AI Video Knockout · every video prompt used in the seven rounds
https://mantis.guru/benchmarks/ai-video-knockout/

===== Round 01 · Real vs AI dance · The choreography prompt (15 s) =====

One continuous 15-second photoreal live-action shot. Animate the exact man in the starting photograph dancing on the same green rooftop. Preserve his face, glasses, short black hair, black T-shirt, loose black trousers, black sneakers with white soles, body proportions and wristwatch. He smiles broadly toward the camera throughout. Keep his entire body including both shoes in frame and preserve the rooftop railing and building behind him. Camera remains in the same frontal position, no cuts or zooms.
0-7 seconds: playful running-man shuffle, knees softly bent, alternating short diagonal cross-steps and backward heel lifts, shifting weight from side to side. His torso twists gently left and right, elbows bent, hands swinging loosely around waist and chest level in counter-rhythm to his feet. He bounces lightly on every step, traveling only a little within the same spot.
7-11 seconds: face the camera squarely, feet spread; continue bouncing and stepping side to side while bringing both palms together repeatedly, sweeping the joined hands across his chest from one side to the other, briefly lifting them beside his shoulder before sweeping them down again. His knees flex with each weight shift.
11-15 seconds: release the hands and return to loose alternating shuffle steps with two small buoyant hops, relaxed swinging forearms, then a low forward heel kick near the end, grinning at the camera. Natural human joint motion and believable planted-foot contact with the turf. Maintain the same person, outfit, daylight and location throughout. No dialogue, no singing, no added text. Light shoe scuffs and outdoor ambience.

===== Round 02 · Spaghetti, now? · The prompt (10 s) =====

Create a photorealistic 10-second deadpan comedy video from the supplied starting frame. Preserve the two distinct men, their faces, clothing, seating, purple-blue studio, desk, laptop, plants and lighting. RIGHT MAN: the man in a dark short-sleeved T-shirt, the only person who retrieves and eats spaghetti. LEFT MAN (Leo): the younger man in a dark zip hoodie, the only person who says "Now?" Never swap their roles or voices. Both remain seated and visible in a continuous medium two-shot, with a very subtle slow push-in that keeps their hands, the bowl and both faces visible. No cuts.

TIMED ACTION AND EXACT ENGLISH DIALOGUE:
0.0-3.8 seconds: The RIGHT man looks casually toward the camera and says in a relaxed conversational male voice, "Hey folks, before we start, I'll eat some spaghetti." While finishing the sentence he glances down, reaches his left hand below the desk and lifts a white ceramic bowl of tomato-sauce spaghetti from below the desk to chest height, on his side of the laptop. His right hand holds a single fork. The bowl physically travels upward from below the desk; it never appears from nowhere.
3.8-5.8 seconds: The RIGHT man winds a small forkful of spaghetti and takes a bite, drawing in one dangling noodle with a brief audible slurp. He stops talking before food enters his mouth. The LEFT man turns his head toward him, looks down at the unexpected bowl, then back at his face; eyebrows knit, nose wrinkles slightly, lips tighten in restrained comic disgust.
5.8-6.8 seconds: The LEFT man alone asks, "Now?" in a dry, incredulous male voice, stressing the absurd timing. The RIGHT man remains silent, chewing.
6.8-10.0 seconds: Hold the awkward comic beat. The RIGHT man gives a tiny matter-of-fact nod while chewing and calmly prepares another forkful, entirely unbothered. The LEFT man leans back a little, looks from the bowl to the camera with a long disgusted, disbelieving stare. End on their contrasting expressions.

Audio: synchronized English speech with correct visible speaker attribution, natural room tone, a quiet ceramic/fork clink and one small spaghetti slurp. Only the two quoted lines; no extra dialogue, narration, music, laugh track or exaggerated sound effects. Restrained believable physical comedy. Keep one bowl, one fork and consistent spaghetti; anatomically correct hands and natural mouth movement. The laptop stays on the desk and does not obstruct the bowl once lifted. Remove the source frame's old subtitle as motion starts; no generated captions, text or watermarks.

===== Round 03 · Café visit · Scene 1 · outside (5 s) =====

Muffin, the black Pekingese on the left, and Missy, the sable-and-white Pekingese on the right, sit outside a dog-friendly café in Hong Kong. Muffin looks up at the paw-print "DOGS WELCOME" sign, then turns to the camera and says in a cheerful, excited voice: "Missy, this one lets dogs in!" Missy wags her tail happily. Slow push-in at dog height. Natural sunlight, realistic fur and natural dog movement.

===== Round 03 · Café visit · Scene 2 · inside (5 s) =====

Inside the cozy café, Muffin, the black Pekingese on the left, and Missy, the sable-and-white Pekingese on the right, sit next to an empty water bowl. A friendly barista's hand comes into frame and pours fresh water into the bowl. Missy looks at the camera and says in a sweet, happy voice: "Best spot in Sheung Wan!" Then both dogs lean in and start lapping the water. Static camera at dog height. Warm sunlight, realistic fur and natural dog movement.

===== Round 04 · Talk Show · Scene 1 · close-up (8 s) =====

A late-night TV talk show. Muffin, the small black Pekingese boy wearing his light-blue harness (no tie, no collar, no clothing), sits behind the pink desk in front of the pink wall. He ducks down behind the desk and pops back up holding a yellow rubber chicken in his mouth. He drops it on the desktop and presses it with one front paw: it gives a loud, comical SQUEAK. Only after dropping the chicken, with his mouth free, does he look straight into the camera and announce in a cheeky, energetic young male voice (a boy's voice, not a woman's or a little girl's), his mouth moving in sync with every word: "Welcome back! Tonight's very special guest is a Xoloitzcuintli!" The rubber chicken stays one single object on the desk. The camera is completely locked off: same framing as the start frame from the first frame to the last, no zoom, no reframing, the framed panel on the wall always visible. Keep the real photographic look of the start frame: a real dog, natural fur and subtle, believable facial movement.

===== Round 04 · Talk Show · Scene 2 · two-shot (5 s) =====

A late-night TV talk show. Two real Pekingese sit side by side behind a pink desk in front of a pink wall, as co-hosts. Muffin, the small black Pekingese boy, is on the left; Missy, the cream Pekingese girl with the dark face mask and light-blue harness, is on the right. Neither dog wears any clothing. Missy turns her head toward Muffin, raises an eyebrow and says drily in a sweet, bright young female voice: "You practised that word all week." Then Muffin turns to the camera, lifts his chin proudly and replies in a cheeky young male voice (a boy's voice, not a woman's or a little girl's): "And it shows." Only one dog speaks at a time, and the other's mouth stays closed. The camera is completely locked off with no zoom or reframing, and both dogs stay fully in frame. Keep the real photographic look of the start frame: real dogs, natural fur and believable movement.

===== Round 04 · Talk Show · Scene 2 · FLUX.3 version (one line trimmed to pass its filter) =====

A late-night TV talk show. Two real Pekingese sit side by side behind a pink desk in front of a pink wall, as co-hosts. Muffin, the small black Pekingese boy, is on the left; Missy, the cream Pekingese girl with the dark face mask and light-blue harness, is on the right. Missy turns her head toward Muffin, raises an eyebrow and says drily in a sweet, bright young female voice: "You practised that word all week." Then Muffin turns to the camera, lifts his chin proudly and replies in a cheeky young male voice (a boy's voice, not a woman's or a little girl's): "And it shows." Only one dog speaks at a time, and the other's mouth stays closed. The camera is completely locked off with no zoom or reframing, and both dogs stay fully in frame. Keep the real photographic look of the start frame: real dogs, natural fur and believable movement.

===== Round 05 · Lightning storm · The prompt (15 s) =====

A cinematic 15-second single continuous take, shot like a big-budget disaster movie: anamorphic lens, shallow depth of field, subtle film grain, a moody teal-and-amber grade. The shot opens exactly on the start frame: a man in a black short-sleeved shirt, glasses and dark jeans crouches under a large old tea tree on a hillside tea garden, one hand on the trunk.
0–3 s: warm golden afternoon light. He smiles at the camera as the camera begins a slow, smooth push-in toward him.
3–6 s: dark storm clouds roll over and swallow the sun, and the whole scene drops to a cold blue-grey. A violent gust of wind thrashes the branches and whips dry leaves past the lens. The first heavy raindrops hit the camera lens and streak down it. He looks up, and the camera tilts up with his gaze to the churning black sky through the branches.
6–8 s: a massive lightning bolt strikes a tree further down the slope. The blinding white flash backlights him into a silhouette for a split second, and the camera shakes from the deafening thunderclap. Sparks and smoke rise from the struck tree.
8–11 s: slow motion at half speed. Raindrops hang in the air, lit by a second flicker of lightning, as he scrambles to his feet, slips in the mud and catches himself with one hand. Then it snaps back to real speed.
11–15 s: the camera turns handheld and tracks alongside him as he sprints down the muddy slope through sheets of rain, one arm over his head. Behind him a huge branch cracks off the tea tree and crashes into the mud where he was crouching, splattering mud onto the lens. The shot ends as he disappears into the rain mist, with lightning flickering over the hills.
Throughout: keep the man's face, glasses and clothes exactly as in the start frame, with a real photographic look rather than CGI. Realistic rain physics, wet fabric clinging to him, mud, volumetric mist and dramatic light changes. Sound: howling wind, pouring rain, a thunderclap with deep sub-bass, the branch cracking, his heavy breathing and splashing footsteps. No music.

===== Round 06 · Stair race + terrace crash · Scene 1 · stair race (5 s) =====

Muffin, the black Pekingese boy (jet-black coat with a small white heart-shaped patch on his chest) on the left, and Missy, the sable-and-white Pekingese girl (charcoal-black face mask and dark ears, silvery-grey mane, white chest and legs, golden-tan patch on her back) on the right, burst off the top step and race each other down the steep stone staircase street, bounding down the steps at full speed, weaving between pedestrians' legs. Muffin's long ears flap and Missy's plumed tail bounces with every leap. Halfway down, Muffin glances back over his shoulder and shouts in a cheeky, energetic young male voice (a boy's voice, not a woman's or a little girl's): "Last one to the café buys the treats!" The camera tracks backward down the stairs in front of them, low at dog height, keeping both dogs in frame the whole time. Bright morning sun, realistic fur and natural, energetic dog movement.

===== Round 06 · Stair race + terrace crash · Scene 2 · terrace crash (5 s) =====

Muffin, the black Pekingese boy (jet-black coat with a small white heart-shaped patch on his chest), and Missy, the sable-and-white Pekingese girl (charcoal-black face mask and dark ears, silvery-grey mane, white chest and legs, golden-tan patch on her back), sprint off the bottom of the stone stairs onto the sunny café terrace. Muffin tries to stop but skids sideways on the smooth stone tiles and bumps into the small round café table in the foreground. The paper coffee cup wobbles and tips off the edge. Missy leaps forward and catches the cup's cardboard sleeve in her mouth just before it hits the ground, not a drop spilled. Muffin lies upside-down on his back on the tiles, looks up at the camera and says sheepishly in a young male voice (a boy's voice, not a woman's or a little girl's): "...I meant to do that." Static low camera at ground level behind the table, keeping the table, the cup and both dogs in frame. Bright morning sun, realistic fur, believable physics.

===== Round 07 · Kitchen commercial · Scene 1 · The show-off =====

Premium live-action dog-food commercial, portrait 9:16. Preserve the exact starting-frame characters, real Hong Kong kitchen geography, golden daylight and tactile materials. Muffin is the black Pekingese boy with white heart chest patch, oversized white chef hat and cream apron; Missy is the sable-and-white Pekingese girl with charcoal mask and ears, silvery mane and white legs. Photoreal fur with gently stylized faces. Keep paws and props anatomically stable. No added captions, no extra logos, no music. Native cinematic sound effects only. One continuous shot. 0-2s: camera dollies forward and slides right past the large foreground pan; the pan rim moves across frame with real parallax while the kitchen remains the same. Rack focus from glossy food on the wooden spoon to Muffin's eyes. 2-4s: Muffin lowers the spoon into the pot, stirs with a confident flourish and steam curls around his hat; Missy watches with amused curiosity. 4-6s: camera settles into a medium close-up; Muffin looks into the lens and says in a cheeky young male boy's voice, 'Watch and learn.' Clear mouth timing. Simmering, metal clink, a subtle steam hiss. No talking from Missy. Do not replace the dolly with a digital zoom.

===== Round 07 · Kitchen commercial · Scene 2 · The save =====

Premium live-action dog-food commercial, portrait 9:16. Preserve the exact starting-frame characters, real Hong Kong kitchen geography, golden daylight and tactile materials. Muffin is the black Pekingese boy with white heart chest patch, oversized white chef hat and cream apron; Missy is the sable-and-white Pekingese girl with charcoal mask and ears, silvery mane and white legs. Photoreal fur with gently stylized faces. Keep paws and props anatomically stable. No added captions, no extra logos, no music. Native cinematic sound effects only. One continuous shot with a deliberate speed ramp. 0-2s: Muffin jerks the single saute pan forward and up, tossing the cooked carrot and sweet potato cubes into one rising arc. Camera tracks laterally from Muffin toward Missy at counter height, keeping pan and airborne food visible. 2-5s: movement becomes half-speed slow motion; camera tilts up with the tumbling food then smoothly down. Muffin's chef hat droops over one eye as he looks alarmed. Missy reaches in and takes the SAME pan handle from Muffin, positioning that pan under the falling cubes. 5-6.5s: speed ramps back to normal and all the food lands inside that pan with a sharp sizzle. 6.5-8s: pan steadies on the stove; camera finishes on their reaction two-shot, Missy composed and Muffin sheepish. No dialogue. One pan throughout, no duplicated utensils, no extra paws, no food teleportation or disappearing food. Sound: restrained whoosh in slow motion, crisp landing and sizzle at normal speed.

===== Round 07 · Kitchen commercial · Scene 3 · The real chef =====

Premium live-action dog-food commercial, portrait 9:16. Preserve the exact starting-frame characters, real Hong Kong kitchen geography, golden daylight and tactile materials. Muffin is the black Pekingese boy with white heart chest patch, oversized white chef hat and cream apron; Missy is the sable-and-white Pekingese girl with charcoal mask and ears, silvery mane and white legs. Photoreal fur with gently stylized faces. Keep paws and props anatomically stable. No added captions, no extra logos, no music. Native cinematic sound effects only. One continuous controlled 45-degree camera arc around the SAME counter, not a scene change. The tray already contains cooked golden kibble at the opening; no food transformation. 0-2s: Missy slides the single baking tray of golden kibble onto the counter with both front paws; it comes to rest and the pieces stay on it. 2-4.5s: camera arcs from Muffin's side toward Missy's side with visible background parallax while Muffin brushes flour from his apron, spreads a paw to present the tray and says in a smug but charming young male boy's voice, 'All my own work.' 4.5-7s: Missy slowly gives him a dry side-eye; Muffin's proud grin becomes a sheepish smile. Rack focus from their faces to the crisp textured kibble on the tray, ending close enough to motivate the product insert. Soft tray scrape and tiny kibble rattle. Only Muffin speaks. Keep Missy's mask and silvery mane intact through the camera arc.

===== Round 07 · Kitchen commercial · Scene 4 · The pour =====

Premium macro tabletop dog-food commercial. Preserve the exact ONE open cream-and-blue Muffin's Kitchen pouch and all its front artwork, navy logo, red recipe stripe, clear window and physical dimensions from the starting frame. Same real stainless kitchen counter and warm backlight. No dogs or hands enter frame. One metal scoop of golden rounded kibble, one open bag. One uninterrupted camera shot. 0-2s: track close alongside the steel scoop as it slowly tips over the clearly open pouch mouth; rack focus from scoop rim to textured kibble. 2-5s: half-speed slow motion as kibble leaves the scoop in a controlled stream, rotating naturally in the backlight; camera moves DOWN beside the stream and arcs slightly around the mouth. Every piece passes INSIDE the bag opening, with correct occlusion at the front lip. No kibble goes through the packaging or spills onto the counter, and the scoop does not refill. 5-6.5s: return to normal speed as last pieces land; camera cranes slightly up to reveal kibble settling inside, then pulls back. 6.5-8s: scoop exits up-left; focus shifts onto the intact printed front, bag remaining upright, mouth open. Golden commercial rim light, real macro depth of field, realistic matte packaging and stable print. Dry crisp kibble patter, a soft pouch crinkle. No dialogue, captions, music, added words, magical particles or additional bags.

===== Round 07 · Kitchen commercial · Scene 5 · The reveal =====

Closing shot of a premium Muffin's Kitchen dog-food commercial, one continuous 6-second portrait shot. Preserve the exact sealed full cream-and-blue pouch and its printed label from the starting frame; logo, chef illustration, red recipe stripe and clear kibble window must not change or warp. Preserve the navy bowl and the two dogs in the background: black Muffin in chef hat and cream apron, sable-and-white Missy with charcoal face mask. Product front stays unobscured. 0-3s: camera makes a slow elegant 30-degree orbit around the stationary pouch, subtle real parallax between bowl, pack and dogs, while a soft highlight travels across the matte material. No spinning or melting packaging. 3-4s: camera eases into a straight-on hero composition and racks focus from the softly smiling dogs to the sharp readable MUFFIN'S KITCHEN logo. 4-6s: lock the FINAL TWO SECONDS into an absolutely stable premium packshot, whole pouch and bowl in frame, dog faces softly out of focus behind. Offscreen voiceover from Muffin in a proud young male boy's voice at 1-2.5s: 'Muffin's Kitchen.' Offscreen Missy in a dry sweet young female voice at 3-4s: 'Mostly mine.' Dogs need not lip-sync the offscreen voiceover. Quiet room tone and gentle packaging rustle, no music, no added captions, floating typography or claims.