Quick decision table
| What the PDF contains | What you notice | Use |
|---|---|---|
| Embedded text | Text selects and copies as meaningful words | PDF Text Extractor |
| Scanned or image-only pages | The page looks like a picture and copied text is empty or meaningless | Image OCR |
| A mixture | Some pages have selectable text and others are scans | Use the method that matches each page; OCR can handle the image pages |
PDF is a container format that may include both text and still images, so a file ending in .pdf does not tell you which kind of content every page contains. The Library of Congress describes PDF as supporting text and still-image content. PDF format overview
What PDF text extraction does
Text extraction reads the document’s existing text layer. It is closer to copying text from a page than to reading the page visually. The extractor can preserve the words that the PDF exposes, but it does not invent a text layer for a photograph or scan.
ZANCTA’s PDF Text Extractor reads pages in a local PDF worker, joins the extracted page text, and lets you search, copy, or download it as a text file. It accepts one PDF up to 50 MB and reports when no embedded text is found.
What OCR does
Optical character recognition, or OCR, analyzes pixels and predicts the characters represented by those pixels. It is useful when a page is a scan, photograph, or image-only PDF. OCR is an interpretation step, so the result can contain mistakes even when the page looks clear.
ZANCTA Image OCR uses Tesseract.js in a browser Worker. English image OCR is free. Premium Local OCR Power adds Hindi, Bengali, Tamil, Spanish, French, and German language packs, plus scanned-PDF OCR up to 20 pages. Images are limited to 20 MB; PDFs use a separate 50 MB limit.
For background on OCR as a recognition process, see Adobe’s OCR explanation and the Tesseract documentation.
How to tell whether a PDF needs OCR
Try selecting a sentence in a PDF viewer and copying it into a plain-text field. Meaningful, complete words usually indicate an embedded text layer. Empty output, one large image-like selection, or unreadable fragments suggest that the page may be image-only.
This is a practical check, not an infallible test. A PDF can contain unusual fonts, broken text positioning, protected content, or both text and image pages. The reliable test is to run a small copy through PDF Text Extractor and inspect whether useful text is returned.
PDF Text Extractor vs OCR
Choose text extraction when: the PDF was exported from a word processor, contains searchable text, or you need to preserve the document’s existing text without recognition guesses.
Choose OCR when: the pages are scans or photos, ordinary extraction returns no useful text, or the PDF contains image pages that need recognition.
Do not assume OCR is always better: it adds an interpretation step and may misread blur, low contrast, unusual type, handwriting, stamps, tables, or mixed scripts.
Which ZANCTA tool should you use?
Start with PDF Text Extractor for a text-native PDF. It does not OCR scanned pages. If the file is an image-only scan, use Image OCR; scanned-PDF OCR is a Premium capability capped at 20 pages.
If you only need to inspect or OCR one page, PDF to Images can render PDF pages as images first. That is a conversion step, not a replacement for OCR.
What can go wrong
Image-only pages: ordinary extraction can correctly return no text because there is no text layer to read.
Mixed PDFs: some pages may extract normally while scanned pages need recognition. ZANCTA’s OCR path probes pages individually and retains embedded text where it is available while recognizing image pages.
Poor scans: low light, blur, decorative fonts, handwriting, stamps, dense tables, tiny type, and mixed scripts can reduce OCR quality. Empty or incorrect output is possible.
Unsupported or damaged files: password-protected, corrupt, or unusual PDFs can fail before either method produces text. Unlock or repair the document first, then retry with a small sample.
Privacy and browser processing
For these implemented ZANCTA workflows, supported file processing occurs in the browser after the required application assets load. The selected PDF is not posted to a ZANCTA processing API. OCR recognition runs in a browser Worker; Premium language data loads only when selected and authorized.
This describes file processing, not an entire offline session. The page still loads HTML, JavaScript, fonts, and other assets. Optional analytics may record tool events without sending the file or recognized text. See how browser OCR works without uploading and the broader local processing guide.
Troubleshooting
1. Try a single page or a small representative file first.
2. Copy a visible sentence from the PDF. If the copied result is empty or unusable, test the file with PDF Text Extractor.
3. If the extractor reports no embedded text, switch to Image OCR for a scan.
4. For OCR, improve contrast and alignment where possible, select the correct language, and expect recognition errors on handwriting, stamps, tables, or blurred images.
5. If a mixed PDF behaves unexpectedly, split or render the relevant pages and test them separately.
Bottom line
Extraction reads text that is already present. OCR recognizes text from pixels. Use the extractor for text-native PDFs, OCR for scans, and treat mixed PDFs page by page. Neither method guarantees perfect output, so inspect the result before relying on it.
Sources
- Library of Congress: PDF format family — PDF content categories and text/image capabilities.
- Adobe Acrobat: What is OCR? — general OCR explanation.
- Tesseract documentation — OCR engine documentation.