AI Audiobook Maker: One Narrator or a Full Cast
AI audiobook maker: paste a chapter, get one narrator or a full cast with a voice per character, plus MP3 and subtitles. Your words stay exactly as written.
最終更新: 2026-10-27
An audiobook maker turns written text into a narrated audiobook. VoiceSmiths does it two ways: one narrator for the whole book, or a full cast where the narrator and every character get their own voice. The text stays exactly as you wrote it; you get an MP3 and line-timed subtitles for each chapter. Paste a passage below to see who the AI finds in it (this needs an account with credits, none used up), or hear a finished chapter first.
生成にはクレジットが必要です · お試しパック $3.90 から · まずはシアターで完成作品を聴いてみてください
What an AI audiobook maker does
| One narrator | Full cast | |
|---|---|---|
| Best for | Non-fiction, literary fiction, lectures | Dialogue-heavy fiction, kids' stories, drama |
| Per job | Up to 50,000 characters | Up to 8,000 characters per paste (longer chapters in parts) |
| Voices | One narrator | Narrator + one voice per character |
| Emotion | Overall tone | Tagged line by line (whispers, laughs, sobs…) |
| You get | MP3 + SRT subtitles | MP3 + SRT + script, and a vertical video |
The single narrator runs in the long-form text to speech mode; the full cast is the audio drama studio. Many people do both: one full-cast chapter to hear the pacing, then decide for the whole book.
Hear a chapter it made
A Christmas Carol, Stave One, straight from Dickens' text: Scrooge, his nephew and the narrator each have their own voice, and no line was rewritten.
The theater has 170+ English episodes made the same way: Poe, Dickens, Aesop and classic Chinese tales.
How to make an audiobook with AI
- Clean the text. Remove page headers and ads. Put dialogue in quotation marks with a "she said" nearby: that is what makes character detection accurate. Respell invented names the way they should sound (Ysolde → Ee-SOL-duh): pronunciation dictionaries only work in the workbench's text-to-speech tab, not in long-form or full-cast renders, and the respelling will show in the subtitles.
- Choose one narrator or a full cast. One voice from start to finish, or the narrator plus a cast the AI picks from 54 curated voices.
- Check the voices. Click any face to hear it and swap what does not fit. For a book, add the first chapter to a series and every returning character keeps the same voice in later chapters; a chapter longer than 8,000 characters goes in as two or more parts of the same series.
- Render chapter by chapter. Credits are charged per 1,000 text characters and refunded if a render fails. Each chapter comes back as MP3 with subtitles. You can't re-render a single line: fixing one means rendering that part again, billed again, so keep parts short.
- Publish. Add intro music or sound effects if you like. Renders are MP3 at 128 kbps. Stores differ on AI narration: ACX (Audible) does not accept it, and Spotify for Authors asks for 192 kbps constant-bitrate MP3 within set loudness targets, so master each chapter (in Auphonic or Audacity, for example) before uploading; see where you can publish an AI audiobook. For an audiobook you can switch off the spoken per-chapter AI notice and disclose once in the opening credits instead.
What an audiobook costs to make
| Characters | Credits (full cast or narrator) | Audio | |
|---|---|---|---|
| One chapter (two parts in full cast) | ~10,000 | 700 | ~12 minutes |
| A 90,000-word novel | ~500,000 | 35,000 | ~9.7 hours |
Standard rate: 70 credits per 1,000 characters for either mode; a single narrator on the faster Flash model is 40. At monthly-plan list prices one 10,000-character chapter (700 credits) comes to about $7.0–9.3 ($2.8–3.7 in the first month at the new-user price).
For a whole book, buy one-time packs rather than a subscription: a 90,000-word novel (about 35,000 credits) takes two 30,000-credit packs, 60,000 credits that never expire, for $499.80, or $199.80 within an hour of sign-up; what is left covers re-renders (budget about 10%) and most of the next book. Monthly-plan credits expire at the end of each billing period, so a whole book on a monthly plan takes several months.
To test first, start with the one-time trial pack (500 credits, about 7,100 characters, never expires; $3.90 for new users in their first hour, list $9.90); Starter is $7.90 the first month for new users, then $19.90/month. See pricing.
Where can I publish it?
Audible's standard ACX channel requires human narration, so AI-narrated audiobooks go elsewhere. At the time of writing, Spotify (through Findaway Voices), Google Play Books and Kobo accept AI-narrated titles with disclosure, and YouTube accepts AI audio when you mark it as synthetic. Store rules change, so check each store's current rules before you upload; you also need the audio rights to the book. The AI audiobook production guide covers the release steps.
One cast, other languages
A full-cast chapter can be adapted into Spanish, Portuguese, Japanese, Korean, French, German, Indonesian or Chinese with the same cast, so one book becomes several audiobooks. See AI audio drama generator.
FAQ
What is the best AI audiobook maker? It depends on the book. A book carried by one voice (non-fiction, literary fiction) needs long single jobs and a steady narrator; dialogue-heavy fiction is better with a full cast that gives every character a voice. Compare four things: whether it rewrites your text, how much it takes in one job, whether you get subtitles, and whether commercial use is allowed.
Is there a free audiobook maker? Not for making one: on VoiceSmiths, finding the cast needs credits in your account (it uses none) and rendering is charged by length. Listening is free: every voice sample and the 170+ English episodes in the theater play without an account. Rendering a chapter costs 70 credits per 1,000 characters, and the one-time trial pack (500 credits; $3.90 new-user price in the first hour after sign-up, list $9.90) covers about 7,100 characters.
Can I import an EPUB, TXT or PDF? EPUB and TXT, yes; PDF, no. In the audio drama studio, Import a whole novel reads a TXT or EPUB in your browser (nothing is uploaded) and splits it into chapters, each with its length and estimated credits; load one at a time (up to 8,000 characters per analysis for a full cast, so a longer chapter is split into parts of one series). DRM-protected EPUBs cannot be read, and a PDF needs its text copied out. A single narrator takes pasted text, up to 50,000 characters per job. Pasting chapter by chapter also keeps chapter files separate, which is what audiobook stores expect.
Can I sell the audiobook I make? Yes, if you hold the audio rights: your own book, a licensed one, or a public-domain classic. Full-cast files carry an AI-generated label in their metadata and by default end with a spoken "This audio was generated by AI" (you can switch it off before rendering); single-narrator files carry no label. Follow your store's rules for disclosing AI narration.
Audiobook maker, generator or creator: is there a difference? No, they are three names for the same kind of tool: text goes in, narrated audio comes out. What differs between tools is whether they offer one voice or a cast, and how much text they take at once.
To turn a whole novel into an audiobook step by step, see novel to audiobook.
Open the studio and cast your first chapter →
Related reading: audiobook cost calculator · can you publish an AI audiobook on Audible?