Which audiobook stores accept AI narration? ACX, Apple, KDP, Spotify

Only some audiobook stores take AI narration, and each one draws the line differently. As read on 2026-10-10: ACX prohibits unauthorized text-to-speech in its titles, Apple Books narrates the book itself and does not accept your audio, Amazon KDP Virtual Voice is an invite-only beta, and Spotify accepts digitally narrated books from a distributor if the description says so.

That matters if you plan to synthesize a book with a text-to-speech API such as Sume TTS 1.0. The tool can make the audio; the store decides whether the audio can be sold there. Check the store before you spend on a full book.

The four store rules, side by side

The table lists only what each store's own page said when fetched. Rules change, so treat it as a dated snapshot and re-read the page before you submit.

ACX: the strictest on AI audio

ACX's submission page states that unauthorized use of text-to-speech, AI or automated recordings in ACX titles is prohibited, and it describes the human-narrated path in detail: MP3 at 192 kbps or higher, 44.1 kHz, one chapter per file, at most 120 minutes per file, a retail sample of five minutes or less, and room tone at the start and end of each file.

The page also sets loudness limits: RMS between -23 and -18 dB, peaks below -3 dB and a noise floor below -60 dB RMS. Those numbers describe what ACX checks on files it accepts. They are not a way around the AI rule. If you do not have explicit authorization, a Sume TTS file is not an ACX submission, however clean it measures.

Apple, KDP and Spotify: three different answers

Apple's digital narration program is the reverse of an upload: authors cannot supply their own audio. Apple generates it, labels it as narrated by Apple Books, and the page lists exclusions such as graphic novels, poetry and cookbooks. Processing is quoted at one to two months and there are no pre-orders.

KDP's virtual voice is also generated by the platform. The eligibility page lists a reflowable ebook with a table of contents, live status, a title under 200 characters, no existing Audible version and no public-domain text. KDP Select exclusivity applies if you are enrolled.

Spotify is the one route where you bring finished audio. Its newsroom post says ElevenLabs-made audiobooks go in through Findaway Voices and the description must begin with a line saying the audiobook is narrated by a digital voice. The post gives no file-format spec, so confirm technical limits with the distributor.

What Sume TTS is good for here

Sume TTS 1.0 is priced at $0.0475 per 1,000 characters and accepts up to 20,000 characters per request. Audio that would run past 1,200 seconds fails with tts_duration_exceeded, so a book is a set of chapter-sized requests rather than one call. Voice cloning is app-only, not in the API.

In practice that fits three jobs the stores above allow or do not touch: a Spotify-style digital-voice edition through a distributor, an author-reviewed sample or trailer, and narration for your own site or newsletter. Each finished chapter can be joined into a reusable file with Timeline audio, which concatenates up to 20 Sume-hosted parts without re-synthesis.

• Plan by chapter: split text so each request stays under 20,000 characters and 1,200 seconds of audio.
• Join takes with Timeline audio concat when a store wants one chapter per file.
• Do not promise ACX, Apple Books or KDP placement from a Sume file; those programs control the voice or require humans.

Where Sume stops

Sume does not give you a store-compliance check. It has no loudness normalizer, no room-tone tool and no ACX validator, so any RMS or peak target has to be met before or after Sume with your own audio editor. Output format fields exist for MP3 and WAV, but this post does not claim any file meets a store's constant-bitrate rule.

For the cost side of a whole book, see cost to make an audiobook with AI, and for chapter splitting see the 1,200-second limit. Confirm each store's current page on the day you publish.