How to Create an Audiobook with AI: The Complete Guide

Updated September 3, 2026

TL;DR: an 80,000-word book is about 480,000 characters — roughly 480,000 credits on LeapFun's standard voices, which fits a Pro month or two-and-a-half Creator months. The work isn't the generation; it's preparation (clean manuscript, pronunciation list), casting, and reviewing paragraph by paragraph. This guide walks the whole path from manuscript to distribution-ready files, including where AI narration is and isn't accepted.

Should you narrate this book with AI?

Honest fit check. AI narration is a strong choice for non-fiction, how-to, business and genre fiction, for multilingual editions, and for backlist titles that would never justify a studio budget. It is a weaker choice for literary fiction and audio drama where the narrator's performance is the product — there, a human narrator or ElevenLabs' most expressive voices still win. Two more constraints: Audible's ACX marketplace does not accept AI-narrated submissions from third-party tools at the time of writing, and some listeners simply prefer human readers. If your distribution plan runs through ACX only, stop here.

Step 1 — Prepare the manuscript

Step 2 — Choose the narrator

Audition with a full paragraph from the book, not a sentence. Listen for how the voice handles a long sentence and a question, and whether it wears well at 1.25× (many listeners speed up). These are real LeapFun samples:

If you're the author and the audience knows your voice, clone it; if you want a voice nobody else has, design one.

Step 3 — Build the Studio project

Paste or upload the manuscript into Audiobook Studio. It becomes a project with chapters and paragraphs you can reorder, so the structure of your audiobook matches the structure of your book. The free tier includes one trial project; unlimited projects, subtitles, music and translation come with Creator.

Step 4 — Cast characters

Assign the narrator once, then give recurring characters their own voices — preset, cloned or designed. Assignments persist across the whole project, so chapter twelve sounds like chapter one. Keep the cast small: a narrator plus two or three distinct character voices reads better than a voice per walk-on part.

Step 5 — Pronunciation rules

Turn your list into per-project rules: spell out how each name or term should sound once and every chapter follows. This is the single biggest quality lever after voice choice, and the one most people skip.

Step 6 — Generate and review

Step 7 — Master and export

Every LeapFun export is loudness-mastered to -14 LUFS, true peak -2 dB, 48 kHz — the streaming-platform spec. Export chapter by chapter (most platforms want separate files) or the whole book, in WAV or MP3, with subtitle files generated from the narration timeline on Creator and above. For retailers, switch to the Audiobook preset: 44.1 kHz, a lower peak ceiling, and a distribution-checks report (RMS, peak, bitrate, file length) measured on the encoded file.

Step 8 — Distribute

Check each platform's current policy on AI narration before you upload; they differ and they change. As of this writing, ACX does not accept third-party AI narration. Aggregators and other retailers each set their own rules and technical requirements (chapter file naming, leading and trailing silence, sample rate, bit rate). Our audiobook distribution prep guide covers the common technical spec.

What it costs, and how long it takes

Book lengthApprox. charactersCredits (standard voices)Credits (Leap Ultra)Fits in
Novella, 30,000 words180,000≈ 180,000≈ 270,000One Creator month
Novel, 80,000 words480,000≈ 480,000≈ 720,000One Pro month (650,000) or 2–3 Creator months
Epic, 150,000 words900,000≈ 900,000≈ 1,350,000Two Pro months or one Team month (1,800,000)

Add 10–20% for re-takes. Generation itself takes hours; reviewing takes as long as listening to the book. Compare that with weeks for a booked narrator and the maths is usually obvious — the question is fit, not cost.

Translated editions

On Creator and above, a finished project can be translated into eight languages — English, Chinese, Spanish, French, German, Portuguese, Japanese and Korean — and re-narrated with the same voice setup. Review the translation with a native reader before you narrate; the pipeline is fast, the proofreading is not.

Frequently asked questions

Can I sell an AI-narrated audiobook?

Yes on any paid LeapFun plan, which include commercial usage rights. Distribution depends on each retailer's policy: ACX does not currently accept third-party AI narration; other platforms and aggregators set their own rules. Check before you produce.

How many credits does an audiobook take?

About one credit per character on standard voices, 1.5 per UTF-8 byte on Leap Ultra. An 80,000-word novel is roughly 480,000 characters — a Pro month covers it, or two to three Creator months. Budget 10–20% extra for re-takes.

Can different characters have different voices?

Yes. Assign preset, cloned or designed voices to the narrator and to individual characters in the Studio project; the assignment holds across every chapter.

How do I stop the narrator mispronouncing names?

Add a pronunciation rule in the project for each name or term. Rules apply to every chapter and survive re-takes, so you fix a word once.

Do I have to regenerate the whole book after an edit?

No. Regenerate the changed paragraph or chapter; earlier renders stay in version history and everything else is untouched.

Can I read my own book in my own voice?

Yes — clone your voice on the Starter plan or above, then assign the clone as narrator. Consent is confirmed and recorded when you create the clone.