Full guide
Why audiobook pronunciation matters to listeners
Audiobook pronunciation is what listeners notice first and forgive last. Readers argue for years about how to say the names in their favorite fantasy series, and the audiobook settles it: once a name is spoken, that is how fans hear it. A name said two different ways in two chapters sounds like an error, because it is one.
AI voices have a particular weakness here. They read common words well, but an invented name, an Irish first name or a town in your made-up kingdom is a guess. FullCastBook has no pronunciation dictionary, so the fix lives in the text: you respell the word where it appears, test it, and regenerate only the lines that need it. This guide shows how, for any genre.
Make a pronunciation list before you generate
Good audiobook pronunciation starts before any audio exists. Go through the manuscript once and list every word a stranger might stumble on: character and place names, invented terms, foreign words, brand names, abbreviations, and words with two readings (lead the metal or lead the verb? a 1/2 that is “one half” or a date?). A chatbot can help find them; can ChatGPT make an audiobook includes a prompt for exactly this. Next to each word, write how you say it.
Settle the hard calls now, as the author. If a name could go two ways, pick one and note why. That list becomes the reference for this book, the next one in the series, and any human narrator you hire later.
Phonetic respelling that TTS reads correctly
A good respelling uses ordinary English chunks a voice already knows. Split the word into syllables with hyphens and build each from familiar sounds: Siobhan becomes shi-VAWN, Niamh becomes NEEV, Persephone becomes per-SEF-uh-nee. In your list, capitals mark the stressed syllable, the way readers' guides do: Hermione (her-MY-oh-nee).
In the line itself, start with the same respelling and listen. Voices differ: some glide over hyphens, some pause at them, and a short block of capitals is occasionally read as letters. If that happens, try a variant such as hermyonee or her my oh nee until the name flows. Plain letters are more predictable than accent marks or IPA symbols, which is the core of phonetic spelling for TTS.
Fix one line, regenerate only that line
You don't have to respell the whole manuscript before you start. In the studio every line is editable: click the text, replace the name with its respelling, and the line is marked as changed. Regenerate that line, listen, and the rest of the chapter stays exactly as it was. Regenerating voices the line again, so it uses credits for that line again.
For a name that appears hundreds of times, respell it in a copy of your manuscript with find-and-replace before you upload, or in the import step, where you can edit a chapter's text. Keep the original file untouched for print and ebook. A bracketed direction such as [say Siobhan as shi-VAWN] can help on voices that take notes, but a respelling in the text is more reliable, because every voice reads the text.
Accents, foreign words and abbreviations
A foreign phrase inside an English sentence is usually read with English habits. If it must sound right, respell it in English sounds (Saint-Étienne becomes san-tay-TYEN) or accept an anglicized reading, as many human narrators do. For a character's accent, don't respell their dialogue into dialect; choose a voice that fits and add a direction instead.
Abbreviations and numbers trip voices more often than names. Write out Dr., Mr. and St. where the meaning matters, regnal numbers (Louis XIV as Louis the Fourteenth), years and dates the way you'd say them, and units (5 km as five kilometers). For an initialism read letter by letter, keep the periods or spaces: F.B.I. or F B I. Fixing these is most of the work when you fix AI voice pronunciation in nonfiction.
Keeping names consistent across a series
Consistent audiobook pronunciation across a series comes from one habit: store the pronunciation list with your series notes and use the same respellings in every book. FullCastBook keeps the cast across a series, so the same voices read the same spellings and a name in book three matches book one. When a new character arrives, test their name in one line before you generate the chapter.
If you change your mind about a name later, the fix is the same: edit the lines that contain it and regenerate them. Search your manuscript for the name first so you know how many lines that is.
Publish a pronunciation guide for readers
Readers search for how to pronounce character names, especially in big fantasy and romantasy series, and fantasy name pronunciation in the audiobook is often what they're checking. Give them the answer: a page on your website or a list at the back of the ebook with each name, its respelling and a short clip cut from your audiobook. It answers fan questions, keeps fan podcasts and videos consistent, and brings those searches to your site.
Keep the reader version friendly: the plain respelling (shi-VAWN) is enough, no IPA needed. Building a whole invented world? Fantasy audiobook narration covers casting it, and AI voice emotion tags covers directing it.
