Guide
EPUB to audiobook: the complete guide
Everything that happens between dropping an EPUB and hearing chapter one, and how to get a better result from the files you already have.
1. Why EPUB is the ideal input
An EPUB is a zipped website: XHTML files for the chapters, a manifest listing them, a spine giving their order and a navigation document naming them. Every piece of structure an audiobook needs is already there. PDF, by contrast, only knows where glyphs sit on a page, and plain text knows nothing at all.
That is why an EPUB converts into narration with an accurate chapter list and why the first chapter starts where the author meant it to, provided the file is reasonably well made.
2. What “well made” means
•
One file per chapter (or clear headings inside a single file). Retail EPUBs and Standard Ebooks editions are like this; a quick Calibre conversion of a Word file may not be.
•
Real headings (h1, h2), not just bold text. Headings become chapter markers.
•
A navigation document that matches the spine. When it does not, headings are used instead.
•
No DRM. Store purchases with DRM cannot be opened by any third-party app.
3. How chapters and front matter are detected
Book2Speech extracts paragraphs and headings in spine order, then classifies each section as front matter (title page, copyright, dedication, contents, preface, foreword), a part divider, body, or back matter (notes, glossary, index, about the author). Contents pages that repeat every chapter title are recognised as a list rather than a chapter. A short language-model pass refines the classification on ambiguous books, and the player suggests the first body chapter as the starting point.
You can always start anywhere; the suggestion just saves you from listening to the ISBN.
4. Choosing a narrator
There are 260+ narrators in 40+ languages, each with a sample. A few rules of thumb: first-person novels sound right in a voice that matches the narrator character; non-fiction benefits from a measured, clear voice; dialogue-heavy fiction from a lively one. The language of the book is detected and matching native voices are listed first. You can change narrator mid-book; already generated chapters are re-narrated on demand.
5. What happens while you listen
Narration is generated a little ahead of your position and streamed, so a 300-page novel starts within seconds rather than after a long render. Audio you have heard is cached, so replaying a chapter or switching devices does not use your allowance again. Your position syncs between the web player and the iPhone app.
6. When a file fails
•
“DRM-protected or encrypted”: the file cannot be read. Get a DRM-free copy from the publisher, or use a different edition.
•
“Too large”: files are limited to 5 MB. Illustrated EPUBs are mostly images; strip them in Calibre (Convert → EPUB → Look & Feel → remove images) and try again.
•
One giant chapter or none: the EPUB has no headings. Add chapter headings in Calibre’s editor or Sigil, or accept the automatic splits.
•
Wrong language voices listed first: the language guess came from the text; just pick the narrator you want.
7. Formats that are not EPUB
MOBI, AZW3, FB2, DOCX, TXT, RTF and HTML are all accepted directly; each is converted to the same internal structure. See the converter pages for what is specific to each.
Questions people ask
Is the result an audiobook file?
It is a listenable book in the web player and the app, with chapters, position and read-along text. A full-book file can be requested from the book page.
How long does a whole book take to narrate?
You do not wait for the whole book: playback starts within seconds and generation stays ahead of you. A novel of 90,000 words is roughly ten hours of listening.
© 2026 Book2Speech