Word to text
Drop a .docx and take the words out of it, as plain text or as Markdown.
The document is read here, not uploaded. This gets you the writing, not the
layout.
Drop a Word document here or click to choose
.docx only — read on your device, never uploaded
Markdown puts a backslash in front of punctuation, so an ordinary full stop is written as a backslash and a dot. Nothing in your writing is then mistaken for formatting. Those backslashes are not in your document and are gone again as soon as the Markdown is rendered.
This takes the words out; it does not rebuild the layout. Pictures, page design, headers and footers, comments and colours are not carried over, and a table arrives as one cell after another rather than as a grid.
How it works
-
Choose or drop a Word document. The first bytes are checked before anything else happens,
so a file that is not really a
.docxis named for what it is straight away rather than after a failed conversion. -
Press Extract. A
.docxis a zip full of XML, and it is unzipped and read on a background thread of this tab, so the page keeps responding and Cancel keeps working even on a document of several hundred pages. -
Switch between plain text and Markdown, copy what you need, or download it as a
.txtor.mdfile.
What survives, and what does not
This is the part most converters leave you to find out for yourself. Extracting text is not converting a document: the words come across and almost everything around them does not. Here is what happens to each thing a Word file can hold — measured against a document built to contain every one of them, not guessed.
| In the Word file | Plain text | Markdown |
|---|---|---|
| Headings | The words, as an ordinary line. | Kept as # and ## headings. |
| Bold and italic | The words, with the emphasis gone. | Kept as __bold__ and *italic*. |
| Bulleted and numbered lists | The words, without the bullets or numbers. | Kept as - and 1. lists. |
| Tables | Every cell becomes its own paragraph, one after another. Rows and columns are gone. | The same. Not turned into a Markdown table. |
| Pictures | Dropped. | Dropped. |
| Headers and footers | Dropped, including page numbers. | Dropped. |
| Footnotes and endnotes | Dropped — their text is not in the file at all. | Kept, collected at the end with links back to the reference. |
| Comments | The comment text is dropped. The sentence it was attached to stays. | The same. |
| Tracked changes | Shown as if every change were accepted: insertions are in, deletions are gone. | The same. |
| Text boxes | The words are kept, in the order they appear in the file. | The same. |
| Fonts, colours and sizes | Gone. Plain text has nowhere to put them. | Gone. Markdown has no colour. |
| Page breaks and page layout | No trace. Nothing marks where a page ended. | No trace. |
| Links | The link text is kept, the address is not. | Kept as [text](address). |
| Turkish and other non-English characters | Intact — ç ğ ı ö ş ü İ and the rest come through unchanged. | Intact. |
Two of those are worth reading twice. Tracked changes come out as though every one of them had been accepted, so a document still under review will hand you the proposed wording rather than the current wording — and the deleted sentences simply are not there to notice. Footnotes are the other: plain text has nowhere to put them and drops them silently, which is why the Markdown option exists and why the tool says so on screen when it finds any.
One thing about the Markdown looks wrong the first time and is not: punctuation comes back
with a backslash in front of it — örnektir\., e\-posta. That is
Markdown's own escaping, and it is there so that a full stop after a number, or a dash at
the start of a line, is not mistaken for a list. None of it is in your document, and none of
it survives once the Markdown is rendered.
Where your document goes
Nowhere. Your browser opens the zip, reads the XML inside it and writes out the text, all in this tab. Nothing is posted, because there is no server of ours that could receive it.
That is worth more here than on a tool for holiday photographs. The documents people want the text out of are contracts, CVs, tenancy agreements and letters from a lawyer — the exact files nobody should be uploading to a stranger's server to have converted, and the exact files most free converters take a copy of on the way through. The honest small print: this page loads, and the ads on it make their own requests, so a network tab is not blank. What you can check is that none of those requests carries your document.
Frequently asked questions
Why will it not take my .doc file?
Because .doc and .docx are two different formats that happen to share most of a name. A .doc is a binary compound file from Word 97–2003; a .docx is a zip full of XML, and it is the second one this tool reads. Renaming the file does not convert it, which is why the tool checks the bytes rather than the extension and tells you when the name is lying. The fix takes half a minute in the program you already have: open the file in Word, LibreOffice or Google Docs and use Save as → Word Document (.docx). The same goes for .rtf, .odt and Apple Pages documents, and each of those gets its own instructions when you drop one here.
What is the difference between the plain text and the Markdown?
Plain text is only the words: no headings, no emphasis, no bullets, nothing but paragraphs separated by blank lines. Markdown keeps the structure that can be written in characters — # for headings, - for list items, __bold__, *italic*, and links with their addresses. There is one difference that is easy to miss: footnotes and endnotes are dropped from the plain text entirely, while Markdown collects them at the end. If your document has notes in it, take the Markdown. The tool says so on screen when it finds any. One more difference worth expecting: the Markdown escapes punctuation with backslashes, so an ordinary full stop is written as \. That is normal Markdown, it is not in your document, and it disappears the moment the file is rendered.
Does it keep the layout?
No, and nothing that only reads text could. Page size, margins, columns, fonts, colours, floating images, headers and footers are all layout, and none of them exists once the words are on their own. A table is the one that surprises people: it comes out as one cell after another, because a run of plain text has no columns to put them in. If you need the layout, you need the document, not the text — keep the .docx.
The document is a scan. Why is the result empty?
Because a scan pasted into Word is a picture of words, not words. There is nothing in the file for a text extractor to find, and the honest answer is an empty result rather than a guess. That case is what optical character recognition is for: our image to text tool reads the words out of a picture, and the document scanner can build a PDF you can search. Both of those also run in your browser.
Can I turn my Word document into a PDF here?
No — this tool extracts text and nothing else. You almost certainly do not need us for that anyway: Word, LibreOffice, Google Docs and macOS all export to PDF locally, on the machine the document is already on, and they keep the layout that any browser-based conversion would have to throw away. Use the program that wrote the file.
How big a document will it take?
Up to 50 MB, and the size of the file is rarely what matters. A .docx is a zip, and the pictures inside it are never unpacked here, so a document that is 40 MB of photographs costs almost nothing to read — measured at 91 milliseconds, against a fixture the test suite rebuilds every run. What costs time is the sheer amount of writing: 300 pages of prose is a file of about 27 KB and takes about two seconds to read on a desktop browser. That work happens on a background thread, so the page stays usable while it runs. The 50 MB line is about holding the file in a phone browser at all, not about the work.
Where does my document go?
Nowhere. The file is read by your browser, unzipped by your browser, and converted by JavaScript running in a background thread of this tab. There is no request that carries it anywhere, because there is no server of ours in the path. Being precise: the page itself loads, and the ads on it make their own requests, so a network tab here is not silent — the claim worth checking is the narrower one, that nothing leaves carrying your document. It matters here more than on most of our tools, because the documents people extract text from are contracts, CVs and letters, which is exactly the category you should not be handing to a stranger.