Skip to the tool
CALCPARK

PDF to Word

Convert a text PDF to an editable Word document in your browser. Paragraphs, heading styles, lists, simple tables and page breaks are rebuilt; nothing is uploaded.

Files stay on your device
Loading the tool…

How to use it

  1. Drop the PDF. Enter a page selection (for example 1-5) if you do not need the whole file.
  2. Choose which structures to rebuild: headings, lists, tables, page breaks, and whether page numbers are dropped.
  3. Press “Convert to Word”. The result shows word, paragraph, heading, list and table counts.
  4. Download the .docx and open it in Word, LibreOffice or Google Docs.

Example

A 4-page meeting minutes PDF (11 pt body, 18 pt title, 13 pt section names, two bullet lists, one 3-column action table) converts to a 24 KB .docxwith one Heading 1, five Heading 2 paragraphs (13 / 11 = 1.18 ≥ 1.08 → level 3 would apply only below 1.25, so these become Heading 2 at 1.25–1.5 only when the ratio reaches 1.25 – here 1.18 gives Heading 3), 31 body paragraphs, 2 bulleted lists and a 3 × 6 table with a bold header row. Page breaks sit where the PDF pages ended.

How the document is rebuilt

pdf.jsextracts every text run with its position, width and font size; the page’s operator list is loaded first so the font objects expose bold/italic. Runs are grouped into lines (baseline tolerance 45 % of the font size), lines into paragraphs (gap under 1.6 × the font size, previous line at least 75 % of the text width, no first-line indent), bullet characters and “1.”-style prefixes into lists, and runs of lines with equal column counts (gaps over 1.5 em) into tables. The docxlibrary then writes paragraphs with Word’s built-inHeading1–Heading3styles, numbering definitions for lists, a 100 %-width table per grid and a page break between PDF pages.

Limits

  • Images and vector drawings are not transferred (the source tool never did; a reliable image placement from pdf.js is out of scope).
  • Multi-column pages interleave; text in headers/footers other than page numbers is kept as paragraphs.
  • Tables need visible gaps between columns; wrapped cells may split into extra rows.
  • Password-protected PDFs must be unlocked first.

Sources: pdf.js – getTextContent API, docx – Document, Paragraph, Table API, ECMA-376 Office Open XML (WordprocessingML).

Questions

Why does the tool say my PDF is scanned?
Every selected page has fewer than 20 characters of real text, which means the pages are pictures. Run them through OCR first – CalcPark's OCR tool at /ocr turns scanned pages into text – then paste or convert that text.
Will the Word file look like the PDF?
The content and structure will match; the appearance will not. Fonts become Calibri 11 pt, headings use Word styles you can restyle, and positions are re-flowed. For a visual copy, insert the PDF pages as images instead.
Are images included?
No. Only text, headings, lists and tables are rebuilt. Use PDF to JPG to extract the images and insert them in Word.