From Idea to Audiobook: How We Use AI to Craft Full-Length Stories

Back to Blog

Let's Be Honest About AI and Storytelling

There is a lot of noise around AI right now. Some people treat it as a magic box that can replace writers overnight. Others dismiss it entirely, insisting that anything AI touches is worthless. The truth, as usual, is somewhere in the middle — and it is more interesting than either extreme.

At Laverty Prod., we have been using AI as part of our creative process for a while now. Our series Nocturnal Tales has over 100 episodes. Each one is a full-length story with narration, sound design, and music. Producing that volume of quality content with a small team would be nearly impossible using traditional methods alone. AI makes it possible — but not in the way most people assume.

This post is a genuine behind-the-scenes look at exactly how we go from a blank page to a finished audiobook. No hype, no hand-waving. Just the actual process, step by step, including the parts that are messy, slow, and very much human.

Step 1: The Spark — Where Every Story Begins

Every story starts with an idea. This is the part that no AI can do for you, and frankly, it is the part that matters most.

Sometimes the spark is a concept: what if a small Irish village had a secret that only came alive after midnight? Sometimes it is a character: a retired detective who cannot stop solving mysteries even in her sleep. Sometimes it is just a feeling — we want to create something that captures the loneliness of a lighthouse keeper, or the tension of a locked-room mystery, or the warmth of an unexpected friendship.

These ideas come from everywhere. From books we have read, conversations we have had, places we have visited, news stories that stuck with us. They come from lying awake at three in the morning with a thought that will not let go. They come from the deeply human experience of wondering "what if?"

This is the foundation of everything we do. The AI does not generate our ideas. We do. And if the idea is weak, no amount of technology will save the story. We have learned this the hard way — there are abandoned outlines on our drives that prove it. A great tool cannot fix a mediocre concept. So we spend real time here, turning ideas over, testing them, asking whether this is a story we ourselves would want to listen to.

Step 2: Drafting the Outline — The Blueprint

Once we have an idea we believe in, we build the outline. This is where storytelling craft comes in, and it is entirely a human exercise.

We sketch the full story arc: beginning, middle, end. We map out character arcs — where does each character start, what do they want, what stands in their way, and how are they different by the end? We identify key plot points, twists, moments of tension, and moments of release. We think about the emotional journey we want the listener to go on.

For a series like Nocturnal Tales, this also means thinking about how individual episodes connect. Are there recurring characters? Does the world have rules that need to stay consistent? Is there an overarching mystery that threads through multiple stories? These are structural decisions that require a human understanding of narrative — what feels satisfying, what feels earned, what will keep someone pressing play on the next episode.

Our outlines are detailed. They are not vague one-liners. A typical outline runs several pages and includes notes on tone, pacing, key dialogue beats, and specific scenes we know we want to hit. Think of it like an architect's blueprint. The more precise the blueprint, the better the building. The same is true for stories.

We also make deliberate choices about what the story is not. What tropes are we avoiding? What expectations are we going to subvert? Where are we going to let silence do the work instead of words? These decisions shape everything that follows, and they are decisions only a human storyteller can make.

Step 3: Teaching the AI Our Voice — The Secret Sauce

Here is where things get interesting, and where our process probably differs from what most people imagine when they hear "AI-assisted writing."

We do not just open a chatbot and type "write me a horror story." That approach produces generic, bland, instantly forgettable content. If you have ever read AI-generated fiction that felt lifeless and formulaic, this is why. The default output of any large language model is, by definition, average. It is a statistical blend of everything it was trained on. Average is not what we are after.

Instead, we have spent months — genuinely, months — training our AI assistant to write in a specific style. Our style. This is an ongoing, iterative process that never really ends.

What does "training" look like in practice? It means feeding the AI examples of writing we admire and explaining why we admire it. It means crafting detailed style guides that describe the voice we want: sentence length, vocabulary choices, how much description versus dialogue, the rhythm of paragraphs, how to handle exposition without slowing the pace. It means writing sample passages ourselves and then prompting the AI to match them.

It also means a lot of failure. The AI will misunderstand instructions. It will latch onto the wrong elements. It will produce something that technically follows the rules but has no soul. And then we adjust, refine, re-prompt, and try again. Over time, the system gets better. The prompts become more precise. The AI starts to "understand" — in whatever mechanical sense that word applies — what we are looking for.

Think of it like training a very fast, very eager apprentice who has read every book ever written but has never actually felt anything. The knowledge is there. The instinct is not. Your job is to provide the instinct, the taste, the judgement. The AI provides the speed and the tireless capacity to generate and regenerate until you get something that works.

Step 4: AI Writes the First Draft — Fast but Rough

With the outline locked and the AI tuned to our voice, we move into drafting. This is where the AI earns its keep in terms of pure productivity.

The AI generates chapters based on our outline, our style guide, and our specific instructions for each section. It is fast. What might take a human writer weeks to draft can be produced in hours. For a production studio putting out regular content — like our weekly Nocturnal Tales episodes — this speed is not a luxury, it is a necessity.

But let us be very clear about something: a first draft is not a finished story. This is true when a human writes it, and it is doubly true when an AI writes it. The first draft is raw material. It is clay that has been roughly shaped but not sculpted.

AI-generated first drafts have predictable strengths and weaknesses. On the positive side, the prose is usually clean, grammatically correct, and follows the outline faithfully. The AI does not get writer's block. It does not have bad days. It does not stare at a blank page for an hour and then go make tea instead of writing.

On the negative side, AI first drafts tend to be emotionally flat in subtle ways. The AI will describe a character's grief without truly conveying it. Dialogue can feel functional rather than natural — characters say what they need to say for the plot, but not the way a real person would actually say it. Pacing can be mechanical, with each scene given roughly equal weight regardless of its dramatic importance. And the AI has a tendency to reach for the obvious choice — the expected metaphor, the predictable plot beat, the safe emotional note.

These are not dealbreakers. They are exactly what we expect. The first draft is not the product. It is the starting point for the real work.

Step 5: Human Review and Editing — Where the Real Work Happens

If you take one thing away from this entire post, let it be this: the editing phase is where the story actually comes to life. This is where the overwhelming majority of our creative effort goes, and it is entirely human.

We read every single line. Not skim — read. Out loud, often, because this content will ultimately be listened to, not read silently. If a sentence sounds wrong when spoken, it does not matter how good it looks on paper.

Here is what we are looking for and fixing during the editing pass:

Dialogue that sounds like real people talking. AI-generated dialogue is often too clean, too on-the-nose. Real people interrupt each other, trail off, say the wrong thing, use pet phrases, speak in incomplete sentences. We rewrite dialogue extensively to make characters feel like individuals rather than mouthpieces for the plot.

Emotional truth. There is a difference between describing an emotion and evoking it. The AI can write "she felt devastated." We need the reader to feel the devastation through specific, concrete details — the way she keeps straightening the cushions on the sofa because if she stops moving she will fall apart. This kind of specificity almost always comes from the human edit, not the AI draft.

Pacing and rhythm. Stories breathe. They speed up and slow down. A chase scene should feel breathless — short sentences, sharp verbs, no time for reflection. A quiet moment of realisation should linger. The AI tends to maintain a steady, even pace throughout. We restructure, cut, and expand to create the dynamic rhythm that keeps listeners engaged.

Consistency. Across a 100-episode series, continuity matters enormously. Does this character's reaction match what we established about them in episode 12? Did we already use this plot device three episodes ago? Is this location description consistent with how we described it the first time? The AI has no memory of previous episodes unless we provide it. We are the continuity department.

Cutting the obvious. The AI's tendency to go for the expected choice means we spend a lot of time asking "is this too predictable?" If the listener can see the twist coming three paragraphs away, we need to rethink it. Sometimes the best edit is to cut a scene entirely and replace it with something the AI never would have written — because it is stranger, bolder, or more specific to these particular characters.

Step 6: The Iteration Loop — Write, Read, Edit, Repeat

Editing is not a single pass. It is a loop, and sometimes a long one.

The process looks like this: we read the AI-generated chapter, mark it up with notes and changes, then go back to the AI with specific instructions. "Rewrite this scene from the antagonist's perspective." "Make this dialogue more confrontational — these characters do not like each other." "This reveal happens too early, restructure the chapter so it lands at the end." "This description is generic, make it specific to 1990s Dublin."

The AI generates a new version. We read it again. It is better in some ways, worse in others. We make more notes. We re-prompt. Sometimes we throw out the AI's output entirely and write a section by hand because we know exactly what it needs to sound like and it is faster to write it ourselves than to explain it to the AI.

A single chapter might go through five to ten rounds of this cycle. For pivotal chapters — season finales, major reveals, emotionally critical scenes — it can be even more. The loop continues until the chapter reads the way it should, until it has that quality where you forget you are reading (or listening) and you are simply inside the story.

This is not a fast process, despite the AI's speed. The AI can regenerate a chapter in minutes, but we need hours to properly read, evaluate, and provide thoughtful feedback. The bottleneck is never the AI's generation speed. It is always the human's ability to evaluate and direct. This is as it should be. The quality of the output is directly proportional to the quality of the human input.

We want to be explicit about something here: the AI is a tool, not the author. We are the authors. The AI generates text. We shape stories. There is an enormous difference between those two things, and it is the difference between content and craft.

Step 7: Narration and Production — Bringing the Story to Life

Once the written story is locked — once every chapter has been through the editing loop and we are genuinely proud of it — we move into audio production. This is a whole second phase of creative work.

Narration is everything in an audiobook. The narrator's voice, pacing, emotional range, and character work determine whether a listener stays for twelve hours or drops off after twelve minutes. We approach narration with the same care we give to writing. We consider the story's tone, the target audience, and the specific demands of each scene.

For dialogue-heavy passages, distinct character voices matter. For descriptive passages, the narrator's pacing and emphasis shape the listener's mental image. For tense, suspenseful scenes — and Nocturnal Tales has plenty of those — the narrator's delivery can make the difference between genuine unease and flat reading.

Beyond narration, we layer in sound design and music. Not gratuitously — we never add elements just because we can. But when a scene is set in a rain-soaked alley, the sound of water on cobblestones creates an immersion that words alone cannot achieve. When a mystery reaches its climax, a carefully crafted musical score can heighten the tension without the listener even consciously noticing it.

Editing and mastering the audio is a technical process with its own demands. We normalise levels, clean up artefacts, balance narration against background audio, and ensure everything meets the technical specifications required by each distribution platform. This is meticulous, time-consuming work, and it is the final polish that turns a good story into a professional audiobook.

Step 8: Distribution — Getting It to Your Ears

The final step is distribution, and it involves more than just pressing "upload."

We publish to YouTube, Spotify, and other major platforms. Each has its own requirements — file formats, metadata structures, content policies, thumbnail specifications. We manage all of it.

But distribution is also about presentation. We create artwork for each episode. We write descriptions that are compelling and accurate — descriptions that help potential listeners decide whether this story is for them, without giving away spoilers. We plan release schedules to build momentum and keep our audience engaged. We think about discoverability: what categories to select, what keywords to include, how to position each story so the right people find it.

For a series like Nocturnal Tales with over 100 episodes, this is a significant ongoing effort. It is not glamorous work, but it is essential. A brilliant story that nobody can find might as well not exist.

What We Have Learned Along the Way

After producing this volume of content with this process, we have drawn a few conclusions that we think are worth sharing.

AI dramatically lowers the barrier to entry. You do not need to be a professional writer with a creative writing degree to create something genuinely good. If you have a strong idea, clear taste, and the willingness to edit ruthlessly, AI gives you the raw material to work with. It democratises storytelling in a real way, and we think that is a good thing.

AI does not replace the need for human creativity. It shifts where the creativity is applied. Less time staring at a blank page, more time shaping, cutting, refining, and making deliberate creative choices. The skills that matter most in this process are taste, judgement, and a willingness to throw out work that is not good enough — including your own and the AI's.

The quality ceiling is set by the human, not the AI. A lazy human using AI will produce lazy work. A skilled, demanding human using AI will produce something they can be genuinely proud of. The tool does not determine the quality. The person wielding it does.

Speed enables ambition. Without AI, a series like Nocturnal Tales would require either a large writing team or a much slower release schedule. AI lets a small team operate at a scale that was previously only available to well-funded studios. That is not about cutting corners — it is about expanding what is possible.

Transparency matters. We are open about using AI in our process because we believe there is nothing to hide. The stories are good because we make them good, through rigorous editing and high standards. The AI is part of our toolkit, alongside our microphones, our DAW, our sound libraries, and our hard-won knowledge of what makes a story work. Tools are not something to be ashamed of. They are something to be skilled with.

This Process Is Not Just for Us

One of the things we are most excited about is that this approach is accessible. You do not need a studio. You do not need a publishing deal. You do not need years of writing experience, although that certainly helps.

What you do need is an idea you care about, the patience to learn how to work with AI effectively, and the honesty to recognise when something is not working and fix it. If you have those three things, you can create a full-length story. You can produce an audiobook. You can put something out into the world that is genuinely yours.

We are not saying it is easy. The editing loop alone will test your patience. Teaching an AI your voice is a long, iterative process. And producing the audio — narration, sound design, mixing — is a skill set in itself. But it is all learnable. Every part of this process can be learned by someone who is willing to put in the work.

The gatekeepers are disappearing. The tools are available. The platforms are open to independent creators. The only question is whether you have a story worth telling — and only you can answer that.