PDF to Word

This tool is for people holding a text-based PDF (a contract, a paper, a notice, an article) who just want the words out so they can rewrite, quote or reformat them. What it does is narrow: it reads each page's text layer with pdf.js, strings the text together, and writes a .docx with the docx library. What it does not do matters just as much: it keeps no layout, fonts, images, tables or columns, and it cannot read scanned pages. The note at the top of the page calls it a simplified version, and that is the right expectation.

How to use it

1. Click "Click to upload a PDF" inside the dashed box and choose one PDF (one file at a time).
2. The page shows "Extracting text and generating the Word document…" and the upload box is disabled while it works.
3. When it finishes you will see "Extracted text from N page(s). Done."
4. Click the "Download xxx.docx" button. The file keeps the PDF's name with .pdf swapped for .docx.

A concrete example: a three-page notice

Say you upload notice.pdf, three pages of normal text. The resulting notice.docx has a bold "[Page 1]", "[Page 2]" and "[Page 3]" line at the start of each page, followed by that page's text. Paragraphs are split after a Chinese full stop, an exclamation mark, a question mark or a line break. Two quirks to expect: pdf.js hands back text in small fragments and the tool joins them with spaces, so Chinese text can pick up stray spaces that you will want to clean with find-and-replace; and a plain English period is not a split point, so an English page often ends up as one very long paragraph.

Supported range and known limitations

• Only PDFs with a real text layer work. A scan or screenshot-style PDF has none, so you get page markers and almost no body text. There is no OCR.
• Original layout is not preserved: images, tables, columns and header/footer styling do not carry over. Table cells come out as ordinary text in reading order, so rows and columns are lost.
• Text order follows how the PDF stores it, so multi-column or unusually built files can come out in the wrong order.
• Encrypted or damaged PDFs fail with: "Parsing failed. Please make sure the file is an unencrypted PDF".
• All body text uses the default font; only the page markers are bold.
• The whole file is processed in browser memory, so very long documents are slow and may stall low-end devices.

The Word file is nearly empty, just "[Page N]" lines. Why?

The PDF has no extractable text layer, most likely a scan or pages saved as images. A quick test: if you cannot select text in a PDF reader, this is the case. Run it through an OCR tool first; this tool cannot.

I see garbled characters or missing letters.

Text is read using the PDF's internal character encoding. Some files use custom font encodings or have had text converted to outlines, and the result is junk or gaps. There is no workaround inside this tool.

Paragraphs are wrong, or a whole page is one block.

Splitting only looks for Chinese full stops, exclamation marks, question marks and line breaks; it does not understand real paragraph structure. English prose, headings and lists without those marks are not split properly and need manual cleanup in Word.

A large file freezes the page or fails. What now?

Every page is read one after another inside a single browser tab, so a long or heavy file uses a lot of memory. Split it first with the PDF Split tool, convert the parts separately, and join them in Word.

Is my PDF uploaded anywhere?

No. See the privacy note below.

Privacy and processing

The code contains no upload request. The file is read by your browser, and both text extraction and .docx generation happen on your machine. Opening the tool loads the pdf.js library, its worker script and the docx library, all shipped with this site; no model is downloaded and your PDF's content is never sent out.