Concept, direction, visual design, performance direction, selection, editing, and final authorship: Mark Chavez.
Cloud Pharaohs began as an experiment in using generative AI to make something deliberately absurd, unstable, and comic. Through animated rap, satire, and speculative design, the film turns toward the growing concentration of wealth and technological power and the desire of an elite class to extend, optimize, and control human life. Beneath the humor, it considers what happens to culture, memory, and ordinary human experience when increasingly powerful technologies are shaped by the ambitions of a privileged few.
I was not interested in producing another demonstration of gleaming seamless artificial imagery. I wanted to discover whether the defects, exaggerations, and unpredictable transformations of these systems could be directed into character, and satire.
The film developed around an older Chicano rapper-philosopher: gaunt, theatrical, overdressed, and physically improbable. Partially inspired by aging chicano actors I’ve recently had the honor of meeting. My rapper, Cipriano, named after a colorful paternal uncle, appears with tinted glasses, elongated limbs, pointed shoes, leather pants, and a beaded turquoise necklace. His body does not behave like a fixed anatomical structure. It stretches, stiffens, mutates, and occasionally seems to be reconsidering its own design. Rather than hiding those inconsistencies, I treated them as part of his performance.
The resulting film uses animated rap and speculative comedy to address the concentration of wealth and technological power, along with the fantasies of optimization, control, and extended life that accompany it. The tone is comic, but the underlying subject is not. Cloud Pharaohs asks what happens to memory, culture, and ordinary human experience when technological futures are increasingly imagined and financed by a privileged few.
Beginning with the absurd
The initial premise was simple: use generative tools against their most familiar promotional logic. Much generative media is presented as proof of technical competence. The image is expected to appear polished, instantaneous, and obedient. Errors are treated as failures to be eliminated, while coherence is offered as evidence that the machine has mastered another area of visual production.
I wanted to begin elsewhere. Absurdity gave me a way to interrupt the standard approach and expectation that generative animation must imitate conventional cinematic realism. It also allowed the system’s instability to become visible without turning the work into a compilation of accidental glitches. The images could be ridiculous, but they still needed intention, structure, and timing.
This distinction became important. Randomness is not automatically absurd, and visual failure is not automatically interesting. An image can be strange and still be inert. The work began to take shape only when I could identify a consistent attitude within the variations: vanity, aggression, philosophical certainty, physical discomfort, and theatrical self-possession. These qualities gradually formed the character at the center of the film.
The absurdity, therefore, did not replace direction. It increased the need for it. Each generated possibility had to be evaluated according to what the character appeared to know, how he occupied the frame, and whether the scene advanced the film’s comic and political tension.
Directing a synthetic performer
I approached the central figure as a performer. His identity depended on more than facial continuity. It emerged through posture, costume, proportion, vocal texture, restraint, and gestural rhythm.
This meant repeatedly specifying how he should sit, stand, look, and move. A raised knee, a hand held near the chin, a rigid shoulder, or an excessively long leg could carry more character than a conventionally expressive face. In several sequences, I asked for minimal movement because generative animation often mistakes performance for constant activity. The character became more convincing when his movement was controlled: a delayed glance, a small shift of the wrist, an awkward pause, or a severe posture held slightly too long.
At other moments, I pushed his performance toward choreography. Bob Fosse-inspired angularity became one point of reference: isolated shoulders, articulated wrists, bent elbows, restrained pelvic movement, and gestures that remain precise even when the body itself is unstable. The aim was not to reproduce a particular dance vocabulary literally. It was to give the generated figure a physical intelligence and a recognizable attitude.
Voice was equally important. The performance required weight and abrasion rather than the frictionless delivery typical of many synthetic voices. Vocal direction involved controlling pace, emphasis, gravel, pauses, and the relationship between spoken language and rap. A technically clean reading could weaken the character if it sounded merely competent. The voice needed to suggest that the figure had lived inside his argument for a long time.
The synthetic performer was therefore assembled across multiple systems, but the performance was not located in any one of them. It existed in the relationship between visual design, voice, music, motion, and editorial timing.
Generative production creates an enormous amount of material that is almost usable. This is one of its defining conditions. A face may be right while the body is wrong. A gesture may begin effectively and then collapse. A costume may hold for several seconds before transforming into another object. A camera movement may introduce energy while destroying the composition.
Working with instability
The central directing question was not how to eliminate every inconsistency. It was about distinguishing expressive instability from visual noise.
Some mutations strengthened the character. His changing proportions made him seem vain, artificial, and slightly ungovernable. Shifts in costume or environment could intensify the film’s speculative atmosphere. An unstable gesture sometimes appeared more psychologically charged than a smooth one. These moments were retained when they contributed to the character’s authority, comedy, or unease.
Other errors had no expressive function. Unmotivated camera drift, loose lip synchronization, excessive facial movement, generic spectacle, or a sudden loss of graphic clarity usually weakened the scene. These outputs were discarded, regenerated, shortened, reframed, or concealed through editing.
This process involved a continuous negotiation with the material. The systems proposed images and movements that I could not have fully predicted, but surprise alone did not determine the result. I selected which accidents belonged to the film and which remained merely accidental.
Constructing the performance across systems
The production moved between language, still-image generation, image processing, video synthesis, voice, music, and editing. My working pipeline included OpenAI ChatGPT, Anthropic’s Claude, ComfyUI, ElevenLabs, and other generative tools. These technologies did not function as a single automated process. They produced fragments with different visual histories, temporal behaviors, and degrees of control.
Language tools were useful for testing descriptions, restructuring prompts, developing lyrics, and clarifying the conceptual direction of individual sequences. Still-image generation established the character, costume, pose, and environment. ComfyUI allowed me to work more deliberately with model combinations and image-to-video processes. LTX 2.3 and Wan models were used to explore movement, speech, performance, and transitions. ElevenLabs provided a space for constructing voice and music, including the tonal and rhythmic qualities that helped define the character.
Moving between these systems was less like operating a single production platform than directing several unreliable departments. Each had its own interpretation of the scene. A successful still did not guarantee successful motion. A convincing voice could change how a gesture needed to be timed. Music could reveal that a shot was too long, too literal, or insufficiently aggressive. Every stage altered the meaning of the material produced in the stage before it.
Prompting was part of this process, but prompting alone does not adequately describe the work. A prompt establishes conditions. It does not make the final decisions about duration, juxtaposition, escalation, repetition, performance, or narrative emphasis. Those decisions accumulated through many cycles of generation, evaluation, revision, and assembly.
Editing as authorship
The film was ultimately constructed in the edit. Generative systems tend to produce self-contained moments: a striking pose, a short transformation, an isolated gesture, or an image that appears complete because it is visually dense. A film requires relationships between those moments.
Editing determined when the character appeared authoritative and when he appeared ridiculous. It established where repetition became musical, where a pause allowed a line to register, and where a visual transformation needed to be interrupted before it exhausted itself. It also allowed discontinuity to become structural. Changes in scale, costume, anatomy, and setting could operate as rhythmic cuts rather than continuity errors.
The rap form was particularly useful because it supported accumulation, return, and variation. Repeated phrases could serve as anchors as the imagery grew increasingly unstable. The music created a temporal discipline that the generated images did not possess on their own. Shots were selected and reshaped according to cadence, breath, accent, and interruption.
This is where the film moved furthest from the AI demo. A demonstration is usually organized around showcasing a tool’s capabilities. It asks the viewer to notice what the system can produce. Cloud Pharaohs uses those capabilities, but it does not treat them as the subject of admiration. The edit redirects attention toward a character, an argument, and a comic world.
Authorship in this context does not require pretending that the systems are neutral instruments. They introduce visual conventions, biases, limitations, and unexpected propositions. My authorship lies in how I engaged those conditions: what I requested, rejected, repeated, altered, performed, sequenced, and finally allowed to remain visible.
Beyond the AI demo
The film’s exaggerated figure occupies a world shaped by fantasies of technological sovereignty. He performs inside a spectacle of wealth and futurity while also mocking it. The title Cloud Pharaohs connects contemporary computational power with an older image of monumental authority: rulers who imagine their systems, monuments, and bodies extending beyond ordinary time.
The film does not present a technological elite through documentary description. It approaches the subject through caricature, rhythm, and inflated performance. Absurdity makes the ideology physical. Bodies stretch. Surfaces become metallic, language becomes grandiose, and environments behave like unstable declarations of power. The central figure appears both seduced by this world and capable of exposing its ridiculousness.
Comedy is essential to this approach. A solemn critique of technology can reproduce the authority it intends to question. Cloud Pharaohs instead allows power to become awkward, vain, repetitive, and bodily. The film’s instability disrupts the fantasy that technological systems are inevitable, or entirely under the control of their owners.
At the same time, the work is concerned with what these futures leave behind. If technological development is organized around private accumulation, life extension, data extraction, and the desires of a small class, then culture and memory risk becoming raw material for systems that do not share the histories they process. The older Chicano figure enters this space carrying another visual and cultural history. He is not simply a spokesman for a position. He is an unstable participant whose presence complicates the universal future promised by technological spectacle.
What comes next
Cloud Pharaohs has opened a direction I want to continue through a small group of character-based films. I’m interested in figures that carry complex arguments without becoming conventional narrators: performers whose bodies, voices, and contradictions allow ideas to emerge through behavior.
One possible continuation centers on Oswaldo, a gruff decolonial scholar with a gravelly voice and a weathered face. A physically imposing character, less theatrically polished than characters in Cloud Pharaohs. An unabashed believer in far-fetched and absurdist ideas. Whose severity shifts the balance between comedy, confrontation, and the reality of his fantastic fictional world, while extending the critique of technological monarchy, concentrated wealth, and controlled futures.
I am also interested in returning to earlier character-based works, including The Adventures of Barty and the Pirate, and reconsidering what their comic structures might become through current production methods. This would not mean replacing the earlier animation with generative imagery. It would mean examining how characters with established histories can be extended without losing the particular timing and design logic that made them legible in the first place. that being a pirate in a 18th & 19th century Southeast Asia.
The broader project is not to produce a continuous stream of AI content. It is to develop a directing practice appropriate to unstable computational media. That practice depends on sustained character development, performance direction, critical selection, and editing. The systems can generate an abundance of images, but abundance is not a point of view.
Cloud Pharaohs became a film when the generated material stopped functioning as evidence of the technology and began functioning as performance. Its absurdity is not an escape from its political concerns. It is the form through which those concerns become visible, rhythmic, and difficult to contain.
Production note: Generative image, video, voice, language, and audio tools used in the production included OpenAI ChatGPT (image generation), Anthropic Claude, ComfyUI, Comfy MCP, LTX 2.3, Wan video models, and ElevenLabs generative audio tools. Concept, direction, visual design, performance direction, selection, editing, musical score, and final authorship by Mark Chavez, produced by Ina Conradi.
Music video
LYRICS
They said: “Relax. The chamber is optional. The contract is eternal. The upgrade is humane.” Then the floor opened up, and the poor carried the cable.
Cloud pharaohs, chrome intestines, buying death with tax exemptions. Cryo sleep, designer veins, children weep while the market gains. Body lease, soul subscription, AI preacher with a soft eviction. They want forever, we want breath — they call it progress when the people starve to death.
Here come the oligarchs, stitched in gold, skin like plastic, blood on hold. Eyes upgraded, hearts outsourced, love replaced by a legal force.
One got a liver from a lottery town; one got a face printed upside down. One bought a monk, one bought a moon, one bought a womb in a boardroom tune.
They put their mothers in the archive bin, then sold “family” as a wellness skin. They got carbon bones and vampire teeth, crying, “We suffer too,” from a bunker underneath.
The masses on shift with their backs bent low, feeding the servers that tell them, “Grow.” AI says: “Be grateful, citizen friend, your poverty is a feature in the transcendence blend.”
Hegel built a room for one: mirror, master, ghost, and gun. But the barrio broke the dialectic shell — no isolated soul survives this hell. Tonalli in the ribcage, heat in the hand, maize in the ashes, blood in the land. The cloud said, “Submit.” The serpent said, “NO!” The people started singing from below.
Now the billionaires glitch when the drums come near, their cheekbones spark and their eyelids smear. One tries to reboot his moral core, but the update fails at the servant door. One cries diamonds into a spoon, one gets lonely on a private moon. One says, “Humanity must evolve,” while twelve unpaid nurses dissolve. They got backup brains in a freezer vault, but still can’t name their original fault. They got clone insurance, Mars debris, and a priest made out of liability. But the masses remember the power line, who held the cable, who mined the spine. Who cleaned the blood from the chrome machine, who fed the god that ate the screen.
Cloud pharaohs, chrome intestines, buying death with tax exemptions. Cryo sleep, designer veins, children weep while the market gains. Body lease, soul subscription, AI preacher with a soft eviction. They want forever, we want breath — they call it progress when the People Starve to Death.
Malinalli laughs in the firewall smoke, turns their empire into a joke. Teotl hums through the broken glass, and the dead refuse to become their past. The rich want heaven with a retinal scan, the poor want water and a place to stand. The rich want time without decay, the poor want dawn without being sold away.
So dance, little bodies, in the server rain, with your cracked-phone prayers and your union pain. The cloud can’t own what the earth still knows: corn breaks concrete when the old root grows. They bought the body. They bought the sky. They bought the coffin and forgot how to die. But Malinalli kept speaking, teotl kept moving, the people kept burning, and the machine kept losing.
