Review & data · guide
What OCR PDF does
OCR PDF reads the text in a scanned PDF and saves it as a plain text file. With Also make a searchable PDF ticked, it also builds a copy of the scan with the recognised words laid invisibly over each page, so it looks like the scan but can be searched, selected and copied. It recognises English, works best on clear printed pages, and reports how confident it is. The recognition engine downloads to your browser the first time you use it, and your document is never uploaded.
How to OCR PDF in your browser
- Choose a scanned PDF Select a PDF made from a scanner or from photos of pages. The first time you use the tool, its recognition engine, about 5 MB, is downloaded to this browser tab.
- Choose pages and output Leave Pages to recognise empty for the whole document, or enter pages such as 1-3, 7. Keep Also make a searchable PDF ticked to get a searchable copy of the scan as well as the text file.
- Read the text and save Select Read the text, check the confidence figure in the result, then save the text file and the searchable PDF. A figure below about 80 means the scan is hard to read, so check important words by eye.
Worked example
Make a photographed typed memo searchable.
- Starting point
- Choose a legible English-language scan. The current OCR release recognises English only.
- What to do
- Select Read the text and inspect the recognised words against the page image, including names and numeric values.
- Result
- Expect a new PDF with a text layer when recognition succeeds. Search for a known phrase to check the result.
If the result is not right
For poor recognition, start with a sharper, straighter English scan. Use a compatible recognition service for other languages, and verify important values against the image.
Limits worth knowing
- Skew, blur, handwriting and language mismatch affect accuracy. Searchability does not imply transcription accuracy.
Questions and answers
Which languages does OCR PDF recognise?
English only for now. It works best on clear printed text, such as a 300 DPI scan; phone photos of creased pages and handwriting are recognised far less reliably. Each result reports a confidence figure so you can judge how far to trust it.
Does OCR change how my scanned PDF looks?
Very little, but the searchable PDF is a new file rather than your original with text added. Each page is redrawn at 200 DPI and stored as a lossless image, with the recognised words laid invisibly over it, so it looks like the scan at normal zoom. A sharper scan loses detail above 200 DPI, the file can be much larger than the original, and links, bookmarks and any pages you did not choose are not in it. Keep the original alongside it.
Should I run OCR on a PDF that already has selectable text?
No. If you can already select or search the words, the text is there, and OCR would redraw each page as an image and swap real text for the recogniser's reading of it. Try selecting a sentence first: OCR is for pages where nothing selects, such as scans and photos of pages. For a PDF that already has text, PDF to Word or PDF to Markdown gets the words out directly.
Why is the recognised text wrong or jumbled?
Usually the scan. Pages that are sideways or upside down are not turned for you, so straighten them with Rotate PDF first; skewed, blurred, faint or low-resolution pages, handwriting and decorative fonts all read poorly. Tables and multi-column layouts can come out in an order that needs tidying. Below about 80% confidence, read the text through; for a phone photo of a page, Scan PDF can clean it up before you run OCR.
What does OCR PDF download, and does my scan leave the device?
The recognition engine and its English language data, about 5 MB, come from filozy.com itself rather than a third-party server, and only when you run a job; the browser's own cache keeps them for next time, and the tool stores nothing else. The scan is rendered and read inside this tab and is never sent anywhere. To see every request, including the engine files, open your browser's developer tools on the Network tab before you run the job.
How long does OCR take on a long document?
Roughly a second a page on a laptop and several seconds a page on a phone, plus the one-off engine download, with a progress bar counting the pages. For a long scan, put only the pages you need in Pages to recognise, such as 40-45; the text file and the searchable PDF then contain just those pages.