Clips
Clips play in order; shape the sequence with the arrows.
Bookmarks
AI edit ledger
Repairs — silence, soften — only clean what she recorded, and need nothing switched on here. Re-voice is different: it puts speech in the book that she did not perform. Turning this on is a disclosure obligation, not just a setting. Every re-voiced passage is recorded in the ledger, written into the audio file's own tags, shown on the clip, and carried onto the public page — and can be undone at any time, because the original take is never altered.
Flags from listeners
Make every chapter sound like the same room on the same day.
Recorded over weeks, chapters drift: the mic moves, the room changes, some days you are louder. This measures every clip, then corrects all of them to one target — −20 dBFS RMS, peaks held under −3 dBFS, noise floor under −60 dB. Those are ACX's numbers, which is what Audible checks on submission.
It also trims each chapter back to where your voice actually starts and stops, then pads every one to the same 0.75s lead-in and 2s lead-out, so one chapter does not open instantly and the next after a pause. Your rights details are written into each file, and any dials or repairs you set are applied here.
Set for the whole book, not per chapter: ACX requires one consistent format across a title, so a single chapter arriving as stereo must not change what the rest is delivered as. This one was recorded on one microphone, so mono is correct.
Your original recordings are never touched. Each run renders new files into its own folder, so an earlier master stays playable and reversible from Runs.
Search every chapter for moments that measure like the ones you flagged.
You cannot re-listen to 200 minutes looking for stumbles. This learns from the marks you already made — the moments you flagged as errors and the takes you starred as keepers — and scans every clip, including ones you have never played, for passages whose measurements resemble your errors and not your keepers.
It hears nothing. It compares level, brightness, dynamics and pacing, so what comes back is suspects, not verdicts, and marginal ones are labelled as such. Each is playable in one click so you decide.
Needs a few keepers first — with nothing starred there is no profile to compare against.
Every chapter, in order, as one continuous WAV.
This is the handoff format — a single uncompressed file for an editor, a proof listener who wants one thing to play, or an archive copy of the whole book.
This is not the ACX submission format. Audible wants one file per chapter, not a merged one. Download the individual mastered chapters from the Clips list for that. This is for handing the book to a person.
The merge runs in this browser and decodes every clip, so a long book takes a while and needs the tab left open.
Submit this collection to the content pipeline.
Publishing writes the collection into studio_content, where the
rest of the site can pick it up — the podcast feed, a course, a product listing. It does not
send anything to Audible or ACX; that submission is still something you do at their end with
the per-chapter files.
Master first. Publishing before a master run hands on the raw takes, at whatever level each was recorded.
What AI is allowed to do to this book.
Review compares measurements against your marks — it never hears audio. Transcription reads each take once and is billed per minute. Voice is the only one that creates audio, so it is off until you arm it here.
Re-voice replaces performance with generated speech in her cloned voice — length locked, words from the transcript, and every use is written to the Edit log as GENERATED. ACX requires that disclosure; get the narrator’s written consent for the clone before enabling. The key is stored in this tenant’s database and is readable by signed-in users of this app.
Pick a collection — or create one.
Each collection is an ordered set of audio clips:
chapters of an audiobook, tracks of an album, lessons of a course.
Video Studio
Coming soon — the audio studio is fully live today.