OCR PDF
Turn a scanned PDF into one you can search, select and copy, or pull the text out as a .txt file. The words are read on your device, so your scan never leaves it.
- Files stay on your device. Nothing is uploaded.
- Free, no sign-up, no daily limit.
- Works with Wi-Fi off once the page has loaded. Check it yourself
Updated by Khokhar Sab
How do I make a scanned PDF searchable?
Press Choose a PDF, pick the language of the text, and leave Save as on Searchable PDF. Press Read text and wait while each page is read; the time left is shown. Download the new PDF. It looks the same as your scan, but you can now search it, select words and copy them.
- Press Choose a PDF.
- Tick the Language of the text (several if the pages mix languages).
- Choose Searchable PDF or Text (.txt).
- Press Read text; press Stop at any time.
- Read the summary, then Download.
How accurate is OCR on a PDF?
On clean print scanned at 300 dpi, our test read every word, even at 6 pt. Accuracy falls when small print is scanned at low resolution: 6 pt text at 100 dpi kept only a sixth of its words. The Tesseract manual says it works best at 300 dpi or more, so scan at 300 dpi when you can.
| Scan resolution | 6 pt text | 8 pt text | 11 pt text |
|---|---|---|---|
| 100 dpi | 16.7% of words | 90% of words | 100% of words |
| 150 dpi | 93.3% of words | 96.7% of words | 100% of words |
| 300 dpi | 100% of words | 100% of words | 100% of words |
Measured with the same engine and English data the tool uses, on a softened, rough JPEG of printed text (five lines, 30 words), counting the true words found in order. Each page took 186 to 689 ms on a desktop computer; phones are slower.
Can it straighten a crooked scan?
Yes. With Straighten tilted scans on, the tool measures each scanned page's tilt, turns it straight, reads the straightened page and puts it in the PDF. This matters: in our test, a page tilted 5 degrees gave 2 words without straightening and all 77 with it. Tilts under 0.3 degrees are left alone.
| Test scan (a 1-page contract at 200 dpi) | Tilt found | Words read, straightened | Words read, as scanned |
|---|---|---|---|
| Tilted 2.5 degrees clockwise | 2.5° | 77 | 76, with "1 October" read as "| October" |
| Tilted 4 degrees counterclockwise | −4.0° | 77 | 78 |
| Tilted 5 degrees clockwise | 5.0° | 77 | 2 |
Measured on 12 October 2026 in Chrome 152 on an i5-8350U laptop, English, automatic layout. The test pages were made by turning a clean page, so real scans with noise may read a little worse.
What is the difference between a searchable and a scanned PDF?
A scanned PDF holds only pictures of pages, so search finds nothing and you cannot copy a word. A searchable PDF has real text as well. This tool keeps the pictures and places each word it reads, invisibly, right where it appears, so search highlights the right spot and copying gives the words.
Can I get just the text out of a scan?
Yes. Set Save as to Text (.txt) and the tool saves the words it reads, page by page, as a plain text file. The text also shows in the result so you can copy it. For PDFs that already have a text layer, PDF to Text is faster because nothing needs reading.
Limits and fixes
- Handwriting is read poorly; the engine is trained on print.
- Sideways or upside-down pages need turning first with Rotate PDF; straightening fixes tilts of up to 7 degrees.
- Straightened pages are saved as new pictures, so the file can grow (our 93 KB test scan became 165 KB).
- Columns: if two columns come out mixed, try Page layout: One column or Scattered text.
- Speed: each page takes about a second on a computer and longer on a phone; long scans may run out of memory on phones, so split them first.
- Signed PDFs: adding text changes the file, so an existing digital signature no longer validates.
- Locked PDFs need their password removed first with Unlock PDF.
Frequently asked questions
Is it free?
Yes, with no sign-up, watermark or page limit.
Is my PDF uploaded?
No. The engine and language files come from this site, and reading happens in your browser.
Does the PDF look different afterward?
No. The pages stay as they were; the text layer is invisible and adds little to the file size.
What does page layout do?
Automatic suits most pages. One column helps with single-column pages that come out jumbled, and Scattered text helps with forms, receipts and labels.
Can it read several languages on one page?
Yes. Tick each language that appears; reading gets a little slower with each one.
Related PDF tools
Learn more
- Why Is My PDF So Large?What takes up the space in a PDF, measured, and how to fix each cause.
- Searchable vs Scanned PDFOne invoice as text, as a scan and after OCR: how to tell which you have, and how to fix a scan.
- How to Scan a Document on iPhone and AndroidFree built-in scanning on iPhone and Android, and how far a phone scan shrinks.
How this tool was built and tested
- Tesseract documentation: Improving the quality of the output: the 300 dpi advice and page layout modes
- Tesseract OCR engine (Apache-2.0): the engine that reads the words
- tesseract.js (Apache-2.0), version 7.0.0: runs Tesseract in the browser
- tessdata_best (Apache-2.0): the trained language data
- pdf-lib, Cantoo fork (MIT), version 2.11.1: writes the invisible text layer