HTML to Markdown Converter

How to use this tool

1. Paste HTML into the box, for example a page's source viewed with "View Page Source" or "Inspect", or choose a .html file. Or press "Try an example".
2. Read the result in the Markdown box. It updates as you type.
3. Press "Copy Markdown", or download the result as a .md file.

What each HTML element becomes

The converter keeps the structure HTML elements describe, headings, emphasis, links, lists, code and tables, and writes it as GitHub Flavored Markdown. Each result below was produced by the converter itself.

## Results

**Done** and *nearly*

~~draft~~ final

[CommonMark](https://commonmark.org/)

Run `npm test`

- milk
- eggs

1. Plan
2. Write

> Measure twice

```js
let a = 1;
```

| Item | Qty |
| ---- | --- |
| Pens | 4   |

one\
two

(removed)

Why convert HTML to Markdown

HTML is built for browsers to render, with opening and closing tags around everything and attributes that have nothing to do with the words themselves. Markdown is built for people to read and write directly: a heading starts with #, a link is [text](url), and most of a page's prose needs no markup at all. That makes Markdown the format documentation sites, READMEs, note apps and most static site generators expect, and the format large language models read most reliably when you feed them a page's content rather than its tag soup.

Pasting rendered text out of a browser tab loses structure: headings become plain lines, a bulleted list loses its bullets, and a table collapses into run-on text. Converting the page's own HTML instead keeps the structure the tags describe, because the converter can see which text was a heading, which belonged to a list and which cells made up a table.

How it works, and why pasted HTML cannot run anything

The text you paste is read by a text parser, hast-util-from-html, which turns HTML into a tree the way a browser's HTML parser would, but it is not a browser: it never creates an actual page element, never fetches an image, and never executes a <script>. The same code already reads the HTML that the DOCX to Markdown tool gets from Word documents. On top of that, this tool deletes every <script> tag, <style> tag and HTML comment before the tree is turned into Markdown, so none of their contents can leak into the result even by accident.

What remains is turned into Markdown following GitHub Flavored Markdown: headings become #through ######, <strong> and <em> become **bold**and *italic*, lists keep their numbers or get a hyphen, and a <table> becomes a pipe table with a header row.

Private by design

The HTML never leaves your device. It is parsed and converted by code already loaded into this page, and the Markdown is shown in the box above. There is no upload, no account and no copy kept on a server, which matters for an internal page, a draft newsletter or anything else you would not paste into a stranger's website.

Limits

• Layout, fonts, colours and anything done with CSS are not converted; Markdown has nowhere to put them. <style> blocks and inline style attributes are ignored.
• <script> tags and their contents are removed entirely, and never run.
• Tables become pipe tables with one line per cell: a cell holding several paragraphs or a line break is joined onto one line, because a Markdown table cell cannot hold either. A table with no <th> cells of its own gets a header from its first row.
• Merged table cells (colspan and rowspan) are not merged back together; each cell is written on its own.
• Forms, embedded video, iframes and SVG are left out, since Markdown has no equivalent for them.
• An <img> with no address to point to (no src) is left out and counted, not guessed at.