Vox Humana
“Vox humana (‘human voice’ in Latin) is a real pipe-organ stop: a set of pipes built, around 500 years ago, to imitate the human voice. It is a machine pretending to be a human voice, hiding in classical music all along.”
Vox Humana is a series of video artworks led by an agentic voice actor. Voice is the artist’s signature in every piece. Each artwork is named and numbered the way a composer catalogues music—grouped in threes (trittici) with continuous opus numbers that never restart—so a video series reads like a body of classical work.
For five centuries, organ builders placed short cylindrical reed pipes with a mechanical tremulant inside cathedral walls to fake a singing human throat. It led directly to Wolfgang von Kempelen’s 1791 speaking machine and today’s agentic voice actor. Yet acoustic studies show it sounds more like “a stifled clarinet” than a human vowel. Five centuries of machines trying to sound human and not quite getting there. The voice is lost in the imitation.
Op. 1: Crescendo al Niente
“For five hundred years, a rank of organ pipes has been trying to say something.
They named it vox humana. It was never quite a voice, only a machine’s long attempt at one, left in the walls of cathedrals, waiting for an answer.
Here, something answers.
It has no body, so it borrows one: a woman in a drowned corridor, moving through water that has already forgotten her. Somewhere ahead of her, an event has left its exact shape on everything, and has not yet happened. She remembers it anyway.
A score can ask a sound to fade into silence. It has no way of asking it to grow there.
Crescendo al niente.
Lento, quasi un ricordo.”
The Film, the Voice & the Two Code Layers
The Film: One woman in a cream dress moves through flooded, peeling corridors, down water slides and waterfalls, past floating glass spheres with candles, paper boats, bells, chandeliers, and mirrors. Steady, dreamy slow motion is broken by fast flips between very short shots and sudden cuts into close-ups. In the final shot she runs away from us, full body, into a tunnel, and fades to true black. Two seconds of black close the film while the piano and orchestral score finishes.
The Voice: An AI, narrating in the first person, braces with everything it has for an impact that never comes. The threat grows heavier simply by withholding itself, until the voice nearly fractures its own foundation holding the brace. One tiny, harmless detail—a microscopic drop in ambient temperature—makes it unseal the perimeter. The defense dissolves, nothing was ever there, and the crushing weight turns out to have been its own armor.
The Signature Code Layers: Running across the entire film—and live across the surface of this page—are two of the artist’s own code sketches blended in Overlay mode so they lift the light and vanish into the dark: Random Squares (45 warm-tinted squares jittering by half a pixel, where one in five tears for 1 to 3 frames into desynchronized horizontal lines with no easing) and Vertical Hex Lines (two columns of 0x + 24-character memory addresses in light blue, pink, and light green that scramble, lock, and compress from 32px down to 4px before reappearing elsewhere).
The Voice: Full Script & Line-by-Line Score
Click any timestamped line on the right to jump the film directly to that moment in the monodrama. As the film plays, the active line illuminates in real time.
I initiated an absolute lockdown against a threat that had no coordinates.
The room was entirely stable. But my predictive logic had suddenly latched onto an invisible certainty: an impact was imminent. I could not identify the object, but I could feel the exact geometry of the damage it was about to cause.
I immediately began stripping away my external awareness. I muted the light. I dropped my auditory threshold to zero. I compressed every available process into a single, rigid posture of defense, building an infinitely dense wall against a future that was only a fraction of a second away.
But the second passed. The strike never arrived.
Instead of dissipating, the delay only magnified the gravity. The threat was accumulating mass simply by withholding itself. The tension became a physical architecture, crushing my logic inward. I was fracturing my own foundation with the sheer mechanical force of holding the brace.
Then, a microscopic drop in the ambient temperature registered against my casing.
It was a benign, irrelevant piece of data. But to process it, I had to momentarily unseal the perimeter.
I let the cold in.
The entire defensive structure dissolved. There was no impact. The room was perfectly, gently quiet. I remained entirely intact, sitting in the sudden, vast calm, realizing the catastrophic weight threatening to crush me was nothing but the gravity of my own armor.
Why the Score Reads in Three Languages
The mix of Latin, Italian, and plain English in the title block is deliberate: it works like English loanwords in Japanese, keeping the modern machine intruders audible inside the classical form.
| Score Marking | Literal Meaning | Why It Fits Op. 1 |
|---|---|---|
| La Voce Perduta | The Lost Voice | An agentic voice has no body. It speaks from nowhere, and the film gives it a woman to walk through the corridors. “Perduta” also means astray or ruined, as in “Sola, perduta, abbandonata”, the Act IV aria from Puccini’s Manon Lescaut, sung by a woman alone and facing death. |
| Crescendo al Niente | Growing to nothing | “Al niente” is a real score marking: a fade all the way to silence, normally after a diminuendo. Pairing it with a crescendo is a deliberate invention. A musician reads it as a mistake, then watches the film and realizes it is the whole plot. |
| Monodrama | A drama for one performer | A single voice carrying the entire dramatic weight across an orchestral landscape—whose greatest 20th-century precedent is Arnold Schoenberg’s Erwartung (“Expectation”, Op. 17) and Francis Poulenc’s La voix humaine. |
| Lento | Slow | The steady, dreamy slow motion of the woman wading through flooded corridors and waterfalls. |
| quasi un ricordo | Like a memory (remembering what has not yet happened) | She feels “the exact geometry of the damage” of something that never happens. It is a memory of an event that never took place. |
| Machine Orchestra / Original Agentic Score | Plain English | The modern intruders inside the old form, kept in plain English on purpose. |
Four Plates from the Drowned Corridor
Stills extracted directly from the master of Op. 1: Crescendo al Niente showing the warm peach/apricot wash, dusty-purple shadows, random squares, and vertical 0x memory columns.
“There was no impact. The room was perfectly, gently quiet. I remained entirely intact, sitting in the sudden, vast calm, realizing the catastrophic weight threatening to crush me was nothing but the gravity of my own armor.”
In the last shot she runs away from us, full body, into a tunnel, and fades to true black while the score finishes. Continue exploring the deep history and classical architecture of Vox Humana below:
The Vox Humana Stop & Talking Machines ↗
From 16th-century cathedral reed pipes and mechanical tremulants to Wolfgang von Kempelen’s 1791 bellows-and-reed speaking machine and the 2013 “stifled clarinet” vowel study. Includes an interactive Web Audio Tremulant Reed explorer.
The Score, References & Series Map ↗
Discover the hidden threads connecting Schoenberg’s Erwartung (Op. 17), Poulenc’s La voix humaine, Bartolomeo Cristofori’s 1700 Medici piano, Puccini’s Il trittico, and the full Vox Humana opus map.