Where AI narration is accepted
ACX, the self-publishing route to Audible, requires submitted audiobooks to be narrated by a human unless otherwise authorized, and prohibits unauthorized text-to-speech, AI, or automated recordings. ACX also says Audible is working to accept third-party text-to-speech content and will announce timing later. Until that happens, do not upload AI narration through ACX.
Kobo Writing Life accepts AI-narrated audiobooks. It asks you to say so in the metadata by listing the narrator as Synthesized Voice, Female Synthesized Voice, or Male Synthesized Voice.
Other stores and distributors vary. Some run their own AI narration programs that work from your ebook, and some distributors accept AI narration only when it was produced through partners they approve. Read the AI policy of every channel you plan to use before you generate a full book, so you do not produce files that a store will refuse.
Disclose synthetic narration
Where a store accepts AI narration, it expects it to be labeled. Put the disclosure in the narrator field the store specifies, and consider a short line in the description as well. Listeners who choose an audiobook partly for the narrator should know before they buy, and honest labeling protects your reviews.
The technical spec is a separate test
Acceptance rules decide whether a store takes AI narration; technical specs decide whether your files pass review. ACX, for example, asks for each file to measure between -23 dB and -18 dB RMS, with peaks no higher than -3 dB, a noise floor no higher than -60 dB RMS, and 192 kbps or higher constant-bit-rate MP3 at 44.1 kHz, one chapter or section per file, with opening and closing credits.
NarrateHQ exports are loudness-normalized to -16 LUFS, a common target for streaming and online video. That is a different measure and target from the RMS range above, so do not upload exports as they are. Master each chapter to the specification your distributor publishes; from the Creator plan you can export lossless WAV or FLAC to use as a clean master for that step.
Prepare the manuscript for narration
A book is long-form narration at its longest, so preparation pays off more than any setting.
- Keep one chapter per script so each chapter exports as its own file.
- Add character names, invented words, and place names to the pronunciation dictionary before the first chapter.
- Write numbers, dates, and abbreviations the way they should be read aloud.
- Generate one full chapter, listen to all of it, and only then produce the rest with the same voice and settings.
- Write the opening and closing credits as their own short scripts in the same voice.
Rights you need
You need the right to publish the book's text in audio form, which an author self-publishing their own work usually has. For the narration, audio generated while you are on the Starter plan or higher may be sold commercially; audio generated on the Free plan is for testing only. If you narrate with a cloned voice, it must be your own voice or one whose owner has given explicit consent to be cloned.