What OCR Actually Does
A scanned page is a photograph. The words are visible to you but invisible to the computer — you cannot search them, select them, or copy them. Optical Character Recognition looks at the shapes in the image and works out which letters they are, turning a picture of text into text.
This runs the recognition engine inside your browser. That is unusual for OCR, and it is the reason this tool exists: the documents people need to make searchable are overwhelmingly the sensitive ones — medical records, legal bundles, historical archives, contracts, identity papers. Every other free OCR service asks you to upload those first.
Getting Accurate Results
OCR accuracy depends far more on the input than the engine. A clean scan reads almost perfectly; a hurried phone photo of a curled page does not.
| What helps | What hurts |
|---|
| 300 DPI or higher when scanning | Low-resolution or heavily compressed images |
| Straight, flat pages | Skew, curl and perspective from handheld photos |
| Strong contrast, dark text on light paper | Faded print, shadows, coloured or patterned backgrounds |
| Standard printed typefaces | Handwriting, decorative fonts, dense tables |
| Selecting the correct language first | Leaving the language on English for a French document |
Recognition covers English, Spanish, French, German, Italian, Portuguese and Russian. The language setting is not cosmetic — it loads a different model and changes which letter shapes and accented characters are expected. Setting it correctly is the single cheapest accuracy improvement available.
Recognition takes time, and that is the trade-off. Because the work happens on your device rather than a server farm, a long document is slower than a cloud service would be. What you get for the wait is that the file never leaves your machine.
If a page comes back as gibberish
Usually one of three things: the language is wrong, the scan is too low-resolution for the engine to distinguish letters, or the page is skewed. Rescan at a higher DPI if you can. If the page is simply rotated, fix it with Rotate PDF Pages first — OCR reads horizontally and a sideways page produces nonsense.
What to Do With the Text
- Copy it straight out for quoting, or to paste into a document you are drafting.
- Search a bundle you could not previously search — the practical reason most legal and medical scans get OCRed at all.
- Feed an editable draft into PDF to Word, which needs a text layer to work with.
- Extract just the words with PDF to Text for a plain file.
Before You Share the Result
Scanned documents are frequently the ones with sensitive detail on them. If you are passing the file on, redact anything confidential first — and note that drawing a black box is not redaction, which this guide explains. Consider stripping the metadata too, since scanners routinely stamp the device name and timestamp into the file.
Frequently Asked Questions
Is my document uploaded for OCR?
No. The recognition engine runs in your browser. The file never reaches a server, which is the main reason to use this rather than a cloud OCR service for anything confidential.
Which languages are supported?
English, Spanish, French, German, Italian, Portuguese and Russian. Select the right one before running — it loads a different recognition model and materially changes accuracy.
Why is it slower than other OCR sites?
Because the work happens on your device instead of a server. That is the trade: you wait longer, and your document never leaves your machine.
Can it read handwriting?
Not reliably. The engine is trained on printed type. Handwriting, decorative fonts and dense tables all produce poor results.
The output is gibberish. What went wrong?
Usually the wrong language setting, a scan below roughly 300 DPI, or a skewed page. Check the language first, then the image quality. If the page is sideways,
rotate it before running OCR.
Does OCR change my original file?
No. It reads the document and produces text output. Your original PDF or image is untouched.