Scanned PDF to Word: OCR, Languages, and Layout Limits
If dragging across a sentence selects a whole page instead of individual words, your PDF may contain photographs rather than text. Renaming it to .docx cannot make those words editable. OCR must recognize the characters before they can become Word paragraphs. PDFFlow reads existing PDF text first and uses OCR for pages without usable text.
1. Choose automatic detection or OCR every page
Automatic mode first extracts existing text and applies OCR to pages with very little text. A page can mix a selectable header with a photographed table; automatic detection may then use the header and miss words inside the image. Choose OCR every page for mixed pages or a scan with an incorrect hidden text layer. This re-recognizes the entire page rather than merging old text with OCR, so compare spelling and numbers with the source.
2. Choose the language in the document
English is the default. Choose English + Korean for Korean or mixed-language documents; website language and document language are separate. The PDF input limit is 100 MB, and at most 30 pages can undergo OCR in one conversion. OCR every page counts all pages toward that limit. Split a longer scan into smaller PDFs first. Engine and language downloads come from jsDelivr on first use, so internet access is needed even though the document stays in your browser.
3. Improve the source before conversion
Enlarging a blurred photograph cannot recover missing character strokes. If you can retake the photo, flatten the paper, hold the camera parallel to it, and avoid glare and shadows. Keep small letters legible when compressing images into a PDF. Do not assume that perspective, curved pages, or shadows are automatically corrected. Handwriting and mathematical notation also need special care and are not promised to convert accurately.
4. Understand tables and multiple columns
The output focuses on editable text and paragraphs. Table cells, two-column reading order, pictures, and original fonts may not be reconstructed. Table and column formatting is in beta and being improved. Compare each column heading with the values below it: readable words in the wrong row can change the meaning. If a table is essential, you may need to create a new Word table and move verified values into its cells.
5. Review the Word file against the scan
Check names, dates, totals, decimal points, minus signs, and reference numbers first. The letter O and digit 0, or lowercase l and digit 1, can be confused even in an otherwise readable sentence. Compare every source page for missing paragraphs and check whether headers or footers have entered the main text. Keep the scan alongside the corrected Word file so another person can verify uncertain passages.
💡Illustrative workflow
For a four-page English notice with one photographed Korean attachment, choose English + Korean. If the attachment shares a page with selectable text, enable OCR every page. Compare the attachment table row by row and check all four pages for missing text.
⚠️Important Guidelines & Caveats
OCR can misread or omit characters. Exact restoration of tables, columns, images, fonts, and complex layouts is not guaranteed. This tool produces DOCX; creating a searchable PDF with a hidden text layer is a different task.
❓Frequently Asked Questions (FAQ)
Q. Is my PDF uploaded for OCR?
A. The document is processed in your browser rather than uploaded to a document-processing server. Downloads of OCR software and language data, and normal website or advertising requests, are separate from the document contents.
Q. Does the Word file just contain pictures of the pages?
A. OCR turns recognized characters into editable paragraphs instead of simply placing a full-page image in Word. That also means the original visual layout is not preserved exactly.
Q. Is Korean recognition as accurate as English?
A. Both are available, but equal accuracy is not guaranteed. Font, image clarity, and character size affect results. Check a representative page from your own document before processing a large batch.
About this guide
This article is maintained against the tools currently available on PDFFlow. Browser and PDF features vary, so preserve the original document and inspect the downloaded result before formal submission.