end to end. The short version: run the OCR step yourself, fix the predictable errors, and export the result as text or EPUB. No step needs an account, and no file leaves your machine.

Scan or real text? Three checks before you OCR

Not every PDF needs this treatment. Many already carry a real text layer, and running OCR on those wastes time for no gain. Three checks settle the question in under a minute.

1. Try to select text. Click and drag across a paragraph. If nothing highlights, you are looking at an image.
2. Search for a common word such as "the". A word you can see on the page returning zero hits means there is no text layer to search.
3. Check the file size. Scanned pages are large images. A 200-page scan-only book often runs past 20 to 50 MB, while a text PDF of similar length stays small.

One more signal: switch on Read Aloud, or a screen reader, and let it run. On a true scan it announces nothing, because there is nothing to announce.

If any check fails, the file needs OCR before comfortable reading is possible at all.

Why your OCR choice is a privacy decision

The reflex move is to search for a free online converter. Adobe Acrobat online, Google Drive OCR, and the various online-convert sites all work, and they share one requirement: you upload the entire document before a single word is extracted. None of them are villains, and for a public flyer the trade may be fine. But an upload is a copy of the document on someone else's server, whatever the privacy policy promises about deleting it later.

Now look at what people actually OCR. Tax forms. Medical records. Unpublished manuscripts. Confidential work documents. The files that most need OCR are disproportionately the files you least want copied anywhere, which makes the local route more than a stylistic preference.

Local tools remove that risk by never creating it. The workflow below works on the copy of the file already on your machine, and the results are written alongside it. It is the same logic as local-first reading: no account, no ads, no feed, and your library stays on your device. The principle covers the whole workflow, not just the conversion step.

How to OCR a scanned PDF with ocrmypdf and Tesseract

For full-length scans, the standard route is ocrmypdf, a popular wrapper that runs the Tesseract engine against PDF files and puts the recognized text back into the file.

Install it through your package manager:

# macOS
brew install ocrmypdf

# Debian/Ubuntu
sudo apt install ocrmypdf tesseract-ocr

Then point it at your scan:

ocrmypdf scan.pdf scan-ocr.pdf

The output keeps the original page images and lays an invisible text layer behind them, so the PDF becomes searchable and selectable. To convert the scanned PDF to text as a standalone file, run a PDF-to-text extractor over the result. pdftotext from the poppler tools does it in one line:

pdftotext scan-ocr.pdf book.txt

Input quality matters more than any setting. Crisp scans produce clean text; soft, skewed, or gray-washed scans produce noise. Start from the best source copy you can get before you blame the engine.

Every command above runs locally. The step where you convert scanned PDF to text is precisely the step you should not outsource to a server.

The zero-install option: Apple's Live Text and Vision

If you read on an iPhone, iPad, or Mac, you already own an OCR engine. Live Text, built on Apple's Vision framework, recognizes text in images and lets you select and copy it, with nothing to download. Processing happens on-device, so the same privacy guarantee holds.

It is the right tool for short documents and pages captured on the go: photograph the page, select the text, paste it where you read. For a 300-page scan, nobody is doing that page by page. Long jobs belong to ocrmypdf.

The OCR cleanup checklist

No engine is perfect on the first pass, and the errors follow predictable patterns. Work through this list before you export:

• Character confusions. "rn" often comes back as "m", so "modern" can turn into "modem", and a lowercase "l" swaps freely with the digit "1".
• Ligatures. The joined "fi" and "fl" pairs sometimes emerge as garbled or wrong characters after recognition.
• Line-end hyphenation. Words broken across lines stay broken once the text reflows, leaving fragments scattered through paragraphs.
• Page artifacts. Running headers, footers, page numbers, and scan noise get captured as if they were body text, and they repeat on every page.

The fix is pattern work: a few search-and-replace passes on the high-risk cases, then spot checks of words that look suspicious. Ten minutes of cleanup here saves frustration in every reading app that touches the file later.

Convert scanned PDF to text: import it into a focus reader

The most portable home for cleaned OCR output is plain TXT or Markdown. Both reflow, both open anywhere, and both keep your options open for the EPUB route below.

This is also the honest place to introduce SReader, limits first. Its PDF import does not perform OCR; the product description says it plainly: "PDF import extracts text; it does not preserve the original page layout, perform OCR, or convert PDF files to EPUB." That is why you ran the steps above before arriving here. TXT, Markdown, and EPUB files import directly, and reading progress is saved on the device so you resume where you stopped.

Once the text is in, you choose how to read it. Word mode shows one word at a time. Line mode keeps a single complete highlighted line centered, a reading rail, while surrounding text stays visible for context. Book mode is an ordinary page. Pace is adjustable in all of them: faster when the material is easy, slower when the ideas need room. Settings, bookmarks, and progress stay on the device, with no account involved. If you want to see the rail before installing anything, there is an interactive demo that runs in the browser.

Scanned document to EPUB: building a reflowable book

Loose text is fine for articles and notes. For a book-length scan, EPUB is the better target, because a scan locks words into fixed page images while an EPUB reflows. Type size, screen size, and the original two-up layout stop mattering; the text adapts.

Three free tools cover the path. Pandoc turns the cleaned text file into an EPUB in one command, pandoc book.md -o book.epub. Sigil opens the result for hands-on editing when the structure needs work. Calibre manages the finished file, organizes your library, and handles any further conversions.

Sort out chapter breaks and structure before you export, so navigation works in any reader. The same care applies when you start from PDFs that already have text; that layout-preserving route has its own guide in How to Convert PDFs to EPUB Without Wrecking the Layout.

The finished EPUB imports straight into SReader on iPhone, iPad, Mac, or Android, with reading progress saved on the device.

When OCR struggles: honest limits

Standard OCR is strong on clean modern print and weak in places you should know about before you blame the tool or yourself.

Handwriting remains largely out of reach for these engines. Old or irregular typefaces and aging print degrade accuracy badly. Multi-column layouts, common in academic papers and old newspapers, can interleave columns in the wrong order and produce sentences no human wrote. Low-DPI photos of pages yield weaker results than proper scans.

The fix is usually upstream: rescan or re-photograph at higher quality rather than fighting bad output. Some archival material will need manual cleanup no matter which engine you point at it. And if your real problem is dense academic PDFs rather than broken scans, that has its own treatment once a text layer exists; see How to Read Dense Academic PDFs Without Visual Overwhelm.

From locked images to comfortable reading

The whole chain fits in one line: detect the scan, OCR it locally with ocrmypdf and Tesseract or Apple's built-in tools, clean the text, export TXT or EPUB, and read at your own pace on any device.

The payoff is largest for readers who were locked out entirely. Students and professionals get searchable, reflowable text instead of pinch-and-zoom squinting. That is a claim about reading comfort, not a medical one.

The privacy thread held for the entire trip. Nothing left your device, and no account was required at any step. Try SReader and bring your newly readable text along.