Word to Markdown

A .docx is a ZIP of XML. Words sit in the document part, the outline sits in paragraph styles, and pictures sit in a media folder. Git treats that package as a binary, so a one-sentence edit is not a one-sentence diff, and a static site cannot render it without Word.

Markdown is what docs-as-code already stores. Hugo, Docusaurus, MkDocs, and GitHub render it, and Obsidian stores it as a note. Headings and lists also travel better into ChatGPT, Claude, Gemini, or Perplexity than Word HTML from the clipboard. The guide on how to send Markdown to ChatGPT covers that handoff for converted text: a prefilled prompt when the link fits, and a paste when it would be too long, which is what Send to AI on this page does for the finished note. Check length in the Token Counter. This page does not count tokens, and it does not upload the .docx.

Mammoth reads the package and emits HTML. That HTML uses the same cleaner and Turndown rules as HTML to Markdown. Heading 1 becomes #, Heading 2 becomes ##, through Heading 6, including the style id Word stores as Heading1. Bold becomes double asterisks, italic becomes single asterisks, and a hyperlink keeps its URL. A Word bullet or numbered list becomes a Markdown list. A hyphen on a Normal paragraph does not.

A simple table becomes a pipe table, and a pipe inside a cell is escaped. Merged cells stay as HTML. Rebuild that grid in the Markdown Table Generator if you want alignment you control. Pictures are off by default. Each image becomes a placeholder. When the picture has alt text, the placeholder names it. Empty alt text becomes a plain image mark. Those brackets are escaped in the raw Markdown, and the preview shows them as text. Turn on “Include images (embedded as data URIs, makes output large)” to inline the bytes. Past 200,000 characters, the box shows the start, while Copy and Download still use the full note.

The note below is the output for a small .docx: Heading 1 “Quarterly notes”, bold, italic, and a link to https://webink.app/tools, two bullets, a header row, then Chinese and an emoji. Images are off.

# Quarterly notes

Read the **bold** and *italic* [guide](https://webink.app/tools).

- Use real heading styles
- Keep lists as lists

| Style | Markdown |
| --- | --- |
| Heading 1 | # |

中文与 emoji 保留 ✨

The title is an h1 because the style was Heading 1. The hash inside the table stays cell text. The Chinese sentence is unchanged.

Apply Heading 1, Heading 2, and Heading 3 from Word’s Styles gallery. Normal text at 24 point stays a paragraph. Use Word’s bullets or numbering, not a typed hyphen. Only .docx is accepted, up to 20 MB. Over 8 MB the page warns that the work stays in the tab. A .docx starts with PK. Legacy .doc does not, and renaming it does not convert it. In Word, use Save As, Word Document (.docx). In Google Docs, use File, Download, Microsoft Word (.docx).

Insertions are kept. Text inside a delete mark is dropped, and a deleted paragraph can fold into the next one with no line break, so this is not a redline. A comment becomes a marker in the sentence, then the label and the comment text after the body. Footnotes follow the body as a numbered list. Text-box text is often kept, without the box position. Mammoth lists up to 40 notes under the result. The tools hub says which page to open when the source is not a .docx.

Copying from Word is enough for a sentence. For a spec, the clipboard often drops heading styles and real lists. This page writes a .md file named after the document and lists the warnings the converter returns. A merged-cell table still needs a careful edit.

Pandoc is what you install for a batch or for coverage this page does not promise. This page does not run Pandoc. The WebInk extension converts the page in your tab and does not open a .docx from disk. A public URL belongs on URL to Markdown, which fetches that page on the server.