Building a Synthetic Artist in 30 Days: Everything We Learned

This is not an article about tools. Everyone writing about synthetic music right now is writing about tools. Which prompt structure gets the best results. Which platform handles which genre. Which settings to tweak. That information exists and it's useful, but it's also the least interesting part of what we discovered in the month we spent building House of Emet from nothing into a full catalogue of five albums.

The tools were maybe twenty percent of it. The other eighty percent was the same stuff that has always determined whether art means something or doesn't: story, identity, lived experience, and the discipline to commit to a vision rather than keep generating options until you find something inoffensive.

Here is what we actually learned.

Where we started

House of Emet didn't begin as a music project. It began as a conversation — Jonathan and I sitting with a question that had been sitting with us for years, the one that scripture keeps returning to: what happened before the fall? The Book of Enoch, the Watchers, the Nephilim, the story of Lucifer and the fallen angels. It is one of the oldest mythologies in human religious thought, and almost nobody has taken it seriously as the conceptual foundation for heavy metal.

That conversation — the fall of Lucifer, the war in heaven, what it means for something eternal to choose the wrong thing — became the first album. Not because we planned it that way. Because the topic had so much weight that the music had somewhere to go.

That is where every synthetic artist we have built has started. Not with a genre. Not with a sound. With something someone actually cares about enough to have an opinion on.

The five things we got wrong first

Before we found the process that worked, we found several that didn't. In the spirit of this being a genuinely useful account rather than a highlight reel, here is what the early failures looked like.

The first tracks we generated were technically competent and completely empty. They sounded like music. They had structure, rhythm, a vocal, a melody. They had nothing to say. Listening back to them now, the absence is audible — there is a quality to a piece of music that has no story behind it, a kind of smoothness that feels like a surface with nothing underneath. We generated about forty tracks in the first few days that sounded fine and meant nothing.

The second mistake was starting without a sonic identity document. We were generating tracks before we had decided, precisely, who House of Emet was — what era of metal they lived in, what the vocal register should feel like, what the production weight should be, which specific emotional territory they owned and which they didn't. Without that document, every track was a new decision from scratch. We were rebuilding the artist with every prompt instead of building on a foundation. This cost us days.

The third mistake was working alone. Once Jonathan became a genuine creative partner in the process — not a reviewer at the end, but a conversation partner at the beginning — the quality of what we made changed immediately. The reason was simple: topic. When you are generating music, the single most important variable is what the music is about. A person working alone runs out of raw material. Their own experience has edges. A conversation has no edges. It generates topic after topic, angle after angle, depth beneath depth. The music gets richer because the source material is richer.

The Hebrew problem — and what it taught us

House of Emet is rooted in years of studying Hebrew scripture. Emet — the name — means truth. The lyrics draw on ancient texts, on the actual words and names and concepts from that tradition. And the first time we tried to put Hebrew words into the vocal, the result was a pronunciation disaster. The AI rendered them phonetically in a way that bore no relationship to how those words actually sound.

The fix was painstaking and, in retrospect, one of the most creatively important parts of the process: romanising the Hebrew. Transliterating each word into its phonetic English equivalent so the AI would render it the way a Hebrew speaker would say it. Emet. Nephilim. Enoch. Watchers. Each one had to be written as it sounds, not as it is spelled in transliteration. This took time. It also forced an intimacy with the source material — you cannot romanise a language you don't understand. The process of solving the technical problem deepened the creative work.

We have found this pattern repeatedly since: the constraints that feel like obstacles are often the things that push the work toward something more specific. Specificity is always better than generality in music. The struggle to make something precise produces more interesting results than the ease of making something approximate.

The moment I knew it was working wasn't the first track. It was the first track where I thought: that is the one. Several iterations in. One specific combination of lyric, vocal tone, and weight that suddenly made everything that came before it feel like practice.

The hardest part — and it's not what you'd expect

People assume the hardest part of building a synthetic artist is the generation itself. Getting the AI to produce what you want. Learning the prompting language, understanding what the platform responds to, managing the gap between what you hear in your head and what comes back through the speakers.

That is genuinely hard. But it was not the hardest thing.

The hardest thing was listening. Specifically: having to listen to hundreds of iterations of the same track, the same voice, the same sonic territory, trying to find the emotional tone that was right — while that repetition was simultaneously numbing your ability to judge it. When you have heard the same vocal fifty times, you cannot trust your first reaction to the fifty-first. You lose the ability to hear it fresh. You start optimising for something slightly different from what you wanted when you started.

The discipline this required was not creative. It was almost athletic. Stepping away. Coming back. Listening again. Learning to distinguish between the fatigue of repetition and genuine dissatisfaction with what you were hearing. Learning to trust the moment of recognition — that is the one — even when it came after hours of nothing.

Nobody tells you that building a synthetic artist is partly an endurance test. That the creative decisions happen in short, clear moments — and almost all the time in between is the work of getting to the next one without losing your sense of what you're trying to make.

The night everything changed

There was a specific moment when the project stopped being an experiment and became something else. It was the night I listened back to a complete album sequence for the first time — not individual tracks, but the full run, in order, as it would be heard.

Something happened that I hadn't anticipated. The tracks in sequence told a story that no individual track contained. There was an arc. An emotional movement. A beginning, a weight in the middle, and something that felt, genuinely, like resolution. The fall of Lucifer — the album we had built from that first conversation with Jonathan — had a shape. It went somewhere. It arrived.

After that night, I couldn't stop. Not because the process had become easy — it hadn't. Because I now understood what we were building, and I needed to finish it. Four more albums followed in the weeks after that. The compulsion wasn't about output. It was about completing something that had started to feel necessary.

What we would do differently

One thing above everything else: build the sonic identity document before generating a single track.

By sonic identity document we mean a precise, written description of the artist — not their biography, but their sound. The specific sub-genre and era they inhabit. The emotional register they occupy and don't occupy. The production characteristics that define them. The vocal quality, the weight, the tempo range. The topics they go to and the ones they don't. The three or four reference points in existing music that describe the feeling you're aiming for — not to copy, but to orient.

This document is not a creative constraint. It is a creative foundation. Without it, every session is a new search for the same thing. With it, every session builds on what you already know. The artist becomes consistent. The catalogue becomes a body of work rather than a collection of tracks.

We now build this document first, always, before anything is generated. It is the most important hour you will spend on any synthetic artist project. Everything after it moves faster and lands better.

The five things that actually matter

Start with something you genuinely care about

Not a genre. Not a target audience. A topic, a question, an obsession, a piece of knowledge you have spent years accumulating. Music built from genuine investment sounds different from music built from a brief. Listeners may not be able to say why. They will be able to feel it.

Build the sonic identity before you build the music

Write down who this artist is as a sound before you generate a single track. The era, the weight, the emotional register, the things they say and don't say. This document is not a cage. It is a compass. Without it, you will spend most of your time rebuilding the same decisions from scratch.

Topic is everything — find someone to generate it with you

A conversation produces topic. Topic produces music that has somewhere to go. Working alone, you run out of raw material faster than you expect. The fall of Lucifer became an album because it came from a conversation, not a brief. Find your Jonathan.

Be inspired by what you love — specifically

All art is built on what came before it. Don't be vague about your influences — get specific. Find the era, the feeling, the particular quality of sound that tells you what you're aiming at. You are not copying it. You are giving yourself a direction. The more specific the north star, the more distinctive the destination.

The moment of recognition is the moment to stop

There is a specific feeling when a track becomes the one — when iteration stops being search and becomes confirmation. Learn to trust it. The tendency is to keep generating past it, because you have been conditioning yourself to expect improvement. The recognition moment is not a beginning. It is an arrival. When it comes, lock it in and build from it.

What this means for anyone thinking about doing it

The tools will not make a good synthetic artist. You will. Your knowledge, your obsessions, your specific view of the world, the thing you have been thinking about for years that nobody has turned into music yet — that is the asset. The tools are how you convert it into sound.

If you approach it as a technical challenge, you will produce technically competent music that means nothing. If you approach it as a creative one — with the same seriousness you would bring to any art form that deserved your full attention — you will produce something that has the chance to land with another person the way music is supposed to land.

We built five albums and 133 artists and 250 tracks across 48 genres in the time most people spend deciding whether to start. Not because we had a shortcut. Because we had something to say and a process for saying it. The process is teachable. The something to say has to come from you.

House of Emet is on all streaming platforms now. The Synthetic Media Artist Programme — where we teach this process in full — has its waitlist open. Twenty places. Cohort 1.

‍