HTML to Markdown Converter

Conversion Reference

Examples

The outputs below come from the tool’s converter: Turndown 7.2.4 with the tables rule of turndown-plugin-gfm 1.0.2 and the options listed under “Output Style”.

A heading, inline code and a link with a title. Input:

<h2>Install</h2><p>Run <code>npm i turndown</code> and see the <a href="https://github.com/mixmark-io/turndown" title="repo">README</a>.</p>

Output:

## Install

Run `npm i turndown` and see the [README](https://github.com/mixmark-io/turndown "repo").

The title attribute is kept as a Markdown link title in quotes.

A table cell with a pipe and a line break. Input:

<table><thead><tr><th>Option</th><th>Values</th></tr></thead><tbody><tr><td><code>--format</code></td><td>json | yaml<br>default: json</td></tr></tbody></table>

Output:

| Option | Values |
| --- | --- |
| `--format` | json \| yaml<br>default: json |

GFM ends a cell at every unescaped |, and a table row cannot span two lines (GFM spec, Tables extension). The converter therefore writes \| and keeps the line break as <br>, so the row still has two cells.

A nested list and an ordered list that starts at 3. Input <ul><li>Step one</li><li>Step two<ul><li>Sub step</li></ul></li></ul><ol start="3"><li>Third</li><li>Fourth</li></ol> gives:

-   Step one
-   Step two
    -   Sub step

3.  Third
4.  Fourth

Output Style

• Headings use # (ATX), bullets use - followed by three spaces, emphasis uses * and strong uses **.
• A <pre><code class=“language-x”> block becomes a ```x fence. The language comes only from a language- class on the inner <code>; a class on <pre> or a data-lang attribute gives a fence with no language. A <pre> without an inner <code> is not fenced: its text is copied with its line breaks, so wrap it in a fence by hand.
• Characters that would start Markdown syntax in plain text are escaped: 5 * 3 becomes 5 \* 3, _a_ becomes \_a\_.
• Whitespace inside paragraphs is collapsed the way a browser renders it.

Limits

• Strikethrough (<del>, <s>), highlight (<mark>) and task-list checkboxes are not converted; only their text remains.
• colspan and rowspan are ignored, because GFM tables have no merged cells. A header cell with colspan=“2” becomes one header column above a two-column body, so fix the header by hand.
• The text inside <script>, <style> and <title> is not removed. Paste only the part of the page you need.
• Input over 1 MB (1,048,576 characters) shows “Input is very large — conversion may be slow.” The conversion still runs in the page.
• An <img> without a src attribute is dropped without a trace. Pages that load images lazily often keep the address in data-src until the image scrolls into view; copy the HTML after the images have loaded, or rename the attribute first.
• Relative links and image paths are copied as written. They resolve against the page where the Markdown is published, not the page you copied from.

To check the result, paste it into Markdown Preview; it shows raw HTML as text, so a <br> in a table cell or a table kept as HTML appears there as tags. To turn Markdown into a Word file, use Markdown to Word.

Common Use Cases

• CMS migration: Export WordPress or Drupal content as HTML, then convert it to Markdown for a static site generator like Hugo or Astro.
• README preparation: Scrape a web page and clean it up as Markdown for a GitHub README or documentation site.
• LLM input: Convert a web page to Markdown to reduce token count and noise before feeding it to a language model.
• Note-taking: Convert email HTML or rich-text snippets to clean Markdown for Obsidian, Notion, or similar tools.