How to Make an Audiobook With AI – Costs, Steps and Where to Sell It

How to make an audiobook with AI: what it costs, how to prepare your manuscript, check the audio and which stores accept AI narration.

We may earn a commission if you buy through links on this page. This never affects our scores.

Most authors never make an audiobook because of the cost. AI narration lowers it, but it doesn’t remove the work. It moves the work from recording to preparing the text, checking the audio and handling distribution. This guide covers what an audiobook costs, how to make one step by step and where you’re actually allowed to sell it.

The cost that stops most authors

Audiobooks are priced per finished hour, meaning the length of the final audio. ACX, Amazon’s audiobook marketplace, has published guidance of roughly $200 per finished hour for narration plus a similar amount for editing and mastering, and typical narrator rates run from about $150 to $400 or more.

ACX’s planning figure is about 9,300 words per finished hour. That makes an 80,000-word novel roughly 8.6 hours and a 50,000-word nonfiction book about 5.4 hours. At $250 to $400 per finished hour, the novel costs around $2,150 to $3,440.

What making an audiobook with AI means

You use text-to-speech software to narrate your manuscript, then produce files that meet a store’s technical rules. There’s no single button. A good result needs three things: a manuscript prepared for the ear, a voice that suits the book and a careful listen-through before you publish.

Step 1: Decide where you’ll sell it

This decides your whole plan. Store rules change, so confirm each on the store’s own policy page. Based on sources published through mid-2026:

  • Audible (through ACX): ACX requires human narration unless authorized, so you can’t upload your own AI-narrated file. Amazon runs a separate virtual voice program that turns eligible Kindle books into audiobooks with its own voices, reported as an invite-only US beta.
  • Spotify: accepts AI-narrated audiobooks with a disclosure. Some sources say it limits this to certain voice providers, so check the current list.
  • Google Play Books: accepts AI narration and offers its own auto-narration with 50+ voices. One reported condition: if an auto-narrated audiobook is for sale elsewhere, it must also be listed on Google Play Books.
  • Kobo: accepts AI narration if you disclose it, by listing the narrator as “Synthesized Voice”.
  • Apple Books: offers its own digital narration for some ebooks. Sources disagree on whether external AI files are accepted.

If you want Audible, a human-narrated edition is the only way in for now. Think hard before signing an exclusive ACX contract, which can lock a title out of other stores for years. Disclose AI narration wherever a store asks.

Step 2: Check you hold the audio rights

If a publisher released your book, your contract may have given them audio rights. If your book uses other people’s material, you need permission to turn it into audio.

Step 3: Prepare the manuscript for the ear

  • Remove page-only material: references like “see figure 2”, footnotes, URLs and tables.
  • Spell out ambiguous items: numbers, dates, abbreviations and symbols.
  • Make a pronunciation list for names, places, brands and invented words.
  • Mark the structure: chapter titles, opening and closing credits and section breaks.
  • Read a chapter aloud yourself. A sentence you stumble over will probably trip the voice too.

Step 4: Choose a voice and a tool

Look for consistency over long stretches, pronunciation controls, chapter and file handling, the cost for a whole book and commercial rights on your plan. Many free plans don’t allow commercial use, so read the terms for the exact plan you’ll use.

One example. ElevenLabs offers an audiobook workflow in its Studio. According to ElevenLabs, you can import a manuscript as EPUB, PDF, TXT, HTML or DOCX, and Studio detects chapters automatically. You can choose from 10,000+ voices, automatically assign voices to characters, use pronunciation dictionaries and export MP3 or WAV per chapter or as a full project. It describes v3 as its most expressive text-to-speech model, covering 70+ languages.

Do the credit math. On ElevenLabs’ standard models one character costs about one credit. A 75,000-word book is roughly 450,000 characters, so about 450,000 credits. As of November 2026 the Creator plan includes 121,000 credits a month and Pro 600,000, so a book fits in one month on Pro but would take several on Creator. The free plan covers about ten minutes and doesn’t allow commercial use. Check how current plans count audiobook projects. See our ElevenLabs credits guide and ElevenLabs review.

If you clone a voice, clone only your own or one you have clear written permission to use. See our guide to cloning your voice.

Step 5: Generate, then listen to every hour

Generating takes far less time than checking. Plan to listen to every finished hour at least once, following the text. Listen for mispronounced names, misread numbers, missing or repeated words, odd emphasis, dialogue where every character sounds alike and tone drift between chapters. Fix problems with your pronunciation list, small text edits or regenerating the section.

Step 6: Master and export to the store’s specification

ACX’s published specification is widely used as a benchmark:

  • MP3 at 192 kbps or higher, constant bit rate, 44.1 kHz
  • Loudness between −23 dB and −18 dB RMS
  • Peaks no higher than −3 dB
  • Noise floor no higher than −60 dB RMS
  • Room tone at the start (0.5 to 1 second) and end (1 to 5 seconds) of each file
  • Each chapter as its own file, with opening and closing credits
  • A retail sample of one to five minutes

AI audio has no room noise, so the noise floor usually passes. Loudness is what people miss, so check it with a meter and normalize if needed. Other stores have their own rules, so check each before you upload.

Step 7: Upload with the right disclosure

Choose stores that accept AI narration, fill in the metadata and disclose the narration method wherever there’s a field. Many authors also say it in the book description. Upload a retail sample that shows how the narration sounds.

What AI narration struggles with

It works best for nonfiction, business and self-help, where steady delivery matters most. Fiction with many characters, strong emotion or subtle humour is harder. Produce one chapter first and ask a few readers to listen. Some listeners dislike synthetic voices, so honest labelling matters.

A quick checklist

  1. Choose stores and check their AI rules.
  2. Confirm you hold the audio rights.
  3. Prepare the text, with a pronunciation list.
  4. Test one chapter with two or three voices.
  5. Check your plan’s commercial license.
  6. Generate, then listen to every hour.
  7. Master to the store’s specification.
  8. Upload with a clear AI disclosure.

Frequently asked questions

How much does an AI audiobook cost?

Mostly a monthly plan for a month or two, compared with thousands of dollars for human narration, but commercial rights and credit limits depend on your plan.

Can I sell an AI audiobook on Audible?

Not by uploading your own AI file through ACX. Amazon runs its own separate program.

Do I have to disclose AI narration?

Several stores require it, and it’s good practice anyway.

Can I use my own voice?

Some tools let you clone it. Clone only your own voice, or one you have written permission to use.

How we researched this guide

This guide is based on store policy summaries published through mid-2026, ACX’s published guidance on rates and audio specifications, ElevenLabs’ published information and pricing, and our tool reviews (checked November 10, 2026). We haven’t run our own trials. Store policies and prices change often, so confirm them on each store’s and tool’s own pages.

Related articles