Generated Sound Is the Material: FocusZen® Code and the Limits of the “AI Music” Definition

What happens when generative audio stops being the finished product and becomes construction material?

The music industry has entered a strange moment. On one side, generative artificial intelligence can produce a track within minutes. On the other, artists are increasingly asking what data these models were trained on, whether their work was used with the consent of rights holders, and what happens to the music ecosystem when streaming platforms can be flooded with thousands of automatically generated tracks. These are legitimate questions. The problem begins when we try to describe every possible use of generative audio with one label: AI-generated music. Does information about the technology used to generate sound really tell us enough about how a particular work was created? FocusZen® Code is an interesting case precisely because it does not attempt to escape that question. We do not hide the use of generative technology. At the same time, we do not present the FocusZen® Code process as traditional composition or musical performance. We call it: Sound Architecture / Attention Engineering. And that distinction does not begin with a name. It begins before the first generation.

SOUND ARCHITECTURE BEGINS BEFORE THE FIRST SOUND

In FocusZen® Code, the process does not begin when the “Generate” button is pressed. It begins with a question: what exactly do I want to build? An intention comes first, followed by a direction and a combination of elements intended to form a specific sonic environment. Only then can that concept begin to be translated into a specification and into language that a generative system can interpret. But between the initial idea and generation, there is another important stage: research. One of the principles behind FocusZen® Code is the deliberate avoidance of building our own sonic language by copying the language of other creators. We do not begin with: “I want something that sounds like X.” We do not attempt to reproduce a specific track, producer, artist, or recognisable signature style. We do the opposite. As far as the available tools, search systems, sources and AI systems allow, we investigate whether the combination we are currently designing already has a clear equivalent, whether a similar configuration already exists within music, and whether the direction is moving too close to something that already has its own recognisable identity. This is not, and cannot be, a mathematical certificate of worldwide originality. There is no complete database containing every recording, experiment and sonic construction ever created that would allow anyone to honestly state: “Nobody in the world has ever done this before.” What we can do, however, is treat originality as a design constraint rather than an assumption. Originality is treated as a design constraint, not an assumption. If our research identifies something too close to the direction of the architecture we are developing, we do not try to move closer to it. We do the opposite. We redesign the assumptions. We change the combination. We remove an element. We introduce another. We alter the relationships between rhythm, space, texture, dynamics, vocals, low-frequency structure and other components. We rebuild the prompt. Then we check again. The objective is not to ask: “How accurately can AI reproduce something that already exists?” The objective is: “How far can we move the architecture away from what already exists while preserving the function it was designed to perform?” This is where the individual language of FocusZen® Code begins to take shape — not from someone else’s work, but from our own concept.

THE PROMPT IS NOT THE BEGINNING OF THE IDEA

In discussions about generative music, the word “prompt” is often presented as a magical sentence entered into a generator: “Make me a song like X.” Generate. Done. In FocusZen® Code, the prompt serves a different function. It is an attempt to translate a previously designed concept into language that a generative system can interpret. That is why it may be rebuilt repeatedly. If it produces an overly predictable result, we change it. If it moves too close to an existing or recognisable sound, we change it. If the generator interprets the relationships between elements differently from the specification, we change it. If the generated material does not fulfil the intended function of the architecture, we change it. The prompt is therefore not a substitute for the design. The prompt is one of the tools used to execute the design.

THE SAME GENERATOR. TWO COMPLETELY DIFFERENT PROCESSES.

Imagine two people using exactly the same generative system. The first enters an instruction, generates a track, selects the result and publishes it. Their process might look like this: PROMPT → GENERATION → PUBLICATION. The second begins before the prompt exists. They define the intended function. They develop the concept. They research existing directions. They examine similarities. If something appears too close to their own developing idea, they redesign it. They build the specification. Only then do they translate that specification into prompt architecture and begin generating sonic material. The first generation, however, is not the finished product. It is the first test of the specification. A more complete FocusZen® Code pipeline therefore looks like this: INTENT → RESEARCH → SIMILARITY CHECK → DIRECTION ADJUSTMENT → SPECIFICATION → PROMPT ARCHITECTURE → MATERIAL GENERATION → EVALUATION → REJECTION / SELECTION → RECONSTRUCTION → REGENERATION → TESTING → REFINEMENT → SOUND ARCHITECTURE. And this pipeline can loop backwards. If the result moves in the wrong direction, we return. Sometimes to selection. Sometimes to the prompt. Sometimes to the specification. Sometimes all the way back to the original concept. This distinction matters because it reveals something that is often lost in the debate around generative audio: the generator exists inside the process. It is not the entire process.

GENERATION DOES NOT EQUAL PUBLICATION

This is one of the fundamental principles of FocusZen® Code. During development, a large number of sonic constructions may be generated. That does not mean that all of them — or even most of them — will ever be published. Material is listened to, compared, rejected, regenerated and reconstructed. From many possibilities, only a small fraction may ultimately be accepted as FocusZen® Code. Generation does not equal publication. Generation produces possibilities. Human selection determines what belongs to the architecture. This distinction becomes particularly important at a time when one of the central criticisms of generative audio is its ability to produce enormous quantities of tracks with almost no practical limit. FocusZen® Code operates in the opposite direction. More generated material does not have to mean more published material. It can simply mean: more material to reject.

SOUND AS CONSTRUCTION MATERIAL

This is where the fundamental difference in our approach to generative audio becomes visible. We do not automatically treat a generated result as a finished work. We treat sound as construction material. Think about architecture. Steel does not design a skyscraper. A brick does not design a house. Glass does not decide where the window should be. Materials have properties, but architecture emerges only when someone defines their function, relationships, position and behaviour within a larger structure. In FocusZen® Code, the material can include synthetic frequencies, sub-bass, rhythm, textures, space, repetition, silence, language, vocals, recorded human voice and digitally generated sonic elements. They do not all have to originate from the same source. The origin of a material does not determine its place in the architecture. Its function does. This is why two FocusZen® Code tracks do not need to belong to the same genre or even sound superficially similar. Jazz, electronic, ambient, techno, gothic, industrial, tribal and experimental structures can use completely different materials. The material changes. The surface changes. The underlying Code can remain.

SOUND ARCHITECTURE IS NOT A GENRE

Sound Architecture in FocusZen® Code is not intended as another name for a musical genre. It describes a method of construction. Each architecture can have its own specification defining, among other things, frequency structure, low-frequency behaviour, rhythm, tempo, predictability and interruption, space, depth, dynamics, density, silence, texture, vocals and the role of language. There is also a physical layer. Low frequencies reproduced through an appropriate sound system can be not only heard but physically perceived. Part of the architecture can therefore take into account how sonic energy behaves not only in headphones but within physical space. This does not mean claiming that a particular frequency automatically produces a specific neurological effect. We make no such promise. We design a sonic experience. Its effects can then be observed, tested and researched.

SOUND. WORDS. PURPOSE.

In FocusZen® Code, text is not treated as a random addition to a generated backing track. Language is one of the materials within the architecture. A word has a position. A moment of entry. Rhythm. Repetition. Context. And function. The objective is not to produce endless simplistic affirmations or tell listeners that positive thinking alone will somehow change reality. FocusZen® is built around a different sequence: Calm → Focus → Action. This is why we repeat: Sound. Words. Purpose. Sound creates the environment. Words can give it direction. Purpose defines why the architecture exists in the first place.

ATTENTION ENGINEERING

This brings us to the second part of the FocusZen® Code definition. Sound Architecture describes the process. Attention Engineering describes the purpose. FocusZen® Code exists within the wider FocusZen® system, whose central concern is attention and the transition from chaos towards deliberate action. Before a particular construction is developed, we can therefore ask: What is this architecture intended to do? This does not mean claiming to treat ADHD, autism, depression or any other condition. Nor does it mean guaranteeing concentration or a state of flow. Claims of that kind would require appropriate research. Attention Engineering is a design premise. We build sonic environments around an intended function related to attention, and we can then observe, test and continue developing how those environments behave. We do not turn a hypothesis into a fact simply because it would sound good in marketing.

THE TRAINING DATA PROBLEM STILL EXISTS

A clear boundary needs to be drawn here. Describing generative sound as construction material does not resolve the dispute over what generative models were trained on. Artists have every right to ask whether their work was used, on what basis, whether they had an opportunity to consent, and how rights and compensation should operate in this new technological environment. That is a real debate. But three different questions are currently very often collapsed into one: TRAINING — What was the model trained on? CREATION — How does a human use the model within their own process? OUTPUT — What is actually present in the result, and does it copy, reproduce or imitate protected material? These three areas are connected. They are not the same question. FocusZen® Code cannot answer on behalf of a technology company when asked how that company constructed its training dataset. We can, however, answer for our own process. This is why we deliberately do not build prompts around artist names or instructions to reproduce a particular existing track. Our direction is the opposite: away from imitation. If the research available to us indicates that a designed combination is moving too close to something that already exists, we do not treat that discovery as an instruction for what to copy. We treat it as information: the design needs to change.

WE ARE NOT FIGHTING FOR THE RIGHT TO BE CALLED “TRADITIONAL ARTISTS”

One of the most limiting assumptions in the current debate is that anyone using generative audio must be trying to become a musician without learning music. We are not. FocusZen® Code was not created to prove that AI is an artist. Nor was it created to diminish the work of people who have spent years learning an instrument, singing, composition or production. We are asking a different question: Can generative sound become material for a different kind of design practice? Not a replacement musician. Not a magic button. Material. Sound technologies have repeatedly changed the ways in which people create and define music. Synthesisers enabled sounds that acoustic instruments could not produce. Sampling changed the relationship between an existing recording and a new composition. Digital production made it possible to construct sonic environments beyond the physical limitations of a traditional studio. Generative audio may open further practices that we have not yet developed adequate language to describe. That does not mean every one of those practices will be valuable. It does not mean everything generated by AI automatically becomes art. And it certainly does not mean legal and ethical questions disappear. It simply means that one technology can be used in many fundamentally different ways.

IS IT STILL “AI MUSIC”?

Someone can look at FocusZen® Code, see that generative audio is involved and say: “That is AI music.” Technically, generative AI participates in the process. We have no intention of hiding that. But the statement “AI was used” still does not tell you who designed the concept, who created the specification, who conducted the research, who decided what not to copy, who changed direction after identifying a similarity, who constructed the prompt, who evaluated the results, who rejected the majority of the material, who determined what belonged to the project, or why a particular construction existed in the first place. The problem, therefore, is not the word “AI” itself. The problem begins when a label describing the technology replaces a description of the process.

BETWEEN TWO EXTREMES

The current debate can very easily collapse into two opposing camps. The first says: AI was used, therefore the human created nothing of value. The second behaves as though entering a prompt and generating hundreds of tracks is, by itself, sufficient evidence of creative authorship. FocusZen® Code does not need to belong to either camp. To the first, we can say: the use of a generative system does not tell you how much human intention, research, design, selection and decision-making surrounds the generation. To the second, we can say: generating material does not, by itself, create Sound Architecture. And perhaps this is where the most interesting question in the entire debate begins. Not: “Is AI good or bad?” Not: “Can AI be an artist?” But: “Where, within a specific process, does human decision-making actually take place?”

WE ARE NOT ASKING FOR A DIFFERENT LABEL. WE ARE SHOWING THE PROCESS.

FocusZen® Code does not expect everyone to accept the terms Sound Architecture or Attention Engineering. Nor are we attempting to remove the information that generative AI is involved. We are doing something much simpler: documenting the method. We show where the project begins. We show where research happens. We show where the decision to move away from similarity happens. We show what the specification is. We show the role of the generator. We show selection. We show rejection. We show reconstruction. And we show why generated material does not automatically become FocusZen® Code. Only then can each person decide whether a single label — “AI-generated music” — is sufficient to describe the entire process. Our answer can be expressed in three sentences: Generated sound is the material. Sound Architecture is the process. Attention Engineering is the purpose. And before any of those stages, there is a human being deciding what should be built in the first place. Sound. Words. Purpose.