The five steps at a glance
- 1
Prepare the manuscript
A final, clean EPUB, DOCX or text file with clear chapter headings.
- 2
Cast the voices
A narrator plus a distinct AI voice for each speaking character.
- 3
Direct key lines
Bracketed notes such as [whisper] and sound tags such as [sigh] where the scene needs them.
- 4
Proof-listen and fix
Listen at normal speed and regenerate only the lines that sound wrong.
- 5
Export and publish
Download an M4B with chapters or MP3 files and upload them where AI narration is allowed.
What you need before you start
Five things decide whether the project takes an afternoon or a month.
- A final manuscript. Edits after narration mean regenerating audio, so lock the text first.
- The audio rights. Your own book, a public domain classic or a title where you hold the audio rights. Ebook rights alone are not enough if you sold audio rights to a publisher.
- A short cast list. Note who speaks most, their age and accent. It speeds up casting and keeps voices consistent across a series.
- Time to listen. Plan to hear every chapter once at normal speed. For a 9 hour book that is a day of focused listening.
- A plan for where it sells. Store rules for AI narration differ, so choose your stores before you choose a file format.
Step 1: prepare your manuscript
Start from the cleanest file you have. An EPUB is best because its table of contents already marks every chapter; the EPUB to audiobook converter turns each entry into a chapter marker. A Word file works too if chapters use Heading 1 or Heading 2 styles, and text-based PDFs can be reflowed with the PDF to audiobook converter.
Remove what should not be read aloud: page numbers, image placeholders, long tables and footnote markers. Spell out symbols a voice may stumble on, such as “No. 3” or “§12”. Add a short list of invented names and how to say them; the pronunciation guide shows how to write respellings an AI voice follows.
Finally, check chapter lengths. Listeners stop between chapters, so a 90 minute chapter is tiring. The free EPUB chapter extractor shows the word count and listening time of every chapter before you generate anything.
Step 2: cast the narrator and characters
When you upload the book, every speaking character is detected from the dialogue tags (“said Alice”, “Holmes replied”) and given a voice of their own, matched on gender where the text makes it clear. A narrator reads everything outside quotation marks, and minor speakers share an “Other voices” role.
Review the cast before you generate. Pick a narrator you would happily listen to for ten hours, then make sure the two or three main characters sound clearly different from each other. Keep the same voices for every book in a series. If you are unsure whether a full cast suits your story, compare it with one voice in full cast vs single narrator.
Same cast in every bookStep 3: direct lines with brackets and sound tags
AI voices read neutral prose well; emotional peaks are where they need help. Type a short note in brackets at the start of a line, such as [whisper], [furious, through clenched teeth] or [trying not to laugh], and the voice performs it. Inline tags such as [laugh], [sigh] and [breathe] land exactly where you put them.
Direct sparingly. Ten good directions in a chapter beat a hundred, because constant notes make delivery sound forced. Focus on reveals, arguments and the last line of each chapter. The full tag list is on the AI voice emotion tags page.
[say with hyped, high energy like a YouTube intro]
Stop scrolling. In the next sixty seconds, I'm going to show you the trick that doubled my channel.
Step 4: proof-listen and fix single lines
Generate one chapter first and listen to all of it at normal speed, with the text in front of you. Note three things: wrong names, wrong speakers and flat delivery on lines that matter. Because every line is its own clip, fixing one means regenerating only that line, not re-recording the chapter.
Once chapter one sounds right, generate the rest of the book. Spot-check each chapter rather than relistening to everything twice, and keep a running list of name spellings so later chapters match earlier ones.
Step 5: export M4B or MP3 and publish
Export one M4B file with chapter markers for audiobook apps such as Apple Books, BookPlayer and Audiobookshelf, or one MP3 per chapter for stores and players that expect separate files. M4B vs MP3 explains which stores want which. Add a square cover at 3000 × 3000 pixels; the audiobook cover size guide covers the specs.
Then pick where it sells. At the time of writing, ACX does not accept AI narration uploaded by authors, while Google Play Books, Kobo and several distributors accept disclosed AI narration under their own rules, and you can always sell direct. Compare current policies in where to sell AI audiobooks.
The Velveteen Rabbit.m4b
6 chapters · 41 min · chapter markers
- M4BOne file
- MP3Per chapter
- ZIPWhole project
Line 48 regenerated, the rest of chapter 3 kept as is
Four ways to make an audiobook, compared
AI narration is not the only route. This is how the options compare for a typical 80,000 word novel, about 9 finished hours.
| Method | Typical cost | Time | Best for |
|---|---|---|---|
| Full-cast AI narration | About $80 to $110 in credits | A day or two, mostly listening | Dialogue-heavy fiction, series, fast releases |
| Single-voice text to speech | Low or free | Hours | Personal listening and rough proofs |
| Record it yourself | Microphone and room treatment | 4 to 6 hours per finished hour | Memoir and nonfiction in the author’s voice |
| Hire a narrator | Often $2,000 to $4,000 or more | Weeks to months | Titles that must go to ACX or need a famous voice |
Narrator rates vary widely. Run your own numbers in the audiobook cost calculator.
How to make an audiobook for free, and the limits
You can try the whole workflow free: importing a book, chapter detection and casting cost nothing, and signing up with Google gets you 100 gift credits (valid for 7 days, about 12 minutes on Standard; usage varies by text and voice tier) to generate a short sample with every voice. That is enough to hear whether AI narration suits your book before you spend anything.
A whole novel needs paid credits, because high quality speech models cost money to run. Free alternatives exist, such as the read aloud features in browsers or open source TTS on your own computer, but they read in one voice and leave chapter files, timing and casting to you. Open source audiobook generators compares the trade offs.
