Word to Markdown Converter

From WordprocessingML to Markdown

A DOCX is an Office Open XML package rather than a single stream of formatted text. Microsoft describes a basic word-processing package as a collection of parts, including a main document part that contains the body. This converter first checks for the expected package structure, then uses Mammoth.js to produce a semantic HTML intermediate. Mammoth’s own guidance favors semantic structure over reproducing Word’s exact appearance, which is the right boundary for Markdown.

The HTML intermediate is parsed in memory and converted to CommonMark blocks and inline elements. Simple tables use the GitHub Flavored Markdown table extension. The intermediate is never mounted raw. The preview is generated from the Markdown and sanitized before display.

Worked example

Suppose a DOCX contains a Project status Heading 1, two paragraphs, a numbered three-step plan, and a 2 × 3 table listing an owner and due date. With hash headings selected, the output begins with # Project status, followed by ordinary paragraphs, 1. list markers, and a GFM table whose first row becomes the header. If the same document contains a floating chart, the chart is not converted; the tool shows a limitation notice so the missing object is not mistaken for a successful export.

What Markdown cannot preserve

Markdown expresses content structure, not Word pagination. Pages, columns, margins, headers, footers, fonts, colors, floating positions, text boxes, charts, SmartArt, equations, macros, OLE objects, forms, comments, and tracked-change history are outside this converter’s output contract. Complex merged tables may lose their intended grid. Review technical or regulated documents carefully, and keep the original DOCX as the authoritative source.

File and privacy limits

The converter accepts .docx, not legacy .doc, .docm, ODT, or PDF. It rejects packages without the expected Office Open XML parts, encrypted ZIP entries, more than 2,000 package entries, over 25 MiB compressed, or over 80 MiB expanded. These limits reduce memory risk but do not repair damaged files.

The DOCX bytes, filename, extracted structure, settings, preview, and Markdown remain in this browser. In the multi-step flow, a random token in this tab’s sessionStorage points to an expiring IndexedDB record for up to two hours. The content is never put in the URL or sent through Livewire. Starting over deletes the tab token and stored record.