Local source conversion

HTML to Markdown Converter

Turn HTML source into readable Markdown for a README, documentation, or a content migration. Keep common document structure, then review what changed before publishing.

Resolves relative links and image paths locally. No page is fetched. Use an HTTP or HTTPS URL without credentials, up to 2,048 characters.

Up to 200,000 UTF-16 characters. Oversized input is rejected, never shortened. Ctrl+Enter or Cmd+Enter converts.

0 / 200,000 characters

.html, .htm, or .txt, up to 600,000 bytes. Loading a file replaces the source field. Nothing is uploaded.

No conversion yet.

Output is plain text. Images and linked pages are never loaded here. Your destination Markdown renderer may load image URLs later.

Ready. Paste HTML or load the example to begin.

How to convert HTML source

  1. Paste an HTML fragment or open a local UTF-8 source file. A full document is accepted; its head is omitted.
  2. Choose GFM for simple tables, task lists, and strikethrough. Supply the original page URL if you need absolute destinations.
  3. Select Convert to Markdown. Check the conversion review, then copy or download the text and inspect it in your destination renderer.

What the converter preserves

Headings, paragraphs, bold and italic text, line breaks, blockquotes, horizontal rules, nested lists, ordinary links, image alt text, inline code, and fenced code blocks are supported. Ordered lists keep their starting number. Code blocks keep indentation and blank lines, with language hints from language-* or lang-* classes. Normal prose whitespace collapses as in HTML; nonbreaking spaces are retained.

GFM converts rectangular tables with inline-only cells. A table without a header receives an empty header so its first data row is retained. Merged cells, nested tables, block content, multiple header rows, and uneven rows become labeled row text with a notice. CommonMark always uses row text. Table cell line breaks become spaces.

Understand conversion losses

This is a bounded source parser, not a full HTML5 browser or a lossless converter. Use balanced, explicit tags for predictable results. Common omitted list, paragraph, and table end tags receive limited recovery; unusual malformed markup may be rejected. CSS, layout, IDs, link targets, event handlers, embedded content, form controls, scripts, and styles do not carry into Markdown. Unknown tags keep their text where possible; SVG, MathML, templates, and active content are omitted with notices.

Common semicolon-terminated entities such as &,  , ©, and valid numeric references are decoded. Unicode spacing in prose uses numeric references so renderers retain it at text edges. Other names and references without semicolons remain literal, with a notice. Compatible adjacent inline styles combine. Uncertain emphasis boundaries retain their text without the formatting and receive a specific notice. Adjacent lists may use alternating markers to preserve separate lists and starting numbers. List numbering resets, reversed lists, and custom list types are reported when they cannot be represented faithfully.

Limits are 12,000 parsed nodes, 64 nested elements, 200 rows and 30 columns per table, 1,000,000 output characters, and a short processing deadline. Reduce or split a document if a limit is reached. No partial result is exported after an error.

Privacy and standards

Tool code processes your source in this browser tab. It does not upload it, store it, or fetch URLs. Pasted markup is never inserted into the page or parsed into browser HTML elements. Clear all removes the current input and output from the interface. The site can still use its shared analytics; this tool adds no tracker.

Output follows the supported constructs in CommonMark and optional GitHub Flavored Markdown extensions. HTML parsing rules cover more cases than this utility. Review external destinations and the final rendered document before publication.