Extract the readable text from HTML with a parser rather than a renderer, so nothing in the markup can run and no image in it is ever requested.
Waiting for markup
Parsed rather than rendered — nothing in the input can run, and no image in it is fetched.
Readable text appears here
Headings and paragraphs keep their breaks, lists keep their shape, and character references are decoded.
HTML to Text runs entirely on your device — nothing is uploaded.
You’re on the Free plan. Compare plans
There are two ways to turn HTML into text, and the difference matters more than the output suggests.
The quick way is three lines: create a <div>, assign the HTML to its innerHTML, read back textContent. It is also an invitation. The moment that element touches the document, any onerror handler or <script> in the pasted markup runs with the page's privileges. Even left detached, assigning innerHTML makes the browser request every image URL in the input — which tells whoever controls those URLs that someone pasted that document, and when.
This tool reads the token stream instead and returns characters. Nothing from the input is ever placed in the page as markup, no URL in it is ever requested, and the result is rendered into a <pre> as text. Paste an email with a tracking pixel in it and nothing is tracked.
The extraction itself tries to keep what a reader would keep. Headings and paragraphs are separated by blank lines, list items get their own line with a bullet, <br> becomes a line break, and the contents of <pre> keep their shape. Script, style and head metadata are dropped entirely, because none of it is readable text. Character references are decoded, so & becomes an ampersand and — becomes a dash.
Links can be collected and listed at the end, numbered, which is what makes a converted email or article still usable — the text of a link tells you nothing about where it went.
Yes, which is the point of building it this way. The input is parsed rather than rendered: nothing is assigned to innerHTML, nothing is inserted into the page, and no URL in the markup is requested. A script tag in the input is text, and a tracking pixel in it is never fetched.
They are dropped entirely, along with head metadata. None of it is readable text, and including it would make the output useless for the thing people usually want it for.
Yes. & becomes an ampersand, — becomes an em dash, and numeric references resolve too. The same parser the HTML Entity tool uses handles it, so the two agree.
Table cells come through as text with their row structure as line breaks, which is readable but not aligned. For a table you want to keep as a table, the Markdown Table Formatter or the JSON to CSV tool is a better fit.
Because the text of a link says nothing about where it points, and that information is usually the reason someone is converting the document. It can be turned off if you only want the prose.
No. A regular expression cannot tell a tag from a less-than sign inside a script or an attribute value, so it both misses real tags and destroys content that only looks like one. This reads the document with a tokenizer.