Extract Text from Scanned PDFs — Free Online OCR

You scanned a contract, a receipt, or a stack of old meeting notes. The result is a PDF — but not the useful kind. You cannot search it, copy from it, or paste it anywhere. The text is locked inside an image. OCR (Optical Character Recognition) is how you break it out. This guide walks you through the process using a free browser tool that never uploads your files to a server.

When You Need PDF OCR

Not every PDF needs OCR. If you can highlight and copy text from a PDF, it already contains real text data — OCR will not help. But if selecting text does nothing, or if "Select All" grabs the entire page as one block, you are looking at a scanned image disguised as a PDF. These are the common cases:

• Scanned contracts and legal agreements — especially older ones sent via fax or physical mail
• Paper receipts and invoices scanned for expense reports or tax filing
• Academic papers from library scanners, particularly pre-2010 publications
• Government forms (immigration, permits, tax returns) scanned at a service counter
• Handwritten meeting notes or whiteboard photos captured as PDFs

How to Extract Text: Step by Step

The entire process takes under two minutes. No account creation, no email required, no software installation.

• Open batch-printer.com/tools/pdf/ocr in any browser — Chrome, Safari, Firefox, or Edge. Works on phone and tablet too.
• Drop your scanned PDF onto the upload area. Choose your OCR engine: Standard (PP-OCR, fast, good for printed text) or Premium (Florence-2, 223 MB download, better for complex layouts and mixed content).
• Click "Run OCR." The tool processes your document entirely in the browser. When finished, copy the extracted text or download a searchable PDF.

Multi-page documents work too — every page is processed sequentially. For a typical 10-page scanned document, expect about 15 to 30 seconds with the Standard engine and slightly longer with Premium.

Tips for Better OCR Results

OCR accuracy depends heavily on the input quality. A clean 300 DPI scan of a typed document will return near-perfect results. A blurry phone photo of a crumpled receipt will not. Here is how to get the best output:

• Scan at 300 DPI or higher. Below 200 DPI, small characters like commas and periods become ambiguous to the engine.
• Keep the document flat and well-lit. Shadows across text confuse OCR engines into seeing characters that are not there.
• Align the document straight. Even a 5-degree rotation can reduce accuracy by 10-15%, especially on dense tables.
• Use the Premium engine (Florence-2) for documents with mixed content — tables alongside paragraphs, stamps over text, or handwriting mixed with print.
• For non-Latin scripts (Japanese, Korean, Arabic, Thai), the Standard PP-OCR engine often performs better because it was specifically trained on multilingual datasets.

Standard vs Premium OCR Engine

The tool offers two engines because no single approach works best for everything. Here is when to use each:

Standard (PP-OCR): Fast, lightweight, excellent for clean printed documents in any language. Loads instantly. Best for: typed contracts, printed receipts, book scans, government forms with clear text. Supports 18 languages out of the box.

Premium (Florence-2): A 223 MB AI model that downloads once and runs locally. Slower to initialize but significantly better at understanding document layout — it knows where columns end, where headers start, and how to handle text that wraps around images. Best for: complex reports with tables and charts, documents with stamps or signatures overlapping text, academic papers with footnotes and multi-column layouts.

Both engines run entirely in your browser. Neither sends your document to any server. The Premium model is cached after the first download, so subsequent uses load faster.

Privacy: Why Client-Side OCR Matters

Most online OCR tools work by uploading your document to a remote server, processing it there, and sending the text back. That means your scanned contract, medical record, or financial statement travels through someone else's infrastructure. Even services with "we delete after processing" policies still had your data on their servers, however briefly.

Client-side OCR eliminates this entirely. The OCR engine runs inside your browser tab. Your PDF never leaves your device — not to our servers, not to any cloud, not anywhere. For documents containing personal data (tax returns, medical records, contracts with confidential terms), this is not a nice-to-have feature. It is the only responsible approach.

You can verify this yourself: open your browser's network tab (F12 → Network), run an OCR job, and watch. Zero outbound requests carrying document data.

What OCR Cannot Do

No OCR tool is perfect, and being upfront about limitations saves you time:

• Heavily damaged documents — water stains, torn edges, or ink bleed make characters unrecognizable to any engine.
• Very low resolution scans — below 150 DPI, the engine cannot distinguish between similar characters (0 vs O, 1 vs l, 5 vs S).
• Complex handwriting — OCR works reasonably well on neat block handwriting but struggles with cursive or highly personal handwriting styles.
• Decorative fonts and stylized text — wedding invitations, certificates with calligraphy, or logos with artistic type are not reliably recognized.
• Mathematical formulas and chemical notation — specialized OCR tools exist for these; general-purpose OCR treats them as garbled text.

For documents that fall into these edge cases, your best bet is to OCR what you can, then manually correct the problem sections. Even partial OCR saves significant retyping time.

Frequently Asked Questions

Can OCR handle handwritten text? Partially. Neat, printed handwriting (block letters) works reasonably well with the Premium engine. Connected cursive or highly stylized handwriting remains unreliable across all OCR tools, not just ours.

Which languages are supported? The Standard engine supports 18 languages including English, Korean, Japanese, Chinese (Simplified and Traditional), Thai, Vietnamese, Arabic, Hindi, German, French, Spanish, Portuguese, Italian, Dutch, Polish, Romanian, and more. The Premium engine is optimized for English and common European languages.

Is it really free? Yes. No account, no trial period, no per-page limit. The tool runs in your browser using your device's processing power, so there is no server cost to pass on to you.

Can I OCR a multi-page document? Yes. Every page is processed in sequence. A 50-page scanned document works — it just takes proportionally longer. For very large documents (100+ pages), consider splitting the PDF first, then OCR each part.

What output formats are available? You get two options: plain text (copy-paste ready) and a searchable PDF where the recognized text is layered invisibly over the original scan. The searchable PDF preserves the visual appearance while making the content findable with Ctrl+F.