Five passes that turn a manuscript into an audiobook script
Do them in this order on a copy of your final file. A 300 page novel takes an evening or two.
- 1
Headings
Make sure every chapter starts with a heading the importer can find, and nothing else looks like one.
- 2
Dialogue
Put spoken words in quotation marks and tag the speaker often enough that each line can be traced to a character.
- 3
Visual only content
Remove or rewrite footnotes, captions, tables and anything that only makes sense on a page.
- 4
Written for the ear
Spell out symbols, abbreviations and numbers that could be read more than one way.
- 5
Directions
Add a few bracketed notes on the lines that carry the most emotion. Leave the rest to the voices.
How each file type is turned into chapters
Files up to 30 MB are read in your browser, and nothing is uploaded until you generate, so you can import, check and fix as often as you like.
| File | How chapters are found | What to check |
|---|---|---|
| EPUB 2 or 3 | From the table of contents (nav or NCX) | Every chapter is listed, with no stray entries such as a map or an image page |
| Word (DOCX) | Split on Heading 1 and Heading 2 styles | Headings use the styles, not just large bold text |
| TXT | Lines such as “Chapter 3”, “Part Two”, “Prologue” or “Epilogue”; a Project Gutenberg header and license are removed | Each heading sits alone on its own line |
| Markdown | Headings # to ###, plus chapter lines as in TXT; front matter, link URLs, images and code are stripped | One heading per chapter, no headings inside the prose |
| PDF (text based) | Reflowed without page numbers or running headers and footers, then split on chapter lines as in TXT | The PDF has selectable text (scans cannot be read); footnotes in the page body are still read |
| Pasted text | Read the same way as a TXT file | Paste one chapter, or the whole book with heading lines |
Text with no headings at all is split into parts of about 4,000 words. Format specific details are on the EPUB to audiobook and Word to audiobook pages. Want to sanity check the chapter sizes too? See audiobook chapter length.
Formatting dialogue so each speaker gets the right voice
In prose, quotation marks decide who speaks. Everything inside them is dialogue; everything outside is read by the narrator. That includes inner thoughts in italics, letters quoted without quotation marks and song lyrics, which is usually what you want.
Dialogue tags decide which character. Speakers are detected from tags such as “said Mara” or “Tom asked”, and each named character gets an AI voice of their own, matched on gender where the text makes it clear. Tag the first line of every exchange, and tag again after any paragraph of action or a long run of back and forth. Long untagged volleys are where a line is most likely to land on the wrong voice, so listen to those first.
Keep names consistent in tags. If Elizabeth is “Elizabeth” in one tag, “Lizzy” in another and “Miss Bennet” in a third, check the cast list after import and give all three the same voice. Up to 10 named characters get their own voices, and the rest share an “Other voices” role.
Writing numbers, symbols and abbreviations for the ear
AI voices read most text correctly. These are the cases where it pays to spell out what you mean.
| In print | In the audiobook script | Why |
|---|---|---|
| No. 3 | Number 3 | Avoids “no three” |
| §12 | section 12 | Symbols are often skipped or misread |
| St. John Street | Saint John Street | “St.” can mean street or saint |
| $5M | five million dollars | Abbreviated amounts are read unpredictably |
| Henry VIII | Henry the Eighth | Roman numerals can come out as letters |
| e.g., i.e. | for example, that is | Listeners hear the meaning, not the Latin |
| www.example.com/books | example dot com | Full web addresses are tiring to hear |
| (see Figure 4) | Cut it | There is no figure in audio |
Invented names and unusual words need their own list. The audiobook pronunciation guide shows how to respell them.
What to remove from the audiobook manuscript
Front matter such as the cover, copyright page, contents, dedication, “also by” lists and the Project Gutenberg license is set aside automatically on import. The rest is up to you.
Remove or rewrite
- Footnote markers and footnotes. Fold an important note into the text, or drop it.
- Image captions, maps, charts and tables. Describe them in a sentence if the story needs them.
- Index, bibliography and long reference lists. Offer them as a companion PDF instead.
- Print instructions such as “see overleaf” or “turn to page 40”.
- Repeated page headers, chapter numbers in the margin or stray page numbers left in a pasted text.
Keep as they are
- Scene breaks: keep your separator on its own line and listen to the first one to confirm the pause.
- Epigraphs and part titles that set the mood, as long as they are short.
- Author’s notes and acknowledgments, ideally at the end.
Italics, emphasis and bracketed directions
Formatting is not heard. Bold, italics, small caps and colored text all come out as plain speech, so a word you italicized for stress may be read flat. Usually the sentence carries the stress on its own. Where it does not, a direction does.
A direction is a short note in square brackets at the start of a line, such as [whisper], [furious] or [like a late-night radio host]. It shapes how the whole line is delivered. Inline sound tags such as [laugh], [sigh] and [breathe] go exactly where the sound should happen; they work on the Standard, Lite and Expressive voice models, not on Premium. The AI voice emotion tags page lists what each one does.
Use directions for reveals, arguments and chapter endings, not every line. Ten in a chapter is plenty. If you plan to hand the same file to a human narrator, keep a copy without brackets.
Audiobook script template: one short chapter, annotated
Each row is one line of a plain text file. Copy the left column into a TXT file or paste it straight into the audiobook generator to hear how it is read.
| Line in your file | What happens |
|---|---|
| Chapter 1: The Lighthouse | Starts a new chapter and becomes its name |
| The fog rolled in before noon, and with it came the boat. | Narration, read by the narrator |
| “Is anyone up there?” Mara called. | Dialogue with a tag; Mara gets her own voice |
| Tom leaned over the rail. | Narration again |
| [whisper] “Keep your voice down,” Tom said. | A direction at the start shapes the delivery |
| “Fine.” She sighed [sigh] and climbed the stairs. | An inline sound tag plays where it sits |
| Chapter 2: Low Tide | Starts the next chapter |
Writing a screenplay or stage play rather than prose? Speaker names, character cues and parentheticals are cast as written, with no quotation marks needed; the AI table read page shows the layout.
