HTML to DOCX Converter

How the conversion works

This tool doesn't rebuild your markup as native Word paragraphs and runs. It packages the HTML you paste into an MHTML part (word/afchunk.mht) inside the .docx container and points document.xml at that part with an altChunk reference — the same technique the original html-docx-js library uses. Word recognizes altChunk and renders the embedded HTML/CSS itself when the file opens, rather than the tool translating tags into OOXML.

Example

Paste this and click Convert:

<h1>Meeting notes</h1>
<p>Attendees: <strong>Alex</strong>, <strong>Priya</strong>.</p>
<ul>
  <li>Approved Q3 budget</li>
  <li>Follow up with legal by <em>Friday</em></li>
</ul>

Open the downloaded converted.docx in Microsoft Word and you'll get an H1, a paragraph with two bold names, and a bulleted list, styled with Word's default Normal template rather than any CSS-driven typography.

Exact supported behavior

• Headings, paragraphs, bold/italic/underline, lists, tables, and inline style="..." attributes render, because they're valid HTML/CSS that Word's MHTML renderer understands directly.
• Images embedded as data: URIs are extracted and packaged as real embedded parts. Images referenced with a normal <img src="https://..."> URL are left untouched — Word tries to fetch them live when you open the file, which fails if you're offline or the host blocks hotlinking.
• The "Preserve styles" switch, when off, strips only inline style="..." attributes using a regex before conversion. It doesn't touch <style> blocks, class attributes, or any other markup.
• Output is always portrait orientation with 1-inch margins. The underlying library supports landscape and custom margins, but this tool doesn't expose those options in the interface.
• Pasted or uploaded HTML is used as-is. There's no sanitization, tag-balancing, or repair pass before conversion.

Known limitations

• Reliable rendering is limited to Microsoft Word (desktop and Word for the web). Google Docs' Word import, LibreOffice Writer, and Apple Pages often show a blank or broken document, because they expect native OOXML content rather than an embedded MHTML chunk.
• No headers, footers, footnotes, tracked changes, table of contents, or embedded custom fonts.
• Very large HTML — especially with many inlined data: images — produces a large, slow-to-open file. The tool doesn't enforce a size limit.
• Malformed HTML, such as unclosed tags or invalid nesting, is passed straight through, so results can be unpredictable.

Test it yourself

Paste this to check that both tables and the styles toggle behave as expected:

<table border="1">
  <tr><th>Item</th><th>Qty</th></tr>
  <tr><td>Pens</td><td>12</td></tr>
</table>
<p style="color:#2563eb;">This line should stay blue with "Preserve styles" on, and turn black if you switch it off.</p>

When not to use this tool

• Your recipients open files in Google Docs, LibreOffice, or a mobile Office app — use a server-side converter such as Pandoc, which writes native OOXML instead of an MHTML chunk.
• You need headers, footers, page numbers, or other native Word layout features.
• You're converting HTML in bulk or from a script — this is a manual, one-file browser tool with no API.
• You actually have a .docx and want HTML instead — that's the reverse direction; use DOCX to HTML.

Privacy

Conversion runs with the html-docx-js-typescript library entirely in your browser. Your HTML never leaves the page, and the .docx is assembled and downloaded locally.

Clean up HTML entities first with the HTML Encoder/Decoder, or check the word count of your source with the Word Counter before converting.