– HTML is cleaned and normalised to well‑formed XHTML (entity replacement, void‑tag handling) so the output passes EPUB validators and works with strict readers.
– manifest, spine, NCX and EPUB 3 navigation document rebuilt to meet current spec requirements; automatic TOC page added when needed.
– images are now always converted to PNG before being added to the EPUB.
These changes fix validation errors that caused some EPUBs to be rejected by stricter readers.
