OCR — Optical Character Recognition — converts the image of text in a scanned PDF into actual, selectable, searchable, and copyable characters. This guide explains how OCR works, when you need it, and how to apply it free in your browser.
Why Do Scanned PDFs Need OCR?
When you scan a paper document, the scanner creates a picture of the page — not a text file. A scanned PDF is essentially a JPEG or TIFF embedded in a PDF container. You can view it, but:
- You cannot select or copy text
- Ctrl+F (Find) returns zero results
- Screen readers cannot read it (inaccessible)
- Search engines cannot index its content
- You cannot fill in form fields or highlight text
OCR reads that image and produces a text layer — an invisible layer of recognized characters laid precisely over the image. After OCR, the document looks identical, but now it has real text underneath that you can interact with.
How to Apply OCR to a PDF Online (Free)
Reduce your PDF file size instantly. No software needed.
Using PDFlexa OCR:
- Go to PDFlexa OCR
- Upload your scanned PDF (drag-and-drop or click Choose a file)
- Select your document's language from the dropdown (24 languages supported)
- Choose output format:
- Searchable PDF — keeps the original appearance, adds invisible text layer
- Plain text (.txt) — extracted text only, no layout
- Click Run OCR
- Download the result
The processing runs in your browser using Tesseract.js. For large PDFs (30+ pages), processing takes a few minutes. No file is uploaded to a server.
Supported languages include: English, French, German, Spanish, Portuguese, Italian, Dutch, Polish, Russian, Arabic, Chinese (Simplified), Japanese, Korean, Hindi, Turkish, and 9 more.
When Is OCR Needed?
| Document type | OCR needed? |
|---|---|
| Paper document scanned to PDF | ✅ Yes — it is an image |
| Photo of a document taken with a phone | ✅ Yes |
| PDF created from a Word/Excel/Google Docs file | ❌ No — already has text layer |
| PDF exported directly from software (invoices, receipts) | ❌ No — already has text |
| Old scanned archive document | ✅ Yes |
| PDF where text is blurry or garbled | ⚠️ OCR may improve it |
| A form PDF you cannot fill in | ✅ OCR + Fill PDF will help |
A quick way to check: try selecting text on a page. If you can highlight a word, the PDF already has a text layer. If the cursor only creates a selection box around a region (like selecting an image area), it is a scanned image and needs OCR.
How OCR Works (Under the Hood)
Turn any PDF into an editable Word document in seconds.
- Page rendering — each page of the PDF is rendered at high resolution (300+ DPI equivalent) as an image
- Pre-processing — the image is de-skewed, de-noised, and converted to grayscale
- Text detection — the OCR engine identifies regions that look like text (columns, lines, words)
- Character recognition — each character region is matched against a trained model for the selected language
- Text layer creation — recognized characters are placed at their detected positions over the original image
- PDF reconstruction — a new PDF is created with the original image plus the invisible text layer
The quality of OCR depends on: scan resolution (higher is better), document quality (clear print vs. faded ink), language match, and whether the text is typed or handwritten. PDFlexa uses Tesseract — the most widely used open-source OCR engine, trained on 100+ languages.
Tips for Better OCR Results
Scan at 300 DPI or higher. Most consumer scanners default to 150 DPI, which is enough to view but produces poor OCR results. Use 300 DPI for text, 600 DPI for very small print.
Scan in grayscale or black-and-white. Color scans are larger and don't improve OCR quality for standard text documents. Grayscale is the sweet spot.
Ensure the page is flat and straight. Curved pages (from book binding), skewed scans, or shadows from a phone camera all reduce accuracy. Most flatbed scanners handle this well; phone-camera scans are the most problematic.
Choose the correct language. Tesseract uses language-specific trained models. A German document processed with English OCR will have poor results on umlauts (ä, ö, ü). Always match the language.
Large caps and printed text OCR better than handwriting. Standard OCR is designed for printed text. Handwriting recognition is a separate (harder) problem that Tesseract handles poorly.
OCR Accuracy: What to Expect
| Document quality | Expected accuracy |
|---|---|
| Clean, typed, flat-scanned at 300 DPI | 98–99% character accuracy |
| Standard office scan at 150 DPI | 92–96% accuracy |
| Old document with ink fading | 80–90% accuracy |
| Handwritten notes | 50–75% (varies widely) |
| Multi-column academic paper | 90–95% (layout detection harder) |
| Non-Latin script (Arabic, Chinese, Korean) | 85–95% with correct language |
For critical documents (contracts, legal records), always review the OCR output for errors — especially proper nouns, numbers, and dates.
OCR vs. PDF to Word: What's the Difference?
| OCR PDF | PDF to Word | |
|---|---|---|
| Output | Searchable PDF (same appearance) or .txt | Editable .docx |
| Best for | Making a scan searchable/accessible | Editing the content in Microsoft Word |
| Layout | Original layout preserved perfectly | Layout may shift in Word conversion |
| Use case | Archive, search, accessibility | Re-use and edit the content |
If you want to edit the scanned content in Word, run OCR first and then use PDF to Word — OCR gives the converter actual text to work with rather than blank image regions.
Frequently Asked Questions
Does OCR change the appearance of my PDF? No. The OCR text layer is invisible — it sits behind the image of the page. The PDF looks identical before and after OCR. The only difference is that text is now selectable and searchable.
Can OCR read handwriting? Standard OCR is trained for printed text and is unreliable on handwriting. Results vary widely depending on how clearly and consistently the handwriting is formed. For important handwritten documents, manual transcription is more reliable.
How many pages can I OCR at once? PDFlexa's browser-based OCR supports up to 50 pages per upload. For longer documents, split the PDF into parts, OCR each part, then merge the results.
Why does the OCR take a long time? Each page is processed individually: rendered at high resolution, analyzed, and recognized. A 20-page document may take 1–3 minutes in the browser. The processing runs on your device — speed depends on your CPU and memory.
Is OCR free? Yes. PDFlexa OCR runs entirely in your browser using open-source Tesseract.js at no cost. No account required, no daily limit on the OCR tool.
Digitizing paper records regularly? See our PDF tools for accountants or PDF tools for researchers for related workflows.