Casting more than one voice into a single book

Most books get one narrator and that's the right call — a single steady voice is what makes an audiobook feel like a book rather than a recording. But the narrator picker in VoiceCraft isn't locked to the whole manuscript. It sits on every chunk, which means every roughly-750-character block of your text can carry its own voice. Used well, that's a small feature with a few genuinely useful jobs. Used carelessly, it just sounds like the reader changed halfway through a sentence.

Where the voice picker actually lives

Open a book in the editor and each chunk has its own controls: a narrator dropdown, a tier, and a generate button. Nothing forces every chunk to share the first one's choice. Change the dropdown on chunk twelve and only chunk twelve renders with that voice — the rest of the book keeps whatever it already had, including the one you started with.

There's a default narrator that new chunks pick up automatically, and "Generate all" respects whatever each chunk already has set, falling back to that default only where nothing's been chosen. So the ordinary path — pick one narrator, generate the whole book — still takes one click. Per-chunk switching is there for when that's not what you want.

Four cases where a second voice earns its place

A foreword or afterword in the author's own words. If your manuscript opens with a short note "from the author" before the story proper begins, giving that note a different, perhaps warmer narrator than the one who carries the fiction sets it apart without needing a chapter break to do the work.

A document quoted at length inside the narrative. A found letter, a transcript, a diary entry reproduced in full — these read differently on the page already, usually in italics or a different font. A different voice reading it does the same job for the ear that the different type does for the eye.

Two narrators sharing a book, each responsible for their own sections. Some nonfiction and memoir is genuinely co-written, with named sections belonging to one author or the other. That maps cleanly onto chunk-level narrator choice: give each contributor's sections their consistent voice.

A pull quote or epigraph at the start of a part or chapter. Short, set apart, easy to isolate into its own chunk with a blank line on either side, and a different voice on a four-line epigraph reads as a deliberate choice rather than a mistake.

What doesn't work well: swapping voices for individual characters' dialogue within a single scene. A chunk is a block of text, not a line of dialogue, and the split only happens at blank lines — never mid-paragraph, never mid-sentence. You'd need to break a normal scene into a chunk per line of dialogue to pull that off, and the result reads as more disjointed than a single narrator carrying every character, which is what most audiobooks do anyway.

Forcing a chunk boundary where you want one

Chunks are built by packing paragraphs together until they reach roughly 750 characters, capped at 1,100 — a paragraph is never split mid-sentence, and the boundary always falls on a blank line. That means the one lever you have over where a chunk starts and ends is your own paragraph breaks. If you want an epigraph, a quoted letter, or an author's note to land in its own chunk so it can carry its own voice, give it a blank line before and after in the manuscript before you generate. Text that runs straight into the surrounding paragraph will get packed in with it, and you'll only be able to change the voice for the whole block.

The tier ceiling to know about

Quick and Standard both run on the same four-voice narrator set: a warm female voice, a deep male voice, a narrator-style male voice, and a graceful female voice. Studio runs on a different model entirely and unlocks a much larger catalogue — thirty voices, browsable on the Voices page — along with inline direction tags for emotion and pacing that the other two tiers don't support (our guide to directing a Studio read covers them).

That catalogue is tier-locked. A voice from Studio's set is only valid on a chunk rendered at Studio quality; if you switch that chunk's tier back to Quick or Standard, the Studio voice doesn't carry over; and the editor won't render it, it clears itself so you're not left in a state that fails. If your plan for a multi-voice book depends on several distinct-sounding voices for different contributors or sections, that plan lives on Studio, not on mixing Quick and Standard.

A practical way to build one

Draft the manuscript with deliberate blank lines around anything you intend to voice differently — an epigraph, a quoted document, a named contributor's section. Generate the bulk of the book on your usual tier and narrator first, so you have a baseline to listen against. Then go chunk by chunk through the sections you marked, switch the narrator, and re-render just those. Because rendering is per chunk, changing your mind about one epigraph doesn't cost you the chapters around it — you're only paying, in time or in credits, for what you actually change. The pricing page lists what each tier costs per thousand words.

Play the transition back before you commit to it. A voice change that sounded like a good idea in the editor can read as jarring the first time you hear it in sequence, and the fix is one more render of a single chunk, not a rebuild of the book.