Built for the indie author who already wrote the book. ACX Narrator Studio turns an EPUB, PDF, DOCX, Markdown or TXT into chapter-wise, spec-mastered audiobook MP3s — with cloned or stock voices, a local Audio Lab, distributor packs and co-narrator royalty statements. We are running a free pilot: download for Mac, sign up inside the app, and narrate your audiobook free for 30 days. We are not charging.
A human narrator is typically $200–500 or more per finished book. The pilot is free for 30 days: download, open the app, and choose Sign up (top right) — your 30-day month starts at that first sign-in, while pilot seats last. Without an account you still get five free generations. No card; nothing renews. Report an issue.
Here is a 30-second sample narrated by the studio from a public-domain short story. The voice, prosody and room tone are the same engine that masters your finished chapter to 192 kbps.
As of 2026 the standard ACX portal requires human narration. Amazon’s authorised AI routes are KDP Virtual Voice and the invite-only ACX Voice Replica beta. INaudio (Findaway), Draft2Digital and Apple Books do not take AI narration from third-party apps either. This studio masters to ACX technical spec as a quality bar and exports for the stores that do take it: Google Play Books, Kobo and selling it yourself, with disclosure. The app shows each store's rule, its source and the date we checked it. We put that warning in the product, in the export gate, and in every pack you write. Read the full policy note →
The app is a pipeline, not a pile of settings. You move left to right down the rail and each stage tells you whether it is done, flagged or still locked.
Drop in EPUB, PDF, DOCX, Markdown or TXT. The splitter finds your chapters — with llama.cpp locally when you have a GGUF, and a heuristic pass when you don’t — then hands them back editable.
Pick a stock narrator or clone from a short reference clip. Assign different voices per speaker with the cast map, and fix stubborn names in the pronunciation list.
Audition every chapter in the app before you commit. Regenerate a single paragraph instead of an entire file when one line lands wrong.
The Audio Lab masters through the FFmpeg built into the app to 44.1 kHz, ≥192 kbps CBR MP3, RMS −23…−18 dB and peak ≤ −3 dB — then tells you exactly which file failed and offers the fix.
One click writes Google Play Books, Kobo and direct-sale folders, each with manifests, a retail sample and the disclosure that store asks for. Stores that would reject AI narration get no folder, and the export says why.
PFH, Royalty Share and RS Plus maths from real finished hours, exported as CSV and a plain-text statement you can send to the other party.
Chapter numbering, loudness targets, retail samples, disclosure files, royalty splits. The dull, decisive things.
Manuscripts, voice samples and audio stay on disk. No upload step, no per-minute cloud bill, no vendor reading your unpublished novel.
Every mastered file is measured — RMS, true peak, noise floor, sample rate, bitrate, length — and failures name the reason.
The app blocks export until you have read the 2026 third-party-AI policy, and writes the policy and each store's rule into every export. We would rather you sell where AI narration is welcome than get rejected.
Finished-hour maths, splits and statements built in, because the second person on the contract always asks for a number.
Zero-shot clone from a short reference clip (beta), or stock voices for dialogue. Cast per speaker and keep it consistent across every chapter.
Names, house voices, recorded permission, and manuscript-check history stay on this Mac and apply to the next book. You do not re-teach them. Chapter audio stays with its title.
macOS 0.2.11 is the current release. Sign in from the app to start your free pilot month. Windows ships a 0.1.0 NSIS installer today; the next Windows build will match Mac entitlements.
ACX Narrator Studio 0.2.11 · notarized DMG · Apple Silicon, 16 GB memory · free pilot month after sign-in.
Download for MacACX Narrator Studio 0.1.0 · NSIS setup · early Windows build. Prefer Mac 0.2.11 until the Windows build catches up.
Download for WindowsDownload the Mac app and choose Sign up inside it: the first sign-in starts a free 30-day month while pilot seats last. Without an account you still get five free generations. Nothing auto-charges. Paid plans come later, and we will ask pilot authors what they would pay first.
Pilot status: free, nothing auto-charges. No finished title yet — we publish the first customer quote the day one ships. Report an issue.
I am Swaroop, a solo developer in Dallas. I built ACX Narrator Studio because I could not find a desktop tool that mastered to the ACX numbers, kept the manuscript on my own machine, and produced the per-distributor packs without a per-minute cloud bill. The studio runs the audio engine locally, blocks the export gate until you have read the 2026 third-party AI policy, and writes a disclosure file into every pack. Read the user guide or follow the build in public on GitHub. Email support@acx-narrator.com — I read it.
We publish customer quotes only after a real audiobook finishes here. None yet — this paragraph will change the day one does. If you finish a title first, the studio asks if you would like to share a short quote, with a book title and a genre you choose.
Download the Mac DMGStill stuck? Email support@acx-narrator.com — a real person reads it.
Not through the standard ACX portal. As of 2026 ACX requires human narration and prohibits unauthorised third-party AI. Amazon’s own authorised routes are KDP Virtual Voice and the invite-only ACX Voice Replica beta. INaudio (Findaway), Draft2Digital and Apple Books do not take AI narration made in third-party apps either. This studio masters to ACX technical spec so your files are clean, and it exports for the stores that do take them — Google Play Books, Kobo and selling it yourself — with disclosure.
No. Since 0.2.1 the audio engine (FFmpeg) and the local writing model runner ship inside the app — there is no Homebrew step and nothing to put on your PATH. You do not need a GPU either: the app uses one when you have one and runs on the CPU when you do not. Two things are optional downloads because of their size, both one click in Settings: the voice engine that narrates in a voice you record, and a writing model for smarter chapter splitting. Until you add the voice engine the app narrates with a macOS system voice and says so on screen.
The manuscript stays in a local SQLite library under your home folder. Sign-in is only for your license account — there is no upload path for your text or audio.
A human narrator is typically $200–500 or more per finished book, plus audition and scheduling time. The pilot is free for 30 days — we are not charging. Paid plans come later.
macOS on Apple Silicon ships as a notarized 0.2.11 DMG (16 GB memory recommended), with a free pilot month after sign-in. Windows x64 has a 0.1.0 NSIS installer on the Download section. Linux is not planned for v1.
Audition before you commit. The studio renders every chapter in the app before you sign it off, and when one line lands wrong you regenerate a single paragraph instead of an entire file. A chapter audio file carries its source text with it, so you can replay the paragraph against the read-aloud and compare. The long-form test is the audit trail: every export pack logs RMS, true peak, noise floor, sample rate, bitrate and length for every chapter MP3, and tells you exactly which file failed and offers the fix.
No. Choose Sign up inside the app and your first sign-in starts a free 30-day pilot month while seats last; without an account you still get 5 free evaluation generations. We are not taking a card. Paid plans come later.
Download the Mac DMG, import a chapter, hear it back in minutes. Five free generations — no card required.