How to make a serial-killer thriller audiobook with AI voices

To make a serial-killer thriller audiobook with AI, you paste your manuscript or import the EPUB into AudioProducer, cast a voice for each character with Auto-Assign, generate chapter by chapter, and export an MP3 you download and own. The decision that separates this genre from every other thriller is structural: most of these books alternate between the investigation and chapters written from inside the killer, and those two tracks have to sound like one book.

This guide covers the two-track casting problem, how to perform the procedural material the genre is full of, the content warning that belongs at the front of a violent book, and what changes across a series. If you are new to the process, our guide on how to make an audiobook with AI covers the mechanics that apply to any genre.

Why this genre is a structure problem before it is a voice problem

A serial-killer thriller is rarely one continuous narrative. It is a clock: an investigator working forward through evidence, a killer working forward through a plan, and a reader who knows things neither of them does. On the page, a chapter break and a chapter heading do all the work of telling you which track you are on.

Audio has no chapter heading a listener is looking at. What a listener has instead is the voice, and that makes the track switch a production decision rather than a formatting one. Get it right and the alternation is the engine of the book. Get it wrong and the listener spends the first paragraph of every other chapter working out whose head they are in, which is exactly the paragraph the genre uses to build dread.

The good news is that this is one of the few genre problems that per-character casting solves almost mechanically. You are not trying to make the book sound frightening. You are trying to make it legible at speed, and the fright takes care of itself.

Casting the two tracks so a listener always knows which one they are on

Assign the investigation narration and the killer chapters to two distinct voices, and choose them for separation rather than for menace. The test is not whether each voice sounds good; it is whether a listener who has stopped paying attention can tell within four words which track has started.

Separation comes from register and pace more reliably than from pitch. An investigative track that runs slightly quicker and busier against a killer track that is slower and more even will read as two different books-within-a-book even at low volume in a car. Audition both on the first two sentences of an actual switch, back to back, rather than on a paragraph from the middle of each.

On the specific question of how the killer's voice should sound, do not reach for sinister. That argument is made in full in our guide to the psychological thriller audiobook, which covers casting a calm, reasonable narrator whose content does the damage; it applies here without modification and this post does not repeat it. What is specific to a serial-killer book is that you are applying that rule to one of two voices, and the calm has to hold up against a second track that is allowed to be urgent.

The other scene worth producing on its own first is the interrogation, and our mystery thriller guide already makes the case for running it before anything else: two people, one of them lying, and no dialogue tags to lean on. In this genre that scene is usually four to six pages of unbroken exchange, which is longer than most authors expect a two-voice scene to hold. Generate it early, listen to it whole, and fix casting there rather than discovering the problem in chapter twenty.

Case files and autopsy reports are not prose, and should not be performed like prose

This genre is full of inserted documents: a case file, a post-mortem summary, a victim list, a transcript of a recorded interview. On the page they are set apart by a font change or a block indent. In audio there is no font, so the delivery has to carry the distinction.

Perform the document material flat. Slightly slower, evenly stressed, with none of the shaping that narrative prose gets. It sounds counterintuitive when you read it, and it works because the contrast against the surrounding narration is free tension: a clinical paragraph delivered without emphasis, sitting between two paragraphs that have plenty, is genuinely unsettling in a way a dramatic read of the same words is not.

Two practical notes. Give each document type a consistent treatment for the whole book, so the listener learns the convention once. And write a short orienting clause into the manuscript before a long insert - naming what the document is and who produced it - because a listener cannot see where the block starts. This is a different problem from crediting real quoted material, which our true crime audiobook guide covers for nonfiction; here the documents are yours, and the question is performance rather than attribution.

A spoken content warning belongs at the front of this book

The genre includes violence, and often sexual violence, and listeners do not get the page's usual warnings: no cover copy in their hand, no flicking ahead. A short spoken note before the first chapter costs you fifteen seconds and is standard practice in commercial audio for this category.

Keep it plain and non-spoilery: what the book contains, stated once, in the narrator's voice rather than as an announcement. Our guide on how to add a content warning to your audiobook has the wording patterns and where to place the file in your chapter order, so it is worth producing that piece first while the casting is fresh.

Producing it as a series without leaking the reveal

Serial-killer thrillers sell in series, and the standing rule across every genre applies: pin your cast list and reuse the same voice for every returning character, book after book, so the recurring detective does not quietly become someone else between volumes.

The genre-specific trap is the killer track in a book whose killer is not revealed until late. If you cast the killer chapters with the same voice you have already assigned to a named suspect, you have told your listener who it is in chapter three. Keep the killer track on a voice that belongs to no one else in the cast, and only merge the two assignments in a later book where the identity is known and the effect is deliberate. Write that decision down in your casting list next to the book it covers, because it is the kind of choice that is obvious while you are making it and invisible six months later.

Working chapter by chapter helps here as well. You can produce and listen to the first switch, and the first document insert, before committing the rest of the manuscript, which is when casting changes are cheap.

What you export and where it goes

A serial-killer book sits inside the broader thriller audiobook family, and the neighbouring genres are worth reading for cast-building: the noir detective guide in particular covers hard-boiled narration that shares a lot of surface with this one. The full guide index lists every production guide we publish.

When the narration is finished, AudioProducer exports a finished MP3 that you download and own. It does not publish, distribute, or host your audiobook anywhere. You take the file and release it wherever you already release your books, following that platform's own submission rules.

The free tier covers 1,200 words a month with no card required, which is enough to run one track switch through two casts and hear whether a listener could tell them apart. Paid plans start from $39.99 per month when you are ready to produce a full book. Create a free account, paste the first killer chapter and the investigation chapter on either side of it, and listen to the switch on headphones before you produce anything else.