EPUB, PDF or DOCX: Which File Makes the Cleanest Audiobook?

Of the 10 formats the audiobook importer reads, EPUB gives the cleanest chapter split, DOCX comes a close second when you use real heading styles, and a text-based PDF is the one to reach for last. The reason is simple. EPUB and DOCX carry their structure with them, and a PDF carries only the look of a page.

Chapters matter more than they sound like they do. Each chapter becomes a unit you can cast and regenerate, then export, and the M4B chapter markers come from that same list. A bad split at import follows you all the way to the finished file.

Best file format for an audiobook: the short answer

If you have the choice, send the EPUB through the EPUB to audiobook importer. If you only have a manuscript, send the DOCX. Keep the PDF for the day nothing else exists, and read on for what to check when it is the only option.

EPUB: chapter titles from the table of contents

An EPUB is a zipped set of web pages with a navigation list attached. The importer reads that list first, the EPUB 3 contents document if there is one and the older NCX file if not, and uses it to name each chapter. Headings from H1 to H3 start a new chapter inside a file. H4 to H6 are treated as sections within one. Public-domain EPUBs from Project Gutenberg are a good example of why the headings matter: they often pack several chapters into one file, and the heading split separates them again.

That arrangement fixes a problem you see constantly in ebooks: one file holding three chapters, or a chapter title set as an image, or a heading faked with a styled paragraph. With a proper contents list the chapter still arrives with its real name rather than 'Chapter 1' and 'Chapter 2' all the way down.

Front matter is the usual leftover. Cover pages, title pages and a printed contents page are dropped when they look like boilerplate, but that is a best guess. Open the first and last chapters in the editor before you generate a single point's worth of audio.

DOCX: headings do all the work

A Word file is a manuscript with styles attached, and the importer cares about one family of them. Microsoft Word ships with built-in styles named Heading 1 through Heading 9, and Google Docs offers Heading 1 through Heading 6. Heading 1 starts a new chapter, and so do Heading 2 and Heading 3. A line that is merely bold, bigger and centred is body text to the parser, however obviously it looks like a title to you. Page breaks do not create chapters either.

1. 1

   Style every chapter title

   Select the title, then pick Heading 1 from the Styles gallery in Word, or Format, Paragraph styles, Heading 1 in Google Docs.

2. 2

   Demote scene titles

   Heading 2 and Heading 3 also start chapters. If a scene break carries a label, use a lower heading level or plain body text so the book is not chopped into 200 pieces.

3. 3

   Delete the printed contents page

   A typed list of chapter names is just more text to narrate. Remove it, or let the importer drop it and check that it did.

4. 4

   Save as DOCX

   From Google Docs use File, Download, Microsoft Word. The older .doc format is not one of the 10 supported formats, so make sure the extension is .docx.

Google Docs can also export an EPUB. It is tempting, but the result depends heavily on how plain your document is, and the contents list may not come through. Check it before trusting it. The DOCX route is the safer one.

EPUB vs PDF for audiobook conversion

A PDF is a page description. It stores glyphs at positions, not paragraphs or chapters, so everything the importer knows about structure is inferred. Chapter breaks are guessed from lines that start with 'Chapter', 'Prologue' or 'Part', or from short all-caps lines, and paragraph breaks are rebuilt from how the lines wrap.

The first requirement is a text layer, and there is no OCR step. A scan is a photograph of text. It will be rejected, and renaming the file changes nothing. Try to highlight a sentence in your PDF viewer; if you can, the text is there. If you cannot, our guide to scanned PDF to audio covers the fix before you upload.

Even a PDF with a good text layer drags its page furniture along. Running headers, page numbers and hyphenated line breaks arrive as text, and the narrator will read them aloud. That is the real price of choosing PDF: you pay in cleanup time. The PDF to audiobook page walks through the cleanup. If a DOCX or the original EPUB exists anywhere, use it instead.

What a wrong heading level costs you

Picture a 24-chapter, 90,000-word novel (3,750 words a chapter) where every chapter has three scene breaks, and the author labelled each scene with Heading 2. The importer sees 24 chapter titles plus 72 scene titles and builds a list of 96 chapters. Nothing is broken, exactly. But now the M4B has 96 chapter markers, the editor is 96 items long, and a listener who skips ahead lands in the middle of a chapter.

Flip it round and the failure is quieter. A Word file with 24 bold titles and no heading styles imports as one chapter of roughly 90,000 words. You can still generate it, but you cannot regenerate chapter 14 on its own, and the exported book has a single marker at the start.

Both fixes take minutes in the source file and hours in the editor, so make them at the source. If you want each chapter as its own file for a store upload, convert chapters into audiobook audio once the list looks right.

TXT, Markdown and the rest

Plain text and Markdown sit in the middle. They carry no styling, so chapters are found from heading-looking lines, and a Markdown file with # markers makes the intent explicit. MOBI and the AZW family work when they are DRM-free. FB2, ODT, RTF and HTML follow the same heading logic as DOCX. All 10 formats end in the same chapter list, so the choice affects cleanup time, not what you can do afterwards.

Fixing a bad EPUB with Calibre or Vellum

If your EPUB loads with the wrong chapters, repair the file rather than fighting the editor. Calibre's conversion tools detect chapters with XPath rules, by default headings that contain words like 'chapter' or 'part', and build a contents list from them. If the input already has a contents list, Calibre uses that unless you tell it not to. Adjust the rules, convert to EPUB again and re-import.

Book-formatting apps such as Vellum can generate EPUB files for you. Whatever tool built the file, open it in an ebook reader and click through the contents list. If the reader shows a clean list of chapter names, the importer almost always will too. Then run the checklist below before you generate anything.

• The file is DRM-free and opens in a normal reader.
• The chapter count in the editor matches the book's real chapter count.
• No chapter is a lone title, a copyright page or a printed contents list.
• Chapter 1 starts where the story starts, not on a cover or a dedication.
• Spot-read chapter 1 and the last chapter for stray page numbers, footnote markers and running headers.

Five minutes here saves points later. Every line you regenerate because of a stray page number spends its characters again, and a calm documentary narrator will read 'Page 47' with exactly the same composure as the first sentence of the book.

• I have a finished ebook

  Upload the EPUB and let the contents list name the chapters.

• All I have is a PDF

  Check for a text layer first, then clean the page furniture.

• I want to control the chapters

  Style your headings, then turn each chapter into its own audio file.

Frequently asked questions