How to turn an EPUB into an audiobook
Updated October 2026
If you have an EPUB, you have the hardest input to an audiobook: clean, chaptered text. What's left is a voice, audio files and, if you want to sell it, packaging that meets each store's rules. Here are the realistic routes in 2026.
1. Human narration ($1,200–$2,800 per book)
Professional narrators typically charge $200–$400 per finished hour, and a 90,000-word novel runs 9–10 finished hours. It's the premium option and the only way onto Audible through ACX, but it's hard to justify for a backlist title with modest sales.
2. Cloud AI voice tools (subscriptions)
Services such as ElevenLabs produce excellent AI narration, priced by the character. Expect a monthly plan large enough for a novel, plus the work of splitting chapters, stitching files and mastering the audio yourself, unless you use a done-for-you service that charges per book.
3. Free AI voices on your own computer
Open-source voice models can now run inside a web browser. Bookvox uses one, Kokoro-82M, to turn a DRM-free EPUB into MP3 files on your own machine, at no cost:
- Open your EPUB on bookvox.app. Chapters come from the book's table of contents; front and back matter are unticked.
- Choose a studio voice and speed, and listen to a few paragraphs to check you like it.
- In the player, choose Download MP3 → Whole document. You get a zip with one MP3 per chapter.
It's slower than a cloud service (a long book can take hours on an older laptop), the voice is good but not a human performance, and there's no pronunciation editor yet for invented names. For proof-listening, beta readers, advance copies and accessibility, it's hard to beat free.
If you want to sell it
Retail audiobook stores publish their own technical requirements. ACX (Audible's production platform), which many others mirror, asks for 192 kbps or higher constant-bitrate MP3 at 44.1 kHz, with RMS loudness between −23 and −18 dB, peaks no higher than −3 dB, and room tone at the start and end of each file. Bookvox's MP3s are speech-optimised (64 kbps mono) and aren't mastered to those specs. Treat them as a listening copy, not a retail master.
Stores also differ on AI narration itself. Audible accepts only its own Virtual Voice program, while Spotify, Apple Books, Google Play and Kobo accept AI narration with disclosure. See the platform-by-platform rules.