Manuscript Studio

Audiobooks

The Narration Studio — AI narration in four stages, per-passage re-rendering, narrator engines, performance direction, quality checking and store-ready delivery.

Manuscript Studio narrates a book with AI voices and packages the result the way audiobook stores expect.

Available on Creator and above. Open it from the bookshelf's menu, or the header's book menu.

The pages

  1. Narrators and voices — the engines, what each plan reaches, and auditioning before you commit.
  2. Performance direction — cues, cast voices, and the auto-tagger.
  3. Rendering and cost — passages, punch-and-roll re-rendering, pricing and cancelling.
  4. Quality checking — catching the two failures that are otherwise silent.
  5. Delivery and stores — what a store-ready package contains, and where you can actually sell it.
  6. Listening — the player that follows you around the app.

Four stages

① Setup — pick the engine and voice, and audition them. ② Tag — optional performance direction, by hand or auto-tagged. ③ Render — synthesise. Priced by length, quoted before you spend. ④ Edit — review, re-render individual passages, check quality.

Like the other studios, these are a running order rather than a progress bar. Moving between them is always free; only the terminal actions gate, and only on real state — no voice selected, nothing to render.

The idea that makes it usable

The book is cut into passages at scene and paragraph boundaries, and each passage is content-hashed.

Only passages whose content changed are re-rendered, and only those are charged. Fix a typo on page 200 and you re-narrate that passage, not the book.

This is punch-and-roll — the technique real narrators use — and it is what makes narrating a book you are still editing affordable rather than a one-shot bet.

The script is the canvas

Unlike the Cover and Promotion studios, which centre on a fixed-aspect artwork you zoom and pan, the Narration Studio's main object is a long scrolling script. So there is no zoom or fit here; it reads like the editor, with the transport where the zoom controls would be.

Each passage is a card carrying its state, and rendered and verified are two different marks — a passage that has never been checked never displays as having passed.