Convert PDF to HTML

Preserved pages or a flowing article

A PDF describes where each element sits on a finished page. Preserved pages keeps those coordinates in one SVG per page, with selectable text and embedded fonts, paths and supported images. Formulas, headers, page numbers and mixed page sizes stay in place. A narrow window scales the page down without rearranging its contents.

Flowing article serves a different purpose. It recovers headings, paragraphs, lists, tables and footnotes so the text can reflow or be reused in a CMS. It does not preserve the page design. Neither mode turns selectable text into a page image.

How Flowing article rebuilds reading order

When the PDF is tagged, Flowing article follows the document's own tag tree. For untagged PDFs it infers order from geometry: columns are found from the gaps, lines are grouped into blocks, and the blocks are read down each column. Heading levels are inferred by comparing text sizes against the measured body size. These are inferred semantics, not a guarantee of an accessible reading order.

Running furniture is handled separately. Headers and page numbers that repeat on page after page are detected early and removed after the tables settle, so the phrase at the top of every page does not get stitched into the middle of your text.

Tables and footnotes in each layout

Preserved pages keeps table cells and footnotes in their drawn positions without adding table or footnote semantics. Flowing article detects ruled tables from their lines and unruled tables from aligned text. It writes table elements with marked headers and joins detected table continuations across pages.

Footnotes come back as a linked pair: the superscript marker in the text points to the note, and the note points back. Notes gathered at the back of the document become one endnotes section. The receipt counts the pairs, and a marker with no matching note is reported instead of invented.

What does not carry, and what to use instead

Preserved pages carries supported images, but refuses drawing features it cannot preserve, including Type 3 fonts, soft masks and text clipping. It does not silently switch layouts. Flowing article leaves images out; Extract images can save them separately. A PDF with no selectable text needs OCR before either conversion.

The check before the download

Preserved pages requires the same non-space character identities and counts on every page as an independent PDF text read. Flowing article requires at least 98 percent retention overall and 90 percent per page after deliberate removals. Both parse the HTML back and check the emitted text sequence. A failed check prevents a download.

Text checks do not prove a pixel match. Different viewers can substitute different fonts in PDFs without embedded fonts and rasterize thin rules or image edges differently. Compare the preview and inspect the saved file in the browser where it will be used.

The other direction, and the other format

To go the other way, from markup to fixed pages, How to convert HTML to PDF covers the reverse conversion in HTML to PDF, with its own strip list and its own count. And when what you actually want is text for an editor or a wiki rather than a webpage, PDF to Markdown runs the same structure recovery and writes Markdown instead of HTML.