Convert PDF to Word (.docx)
Extract a text-based PDF into an editable Word .docx with paragraphs, detected headings and page breaks preserved. No upload, no sign-up.
- Runs in your browser
- Files are never uploaded
- Free, no account
What this converter does — and does not — preserve
It rebuilds the text of a text-based PDF as editable Word paragraphs, with detected headings and one page break per PDF page. It does not reconstruct page layout, columns, tables, images, headers, footers or fonts, and it cannot read scanned pages. If you need a pixel-faithful copy, keep the PDF.
How to convert a PDF to Word
- Drop the PDF onto the box above. It is read locally with pdf.js — nothing is uploaded to a conversion service.
- Let the text extraction finish. You will see the word, paragraph and heading counts as soon as it is done.
- Decide whether to detect headings and whether to keep one page break per PDF page.
- Check the preview — it shows the document exactly as it will be structured in Word.
- Press Create .docx and download the file.
What is preserved, and what is not
Being straight about this matters more than the conversion itself. A PDF is a description of marks on a page; a Word document is a flowing structure of paragraphs and styles. Turning one into the other is a reconstruction, not a translation, and every converter that promises perfect fidelity is overselling.
| Element | Result | Why |
|---|---|---|
| Body text | Preserved, in reading order | Extracted directly from the PDF's text layer |
| Paragraph breaks | Preserved | Inferred from line spacing and indentation |
| Headings | Detected heuristically, styled as Heading 1 / 2 | PDFs carry no structure, so line shape is the only signal |
| Page breaks | Optional, one per PDF page | Page boundaries are known exactly |
| Bold, italic, colour, font sizes | Not preserved | Requires mapping every embedded font back to a style |
| Tables | Flattened to text | PDFs draw table borders as lines, with no cell structure |
| Images and charts | Not included | The text layer is extracted; graphics are separate objects |
| Columns, headers, footers | Flow into the body text | They are just text at particular coordinates |
| Scanned pages | Not convertible | There are no characters to extract — OCR is required |
The upshot: this tool is excellent when you want the words in an editable form — a report to revise, a letter to reuse, a policy to quote from. It is the wrong tool if you need a Word file that looks exactly like the PDF. For that, find the original document.
How heading detection works
Since the PDF carries no outline, the converter judges each extracted paragraph by its shape. A line becomes a heading when it is short (under about ninety characters and twelve words), does not end in a full stop, comma, semicolon or colon, and matches one of these patterns:
- a numbered section prefix such as
3.or3.2— the deeper the numbering, the lower the heading level; - a run of capitals, as in
TERMS AND CONDITIONS; - an opening word like Chapter, Section, Part or Appendix.
The result is a document whose navigation pane and automatic table of contents actually work. If the heuristic misfires on your document, turn it off and style the headings yourself — it is one click in Word.
Is the .docx a genuine Word file?
Yes. A .docx is a ZIP archive containing an Open Packaging
Conventions structure, and this tool writes the whole thing: the content-type
manifest, the package relationships, the main document part, a stylesheet
defining Normal, Body Text and the three heading styles, the document's own
relationships, and both sets of document properties. It is not a renamed
HTML file or an RTF, so Word, LibreOffice Writer, Apple Pages and Google Docs
all open it natively and let you edit and re-save normally.
Why doing this locally matters
The documents people convert to Word are contracts, invoices, medical letters, tax returns and legal filings. Every mainstream online converter uploads those to a server, processes them there and promises to delete them later. Here, extraction runs in pdf.js and the ZIP is built with JSZip, both inside this browser tab. There is no upload, no retention policy to trust, no account, no page limit and no watermark — and the tool works with your network disconnected, which is the easiest way to verify the claim.
Questions
Will my layout, tables and images come across?
No — and no browser-only tool can honestly promise otherwise. This converts the text layer into clean, editable paragraphs. Columns, tables, images, headers and footers are not reconstructed; you get the words, correctly ordered and ready to edit.
Does it work on scanned PDFs?
No. A scan holds no characters, only an image of them, so there is nothing to place in the document. The tool detects this and tells you instead of handing you an empty file.
Is the .docx a real Word file?
Yes. It is a proper OOXML package — [Content_Types].xml, relationships, styles and document parts, zipped as a .docx — so Word, LibreOffice, Pages and Google Docs all open it natively.
How are headings detected?
By shape, since PDFs carry no structure: short lines with no terminal punctuation, all-caps lines, and numbered section prefixes such as "2." or "2.1" become Heading 1 or Heading 2. You can restyle anything in Word afterwards.
Is the file uploaded to a converter service?
Never. Extraction runs in pdf.js and the .docx is zipped locally with JSZip, which is what makes it safe for contracts, medical letters and anything else confidential.