PDF OCR

Read the words off a scan and hand them back as a searchable PDF or a text file.

Hybrid processingPDFWorkingNo account needed

What this tool does

A scanned page is a photograph of writing: the letters are pixels, so searching, selecting and copying all fail on it. PDF OCR runs Tesseract over each page through ocrmypdf, then writes what it read back into the document as an invisible text layer positioned over the picture. The scan is still what you see on screen, and Ctrl+F starts working.

What comes back depends on what you send and what you ask for. A PDF returns either that searchable PDF or a plain text file carrying the words the engine read. An image — JPG, PNG, TIFF or BMP — always returns text, because there is no such thing as a searchable photograph. By default, pages that already contain real text are skipped instead of being read again. That is the right behaviour for the searchable output, and a trap for the text one: a skipped page contributes none of its words, so a half-scanned document exported as text arrives with gaps where its born-digital pages were.

Recognition is a reading, not a transcript, and that is the honest limit of every OCR engine. Clean printed text scanned at around 300 dpi comes back very close to the page; faxes, phone snaps taken at an angle, dot-matrix print and tight multi-column layouts come back with mistakes. Handwriting is largely beyond Tesseract. Column and table structure is not rebuilt either — the text layer follows reading order, so a table arrives as lines rather than cells. Proofread names, totals and reference numbers before anyone relies on them.

How to use it

  1. Add the scan. A PDF, or a JPG, PNG, TIFF or BMP image, each up to the 200 MB upload ceiling. Queue several and every one of them is recognised — each becomes its own upload and its own job, run one after another, so a batch costs the sum of its parts in both time and file allowance.
  2. Pick the language. English, French, German, Spanish, Italian, Portuguese and Dutch are the languages installed here. Choose one per job, and choose the right one — French read as English produces noise rather than a few typos.
  3. Choose the output. A searchable PDF keeps the scan and adds the invisible text layer. Plain text drops the pages and returns the words alone. An image input returns text either way.
  4. Decide about pages that already have text. Leave Re-OCR pages that already have text switched off unless you know the existing layer is wrong. Straighten skewed scans is on by default and rescues pages that went through a feeder crooked, at the cost of rewriting the image of every page it reads.

Processing & privacy

Hybrid processing. Smaller jobs can run locally; heavier jobs may use secure server processing.

Common use cases

A contract or invoice that arrived as a scan

Someone scanned it and emailed a picture. A text layer means you can find the clause or the amount without reading every page again.

Old paper records going into an archive

Once each scan carries its words, the collection is searchable by content instead of by whatever the filenames happen to say.

A photo of a page, a notice or a receipt

Point the camera, upload the JPG, and the words come back as a text file you can paste into notes or a ticket.

A report where only some pages are scans

Ask for the searchable PDF here. Pages that already have text pass through untouched, so recognition is spent on the inserted scans and the rest of the document keeps its original characters.

Questions

Is my file uploaded?
Yes. Tesseract and ocrmypdf run on the processing service, so the scan travels there over an encrypted connection, is recognised, and is then deleted along with the result and the job record when the retention window closes — six hours after the job ran, unless an administrator has changed that. Nothing about this tool runs in the browser tab, whatever the size of the file.
How accurate is the result?
It depends almost entirely on the scan. Sharp printed text at 300 dpi is usually near-perfect; low-resolution faxes, skewed phone photographs and unusual typefaces produce errors, and handwriting mostly fails. Treat the output as a draft to check, not a certified transcription.
Does it change how my pages look?
Slightly, in the default setup, because Straighten skewed scans is switched on: any page the engine reads is written back from the straightened image, which is exactly what that setting is for but is still a real change to the page. Turn it off and the searchable output keeps your scan untouched under invisible words. Re-OCR pages that already have text goes much further, rasterising every page — text that used to be genuine becomes an image with the engine's guess laid over it.
Can I recognise two languages at once?
No. One language is passed to the engine per job, and only the seven with training data installed on this server are offered. Anything else is refused rather than quietly falling back to English.
My PDF already has selectable text — should I use this?
Probably not. If you can select a sentence in your viewer, the words are already there and PDF to Text will pull them out without recognition, keeping the original characters instead of a reading of their shapes.
Is the output a PDF/A archive file?
No, it is an ordinary PDF with a text layer added, and this server cannot turn it into one. Archival output from ocrmypdf needs Ghostscript, which is deliberately not installed here, so the job runs with plain PDF output — and PDF/A Converter refuses its jobs over the same missing engine rather than relabelling a file. An archival copy has to be produced somewhere else.
How large or long a job can it take?
The upload limit is 200 MB and a job is stopped after ten minutes. There is no page cap, but recognition is slow work — a few seconds a page on one core — so a long scan can reach the time limit instead; splitting it into shorter documents is the way through.

Related tools

  • Merge PDF — Combine several PDFs into one document in the order you choose.
  • Split PDF — Split one PDF into separate documents by page range.
  • Compress PDF — Reduce PDF file size while preserving the quality that matters.
  • PDF to JPG — Export every page of a PDF as a JPG image.
  • JPG to PDF — Turn one or many JPG photos into a single PDF.
  • Protect PDF — Add a password and set permissions on a PDF.