Can you make an audiobook with AI?
Yes. You can narrate an entire book with an AI voice: prepare the text, generate the narration chapter by chapter with a realistic text to speech tool, then assemble and master the audio. With Speechma you choose from 1,700+ voices in 75+ languages and locales, control speed and pitch, and download every part as MP3, free and with commercial use included.
The bigger question is where you can publish it. ACX, the platform that distributes to Audible, states that unauthorized use of text-to-speech, AI or automated recordings in ACX titles is prohibited and that audiobooks must be narrated by a human unless otherwise authorized. Other stores and direct-sales options have their own rules, which change often. Always read the current policy of the store you plan to use before you start.
Step-by-step: narrate your book with AI
- Prepare the manuscript for listeningRemove page numbers, footnote markers, image captions and “see Figure 3” references. Spell out abbreviations, symbols and years the way a narrator would say them.
- Choose a narrator voicePreview several voices in Speechma with the same paragraph. For thrillers, try a lower, steady voice; for romance or memoir, a warm one. Save your pick to favorites.
- Split chapters into 2,000-character partsBreak each chapter at paragraph ends, never mid-sentence. A 3,000-word chapter is roughly eight or nine parts. Label them Ch01-01, Ch01-02 and so on.
- Generate and review each partPaste each part, keep speed and pitch identical across the book, enter the code and click Generate Audio. Listen for mispronounced names and regenerate those parts with phonetic spellings.
- Download everythingUse “download all” to save the parts as a ZIP. Your history is stored privately in your browser, so download before clearing it.
- Join parts into chapter filesIn a free audio editor such as Audacity, place the parts end to end, add a short beat of silence at the start and end of each chapter, and export one MP3 per chapter.
- Master and check levelsNormalize volume so all chapters sound equally loud and remove clicks at the joins. If a store publishes technical specs, check your files against them.
- Publish where AI narration is allowedSell or share through platforms whose current rules allow digitally narrated audiobooks, your own website, a membership page, or a podcast feed for serialized fiction.
Audiobook file specs: a useful benchmark
Even if you don’t publish on ACX, its published technical specs are a solid quality target for any audiobook.
| Spec | ACX requirement | How to get there |
|---|---|---|
| Volume (RMS) | Between −23 dB and −18 dB | Normalize or use a loudness tool in your editor |
| Peak level | Below −3 dB | Apply a limiter at −3 dB |
| Noise floor | Below −60 dB RMS | AI narration is naturally clean; avoid adding noisy music |
| File format | MP3, 192 kbps or higher CBR, 44.1 kHz | Export settings in your audio editor |
| File structure | One chapter or section per file, max 120 minutes each | Join Speechma parts into chapter files |
| Room tone | Short silence at start and end (under 5 seconds) | Add 1–2 seconds of silence in your editor |
Specs from ACX’s audio submission requirements. ACX does not accept unauthorized text-to-speech or AI narration; these numbers are shown as a quality benchmark only.
How long will my AI audiobook be?
Narrators average roughly 150–160 words per minute, so you can estimate runtime from your word count: a 60,000-word novel lands around six and a half hours, a 25,000-word nonfiction guide around two and three-quarter hours, and a 1,500-word children’s picture book about ten minutes. Slightly slower speed settings add time but help with dense nonfiction.
For planning parts: 2,000 characters is about 300–350 words, so a 60,000-word book needs around 180–200 conversions. That’s very doable over a few evenings if you work chapter by chapter and keep a simple checklist. Name every file consistently (Ch07-03.mp3) so nothing ends up in the wrong order when you join them.
Plan your opening and closing too. Most audiobooks begin with the title, author and narrator credit (“The Lighthouse Keeper, written by Jane Doe”) and end with a short closing line such as “The End.” Generate these as separate short files.
Tips and troubleshooting for AI audiobooks
- Character names sound wrong. Create a pronunciation list (“Siobhan” → “Shi-vawn”) and replace names with phonetic spellings in the narration text only.
- Dialogue all sounds the same. Keep one narrator voice for consistency, or for kids’ books use a second voice for a key character. See kids story narration.
- Chapters sound slightly different in volume. Normalize all chapter files to the same level in one batch at the end.
- Splitting a long book takes forever. SpeechmaPRO.com converts 10,000 characters at once, about five times fewer parts, with no captcha or ads.
- You want the book in your own voice. Record a 15–20 second sample and use Voice Cloning (up to 1,000 characters per generation, 10 languages). Only clone a voice you have the right to use.
- Long pauses feel too short. Add a pause tag such as
[break 2s]or[break 3s]at scene breaks, or use the pause buttons in Voice Effects.
Frequently asked questions
Can I publish an AI-narrated audiobook on Audible?
Is it free to make an audiobook with Speechma?
How long does it take to narrate a book with AI?
Which AI voice is best for audiobooks?
Can I make an audiobook in Spanish or another language?
What file format should audiobook chapters be?
Make better voiceovers, faster
- New voices and features the day they launch
- Script templates and tricks that make AI voices sound human
- Creator stories, community picks and early access
