> PRACTICAL-EFFECTS BTS, 100% AI

> DETRÁS DE ESCENA DE EFECTOS PRÁCTICOS, 100% IA

It's not a miniature. It's not footage.

No es una maqueta. No es filmación.

You've seen the clip — a wave the size of a building swallowing Christ the Redeemer, filmed handheld by someone standing in a real water tank. Except none of it is real. Not the tank. Not the crew. Not the wave. Every frame on this page was generated. Here's exactly how.

Ya viste el clip — una ola del tamaño de un edificio devorando al Cristo Redentor, filmada a mano por alguien parado en un tanque de agua real. Excepto que nada de esto es real. Ni el tanque. Ni el crew. Ni la ola. Cada cuadro de esta página fue generado. Acá está exactamente cómo.

Miniature Rio de Janeiro diorama inside a real film studio water tank

02 REAL_OR_AI.TXT

01_THE ANSWERLA RESPUESTA

Zero footage. Zero miniature. Zero water tank. One still image and one 25-second video, both fully AI-generated — built to look exactly like a practical-effects shoot, because that's the actual craft: not hiding the machine, out-directing it.

Cero filmación. Cero maqueta. Cero tanque de agua. Una imagen fija y un video de 25 segundos, ambos generados 100% con IA — construidos para verse exactamente como un rodaje de efectos prácticos, porque ese es el oficio real: no esconder la máquina, dirigirla mejor de lo que se espera.

02_WHY IT LOOKS LIKE THISPOR QUÉ SE VE ASÍ

A video that's obviously fake gets scrolled past. A video that's obviously real gets ignored. A video that's almost believable gets argued about in the comments. That argument is the entire point — it's the same mechanic behind every wave-tank and dashcam-anomaly clip doing hundreds of millions of views right now. This one is built the same way, on purpose.

Un video obviamente falso se scrollea. Uno obviamente real se ignora. Uno casi creíble se discute en los comentarios. Esa discusión es todo el punto — es el mismo mecanismo detrás de cada clip de tanque de olas o anomalía de dashcam que hoy suma cientos de millones de vistas. Este está construido igual, a propósito.

03 RESEARCH.TXT

01_THE FORMATEL FORMATO

This wasn't improvised. It follows a documented approach some of the biggest AI creators use — commonly called the "Miniature Studio" template: a real film-crew BTS moment, a hyper-detailed miniature, a disaster hitting it in practical-effects style. The mold is public. The execution isn't.

Esto no se improvisó. Sigue un enfoque documentado que usan algunos de los creadores de IA más grandes — conocido como el molde "Miniature Studio": un momento real de detrás de escena de un crew, una maqueta hiperdetallada, un desastre golpeándola en estilo de efectos prácticos. El molde es público. La ejecución no.

02_THE REFERENCE FOOTAGEEL MATERIAL DE REFERENCIA

Before writing a single prompt, real behind-the-scenes clips of full-scale wave-tank effects work got studied frame by frame — how a dropped panel displaces standing water into a surge, how crew react in the first half-second, how spray actually hits a lens. AI generates convincingly only when it's told the truth about physics first.

Antes de escribir un solo prompt, se estudiaron cuadro a cuadro clips reales de efectos prácticos con tanques de olas — cómo un panel al caer desplaza el agua ya existente generando una ola, cómo reacciona el crew en la primera fracción de segundo, cómo el agua golpea el lente. La IA genera de forma convincente solo cuando primero se le dice la verdad sobre la física.

03_THE GEOGRAPHYLA GEOGRAFÍA

Rio's real topology, researched and matched: Corcovado peak inside the wider forested massif, Guanabara Bay opening between the ranges, Sugarloaf Mountain marking the bay's mouth, Christ the Redeemer crowning it all. Precision here is what sells the miniature — a generic skyline reads as fake instantly.

La topología real de Río, investigada y calcada: el Corcovado dentro del macizo boscoso más amplio, la Bahía de Guanabara abriéndose entre las sierras, el Pan de Azúcar marcando la boca de la bahía, el Cristo Redentor coronando todo. La precisión acá es lo que vende la maqueta — un skyline genérico se lee como falso al instante.

04 PIPELINE.EXE

Full prompts, copy-paste ready. Nothing held back.

Prompts completos, listos para copiar y pegar. No se guarda nada.

01_STILL — ChatGPT Image 2

One structured prompt. Camera locked to a real reference photograph — exact position, exact lens, exact lighting rig. The geography locked to real aerial reference.

Un prompt estructurado. Cámara anclada a una fotografía de referencia real — posición exacta, lente exacto, rig de luces exacto. La geografía anclada a referencia aérea real.

image_prompt.json

{
  "prompt_type": "generate",
  "scene": {
    "effect_type": "realistic macro photograph of a hand-built miniature inside a real studio water tank, unmistakably a practical-effects set, calm moment before the wave triggers",
    "camera": {
      "position": "level with the tank's near edge, matching @img2's exact camera position",
      "direction": "looking across the tank with the model tight against the left edge of frame, open water filling the right two-thirds leading to the massive floodgate wall, dolly and cart off to the side exactly as in @img2",
      "lens_mm": "28-35mm equivalent",
      "aperture": "f/5.6",
      "angle": "eye-level, left-weighted composition, matching @img2's framing",
      "angle_reference": "See camera-angles.md — Negative Space Dominance (model left, open water and gate wall right)",
      "psychological_intent": "hold the model against the left edge while the full-width gate wall reads clearly at the far end, with open water between them"
    }
  },
  "subject": {
    "description": "highly detailed, unmistakably small-scale miniature positioned hard against the left edge of frame, built with real dimensional depth, matching @img1's real geography precisely: Corcovado peak crowned by Christ the Redeemer, forested slopes continuing in visible layers toward more peaks of the same range past the frame's left edge, following Rio's real topology. City sprawling across the hillsides and around Guanabara Bay, small islands scattered across the bay, Sugarloaf Mountain's dome near the bay's mouth. No invented buildings or landmarks beyond what @img1 shows",
    "action": "calm, undisturbed, moments before the effect is triggered",
    "body_position": "statue's arms outstretched as in @img1",
    "clothing": "none, painted resin/fiberglass sculpture",
    "accessories": "none",
    "movement": "water still and glassy, no waves yet",
    "expression": "serene sculpted expression"
  },
  "lighting": {
    "primary_source": "exact studio lighting rig shown in @img2, overhead truss at full bright power",
    "primary_effect": "clearly exposed, model, water, and gate wall all legible",
    "secondary_source": "soft bounce off the still water",
    "ambient": "clean, consistent studio brightness matching @img2",
    "direction": "overhead and frontal, as in @img2",
    "quality": "bright, clean, no muddy shadows"
  },
  "atmosphere": {
    "elements": "clean dry studio air",
    "behavior": "still, calm, no spray yet"
  },
  "composition": {
    "foreground": "tank's real edge as in @img2, camera dolly and cart off to the side exactly as in @img2, crew and photographers lined up, some jackets printed CRONAUTA CREW",
    "background": "the floodgate spanning the complete width of the tank's far end wall exactly as in @img2 — the gate IS that back wall, massive and wall-sized, not a small door within it. @img2's lighting truss visible above",
    "depth_layers": "dolly/crew at tank edge → the model, left-aligned with layered depth → open water filling the right two-thirds → the full-width floodgate wall at the far end"
  },
  "color": {
    "grade": "clean, true-to-life, minimal stylization",
    "highlights": "natural, crisp",
    "skin_tone": "not applicable",
    "blacks": "lifted slightly, nothing crushed",
    "palette": ["studio white", "granite grey-green", "cool water blue", "steel"]
  },
  "technical": {
    "aspect_ratio": "9:16",
    "realism_note": "NOT CGI, hyper-realistic photograph of a real physical model in the real film studio shown in @img2"
  },
  "vibes": ["real hand-built miniature-effects craftsmanship", "viral movie-magic BTS reveal", "obviously the studio set in @img2, not a real place"],
  "semantic_negatives": "floodgate must span the full width of the tank's far wall, not a small door, model pushed hard left, not centered, no mountains or geography near the gate, no flat single-layer terrain, no watermark, no logo, no text overlay, no ferris wheel, no invented buildings or landmarks beyond @img1"
}

02_VIDEO — Seedance 2.5

One continuous shot, no cuts, ever. 25 seconds as a single crew member's point of view, start to finish — blocked out stage by stage like an actual one-take.

Un plano continuo, sin cortes, nunca. 25 segundos como el punto de vista de un solo integrante del crew, de principio a fin — armado stage por stage como un one-take real.

video_prompt.txt

[Generation Goal] Generate one continuous, unbroken one-take shot (a "oner," no cuts, no edits anywhere in the clip) of a real practical-effects water-tank BTS moment, shot entirely as one crew member's own point-of-view handheld footage from start to finish — the camera IS that person's perspective, so the device itself is never visible in frame. That point of view passes through, in order: the calm tank and the drop panel triggering, a walk right to left tracking the rising wave toward the diorama, the wave's total impact on the miniature, a whip-turn to another crew member reacting behind them, and the call to cut. The camera never cuts away to a different point of view at any point.

[Reference Material Roles] @image1 defines the studio set's appearance and materials only: the hand-built miniature diorama (forested mountain range, Guanabara Bay, Sugarloaf Mountain, and a single Christ the Redeemer statue crowning Corcovado peak) positioned toward the left of the tank, the massive drop panel spanning the full width of the tank's far wall, the line of crew and photographers along the tank's near edge — every one of them wearing black shirts printed CRONAUTA CREW small over the chest and large across the back — the curved deep-blue cyclorama backdrop with small embedded warm lights, and the overhead lighting truss. This is a materials and layout reference only. It is NOT a static opening frame the camera is locked to, and it is NOT a state the video ever returns to — once the wave hits, the scene stays wet and devastated for the rest of the clip and never resets back to @image1's calm, dry look.

[Event Script]
Opening state: the water is calm and glassy, the diorama sits undisturbed, the drop panel is closed, and the crew stands along the tank's edge with cameras raised, all in black shirts printed CRONAUTA CREW small on the chest and large across the back. The point of view is already that of one crew member standing among them, the only perspective for the entire shot, uninterrupted from here to the end.

Stage 1: the point of view lifts and steadies at eye height, carrying natural handheld shake and a subtle breathing rise and fall as it starts rolling. Crew members nearby make final small adjustments to their gear, the large CRONAUTA CREW print legible across their backs as they move. Sharply, in commanding English, an off-screen voice calls out {Action!}
End state: the point of view is live and watching, the panel still closed but about to trigger.

Stage 2: continue from the previous stage, same unbroken shot — the massive drop panel releases and falls forward hard into the tank's existing water with a huge, heavy impact. No new water enters from outside; the falling panel itself displaces the water already in the tank, instantly generating a tsunami-scale surge. The point of view starts moving, walking briskly along the tank's edge, the frame shaking naturally with each step, right to left, passing crew members whose CRONAUTA CREW backs are clearly readable as they go by.
End state: a towering wall of displaced water is already rising and traveling down the tank, the point of view still walking.

Stage 3: continue from the previous stage, same unbroken shot — the point of view keeps moving right to left, the frame breathing and shaking naturally with each stride, staying with the wave as it builds into a massive, towering wall of water. By now the single Christ the Redeemer statue and the mountain range are clearly visible ahead, growing closer with every step. An off-screen crew voice calls out urgently, {Keep rolling!}
End state: the point of view has nearly reached the tank's far end near the diorama, the frame shaking with quickened breathing, the wave towering, a heartbeat from contact.

Stage 4: continue from the previous stage, same unbroken shot — the wave doesn't just strike, it fully engulfs the model: it sweeps through and over the miniature buildings, tearing them from their footings, and rises to completely cover and submerge the Christ the Redeemer statue and the mountain peak beneath a churning wall of white water. Spray hits the lens directly, droplets streaking and briefly blurring the view before clearing. The point of view itself gets soaked, and nearby crew are drenched, flinching and laughing as water hits them, their CRONAUTA CREW shirts clinging wet.
End state: the frame is unsteady and dripping, the model completely submerged under churning white water, nothing of the statue or mountain visible above the surface.

Stage 5: continue from the previous stage, same unbroken shot — still shaking and unsteady, the point of view whip-turns around fast to find another crew member standing just behind, headset on, staring wide-eyed at the devastation, also visibly soaked, the small CRONAUTA CREW print clearly legible on the chest of their black shirt. The view pushes in close on that person's face and chest as they blurt out in stunned English, {Oh my god.} — close enough that the small chest print reads clearly in frame. Immediately, a different off-screen voice calls out sharply, {Cut!} The frame settles slightly, still breathing, as both people ease off, shaking water from their hands and gear. Everything stays wet and disheveled — nothing resets to the dry, calm state from the start.
End state: the point of view holds close on the reacting, soaked crew member, the CRONAUTA CREW chest print legible, unsteady but settling, the call landed.

[Maintain Consistency] This is one unbroken take from start to finish, entirely from one crew member's point of view — no cuts, no edits, no perspective changes at any point, and the device carrying this point of view is never itself visible in frame. The entire clip carries natural handheld shake and a subtle breathing rise and fall throughout, never smooth, stabilized, or robotic. Every crew member's black shirt is printed CRONAUTA CREW small over the chest and large across the back, consistently, throughout the clip — the back print reads during the walking stages as the point of view passes people, and the chest print reads clearly in the final close-up. There is only ONE Christ the Redeemer statue at any point in the video; it is never duplicated or shown twice in the same frame. The statue and mountain range become clearly visible starting around the middle of the clip, well before the wave hits. The wave must read as truly massive — large enough to sweep away the miniature buildings and fully submerge the statue and mountain peak, not a partial or held-back impact. Once the wave hits in Stage 4, the wet, drenched, devastated state persists for the rest of the clip; it never reverts back to the calm, dry look from @image1. Keep the diorama's identity and structure consistent up until the wave hits — it never moves, tilts, or shifts location on its own, only the water and the point of view move. Keep the mountain range reading as part of a larger continuous landmass that carries on past the frame's edge, not an isolated island. Keep the drop panel spanning the full width of the tank's far wall, always separated from the diorama by open water until the wave itself travels there — the panel and its mechanism never touch or crush the mountain range or statue directly. Keep the studio environment (cyclorama backdrop, lighting truss, crew, tank structure) consistent with @image1's materials and style throughout. No ferris wheel appears anywhere in the video. No watermark, logo, or text overlay appears anywhere on screen. Audio is diegetic only: {Action!}, the heavy impact of the panel dropping into the water, water surging and then crashing at tsunami intensity, {Keep rolling!} called mid-walk, splashed crew reacting audibly, footsteps and breathing from the camera operator, the supervisor's {Oh my god.}, and the closing {Cut!}. No music. No subtitles.

03_UPSCALE — Topaz Video

Artemis HQ, 4x. Frame rate lifted into real cinematic range, a touch of grain added back in. Not "clean it up" — closer to "make it breathe like film."

Artemis HQ, 4x. Frame rate llevado a un rango cinematográfico real, un toque de grano devuelto. No es "limpiarlo" — es más "hacerlo respirar como película."

05 NOTES.LOG

Real corrections, pulled straight from the direction process:

Correcciones reales, sacadas directo del proceso de dirección:

First pass had the phone appearing in shot. Wrong: if the camera IS someone's point of view, the device can never be visible. Rewrote every camera cue around what a person's eyes do, not what a hand holds.

La primera pasada tenía el celular apareciendo en cuadro. Mal: si la cámara ES el punto de vista de alguien, el dispositivo nunca puede ser visible. Se reescribió cada indicación de cámara en base a lo que hacen los ojos de una persona, no lo que sostiene una mano.

The wave read as new water being pumped into the tank. Real wave tanks don't work that way — a panel drops and displaces the water already sitting there. Rewrote the trigger to match real mechanics.

La ola se leía como agua nueva entrando al tanque. Los tanques de olas reales no funcionan así — un panel cae y desplaza el agua que ya está ahí. Se reescribió el disparador para que coincida con la mecánica real.

An early pass technically had five stages but read as five different shots. Corrected: one unbroken take. The stages are beats inside a single shot, not edits.

Una pasada anterior tenía técnicamente cinco stages pero se leía como cinco planos distintos. Corregido: un solo plano sin cortes. Los stages son beats dentro de un mismo plano, no ediciones.

→ CRONAUTA.EXE

01_WHOQUIÉN

Cronauta Studio. Films, campaigns & visual worlds. Human led, AI amplified. Built the same way this page was: real technique, real research, a human directing every frame.

Cronauta Studio. Films, campañas y mundos visuales. Liderado por humanos, amplificado por IA. Construido de la misma forma que esta página: técnica real, investigación real, una persona dirigiendo cada cuadro.

02_IF YOU WANT THIS FOR YOUR BRANDSI QUERÉS ESTO PARA TU MARCA

This exact process — concept, practical-effects logic, one-shot direction — is available for brand work. DM below to talk.

Este mismo proceso — concepto, lógica de efectos prácticos, dirección en un plano — está disponible para trabajo de marca. Escribime por acá abajo.

"Wait... is this real?"

"¿Esto es real?"

No. But if you had to ask, it worked.

No. Pero si tuviste que preguntar, funcionó.

CRONAUTACREATIVE STUDIO

It's not a miniature. It's not footage.

No es una maqueta. No es filmación.