Local-first desktop studio: manuscript → chapter audio → ACX-spec packs and podcast feeds. We are running a free pilot: download the Mac app, choose Sign up inside it, and you get 30 days free while pilot seats last. No card, nothing renews. Without an account you still get 5 free generations.
macOS 0.2.11 for Apple Silicon (M1 or later, 16 GB memory). An early Windows 0.1.0 build is also available.
Download the Mac app, move it to Applications, and choose Sign up inside the app to start your free 30-day pilot while seats last.
The Projects page shows the same steps as a checklist.
No. You get five evaluation generations without an account, or a free 30-day pilot after signing up inside the app while seats last; no card is required and nothing auto-renews.
| Plan | What you get | How you get it |
|---|---|---|
| Free | 5 evaluation generations. No commercial packs or podcast-feed export | Default on install |
| Pilot month | Everything, for 30 days. No card. Does not auto-renew | Sign up inside the app (while seats last) |
| Paid plans | Not sold yet | We will ask pilot authors what they would pay first |
If something goes wrong: Settings → Get help → Create diagnostic file and email it to support@acx-narrator.com. It contains logs and numbers only — never your manuscript or audio.
For AI narration made in this app, including a clone of your own voice, export creates Google Play Books, Kobo, and direct-sale packs only, with digital-voice disclosure. The app's current store rules exclude standard ACX/Audible, INaudio (Findaway), Draft2Digital, and Apple Books for this audio, so those destinations get no pack. Spotify for Authors eligibility is unverified; no Spotify pack is generated. Read NOT_INCLUDED.txt for omitted destinations and STORE_RULES.txt for each rule, its source, and its check date. Follow the disclosure instructions in each pack: Google uses “Synthesized voice”; Kobo uses “Synthesised voice” plus a digital-voice note in the description. Direct sales also need that disclosure on the sales page. Meeting ACX technical audio specifications does not make an AI-narrated book eligible for standard ACX submission.
Work from a copy of your manuscript, then check the imported chapter boundaries and text before generating audio. After importing, read the chapter list and inspect the text at the start and end of each chapter. Look for a table of contents treated as narration, repeated page headers, missing paragraphs, and chapter headings attached to the wrong section. Fix those before spending time generating audio.
A PDF needs a selectable text layer. If you cannot select and copy a paragraph from it, run OCR in a separate tool or use a DOCX or TXT version instead. Even selectable PDF text can arrive with broken lines or misplaced page numbers. Compare the imported text with the original; the app cannot recover words that were absent from the imported file.
If one chapter becomes several short entries, select them and use Merge selected. If a section is too long, use Split at midpoint, then check that the new boundary lands in a sensible place. Give each chapter a clear title so you can identify it during listening and export.
Choose an included narrator or record your own voice with consent, then test names in Studio → Names & pronunciations. Choose a quiet room, speak naturally, and keep background music and other speakers out of the reference recording. A sample with noise or competing speech gives you a poor starting point. Listen to the saved sample before using it for a whole book.
In Studio → Names & pronunciations, enter the written name and how it should sound. For example, the app shows Siobhan → Shiv-awn as a respelling example. Generate a passage that contains the name, listen, and adjust the spelling if necessary. Also test place names, initials, and words that have more than one pronunciation in your manuscript.
Choose whether a pronunciation belongs to this book or every book on this Mac. A book-specific spelling wins when it differs from the shared one. Shared names and default narrator choices help with a second title, but they are stored locally; signing into an account does not move that library to another computer.
Yes. Generate and review one chapter with your chosen voice before committing time to the rest of the book.
Pick a chapter containing ordinary narration, dialogue, and at least one difficult name. Generate it with the voice you intend to keep. Listen while following the source text, including the opening and closing sentences. Check for missed words, repeated lines, unexpected pauses, changes of speaker, and a voice that becomes tiring over several minutes.
Use the guide's rough generation estimate for planning, not as a deadline. At 2–6 times the audio length, a ten-minute chapter would take roughly twenty to sixty minutes. Your Mac, voice engine, chapter length, and any retries affect the actual time. Measure your own first chapter before setting aside time for the rest of the book; a short successful preview is not a full-book quality test.
When a check flags a mismatch, review the passage against the manuscript. Keep-best this chapter runs repair on the chapter and keeps the better take; it is not an editor for splicing together independently recorded snippets. Listen again after repair. If the check is unavailable or the chapter remains flagged, treat it as unfinished rather than assuming the audio is ready to publish.
Open Audio Lab after you have a take you want to keep. Mastering changes the audio's technical levels; it does not correct a wrong name or replace a skipped sentence. Fix narration problems first, then run the mastering and quality checks. Master all + sample creates chapter masters and a retail sample; Renormalize addresses common loudness failures.
The technical targets are 44.1 kHz, at least 192 kbps CBR MP3, RMS between −23 and −18 dB, and peaks no higher than −3 dB. Review the verdict for each file instead of checking only the first chapter. A file can meet loudness targets and still need another listening pass for pronunciation or missing text.
After export, open the output folder and play the first and last chapter files. Check their order, titles, and endings, and listen to the retail sample. Read the included manifests and disclosure files. Technical compliance does not guarantee a distributor will accept AI narration: check that destination's current rules before uploading, including when the voice is cloned from your own recording.
Read the error message and check the relevant manuscript, engine, or export setting below; if you remain stuck, create a diagnostic file in Settings → Get help and contact support.
For support, note your app version, Mac model, the step that failed, and the exact message. Create a diagnostic file from Settings → Get help and review it before attaching it to your support email. Keep copies of your original manuscript and finished export folders somewhere separate from the app's local library; the account connection is not a backup of your book or recordings.
Full docs live in the repo: USER_GUIDE.md · DEMO_GUIDE.md