ACX Narrator Studio turns an EPUB, PDF, DOCX, Markdown or TXT into chapter-wise, spec-mastered MP3s — with cloned or stock voices, a local Audio Lab, distributor packs and co-narrator royalty statements. A human narrator costs $200–500 a book. Pro is $49/month.
Five free generations. Sign in when you want Pro — no card required to start.
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. This studio masters to ACX technical spec and is built for wide distribution with disclosure — INaudio/Findaway, Draft2Digital, Spotify for Authors, Apple, Google and Kobo. 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 an acx-spec pack plus INaudio/Findaway, Draft2Digital, Spotify for Authors and Apple/Google/Kobo folders, each with manifests, a retail sample and the disclosure file.
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 No upload step, no per-minute cloud bill, no vendor reading your unpublished novel.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 ships ACX_AI_POLICY.txt inside the pack. We would rather you distribute wide 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.
Projects, chapter state and settings persist in a local SQLite library, so book two starts where book one left off.
macOS 0.2.3 is the current release (5 free generations). Windows ships a 0.1.0 NSIS installer today; the next Windows build will match Mac entitlements.
ACX Narrator Studio 0.2.3 · notarized DMG · 5 free generations on first launch.
Download for MacACX Narrator Studio 0.1.0 · NSIS setup · early Windows build. Prefer Mac 0.2.3 until the Windows build catches up.
Download for WindowsPro is billed through Clerk at $49/month or $300/year. Free accounts get five evaluation generations and cannot create commercial export packs.
Sign in to sync Pro. See the user guide. Support: iamk.swaroop@icloud.com.
We publish customer quotes only after a real audiobook ships with the studio. None yet — this line will change when the first one does.
We publish customer quotes only after a real audiobook ships with the studio. None yet — this line will change when the first one does.
We publish customer quotes only after a real audiobook ships with the studio. None yet — this line will change when the first one does.
Still stuck? Email iamk.swaroop@icloud.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. This studio masters to ACX technical spec so your files are clean, and it points you at wide distribution — INaudio/Findaway, Draft2Digital, Spotify for Authors, Apple, Google and Kobo — 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. Pro is $49/month or $300/year, and you can re-render a chapter at 2am when you spot a typo.
macOS on Apple Silicon ships as a notarized 0.2.3 DMG with 5 free generations. Windows x64 has a 0.1.0 NSIS installer on the Download section. Linux is not planned for v1.
No. You get 5 free evaluation generations. Pro is optional and billed through Clerk at $49/month or $300/year. Cancelling keeps access through the paid period, then for seven offline days.
Download the Mac DMG, import a chapter, hear it back in minutes. Five free generations — no card required.