How to convert a scanned PDF to editable text
A scanned PDF looks like a document, but to a computer it's a stack of photos. You can't search it, copy from it reliably, or give it to an AI model as text. To fix that you need OCR, optical character recognition, which reads the letters in an image and turns them into real text.
How to tell if your PDF is scanned
Open it and try to select a sentence with your mouse.
- If you can highlight individual words, the PDF already contains text. It converts instantly, without OCR.
- If the whole page highlights as one block, or nothing highlights, it's a scan.
Many PDFs are a mix: typed pages with a few scanned annexures. Docuflat checks each page separately and only runs OCR on the pages that need it.
Convert it in three steps
- Choose the language. In the OCR language box, pick the language of the document. For documents that mix two languages, such as Hindi and English, choose both (up to three).
- Drop the PDF onto the converter.
- Download or copy the result. Pages converted with OCR are labelled
(OCR).
OCR runs entirely in your browser, so the document never leaves your computer. The first time you use a language, your browser downloads its OCR data; after that it's cached and starts straight away.
What affects accuracy
| Factor | Effect |
|---|---|
| Scan resolution | 300 dpi is the sweet spot. Below 200 dpi, small letters and punctuation are lost. |
| Contrast | Black text on white converts best. Faint photocopies and coloured paper reduce accuracy. |
| Straightness | Tilted pages and curved book spines cause errors at the edges. |
| Language setting | Choosing the right language matters more than anything else for non-English text. |
| Fonts | Printed text works well. Handwriting, stylised fonts and stamps usually don't. |
For a full checklist, see 10 ways to get more accurate OCR.
Always check the important details
OCR is very good on clean printed pages, but it isn't perfect. Before you rely on the output, compare against the original for:
- Names, addresses and ID numbers
- Amounts, dates and percentages
- Clause and section numbers
A quick way to check is to paste the OCR text into an AI assistant along with the question "List every number, date and name in this text", then compare that list against the original.
Scanned documents in other languages
OCR works in more than 100 languages, including all major Indian scripts. See the pages for Hindi, Bengali, Gujarati, Tamil and other languages for language-specific tips.