No, you cannot use an AI voice for an audiobook submitted to ACX. The platform's submission requirements state that audiobooks must be narrated by a human, and that "unauthorized use of text-to-speech, AI, or automated recordings in ACX is prohibited."[1] Files submitted with AI narration are rejected. There is one narrow exception: ACX's own narrator voice replica programme, which operates under a separate opt-in agreement between narrators and Audible. Third-party AI narration tools remain prohibited regardless of output quality.
If your question is about audio processing tools (noise reduction, loudness adjustment, mastering), those are a different category entirely. Post-production software is permitted and expected. The prohibition is on AI-generated performance, not AI-assisted production. More on that distinction below.
What Does ACX's Narration Policy Actually Say?
ACX's official help documentation is direct: audiobooks must be narrated by a human unless otherwise authorised by ACX.[1] Text-to-speech recordings are not accepted. The policy exists because Audible listeners expect human narration, and the purchase expectation is built around that.
ACX places responsibility on both sides of the marketplace. Authors are expected to verify that any narrator they hire through the ACX marketplace is providing genuine human narration, not a text-to-speech output. The platform acknowledges that synthetic narration can be difficult to detect after full production, which is why authors are advised to listen carefully at the audition stage and again before paying for a completed project.[1]
The prohibition covers all current AI narration tools: ElevenLabs, Murf, Speechify, Descript's voice generation features, and any similar service. Submitting a file narrated with these tools to ACX is a policy violation, not a grey area.
What Is the ACX Narrator Voice Replica Programme?
In 2024, ACX launched a US-only beta programme that allows human narrators to create AI-generated replicas of their own voices.[2] This is the one exception to the blanket prohibition, and it is narrow: it applies only to narrators who opt into the programme by submitting sample recordings to ACX, which then generates a voice model.
Under the beta, narrators using voice replicas retain control over which projects they accept. They remain involved in reviewing and editing the final output for pronunciation and pacing. Titles narrated with a voice replica are labelled as such in the narrator field on the audiobook's listing and detail page. As of July 2025, participating narrators can receive per-finished-hour fees, royalty shares, or a blended structure.[2]
What this programme does not do: it does not open ACX to third-party AI narration tools. An author who uses ElevenLabs to narrate their own book cannot point to the voice replica programme as a precedent. The programme is specifically for professional narrators creating replicas of their own voice, operating within ACX's own production infrastructure. Narrators interested in the beta can contact acxbeta@acx.com.
What Counts as AI-Generated Narration Versus AI-Assisted Production?
This distinction matters for audiobook producers who use modern software. The rule is about the performance, not the processing.
AI-generated narration (prohibited on ACX): Using any text-to-speech or voice synthesis tool to produce the spoken audio. This includes generating narration with ElevenLabs, Murf, Play.ht, Amazon Polly, Google Cloud Text-to-Speech, or similar services, regardless of whether you then edit the output in a DAW.
AI-assisted production (permitted): Using software that applies machine learning to your human-recorded audio during editing or mastering. This includes:
- Noise reduction tools such as iZotope RX, Adobe Podcast Enhance, and Accusonus ERA
- De-clicking, de-breathing, and mouth noise removal plugins
- Room correction and reverb removal tools
- Loudness normalisation and peak limiting (including tools like ChapterPass that automate mastering to ACX specs)
- AI-powered tools within DAWs that process existing recordings rather than generating new speech
The distinction is: did a human voice perform the narration? If yes, then processing that recording with whatever tools you choose is permitted. If the spoken audio was generated by software, it is prohibited regardless of how much post-processing was applied.
What Is Amazon's Broader AI Content Policy Through KDP?
Kindle Direct Publishing (KDP) is Amazon's ebook self-publishing platform. Its AI content policy is sometimes cited in discussions about audiobooks, but it governs a different product. KDP's policy covers ebooks, not audiobooks. ACX governs audiobooks, and its narration policy is more restrictive than KDP's general content stance.
KDP requires authors to disclose when a book contains AI-generated text, images, or translations.[3] The disclosure is made to Amazon during the upload process and is not currently shown publicly to readers. KDP distinguishes between "AI-generated" (content produced by an AI tool, requiring disclosure) and "AI-assisted" (your own content refined with AI tools, no disclosure required).[3] KDP does not prohibit AI-generated content outright for ebooks.
ACX does not have an analogous disclosure-and-permit system for AI narration. The prohibition is categorical, not conditional on disclosure. Submitting AI-narrated audio with a disclosure note does not make it compliant.
How Do Other Audiobook Platforms Handle AI Narration?
The audiobook distribution landscape is splitting on this issue. ACX/Audible is the most restrictive major platform. Others have moved to permit AI narration with varying requirements.
Findaway Voices / Voices by INaudio: Findaway (rebranded as Voices by INaudio in August 2025) accepts AI-narrated audiobooks if disclosed during the publishing process.[4] In early 2025, Findaway began accepting audiobooks narrated with ElevenLabs for distribution to select partner platforms. Not all retailers within Findaway's network accept AI-narrated titles, so distribution reach is narrower than for human-narrated titles. AI-narrated titles are labelled in metadata across Spotify and other participating retailers.
Author's Republic: Author's Republic is testing distribution of AI-narrated audiobooks, specifying that this covers distribution only, not production.[5] The platform follows the Audio Publishers Association's labelling guidelines for AI narration. Their narrator marketplace continues to feature only human professional voice actors.
Google Play Books: Google operates an auto-narration programme that generates audiobooks from ebook files using AI voices.[6] Publishers in six countries (US, Canada, UK, Spain, Australia, and New Zealand) can create and sell auto-narrated audiobooks directly through Google Play. Google's programme is for books already sold as ebooks on Google Play; it is not a general-purpose submission route for externally produced AI narration.
Apple Books: Apple offers a digital narration programme that creates AI-narrated audiobooks from ebook files, but the process requires working through Apple's preferred partners and takes one to two months.[7] Apple retains the rights to the AI-narrated version for six months after the ebook is withdrawn. The service is available for English-language titles only.
Kobo Writing Life: Kobo does not publish a specific prohibition on AI narration as of early 2026, and the platform has accepted AI-narrated titles.
The practical result: if you want to use AI narration, Audible is not available to you through the standard ACX route. You can distribute through Findaway Voices to Spotify, Kobo, and select other retailers. You cannot reach Audible's customer base, which represents the majority of English-language audiobook sales.
What Post-Production Tools Are Still Allowed When Mastering for ACX?
ACX's prohibition is on AI-generated narration, not on the software used to process human recordings. The full post-production toolkit remains available.
Noise and artefact removal: iZotope RX, Adobe Podcast Enhance, CEDAR, Accusonus ERA, and similar tools that reduce room noise, clicks, pops, and mouth noise from existing recordings are all permitted. These tools process human-narrated audio rather than generating speech.
Loudness mastering: Achieving ACX's required RMS loudness of -23 to -18 dBFS, true peak below -3 dBFS, and noise floor below -60 dBFS involves gain adjustment, compression, and peak limiting. This processing can be done manually in any DAW or automated with tools like ChapterPass, which handles the complete mastering chain in the browser without requiring software installation. The first file is free.
Format conversion: Converting your edited files to the required MP3 format at 192 kbps CBR, 44,100 Hz, mono is a standard mastering step. Tools including Audacity, Adobe Audition, Logic Pro, and purpose-built mastering tools all handle this.
Breath and silence editing: Removing or reducing breath sounds between sentences, adding room tone silence at chapter heads and tails (ACX requires 0.5 to 1 second at the start and 1 to 5 seconds at the end), and cleaning up pacing are all part of standard audiobook editing. None of this conflicts with ACX's narration policy.
For a complete walkthrough of the mastering process for ACX, see the audiobook mastering complete guide and the ACX audio requirements guide.
What Happens If You Submit AI-Narrated Audio to ACX?
Submission is rejected. ACX reviews audiobook submissions both through automated quality checks and human review. The automated layer catches technical spec failures (loudness, peak, format). The human review layer assesses audio quality and consistency, which is also where narration authenticity issues would be identified.
If AI-narrated audio is detected after a title has already been published, the consequences extend beyond rejection of the file. ACX's terms of service give Audible grounds to remove the title, suspend the associated account, and claw back royalties already paid. Using AI narration without authorisation is a breach of the publisher agreement, not just a quality failure.
Detection methods include audio fingerprinting, listener reports, and pattern recognition. The consistency of AI voices, their characteristic artefacts, and the absence of natural human variation can flag files for closer review. This is not a reliable game of chance.
The safest course is straightforward: if you want your audiobook on Audible, hire a human narrator or narrate it yourself.
Human narration sorted. Now for the technical specs. ChapterPass masters your ACX files to every required specification automatically. Upload your chapters and get submission-ready files without the guesswork. First file free.
Frequently Asked Questions
Can I use ElevenLabs or other AI voice tools for an ACX audiobook?
No. ACX prohibits the use of any text-to-speech, AI, or automated narration tool unless specifically authorised by ACX. ElevenLabs and similar services fall under this prohibition. Files narrated with these tools will be rejected. If you want to distribute AI-narrated audiobooks, Findaway Voices (Voices by INaudio) and Google Play Books are alternatives that permit it with disclosure.
Does ACX allow AI tools for audio editing and mastering?
Yes. ACX's prohibition is on AI-generated narration, meaning software that produces the spoken voice. Post-production tools that process your existing human-recorded audio are permitted and standard practice. This includes noise reduction plugins, loudness mastering tools, de-click and de-breath processors, and automated mastering services like ChapterPass.
What is ACX's narrator voice replica programme?
ACX offers a beta programme where human narrators can create AI-generated replicas of their own voice by submitting sample recordings. Narrators in the programme retain control over which projects use their voice replica and remain involved in reviewing the final output. This programme is opt-in for narrators only. It does not permit authors to use third-party AI narration tools. Titles narrated with a voice replica are labelled on the Audible listing.
If I disclose that my audiobook uses AI narration, will ACX accept it?
No. Unlike KDP's ebook policy, which permits AI-generated text with disclosure, ACX has no disclosure-and-permit system for AI narration. The prohibition is categorical. Disclosure does not make an AI-narrated submission compliant.