PDF OCR
Read text out of scanned PDFs that have no text layer. Twelve languages, entirely in your browser.
How to Use
- 1. Upload the scanned PDF you want to read.
- 2. Choose the language the document is written in.
- 3. Start the scan and wait while each page is recognised.
- 4. Copy the extracted text or download it as a .txt file.
When a PDF Has No Text to Extract
There are two kinds of PDF that look identical on screen. One stores its text as characters, so selecting a paragraph highlights words and copying it works. The other is a picture of a page — produced by a scanner, a phone camera or a fax — where the text exists only as pixels. Select a line in that second kind and nothing happens, because there is nothing there to select.
OCR is what bridges the two. Each page is rendered at high resolution and then analysed character by character, rebuilding the words from their shapes. That is a much heavier job than reading an existing text layer, which is why this tool takes seconds per page where a plain extractor is instant, and why it is capped at fifteen pages.
Accuracy depends almost entirely on the source. A clean 300 DPI scan of printed text usually comes back above 90% confidence; a phone photo taken at an angle in poor light may come back unusable. Raising the render quality helps, selecting the correct language helps more, and neither can recover detail the original scan never captured.
If your PDF does have a text layer, use the faster Extract Text from PDF instead. Related: Image to Text (OCR) and PDF to JPG.
Frequently Asked Questions
What is the difference between this and a normal PDF text extractor?
A text extractor reads text that is already stored in the file. This tool renders each page as an image and recognises the characters visually, which is the only thing that works on a scanned document or a photographed page, where there is no text layer to read.
Is my PDF uploaded anywhere?
No. Pages are rendered and recognised inside your browser. The document never leaves your device — the only thing downloaded is the OCR engine and language model, and that is code coming in rather than your file going out.
Why is it limited to 15 pages?
OCR is heavy work. Each page is rendered at high resolution and then analysed character by character, which takes several seconds and a good deal of memory. Capping the count keeps a long document from freezing the tab.
How do I get the most accurate result?
Start from the highest-quality scan you have, keep the render scale at 2 or higher, and pick the right language before you start. OCR struggles most with skewed pages, low contrast, heavy compression artefacts and decorative or handwritten type.
Does it preserve the original layout?
Not columns or tables. The output is plain text in reading order with a page marker between pages, which suits searching, copying and re-editing rather than reproducing the original formatting.
Was this tool helpful?
Guide
How to extract text from a scanned PDF with OCR
Some PDFs are really photographs of paper. They came from a scanner, a phone camera or a fax, and the words exist only as pixels, so you cannot select, search or copy them. OCR, short for optical character recognition, looks at each page as an image and rebuilds the text from the shapes of the letters.
This tool runs that recognition in your browser, one page at a time, for up to 15 pages. You choose the language, a render quality, and press a button. It shows the text per page with a confidence score, then lets you copy everything or save it as a plain text file.
Step by step
- Drop a scanned PDF on the upload area or click to choose it. The tool shows the page count; if there are more than 15 pages it notes that only the first 15 will be scanned.
- Select the Document language from the list: English, Hindi, Spanish, French, German, Portuguese, Italian, Dutch, Russian, Arabic, Chinese (Simplified) or Japanese.
- Adjust Render quality if needed. The slider runs from 1x to 4x in half steps and starts at 2x. Higher is more accurate but slower.
- Press Extract text. A status line first says it is loading the OCR engine, then reports each page as it reads it, with progress. Use Stop to end early.
- Review each page's text and its confidence badge. Use Copy all, or Download .txt to save a file named like yourfile-ocr.txt.
When this is useful
- Pulling the text out of a scanned letter or notice so you can search or quote it.
- Digitising a printed page of an old government order or certificate for re-typing.
- Reading a Hindi document that was scanned as an image, by choosing Hindi as the language.
- Copying a table of figures from a scanned report into notes.
Worked example: A 20-page scanned notice
You load a 20-page scanned file. The tool says 20 pages and that the first 15 will be scanned. At the default 2x scale, an A4 page is rendered at about 1190 by 1683 pixels before recognition. After the run you might see 15 pages, a word count and an average confidence. To handle pages 16 to 20, cut them out with Split PDF using the range 16-20 and run the resulting 5-page file through the tool again. The text for each run downloads as notice-ocr.txt, and the first line of each page section reads --- Page 1 ---.
Tips for better results
- Choose the correct language first. It matters more than any other setting, because the recogniser compares shapes against that language's letters.
- Begin at 2x. Try 3x for small print or faint scans, but expect each page to take noticeably longer.
- Treat the confidence badge as a hint. Green means 80% or more; amber means check this page by eye.
- Pages are separated by lines like --- Page 3 ---, which makes the copied text easy to split.
Common mistakes
- Using a PDF that already has selectable text. OCR is slower and less accurate than reading the existing text, so try copying text from the file first.
- Expecting a searchable PDF back. The output is plain text only, in reading order, with no columns, tables or original layout.
- Mixing languages. Only one language is used per scan, so English words inside a Hindi page may be misread. Review those parts.
- Feeding a skewed, shadowed or low-contrast photo. No setting can recover detail that the picture never captured. Rescan flat and well lit.
Your files and privacy
Your PDF is rendered and recognised inside your browser and is not uploaded. The tool does download the OCR engine and the language data from the cdn.jsdelivr.net content network when you start a scan, so an internet connection is needed on first use, and the library may keep that data cached in your browser.
Frequently asked questions
Why is there a 15-page limit?
Every page is drawn at high resolution and analysed, which takes several seconds and a lot of memory. The cap stops a long document from freezing your tab. Split longer files and process them in parts.
How accurate is it?
It depends almost entirely on the source. A clean, upright scan of printed text usually reads well, while handwriting, decorative fonts and tilted phone photos do badly. Check the confidence figure and read amber pages by eye.
Does it keep the layout of the original?
No. You get plain text in reading order with a page marker between pages. Columns, tables, images and fonts are not reproduced, which suits searching and re-editing more than recreating the document.
Why did the scan fail or return nothing?
The engine is downloaded on first use, so a failed connection stops the run. If you see a no text recognised message, raise the render quality, check you chose the right language, or confirm the PDF is not blank.
Can it read handwriting?
It is built for printed text. Handwriting is not recognised reliably, so treat output from handwritten pages as rough at best.
Guide reviewed 2026-10-11 by the useToolify team.
More Tools
Related PDF & Docs
Rotate PDF
Protect PDF
Resize PDF in KB
Images to PDF
You Might Also Like