How to convert HTML to Markdown on iPhone

HTML is a delivery format. Markdown is a working format. If you have saved an article, exported a note from an app that only writes HTML, or received a formatted document and want to edit it rather than admire it, Markdown is what you want: plain text you can open in any editor, put in version control, paste into a notes app, or feed to anything else.

The alternative — copying rendered text out of a browser — loses every heading and list and gives you a soup of paragraphs. Converting keeps the structure as syntax you can see and edit.

Convexy parses the HTML into a styled document, then reads that document back out as Markdown. That second step is a reconstruction, not a lookup: a styled document knows that a run of text is 24-point and bold, but not that it was originally an <h2>. So heading levels are inferred from how much larger the text is than the document's body size, and lists are recovered from their list structure and markers.

In practice this works well, because that inference is exactly what a human reader does. It also means the honest guarantee is a two-part one:

• Text is never lost. Every word in the document comes through.
• Structure may be. Anything the reconstruction cannot recognise degrades to an ordinary paragraph — with its inline formatting intact.

• Headings — as #, ##, ###, based on their relative size in the document.
• Bold and italic — as **bold** and *italic*, including both together.
• Bullet and numbered lists — as - items and 1. items, with numbering preserved.
• Links — as [text](https://example.com), with the target URL intact.
• Monospaced text — as `inline code`.
• Paragraph structure — blank lines between paragraphs, as Markdown requires.

Characters that mean something in Markdown are escaped on the way out. A literal asterisk in your prose is written as \*, so it survives as an asterisk instead of accidentally turning half the sentence italic when the file is next rendered. A line that happens to start with # or - is escaped too, so it does not silently become a heading or a bullet.

• Tables. A table flattens into paragraphs of its cell text. The grid is gone.
• Images. Pictures are not written out as Markdown image references; the file has no image in it, and the reference is not recreated.
• Nested lists. Sub-items are flattened to the top level rather than indented.
• Code blocks. Preformatted blocks come back as monospaced text, not as fenced ``` blocks — you get the code, but you may want to re-fence it.
• Strikethrough, colours, fonts, alignment, and CSS in general. Markdown has no way to express them, so they are dropped rather than faked.

None of that is recoverable by a better setting; it is what happens when you move from a format that can express anything to one that deliberately expresses very little. That constraint is why Markdown is pleasant to edit.

Markdown converts back to a complete standalone page — see Markdown to HTML. Round-tripping HTML to Markdown and back will not reproduce the original file byte for byte, and nothing that works this way ever will: the second file only contains what Markdown was able to express.