Character Voicing in AI Audiobooks: Multi-Voice Fiction
Updated September 3, 2026
TL;DR: one narrator plus a small, distinct cast reads better than a voice per walk-on. In Studio you assign voices per paragraph and the assignments hold across the book; on Ultra voices, tags such as [angry] or [whispers] handle the line-level emotion. Cast for contrast, direct for consistency, and keep the narrator neutral.
Full cast or single narrator?
Most audiobooks are single-narrator with the narrator doing light character colour. Full-cast productions suit dialogue-heavy fiction, YA, and anything with a clear ensemble. AI makes full cast cheap — the cost is your review time, not credits — so the decision is about the book, not the budget.
Casting for contrast
- Narrator: a neutral, even voice the ear can rest on — Stella or Carl.
- Two to four principal characters with clearly different timbres: age, gender, pace, accent.
- Minor characters: let the narrator read them, or share one "utility" voice.
- Antagonists and children are where designed voices shine; describe them rather than hunting the library.
Assigning in Studio
- Mark the dialogueConsistent quote marks and speaker tags in the manuscript make the split fast.
- Assign per paragraphSet the narrator on the project, then assign a character voice to dialogue paragraphs. The assignment persists in every chapter.
- Direct the lineOn Ultra voices, add [angry], [gentle], [surprised] or [whispers] before the line, or give each character its own director's note.
- Review scenes, not paragraphsListen to a whole scene for balance; a character voice that's slightly louder or faster than the narrator reads as shouting.
Keeping the cast consistent
- One project per book; don't recreate voices between volumes — reuse them from My Voices.
- Write the cast list (voice, director's note, tag habits) into the project notes.
- Add pronunciation rules for character names before generating; they apply to every voice.
- When you change a character's voice, regenerate every earlier line — listeners notice.
What not to do
- Don't give the narrator emotion tags on narration; reserve them for dialogue and the occasional dramatic beat.
- Don't use extreme tags on every line — [angry] once lands; six times reads as noise.
- Don't cast two voices with similar timbre for characters who share scenes.
Frequently asked questions
How many voices should a novel have?
A narrator plus two to four principals is the sweet spot. Beyond that, listeners lose track and review time balloons.
Can a character voice have a different accent?
Yes — pick or design a voice with that accent, or on Ultra voices set the accent in the character's director's note.
Do emotion tags work for every character voice?
On Leap Ultra voices, yes. Other tiers don't interpret tags, so cast dialogue-heavy characters on Ultra.
What does full cast cost?
The same credits as single narrator — billing is per character generated, not per voice. Designed voices cost 2,500 credits each to create.