Skip to content
PzPDFzy

OCR PDF

Recognise the characters in a scanned document so you can copy and search them.

Drop your file here

Nothing is uploaded — your file is processed on this device.

Options for OCR PDF

The recognition model for this language is downloaded once, then cached by your browser.

Higher numbers look sharper and create larger files.

Recognition runs in your browser. The language model (a few megabytes) is downloaded from a public model host the first time you use a language; your document itself is never uploaded. The result is a text file — this version does not write an invisible text layer back into the PDF.

How to ocr pdf

  1. 1Upload the scanned PDF.
  2. 2Pick the language of the text and the recognition resolution.
  3. 3Run OCR, then read the text on screen or download it.

About this tool

A scanned PDF looks like a document but behaves like a photograph: you cannot search it, copy from it, or convert it, because there are no characters in the file — only pixels arranged to look like characters. Optical character recognition solves that by reading the shapes and working out what they say.

Each page is rendered at the resolution you choose and passed to a recognition engine running inside your browser. 200 DPI is the sweet spot: enough detail for accurate recognition without making each page slow to process. The first run downloads a language model of a few megabytes, which your browser then caches for next time.

Accuracy depends heavily on the source. A clean, straight, well-lit scan of printed text typically comes out near-perfect. Skewed pages, faint photocopies, unusual fonts and handwriting are much harder — always read the result before relying on it.

What it's good for

  • Making an old scanned contract searchable.
  • Copying figures out of a scanned invoice or statement.
  • Turning a scanned book chapter into editable text.
  • Preparing a scan so conversion tools can work with it.

Features

  • Seven languages

    English, Spanish, French, German, Italian, Portuguese and Dutch.

  • Progress feedback

    You can see which page is being read.

  • Resolution control

    Raise the DPI for small or faint print.

  • Document stays local

    Only the language model is fetched, never your file.

Questions people ask

Why is the recognition wrong in places?
OCR is a prediction. Faint, skewed or unusual text reduces accuracy — try a higher DPI or a better scan.
Can it read handwriting?
Not reliably. The engine is trained on printed text.
Do I get a searchable PDF back?
You get the recognised text as a file. Converting that text back into a PDF is possible with Text to PDF.
Why is the first run slow?
The language model downloads once. Later runs reuse the cached copy.

More ocr tools