The first ACX question is not loudness. It is eligibility.
ACX does not accept arbitrary externally generated AI narration through its standard upload workflow. It does run an invitation-only U.S. Narrator Voice Replicas beta inside ACX, with participating narrators reviewing and controlling the work.
Audible separately offers integrated AI narration to selected publishing partners. Neither program means an author can generate an audiobook in Vois and upload it through the normal ACX path. Vois remains a practical local studio for destinations and programs that explicitly accept the final files you create.
Choose the distribution path before you produce
Make the platform decision before you cast a voice or generate a chapter.
| Your goal | Practical path |
|---|---|
| Standard ACX upload from Vois | Not supported for externally generated AI narration; use a human narrator or an ACX-approved program. |
| Publish an AI-narrated audiobook elsewhere | Confirm the destination's current written policy, disclosure, and delivery requirements before production. |
| Produce a local draft or listening copy | Use Vois to create a reviewed audio version, then decide separately whether and where it may be distributed. |
Policies and disclosure requirements can change. Keep a saved copy of the destination's rules with the project and check them again before delivery.
Where Vois fits in an audiobook workflow
For an outlet that accepts AI narration, Vois gives you a complete local production path:
- Create a project for the book and add the approved manuscript in chapters.
- Audition a narrator with both an emotional scene and explanatory prose.
- Use the pronunciation dictionary for names, places, invented terms, and abbreviations.
- Generate chapter by chapter, then listen for pacing, emphasis, and continuity.
- Regenerate only the passages that need a correction.
- Use the mastering and export controls that match the actual destination.
The benefit is not a claim that the first render is final. It is a shorter, local review loop. You can hear a chapter, correct a word or pacing issue, and keep the rest of the project organized while you revise.
Cast for long-form listening
An audiobook listener spends hours with one narrator. Do not select a voice from a short sample alone. In Vois, test several minutes of the real manuscript, including dialogue, technical passages, names, and quieter transitions.
For fiction, make sure the voice can hold a scene without overplaying it. For nonfiction, look for clarity and restraint. In either case, lock the approved narrator and pronunciation choices before generating the full book, then document any reason to change them.
Use a voice clone only when you have clear permission from the voice owner. A recognizable voice without consent creates a rights issue before distribution policy even enters the picture.
Review every chapter as an editorial asset
AI narration removes microphone noise and recording-room logistics. It does not remove editorial work. Listen for a misread character name, a list that needs a clearer pause, an acronym that sounds wrong, or a sentence that should be rewritten for the ear.
Keep the source manuscript, approved narration script, pronunciation entries, and final chapter export connected in the Vois project. When a manuscript changes, review the changed section and regenerate it deliberately rather than assuming an old chapter remains accurate.
Match the master to the real destination
ACX has technical requirements for human-narrated submissions, including loudness, peak level, noise floor, head and tail room tone, and file format. Always read ACX's current specification before submitting a human-narrated book. A technical target is not a substitute for platform eligibility.
Vois includes an ACX export preset for those technical targets. It can be useful when a human-produced project needs that format, or as a reference point while comparing delivery specifications. For an AI-narrated title, choose the export settings required by the distributor that has actually accepted the book and verify them against that distributor's current documentation.
Before any upload, check:
- the narrator and distribution rights;
- the exact approved manuscript and chapter order;
- pronunciation, pace, and chapter transitions;
- the destination's current technical requirements;
- any required AI-narration disclosure; and
- the final files by listening, not metadata alone.
Sources
- ACX: Now in Beta, Narrator Voice Replicas on ACX
- Audible: AI Narration and Translation for Publishers
A clear answer is better than a failed submission
Trying to master externally generated AI narration into standard ACX eligibility wastes time because the blocker is the submission pathway, not audio engineering. Use a human narrator when standard ACX distribution is non-negotiable. Use Vois only when the chosen destination or approved program explicitly accepts the files you plan to create.
Start with one chapter and the destination's actual rules. For the broader production path, explore Vois for audiobooks and Get started when the platform you chose accepts the narration you plan to make.
The Vois Team