Convert DOCX to TXT

Keeping the words, discarding everything else

A .docx is a ZIP archive of XML, and the text lives in one document part inside it. Extracting it is therefore reliable in a way that pulling text out of a PDF is not — the words are stored as words, in reading order, with no glyph positions to reconstruct and no ambiguity about which column comes first. Whatever else this conversion loses, it does not lose or reorder the text.

What plain text cannot hold is everything else, by definition: bold and italic, fonts and sizes, colours, page layout, headers and footers. Structure is preserved where plain text has a convention for it — paragraphs stay separate, list items keep their markers and indentation, and table cells are separated so that rows remain readable rather than collapsing into one run-on line. Footnotes are gathered rather than dropped, since the text of a footnote is still text.

Pictures, charts and SmartArt are left out — they contain no text to extract, and a placeholder helps nobody. Equations come out in a simple linear form such as a/b + x^2 rather than as symbols, which is the honest compromise: readable, unambiguous, and not pretending to be typesetting.

Two practical notes. Tracked changes and comments are not included — what you get is the document as it currently reads, not its revision history, which is worth knowing if you were hoping to recover a deleted passage. And legacy .doc files must be saved as .docx first, because the binary format from 1997 is a completely different thing that happens to share a purpose.

When you want something else

Pandoc converts a .docx to text or Markdown with far more control: pandoc in.docx -t plain -o out.txt for plain text, or -t markdown to keep the headings, emphasis and lists as readable markup — which is usually what people actually want when they ask for "just the text". It also extracts the images to a folder rather than discarding them.

For pulling text out of hundreds of documents, Python's python-docx and docx2txt give you the paragraphs programmatically, so you can filter, search or index them rather than reading a file. And for getting at tracked changes or comments specifically, neither a converter nor a script beats opening the document — that information is in the file and is simply not part of "the text".