AI Audiobook Narration: How to Create One in 2026
AI audiobook narration step by step: prepare your manuscript, pick a voice, fix pronunciation, check audio specs and choose where to publish in 2026.
AI audiobook narration means using text to speech to read a finished manuscript aloud, chapter by chapter, instead of hiring a human narrator. It works best for books where one steady voice carries the text, and it needs three things: a clean manuscript, a voice that suits the book, and audio that passes each store's technical rules.
How long will your AI audiobook be?
ACX, Audible's marketplace for narrated books, says most performers narrate about 9,300 words per hour. Its guide tells authors to divide the word count by that figure to estimate the finished length (ACX: how long will my narrated audiobook be). That works out to about 155 words per minute. An 80,000-word novel comes to roughly 8.6 finished hours (80,000 divided by 9,300).
ACX also estimates human production at 6.2 hours of work per finished hour, made up of about 2 hours of recording, 3 hours of editing and 1.2 hours of quality review. It puts new narrators at 5 to 7 hours per finished hour. Those figures are a human baseline. No source gives a percentage saving for AI narration, so this guide does not offer one. Your real time saving depends on how much you listen back and fix.
AI audiobook narration: the step-by-step workflow
Every project follows the same order. The table shows the six steps and what to check at each one.
| Step | What you do | What to check | |---|---|---| | 1. Prepare the text | Export a clean manuscript with real chapter headings | Footnotes, tables, numbers and names | | 2. Pick a voice | Audition the voice on a passage that includes dialogue | Pace, tone and pronunciation of names | | 3. Generate by chapter | Convert each chapter as its own file | Each file starts with its section header | | 4. Listen and fix | Play every chapter at normal speed | Odd breaks, skipped sentences, wrong names | | 5. Check loudness and format | Measure levels and export the file type the store needs | Loudness, peak, noise floor, bit rate, channels | | 6. Publish | Upload where the store accepts your audio | Audio rights, labels and the store's disclosure rules |
Generate one file per chapter from the start. ACX requires one section or chapter per file, so splitting by chapter early avoids rework later. If your tool caps the length of one request, split long chapters into smaller parts and join them in your editor.
Prepare the manuscript so the narration sounds right
Most awkward reads come from the text, not the voice. Fix these before you generate any audio:
- Numbers and dates. Decide how each figure should be read. If the engine guesses wrong on a test passage, rewrite the figure the way you want it spoken.
- Footnotes. Put short footnotes into the text, or remove them. Long footnotes usually read badly aloud.
- Tables. Simple tables read better than complex ones. Rewrite a complicated table as sentences.
- Names and invented words. Build a short pronunciation list for characters, places and made-up terms. Test each entry on a sample before you generate the whole book.
- Chapter headings. Keep them as real text, not images, so they sit where the listener expects them.
Choose the voice and delivery
A steady single narrator suits exposition, memoir and non-fiction. Dialogue-heavy fiction may need separate character voices. OndaLabs offers two tools for this. The dialogue tool voices several speakers in one script, and custom voices give one narrator the same sound across a series. You can clone a voice from a recording you have permission to use, or design one from a text description for 1,500 credits. The Creator plan includes 3 custom voices, and Pro includes 10.
Delivery is controlled by describing it in plain words and by adding supported expressive tags to the script. Describe the pace and emotion you want, add tags where they help, and then listen to the result. The Free plan runs Onda Lite. Onda Flash, the most expressive model, is available on paid plans.
For a fuller guide to picking a voice, read how to choose the best AI voice for narration.
Check the audio before you publish
Store rules differ, but ACX's submission requirements are a useful reference for the audio files themselves. The table lists the main ACX values, last updated April 15, 2026.
| Measure | ACX requirement | |---|---| | Average level (RMS) | Between -23 dB and -18 dB for every file | | Peak | No higher than -3 dB | | Noise floor | No higher than -60 dB RMS | | Format | MP3, 44.1 kHz, 192 kbps or higher, constant bit rate | | Channels | All files mono, or all files stereo | | Length per file | One chapter or section, 120 minutes or less | | Room tone | 1 to 5 seconds at the start and end of each file |
Source: ACX audio submission requirements
OndaLabs does not state that its output meets these specifications or any store's rules. Measure the levels and check the format in an audio editor before you upload.
Where can you publish an AI-narrated audiobook?
Each store has its own program and rules, and some accept only certain kinds of audio. As of October 2026:
- Spotify for Authors. Its audiobook creation tools are US-only and English-only, with a limit of five audiobook projects per account. The narration uses a digital voice, and titles are labelled "Narrated by Digital Voice." Spotify says most audiobooks publish within 72 hours. The Spotify audiobook creation tools page explains the workflow.
- Google Play Books. Auto-narration covers English, French, German, Hindi, Italian, Portuguese (Brazil) and Spanish for EPUB books on sale on Google Play. You must own the audio rights. The Google Play Books partner help page lists the editing options, including the narrator voice, speed and pronunciation fixes.
- Amazon KDP Virtual Voice and Apple Books. Both have their own digital-narration routes with tight eligibility. KDP's is an invite-only beta limited to the US marketplace, and Apple's is English-only.
ACX does not accept unauthorized AI narration, so a standard ACX upload cannot carry an AI-narrated book. For the full comparison of each store's rules, read AI narrated audiobooks on Audible, KDP and other stores.
Rights, disclosure and listener expectations
Your manuscript stays yours. The US Copyright Office's January 2025 report says that using AI tools to assist human creativity does not change copyright in the output. It also says that prompts alone do not give a user enough control to be the author of AI output (US Copyright Office, Part 2 report). Do not promise buyers that the generated audio file itself is protected by copyright.
Audio rights are a separate question. Google, Apple, KDP and INaudio all require that you hold the audio rights to the book. Where a store requires a label, use it.
Listeners are cautious. The Audio Publishers Association's 2026 results, published June 5, 2026, show that AI-narrated titles made up 0.03% of 2025 US audiobook revenue. Willingness to try AI narration fell from 70% in 2025 to 61% in 2026, and 16% of listeners have heard an AI-voiced audiobook (APA 2026 results via InfoDocket). Make sure your listing says clearly what the listener is hearing.
How to do it in OndaLabs
OndaLabs is a text-to-speech studio, so it produces the narration audio. It does not publish to stores, and it makes no claim about store approval or store integrations. You take the files and upload them where each store accepts them.
Credits are the cost unit. On Onda Lite, one character costs one credit. On Onda Flash, one character costs 1.5 credits. For example, 1,000 characters cost 1,000 credits on Lite and 1,500 credits on Flash. Count your manuscript's characters yourself before you choose a plan. Voice design or cloning costs 1,500 credits.
| Plan | Monthly price (USD) | Monthly credits | What is included | |---|---|---|---| | Free | $0 | 2,000 (about 2 minutes) | Onda Lite, expressive tags, dialogue, speech to text, 5 saved clips | | Starter | $9 | 170,000 (about 189 minutes) | Onda Flash, image generation, commercial use | | Creator | $22 | 400,000 | 3 custom voices, music generation | | Pro | $49 | 900,000 | 10 custom voices | | Business | $199 | 3,750,000 | 30 custom voices, priority support |
Yearly billing is 10 times the monthly price. Top-ups are 100,000 credits for $5 or 500,000 credits for $24. Commercial use is included from the Starter plan, which matters if you plan to sell the audiobook. Check OndaLabs pricing for current plan details, then test one chapter in the text to speech tool before you commit to a full book.
Frequently asked questions
How do I turn my book into an audiobook with AI?
Prepare a clean manuscript, split it by chapter, choose a voice and generate each chapter as its own audio file. Listen to every file, fix mispronounced names, and check the levels and format against the store's rules. Then upload the files where the store accepts your own audio. The workflow table above shows the order.
How many words is an hour of audiobook?
ACX says most performers narrate about 9,300 words per hour, which is about 155 words per minute. Divide your word count by 9,300 to estimate the finished length. Treat that as a planning figure, because your voice and delivery will change the real length.
Can I sell an AI-narrated audiobook?
It depends on the store. ACX requires human narration unless it is authorized, and it prohibits unauthorized AI recordings. Other routes accept AI narration under their own programs and labels, and the store rules post lists them. Commercial use is included on OndaLabs Starter and above.
Do I need the audio rights to my book?
Yes, for the stores that narrate or accept audio from you. Google, Apple, KDP and INaudio all require that you hold the audio rights. If you have a publishing contract, check its audiobook and AI clauses, and speak to a lawyer about your specific deal.
How do I fix mispronounced names in AI narration?
Make a pronunciation list before you generate the book, and test each name in a short sample. If a name still reads wrong, respell it in the text or adjust the pronunciation in your tool, then regenerate that chapter. Listen to the full chapter again before you export it.
Is AI narration good enough for fiction?
Listener research is mixed. Some surveys show that many listeners still prefer human narration, while a vendor-released study found an AI multi-voice version rated higher for character-driven fiction. Test a sample passage from your own book with listeners in your genre before you decide.