How to OCR PDFs: Extract Text from Scanned Documents

Open a scanned contract, try to search for a clause, and nothing happens. Try to copy a paragraph, and you get nothing. That is because a scanned PDF is not really a document—it is a photograph of one. To the computer, the page is a flat image with no idea that those black shapes are letters. OCR, or Optical Character Recognition, is the technology that teaches the computer to read those shapes and turn them back into real, searchable text.

This guide explains how OCR works, when you need it, how to get the most accurate results, and how to convert a scanned PDF into a searchable one using BananaPDF OCR PDF. By the end you will be able to make any pile of scans fully searchable and ready to reuse.

What Is OCR?

Optical Character Recognition analyzes the image of a page, detects the regions that contain text, and matches the shapes to characters in a language model. The result is machine-readable text that can be searched, selected, copied, and indexed. Crucially, a good OCR process does not replace your page—it keeps the original scan as the visible layer and tucks the recognized text invisibly behind it. The document looks identical; it just gained a brain.

Scanned PDF vs. Text PDF: Know the Difference

There are two very different kinds of PDF that look the same on screen:

• Text-based PDF: Created digitally (exported from Word, a browser, or design software). The text is real—searchable and selectable from the start.
• Image-based (scanned) PDF: Created by a scanner, camera, or photo-to-PDF conversion. The "text" is just pixels in an image. Searching finds nothing.

The quick test: try to select a word with your cursor. If you can highlight it, the PDF already has text. If the cursor refuses to grab anything, you have an image-based PDF that needs OCR.

Why OCR Matters

• Searchability: Find a name, date, or clause across hundreds of archived pages in seconds.
• Copy and reuse: Pull quotes, figures, or addresses out of a scan without retyping.
• Accessibility: Screen readers can voice OCR'd text for visually impaired users—an image cannot be read aloud.
• Editing: Extracted text can be moved into editors or spreadsheets for further work.
• Compliance and archiving: Searchable archives are far easier to audit and retrieve than image dumps.

Step-by-Step: OCR a Scanned PDF

1. Open the tool. Go to /tools/ocr-pdf and upload your scanned PDF (or an image-based file).
2. Let it process. The engine scans each page, detects text regions, and recognizes the characters.
3. Download the searchable PDF. The output looks identical to the original but now carries a real text layer.
4. Verify. Open it, press Ctrl+F (Cmd+F on Mac), and search for a word you can see on the page—it should jump straight to it.

For the highest accuracy, start from the best-quality scan you have rather than a compressed or low-resolution copy.

How to Get the Most Accurate Results

OCR accuracy depends far more on the source image than the software. To maximize it:

• Scan at 300 DPI or higher. Low resolution blurs character edges and confuses recognition.
• Keep pages straight. Skewed or rotated scans hurt accuracy—straighten them first.
• Maximize contrast. Dark text on a clean white background reads best; faint or yellowed pages perform worse.
• Avoid shadows and glare. When photographing instead of scanning, use even lighting and hold the camera parallel to the page.
• Use standard fonts. Common printed fonts are recognized far more reliably than decorative or unusual ones.

If your scan is crooked or pages are out of order, fix that with Organize PDF before running OCR.

Starting From Photos or Images

Often the "scan" is actually a phone photo. The workflow is simple: first convert your images into a PDF with JPG to PDF, then run that PDF through OCR PDF to make it searchable. This two-step path—capture, then recognize—turns a stack of photographed pages into a clean, searchable document.

The Limits of OCR

OCR is powerful but not infallible. Set expectations correctly:

• Handwriting is hard. Standard OCR targets printed text; cursive and messy handwriting usually fail.
• Poor scans yield poor text. Garbage in, garbage out—a blurry fax will produce errors.
• Complex layouts can confuse it. Multi-column pages, tables, and mixed graphics may need review.
• Always proofread critical text. For legal or financial figures, verify recognized numbers against the image.

Even at 99% accuracy, a 1,000-word page can contain a handful of errors—worth a quick check on anything important.

Common OCR Use Cases

Legal: Make scanned contracts and case files searchable so a clause can be found instantly across thousands of pages.

Accounting: Turn scanned invoices and receipts into searchable records, then extract amounts and dates for bookkeeping.

Research and academia: Convert scanned journal articles and old books into searchable, quotable text.

Business archives: Digitize filing cabinets into a searchable knowledge base instead of a folder of unsearchable images.

Personal admin: Make scanned IDs, warranties, and medical records findable by keyword.

An Efficient OCR Workflow

1. Capture or scan at high quality (300 DPI, straight, well lit).
2. If starting from photos, convert with JPG to PDF.
3. Straighten and reorder pages with Organize PDF if needed.
4. Run OCR PDF to add the searchable text layer.
5. Compress the result for storage or email if the file is large.
6. Split or merge the searchable output as your archive requires.

Languages and Multilingual Documents

Modern OCR supports dozens of languages, and recognition quality improves dramatically when the engine knows which language to expect. The reason is that OCR uses a language model to resolve ambiguous characters—deciding whether a shape is a zero or the letter O, for example—based on what words are plausible in that language.

A few practical points for multilingual work:

• Match the language: Recognizing a French document with an English model introduces accent errors; choose the correct language where the tool allows it.
• Mixed-language pages: Documents that switch between languages are harder; expect more errors at the boundaries and proofread those sections.
• Non-Latin scripts: Arabic, Chinese, Cyrillic, and others are well supported by leading engines but benefit even more from high-resolution, clean scans.
• Special characters: Currency symbols, accented letters, and ligatures are common error spots—verify them in financial or formal text.

Verifying and Using OCR Output

OCR gets you most of the way, but the last step is yours. After processing, spend a moment confirming the result is fit for purpose:

1. Spot-check searches: Search for several distinctive words from different pages to confirm the text layer covers the whole document, not just the first page.
2. Proofread critical figures: For contracts and invoices, read the recognized numbers against the image—a misread "8" as "3" in an amount matters.
3. Copy a paragraph: Paste it into a text editor to see the raw recognized text and catch systematic errors.
4. Re-OCR if needed: If accuracy is poor, the source scan is usually the cause—rescan at higher resolution rather than blaming the engine.

Once verified, your searchable PDF can feed archives, search systems, and accessibility tools with confidence.

Turn Scans Into Searchable Documents

A scanned PDF you cannot search is a dead end—a picture pretending to be a document. OCR brings it to life, adding the invisible text layer that makes every word findable, copyable, and accessible. Start from a clean, high-resolution scan, run it through a reliable OCR tool, and verify the result, and you transform unusable images into a genuine, searchable archive.

Upload your next stack of scans to BananaPDF OCR PDF and make them searchable in moments. Pair it with JPG to PDF for photographed pages and Compress PDF for tidy storage, and your paper archive finally behaves like the digital documents it should be.