A manuscript formatted for print or for a fixed-layout PDF carries a lot of assumptions that don't survive the move to an ebook: a fixed page size, page numbers meant for a table of contents, headers repeating the book's title on every page, and hyphenation tuned to a specific column width. None of that is wrong for a printed page. All of it becomes noise the moment the same content needs to reflow across phone screens, tablets, and e-readers of different sizes, which is exactly what ebook formats like EPUB are designed to do.
Why PDFs make bad ebook source files
A PDF fixes text at exact coordinates on a fixed-size page. An EPUB, by contrast, stores flowing text and lets the reading device decide how to lay it out at whatever font size and screen width the reader has chosen. Converting from one to the other is not a file format change so much as a structural one: you are going from "text glued to specific positions on a specific-sized page" to "text organized into paragraphs, headings, and chapters with no fixed position at all."
Because of this mismatch, naive PDF-to-ebook converters tend to carry over exactly the artifacts that only made sense on a fixed page: hyphens inserted because a word hit the right margin of a page that no longer exists in the target format, page numbers and running heads that repeat every few hundred words with no page boundary to justify them, and line breaks that were only ever meant to wrap text at 6 inches wide.
Step 1: extract and clean the text
Start by getting a clean text or Markdown version of the manuscript's content, with the print-specific artifacts already removed. This means rejoining hyphenated words so they read as complete terms regardless of where a line happened to wrap on the original page, merging wrapped lines back into full paragraphs, and stripping repeated titles and page numbers that would otherwise appear as stray text every few paragraphs in the final ebook. Doing this with the text cleaner before touching any conversion software means the converter is working from prose, not from a page layout pretending to be prose.
Step 2: rebuild structure with Markdown
Once the text is clean, the next problem is that plain text has no concept of chapter breaks, headings, or lists, all of which a PDF only communicated through visual formatting (a larger font, extra whitespace) that plain extraction throws away. Marking chapter titles as Markdown headings and converting bullet or numbered lists back into proper list syntax gives an ebook conversion tool the structural signals it needs to build a working table of contents and consistent chapter styling, rather than one long undifferentiated block of text.
Step 3: handle front matter and footnotes separately
Title pages, copyright notices, dedications, and footnotes behave differently from body prose and are worth pulling out and reformatting by hand rather than running through the same automated cleanup as the main text. Footnotes in particular are a common failure point: in the source PDF they are visually separated at the bottom of a page, but a naive extraction interleaves them with body text at the point where the page happened to break, not at the point where the footnote marker actually appears in the sentence. For a manuscript with heavy footnoting, it is usually faster to extract the body text and footnotes as separate passes and reassemble the reference markers manually.
Step 4: convert Markdown to your target format
With clean, structured Markdown in hand, converting to EPUB, MOBI, or a web-based reading format becomes a mechanical last step, since dedicated conversion tools handle that transformation reliably once they are not also being asked to guess at paragraph boundaries or strip artifacts. The hard part of ebook conversion was never the file format conversion itself. It was getting from "a fixed page description that happens to contain your book" to "your book," and that work happens before any converter sees the file.
The field guide covers the extraction mechanics behind each of these cleanup steps in more depth, including exactly how hyphen rejoining, line merging, and running-head detection are decided, which is useful background if you are processing more than one manuscript.