OCR PDF

New

Make scanned PDFs searchable and selectable.

On a secure server

Text recognitionChecking…

OCR PDF runs on our secure server

Add a scan to OCR, or browsePDF, JPG, or PNG

Recognition settings

Languages (up to 4)

Select every language printed on the page — a page that mixes Bengali with English technical terms needs both. The wrong language does not produce an error, it produces plausible-looking gibberish.

Output

Add an invisible, selectable text layer under the original scan. It looks identical, but the text becomes selectable and searchable.

Redo existing OCRPages that already have text keep it, and only the pages without text are recognized. Turn this on when a PDF's existing OCR text is poor: that text is removed and those pages are recognized again. Text typed into the document is kept either way. PDFs only.

This request: English → Searchable PDF

Checking our secure server…

About this tool

Run optical character recognition to add a searchable, copyable text layer to scanned PDFs.

Files are processed on a secure server, which deletes the upload when the job ends and serves the result for a limited time.

More details

A scanner or phone camera saves each page as a picture, so Ctrl+F finds nothing and no sentence can be copied out. OCR PDF reads the lettering and returns it as real text. For a PDF, your pages stay as they are and an invisible text layer is laid underneath, so the document looks the same but becomes searchable and selectable. Send a JPG or PNG instead, such as a photographed receipt or a screenshot of a form, and any PDF you ask for comes back as a single page holding that image with the recognised words beneath it.

Recognition needs a native OCR engine, so this online tool does not run in your browser. Your file is uploaded over TLS to pdf.xyz's secure server, which runs OCRmyPDF on PDFs and Tesseract on images. The input is deleted when the job finishes, the download link expires after 60 minutes, and a daily storage rule then erases the stored result. One scan goes in per run, at most 100 MB.

Of every choice on the page, the language is the one most likely to spoil a result. 34 are on offer, with Bengali, Arabic, Chinese (Simplified), and Hindi among them, and a single job can combine up to 4. Picking the wrong one does not produce an error: the engine matches shapes against the alphabet it was given and returns fluent-looking nonsense.

When to use it

Search an archive of scanned letters
Old correspondence, deeds or minutes digitised as images become findable by name, date or phrase with a PDF viewer's search once each file carries a text layer. The tool takes one file per run, so work through a folder in turn.
Copy a paragraph out of a photographed page
Choose Text sidecar and you get only the recognised words in a .txt file, ready to paste into an email or document without retyping from the photo.
Read a mixed-language document
A form printed in Bengali with English headings, or an Arabic contract with French clauses, needs every language on the page selected. Up to 4 can be selected for one run.
Finish a PDF that is only partly scanned
When typed pages and scanned pages sit in the same file, just run it: pages that already have text are skipped and kept as they are, and only the image pages are recognised. A scanned page that already carries any text, such as a digitally added stamp or header, counts as having text and is skipped too. Tick “Redo existing OCR” only when an earlier OCR text layer is poor and should be recognised again; text typed into the document is kept either way.

What this tool can do

Languages: 34 recognition languages advertised, up to 4 per job — our server's own live report governs which packs are actually installed.

How it works

1Select the languages printed on the page
Under Languages, English is selected to start with. Add every language the document actually uses, up to 4; to swap one out once you are at the limit, deselect a language first.
2Choose what you want back
Under Output, pick Searchable PDF, Text sidecar (a .txt file), or Searchable PDF + text. Pages of a PDF that already have text are skipped and kept as they are; tick “Redo existing OCR” only if the PDF's existing OCR text is poor and should be recognised again.
3Add the scan and run it
Drop one PDF, JPG or PNG into the Add a scan to OCR area and press Recognize text. That area takes a file as soon as the page loads; the button stays off until the page confirms the OCR server is available, and your file is sent only when you press it.
4Download and check the result
Each file appears as a download button named after your original, with -ocr added to the PDF, above a Recognition summary of what was produced. Save each file within 60 minutes of the job finishing, because its link expires then.

Common questions

Why does my OCR output look like gibberish?

Usually the selected language does not match the page. Bengali read as English comes back as Latin characters that look plausible and mean nothing, with no error. The file stays loaded after a run, so change the language selection and press Recognize text again. Blurry, skewed or low-resolution scans cause the same symptom, and neither engine is told to straighten or auto-rotate pages, so turn a sideways PDF page upright with Rotate PDF before recognising it.

Does OCR change how my PDF looks?

The Searchable PDF output keeps your pages and adds hidden text underneath them, so it looks like the scan you uploaded. The words are not made editable on the page. To rewrite them, run the OCR'd file through PDF to Word, which turns its new text layer into plain paragraphs in a DOCX.

Can I OCR a photo or screenshot?

Yes, JPG and PNG are accepted. Images are read by Tesseract, whose per-word scores give the Recognition summary a Confidence row, plus a low-confidence warning when at least 25 words were recognised and their scores come out poor. PDFs go through OCRmyPDF, which reports no confidence, so that row is absent for them.

Which languages are supported, and can I mix them?

The menu lists 34, from English and Spanish to Bengali, Arabic, Chinese (Simplified), and Hindi, and one job can combine up to 4. If our server ever lacks the recognition pack for a language you selected, the job stops before recognition starts and the error names that language, instead of blaming your file.

What happens to my file on our secure server?

It is uploaded over TLS, recognised on pdf.xyz's secure server, and the input is deleted when the job ends. You have 60 minutes to download the result before its link expires, and a daily storage rule clears the stored output after that. The tool never asks you to create an account.

How big a scan can I process, and how many at once?

One file per run, up to 100 MB. For the next document, press Remove beside the current file name, add the new scan and run again: your languages and output choice stay as you left them until you reload the page. Size is not the only limit, since every page takes recognition time and a job has a time limit on our secure server, so cut a very long scan into parts with Split PDF first.

How large can my file be?

Up to 100 MB per file.

Which languages can OCR recognize?

34 languages are advertised, with up to 4 selectable per job. Our server reports the packs it actually has installed, and the page's live status reflects that report.