Extract text
PDF to text vs OCR: which one do you need?
By DeckleUpdated
Use PDF to text when the PDF already has selectable text and you want the words in a TXT file. Use OCR when the pages are scans or photos with no text at all: it recognises the words and adds them to the PDF as a searchable layer. The quick test is to try selecting a sentence in a viewer. If it highlights, extract. If it does not, run OCR first, then extract from the recognised PDF if you still need a text file.
Pages become plain text
Selected PDF pages
Order chosen in the review
UTF-8 text
Words without page formatting
Page boundary: U+000C or a blank line
Which one you need
| Your PDF | Your goal | What to do |
|---|---|---|
| Text highlights when you drag | Words in a TXT file | PDF to text |
| Text highlights when you drag | Search inside the PDF | Nothing; it is already searchable |
| Scanned, nothing highlights | Search inside the PDF | OCR to a searchable PDF |
| Scanned, nothing highlights | Words in a TXT file | OCR, then PDF to text |
| Mixed typed and scanned pages | Words from every page | OCR the scanned pages, then PDF to text |
| Already has an OCR layer | Words in a TXT file | PDF to text, then proofread |
What each one does
PDF to text reads characters the PDF already stores and writes them to a UTF-8 TXT file. It does not look at images, so it cannot recognise anything new. It is exact: what the PDF stores is what you get, with no recognition errors added. You choose which pages to include, in which order, and whether to mark where each page ended.
OCR (optical character recognition) looks at a picture of each page, identifies the letters and writes an invisible text layer behind the image. The page looks the same, but you can now search it, select words and copy them. It is a recognition step, so it can misread characters, particularly on faint or skewed scans. Deckle's PDF to text does not run OCR, so this step needs a separate OCR tool.
The two are not competitors. OCR creates text; extraction exports it.
How to extract text from a PDF that already has it
- Open PDF to text and choose your PDF, or drop it onto the page.
- Keep All pages or choose Choose pages and type the pages you need.
- Under Between pages, choose Mark where each page ended or One continuous document.
- Choose Extract text.
- Download the TXT file and check a sentence against the PDF.
If the result screen's Verified: line says there was no selectable text to compare, the pages are images. That is your sign to run OCR first.
When you have a scan and want a text file
Run OCR on the scan to make a searchable PDF, then open that PDF in PDF to text. Proofread the result: OCR mistakes travel into the TXT file unchanged. Names, reference numbers, amounts and dates are the parts most worth checking. Native PDF text vs OCR text lists the misreads to look for.
What each output is good for
| TXT from PDF to text | Searchable PDF from OCR | |
|---|---|---|
| Looks like the original page | No | Yes |
| Searchable | Yes, in any editor | Yes, in any PDF viewer |
| Easy to paste and edit | Yes | Copying works, editing does not |
| File size | Small | Similar to the scan, plus the text layer |
| Keeps signatures, stamps and photos | No | Yes |
| Good for a script or analysis | Yes | Needs extracting first |
If you are deciding which file to keep for the long term, searchable PDF vs TXT file goes through that choice.
Tips
- Test more than the first page. A cover can be an image in an otherwise typed document, and a typed document can end with scanned pages.
- Do not OCR typed pages. They already hold exact text, and recognition can only add misreads.
- Proofread OCR text before relying on it. Check numbers digit by digit.
- Keep the scan. If recognition was poor, you can try again with a better language setting or a cleaner scan.
Frequently asked questions
Is PDF to text the same as OCR?
No. PDF to text exports text the PDF already has. OCR recognises text from images and adds it to the PDF.
Can PDF to text read a scanned PDF?
Only after OCR. On an image-only page there is no text for it to read, so the output is empty.
Which is more accurate?
Extraction from native text, because nothing is recognised. OCR accuracy depends on the scan's quality, language and fonts.
Do I need OCR to search a PDF?
Only if the PDF is a scan. A PDF exported from software is searchable already.
Related tools
- PDF to text to export selectable text to a TXT file
- Split PDF to separate scanned pages before OCR
- Merge PDFs to rejoin pages after recognition