MiniMax H3 PromptsFree Hailuo 3 prompt library — real videos with their full prompts

integrated_multimodal_description: [Shot 1] Live-action, cinematic and photorealistic. The young female hiker shown in <Picture 1> remains on the narrow mountain ridge, preserving her exact facial identity, natural skin texture, brown hair, red insulated hiking jacket, backpack straps, body proportions, front-facing orientation, lighting, background mountains and initial composition.

The camera begins in the close head-and-shoulders framing established by <Picture 1>. She remains unaware of the camera and keeps her eyes fixed on one distant point in the valley ahead. A light mountain wind moves only a few loose strands of her hair and slightly shifts the fabric around her collar.

She takes one natural breath, gently raises her chin and loudly calls into the valley. The young woman with a clear, strong, natural American voice (S1) shouts with joyful energy: <d>[English] Helloooooooo!</d>

Her lips, jaw, cheeks, throat and chest move naturally with the sustained call. Her arms remain relaxed beside her body. Her feet, hips, shoulders and head remain oriented in the same direction; she does not rotate, step, turn, or follow the camera with her eyes.

As she begins shouting, the camera pulls out with large amplitude at slow speed while pedestaling upward with small amplitude at slow speed. It travels directly backward along the same optical axis, progressing continuously from a close portrait to a chest-up frame, then to a full-body view, and finally to a wide view of the ridge and surrounding valley.

The camera does not orbit, truck sideways, pan around her, or use a digital zoom. Realistic parallax reveals the steep grassy slopes, distant layered mountains and deep valley while the horizon remains stable.

As the camera moves farther away, her direct voice becomes quieter and more distant. After a short natural delay, the final syllable reflects first from the left mountain wall and then more softly from the right side of the valley, creating separated stereo echoes that gradually become quieter and more diffuse.

She completes the call and remains standing in exactly the same place and orientation. By the end of the shot, she appears as a small red figure on the vast green ridge while the final distant echo fades naturally across the valley.

Create a polished 6-second, 16:9 photorealistic racing-game car-selection sequence for the fictional game APEX RIDGE.

REFERENCE MAP
[REF_1]: VULCAN S1 — crimson red track-focused supercar.
[REF_2]: TITAN RS — matte silver high-performance GT.
[REF_3]: KAIRO X — yellow electric hypercar.
[REF_4]: complete APEX RIDGE car-selection interface, visual-layout reference and final metallic-blue AURELIS R9 appearance.

GLOBAL CONTINUITY
Recreate the premium dark showroom and game UI from [REF_4]. Keep the exact title “APEX RIDGE” and subtitle “SELECT YOUR CAR” visible and perfectly stable throughout. Preserve the same camera position, interface layout, showroom lighting, vehicle scale and perspective. Large white left and right navigation chevrons remain visible beside the active vehicle. The bottom carousel always contains exactly four vehicle cards in this order: VULCAN S1, TITAN RS, KAIRO X, AURELIS R9. The specifications panel always uses the labels SPEED, ACCELERATION, BRAKING and HANDLING. Use clean, readable English UI text only. No real-world brands, badges or logos.

TIMELINE — EXACTLY 6.0 SECONDS

0.0–1.5 seconds — VULCAN S1
Open on the red VULCAN S1 from [REF_1] as the large central vehicle. Its bottom card gains a sharp red-white selection outline. The right chevron gives a subtle luminous pulse. The four specification bars animate rapidly to their VULCAN S1 values. Add a restrained turntable movement of only a few degrees and realistic moving reflections across the red bodywork.

1.5–3.0 seconds — TITAN RS
At exactly 1.5 seconds, a crisp right-arrow click triggers a fast horizontal carousel slide with elegant motion blur. The red vehicle exits left and the silver TITAN RS from [REF_2] enters from the right, settling in the identical central position. The TITAN RS card becomes highlighted, the previous card returns to neutral, and all four specification bars smoothly update. Preserve body design and silver color exactly.

3.0–4.5 seconds — KAIRO X
At exactly 3.0 seconds, the right chevron pulses and clicks again. Perform the same quick premium carousel transition. The silver GT exits left and the yellow KAIRO X from [REF_3] enters from the right. Highlight only the KAIRO X card. Animate the stats to new values and sweep a thin light reflection across the yellow body without changing its shape.

4.5–6.0 seconds — AURELIS R9 FINAL SELECTION
At exactly 4.5 seconds, trigger the final right-arrow click. The yellow car exits left and the metallic-blue AURELIS R9 shown in [REF_4] enters from the right and locks perfectly into the hero position and design shown in [REF_4]. Highlight the AURELIS R9 card with a vivid electric-blue border, reveal the exact label “SELECTED”, and animate the final stats to SPEED 92, ACCELERATION 95, BRAKING 88, HANDLING 91. The blue CONFIRM button illuminates and gives one controlled pulse. Finish on a clean, stable hero frame of the selected blue car with the full interface readable.

MOTION AND CAMERA
Locked front three-quarter showcase camera; only subtle showroom parallax and small turntable movement. Carousel transitions are fast, precise and game-like, never cinematic scene cuts. Every car remains sharply recognizable and consistent with its corresponding reference image. No camera shake, no zoom jump, no morphing and no change of environment.

AUDIO
Premium racing-game UI sound design: one soft navigation click and short aerodynamic whoosh at 1.5, 3.0 and 4.5 seconds; a distinct half-second engine-preview sound for each car; finish with a clean electronic confirmation tone when AURELIS R9 becomes SELECTED. Keep audio tight and synchronized. No dialogue and no voice-over.

NEGATIVE CONSTRAINTS
No extra cars, no missing cars, no duplicated vehicles, no malformed wheels, no model morphing, no color changes, no illegible text, no misspelled labels, no flickering UI, no random icons, no real manufacturer marks, no racing footage and no transition away from the selection showroom.

TITLE: APEX RIDGE - DOWNHILL TELEMETRY RUN

FORMAT
Exactly 15 seconds, 16:9 widescreen, ultra-photorealistic AAA racing-simulator gameplay with native synchronized stereo audio. Fast, premium motorsport presentation with multiple camera viewpoints, physically credible vehicle dynamics and strict spatial continuity. The entire sequence takes place on a closed competition road with no public traffic.

OMNI REFERENCES
- Use [REF_1] as the exact exterior vehicle reference. Preserve the same fictional cobalt-blue supercar body, proportions, headlights, carbon details, black wheels, copper brake calipers and rear wing in every exterior view.
- Use [REF_2] as the exact cockpit reference. Preserve the steering wheel, carbon dashboard, copper stitching, paddle shifters, windshield geometry and black racing gloves.
- Use [REF_3] as the exact mountain-road, lighting and downhill-action reference. The car must always travel from the high mountain pass DOWN toward the valley through successively lower hairpins. Never depict uphill travel.

PERSISTENT RACING HUD - VISIBLE FOR ALL 15 SECONDS
Keep one stable, clean, semi-transparent racing-simulator telemetry overlay visible in every shot, including cockpit, chase, wheel and aerial views. The HUD remains anchored to the screen safe area and never disappears during camera cuts.
- Lower right: large digital SPEED in km/h, current GEAR, curved RPM tachometer and redline.
- Lower left: horizontal BRAKE and THROTTLE input bars, with brake in red and throttle in green.
- Upper left: compact downhill route map with a moving position marker and the next hairpin highlighted.
- Upper right: running sector timer and delta.
- Small central lower strip: steering input, ABS and traction-control activity.
Telemetry must update logically and continuously with the driving: speed falls under braking and rises on exit; RPM drops with each downshift and climbs under acceleration; gear changes sequentially; brake and throttle never activate fully at the same time. Keep all indicators crisp, readable, correctly spelled and visually consistent. No flicker, random values, duplicated widgets or HUD redesign between shots.

DRIVING AND CAMERA TIMELINE
0.0-3.5 seconds - COCKPIT VIEW
Begin directly inside the cockpit from [REF_2], descending rapidly toward a tight right-hand hairpin visible below. Speed reads approximately 178 km/h in 5th gear at high RPM. The driver brakes firmly; the red BRAKE bar rises, the car's nose loads forward naturally, and the speed decreases smoothly. Two synchronized paddle downshifts: 5th to 4th, then 4th to 3rd, each with a realistic rev-matched engine blip. Gloved hands rotate the wheel progressively right. No camera shake beyond subtle chassis vibration.

3.5-7.0 seconds - REAR THREE-QUARTER CHASE VIEW
Cut to a low rear chase camera that preserves the exact car from [REF_1]. The car enters the downhill right-hand hairpin in 3rd gear at approximately 92 km/h. Brake lights glow during trail braking, the outside suspension compresses, tires follow the asphalt without sliding, and the car clips the inside apex cleanly. The road and valley remain visibly lower ahead. The HUD stays fixed on screen and continues updating through the cut.

7.0-10.0 seconds - FRONT-WHEEL / FENDER VIEW
Cut to a stabilized camera mounted just behind the front-left wheel, showing the spinning black wheel, copper brake caliper, damp asphalt and barrier rushing past. The driver releases the brake; the red bar falls to zero and the green THROTTLE bar rises progressively. Speed increases from roughly 96 to 128 km/h. Shift cleanly from 3rd to 4th as the RPM reaches the shift point. Show realistic tire deformation, suspension recovery and restrained motion blur while the car body remains stable.

10.0-13.0 seconds - HIGH AERIAL TRACKING VIEW
Cut to a high cinematic tracking shot revealing the car descending from the completed hairpin toward the next lower switchback, matching [REF_3]. The camera moves parallel to the slope while preserving screen direction and exact road geography. The cobalt-blue car accelerates through the short downhill straight, with the valley far below and several lower curves visible ahead. Telemetry remains readable and correctly synchronized.

13.0-15.0 seconds - HOOD VIEW AND SECTOR RESULT
Cut to a low hood-mounted forward view. The road drops sharply toward the next left-hand hairpin. The car reaches approximately 151 km/h in 4th gear, then the driver begins braking at the final half-second. Show a brief, tasteful HUD notification: “SECTOR 2 - PERSONAL BEST” while all core telemetry remains visible. End in motion with the engine note falling under braking, ready to continue into another clip.

VEHICLE PHYSICS
Authentic high-performance simulation: correct downhill acceleration, progressive braking distance, forward weight transfer, lateral suspension load, stable tire contact patches, sequential paddle shifts, realistic steering angle, physically coherent wheel rotation and consistent road grip. The driver follows the racing line and remains fully in control. No drifting, jumping, collision or arcade behavior.

AUDIO
Native synchronized stereo game audio with no music and no narration. Detailed naturally aspirated high-performance engine, intake resonance, exhaust reflections from rock walls, rev-matched downshift blips, paddle clicks, transmission whine, tire scrub under load, brake hiss, suspension movement, gravel ticks beneath the chassis, wind pressure and subtle cockpit vibration. Audio perspective changes naturally with each camera: enclosed and detailed inside the cockpit, louder exhaust in chase view, mechanical tire and brake detail at the wheel, wider mountain echo in the aerial view. No audio cuts, distortion or artificial looping.

VISUAL STYLE
Ultra-realistic premium racing game capture, physically based materials, ray-traced reflections, realistic damp asphalt, natural cold sunrise, thin alpine mist, restrained cinematic contrast, believable motion blur, sharp vehicle identity and stable mountain geography. Smooth professional cuts with no teleportation.

NEGATIVE CONSTRAINTS
No real car manufacturer, existing racing-game branding, logos, badges, license-plate text, watermark, subtitles, public traffic, spectators, pedestrians, police, road debris, crash, collision, drifting, uphill motion, reversed travel direction, changing car design or color, changing cockpit, extra hands, malformed fingers, steering-wheel deformation, impossible wheel rotation, floating tires, incorrect reflections, warped barriers, discontinuous road, impossible camera positions, HUD disappearance, HUD flicker, unreadable telemetry, misspelled labels, random speed jumps, impossible gear changes, simultaneous full brake and throttle, duplicated UI, arcade effects, nitrous flames or music.

Use **@Image1** as the exact visual reference for the finished kitchen. Preserve the same room layout, island, cabinets, appliances, pendant lights, stools, countertops, flooring, lighting, materials, and camera composition.

Create a **10-second ultra-realistic POV pressure-washing transformation video** in **vertical 9:16**.

## Camera

First-person perspective from a worker holding a powerful pressure-washer wand. Two realistic hands remain visible throughout. Keep the kitchen geometry and camera direction consistent with **@Image1**.

Use energetic handheld movement, realistic recoil from the water pressure, fast sweeping motions, close-up spray passes, water droplets on the lens, mist, splashes, and subtle motion blur.

## Scene 1 — 0–2 seconds

The entire kitchen is completely covered in thick black dirt, soot, grease, mud, and grime. The cabinets, island, countertops, appliances, stools, walls, pendant lights, and floor appear almost entirely black.

The worker raises the pressure-washer wand and activates it. A powerful jet of water blasts the front edge of the island, immediately cutting a clean bright line through the dirt.

## Scene 2 — 2–7 seconds

The worker rapidly pressure-washes across the kitchen in continuous sweeping movements.

Black grime peels away in satisfying strips from the island, cabinets, refrigerator, countertops, backsplash, cooker hood, stools, and floor. Dirty water flows downward and spreads across the floor while mist and dark particles fill the air.

Use speed ramps and quick whip movements between surfaces while every cleaned area accurately reveals the original kitchen from **@Image1**.

## Scene 3 — 7–10 seconds

The final spray clears the remaining dirt from the floor and island. The worker lowers the pressure washer as the mist settles.

Reveal the kitchen completely clean, dry, illuminated, and identical to **@Image1**. Finish with a smooth cinematic pullback showing the entire spotless kitchen.

Ultra-realistic water physics, satisfying cleaning transformation, detailed grime removal, realistic reflections, no people besides the POV hands, no text, no subtitles, no logo, no watermark.

Jim and Dwight from The Office discuss autonomous coding agents

A 15-second photorealistic cinematic video captured as one continuous unbroken shot on a rain-soaked downtown street at blue hour.

A woman in a bright yellow raincoat rides a red bicycle quickly toward the camera while carrying a sealed paper coffee cup in one hand. [low-angle tracking shot]

She brakes sharply beside a glass storefront. The rear tire throws a realistic arc of rainwater across the pavement. The coffee cup slips from her hand, rotates once in the air, and she catches it cleanly without spilling. She then dismounts from the bicycle in one smooth, physically believable movement.

[smooth 180-degree camera orbit]

Maintain exactly the same woman, face, hairstyle, yellow raincoat, red bicycle, and coffee cup throughout the entire shot. Her body must move with realistic weight, balance, momentum, and contact with the bicycle and ground.

Behind her, a bright neon storefront sign remains clearly readable and correctly spelled throughout the shot:

“HAILUO 3.0 — ONE TAKE”

[slow cinematic push-in]

She looks toward the camera, smiles slightly, and says naturally:

“Okay... that was impressive.”

Photorealistic skin, believable hands, accurate reflections in the wet pavement and storefront glass, natural rain physics, realistic cloth and hair movement, cinematic lighting, shallow depth of field, and detailed environmental motion.

Native stereo audio: steady rainfall, bicycle chain movement, tire skid through water, coffee cup catch, distant city traffic, and perfectly synchronized dialogue.

No cuts, no scene transitions, no morphing, no identity changes, no wardrobe changes, no extra limbs, no disappearing objects, no misspelled text, and no background music.

A 15-second video that begins as a photorealistic cinematic dinosaur adventure and smoothly evolves through multiple visual styles without losing story continuity.

Scene begins in a misty prehistoric jungle at sunrise. A massive realistic Tyrannosaurus rex slowly steps into frame, its skin textured with scars, dust, and morning moisture. The camera tracks low beside its feet as the ground trembles with each step. Cinematic live-action style, natural lighting, realistic weight, realistic muscle movement.

[match cut transition]

The T. rex looks up. Its eye reflects the night sky, and the camera flies into the reflection, transitioning into outer space. The dinosaur is now floating beside a sleek silver spacecraft shaped like a fossilized bone. Earth is visible far behind it. Epic cinematic sci-fi style, dramatic lens flares, realistic stars, slow zero-gravity motion.

[liquid morph transition]

The scene transforms into a vibrant 2D animated style. The T. rex now wears a tiny astronaut helmet and comically pedals a rocket-powered bicycle across Saturn’s rings. Bright colors, expressive animation, smooth exaggerated motion, playful timing.

[paper rip transition]

The style changes into stop-motion clay animation. The dinosaur lands on a small asteroid, where tiny clay alien creatures offer it a glowing meteor sandwich. Tactile clay texture, handmade miniature sets, charming imperfect movement, visible stop-motion feel.

[glitch transition]

The style shifts into high-energy anime. The T. rex launches through a neon wormhole, roaring silently as planets streak past in colorful speed lines. Dynamic camera movement, intense anime lighting, dramatic perspective, cosmic scale.

[cinematic dissolve transition]

Final scene returns to photorealistic live-action cinema. The T. rex stands calmly on the moon in an astronaut helmet, looking at Earth. A small flag beside it clearly reads:

“HAILUO 3.0 TEST”

The camera slowly pushes in as the dinosaur blinks, exhales fog inside the helmet, and gently taps the moon dust with one claw.

Maintain the same dinosaur identity and recognizable silhouette across every style. Each transition should feel intentional, smooth, and visually creative. No random scene cuts. No disappearing dinosaur. No extra limbs. No unreadable text. No misspelled words. No chaotic camera shake. Make the final result feel like one connected mini-adventure through time, space, and animation styles.

A 9:16 vertical family-confrontation scene with grounded live-action performances, set in a Chinese family home or small restaurant. Use warm interior light, red decorations and calligraphy in the background, shallow depth of field, intense emotion, and tight pacing.

Performance: natural short-form drama, never theatrical. Qin Haoxuan argues back with anger, hurt, and urgency. The older woman questions him in a sharp, forceful, relentless tone. Build the confrontation steadily.

Shoot mainly in medium-close shots with frequent shot/reverse-shot cutting. Keep the setting lived-in and realistic. No sci-fi, period costume, or animation styling. Do not show subtitles, added text, platform watermarks, or stickers.

Create a fast-paced 15-second, 16:9 anime opening-title sequence for an original series titled “MINIMAX.” Use Image1, Image2, Image3 as the overall visual, ensemble and color-style references. Use Image4 for the pink-haired swordswoman, Image5 for the red-haired swordswoman and Image6 for the black-haired swordswoman. Preserve each character’s face, hairstyle, outfit, proportions, colors and weapon design.

Use the supplied audio Audio1 as the exact timing reference. Synchronize every cut, character action, camera movement, graphic transition and title reveal to its beats, accents, fills, rises and final hit. Do not create a separate soundtrack. Add only subtle sword swishes, cloth movement and transition impacts under the reference audio.

VISUAL STYLE

A real high-energy anime opening combined with premium motion graphics. Use black, crimson, wine red, dusty pink, pale ivory and cool gray-blue. Combine angular split screens, diagonal masks, torn-paper shapes, manga-style framing, ink textures, petals, crimson moon forms, torii silhouettes, temple rooftops and sharp graphic lines.

Keep the sequence kinetic and animated, not a slideshow. Characters must run, turn, pivot, slide, draw their swords, change stance and react to momentum. Hair and clothing must remain in motion. Use tracking shots, whip pans, rapid push-ins, low angles and short camera orbits.

Never show multiple copies of the same character in one shot. A shot may feature one character alone or all three together naturally. No clone effects, repeated character fragments or layered duplicates.

SEQUENCE FLOW

[0.0–1.5s]
Open instantly on the first beat. A crimson line cuts through black as petals and white fragments burst outward. Pink, red and black graphic panels snap into place while a crimson moon quickly assembles in the background.

[1.5–4.3s]
Feature the pink-haired swordswoman alone. She runs into frame, plants one foot, turns and draws her katana in one continuous agile motion. Track beside her, then push toward her face. Cut between her eye, hand, blade and full-body movement on the audio accents. Use pink petals, red lines and diagonal panel cuts.

[4.3–7.2s]
Switch to the red-haired swordswoman on a strong musical hit. Use a low-angle tracking shot as she moves forward, pivots and performs one powerful controlled sword motion. Her long hair and garments sweep with momentum. Place a large crimson sun disc behind her while black-red panels strike into frame on the heavier beats.

[7.2–10.2s]
Feature the black-haired swordswoman alone. Begin with a moving close-up of her crescent earring and eyes, then orbit around her as she turns, changes stance and redirects her blade. Keep her hair and sleeves flowing. Use dark negative space, moon-shaped masks and precise cuts synchronized to smaller rhythmic details.

[10.2–12.7s]
At the main musical peak, reveal all three characters together in one shared scene. They enter through real movement and form a strong triangular composition. Use a fast camera push or short orbit as the crimson moon, torii silhouette, petals and graphic debris align behind them. Show each character only once.

[12.7–15.0s]
During the final phrase, alternate between very short moving close-ups of the three characters, one character per shot. Show an eye turn, hand gripping a sword, flowing hair or a blade catching light. On the final major hit, reveal the title “MINIMAX” through sharp diagonal masks, sword-line wipes and fractured graphic panels.

End on a clean ensemble composition with all three characters together. Keep subtle motion in their hair, clothing, petals and light reflections.

TRANSITIONS
Use beat-synchronized slash cuts, angular panel snaps, moon-mask wipes, petal streaks, torn-paper shutters, ink-impact cuts, foreground wipes and oversized title masks. No soft dissolves, slow fades, liquid morphing or random transitions.

TYPOGRAPHY
Use only the title “MINIMAX.” Keep it sharp, elegant, premium and fully readable. No credits, extra text, random symbols or broken lettering.

FINAL RESULT
A fast, polished and rhythmically precise anime opening with expressive character motion, strong graphic editing, strict visual consistency and no duplicated characters within the same shot.

REFERENCE USE
Images 1 and 2 lock face, hair and wardrobe identity only. Do not copy their frontal pose, centred framing or lens gaze. Image 3 locks apartment geometry, light direction and materials only. Recompose the actors according to the screen directions below. Identities remain stable.

IDENTITY LOCKS
YEON-SU: Korean woman, 28. Loose unstyled dark hair tucked behind her anatomical right ear only; left ear covered. Oversized heather-grey knit, sleeves past both wrists, loose charcoal lounge trousers, bare feet, no jewellery.
JAE-HYUN: Korean man, 32. Dark knee-length coat stays on. Plain shirt open by exactly one button. Watch on anatomical left wrist. Keys handled only by anatomical right hand.
Never mirror these details. No wardrobe change.

SCENE
Three years into a relationship. The argument peaked before this clip. Neither is trying to win. YEON-SU needs him to deny he is leaving but cannot ask directly. JAE-HYUN has waited to be actively chosen and is ashamed that he needed it. His answer is an admission, not an attack. Her last line is not a reply; she discovers that she treated his staying as permanent while he experienced it as a choice she never made.

DIALOGUE
Spoken Korean only. Contemporary Seoul banmal between a long-term couple, with dropped subjects, intimate volume and natural consonants. Never translate, speak, subtitle or display English.

YEON-SU: "또 갈 거야?"
JAE-HYUN: "한 번쯤은... 가지 말라고 할 줄 알았어."
YEON-SU: "안 갈 줄 알았어."

SCREEN GEOGRAPHY
JAE-HYUN is frame-left facing screen-right. YEON-SU is frame-right facing screen-left. Singles leave look-room toward the unseen partner. Matched off-camera eyelines; never look into the lens. In the wide they are three metres apart. The table stays on YEON-SU's anatomical right so her right hand can hold it. The entry shelf stays beside JAE-HYUN's right hand. Fixed marks, but natural breathing, a swallow, finger pressure and tiny posture settling are allowed. No walking, sitting, approaching or contact.

SHOT LIST
[0.0-2.8] LOCKED WIDE TWO-SHOT, 35mm, eye level. Entry frame-left, living room frame-right. JAE-HYUN stands just inside the door, coat on. YEON-SU stands beside the table three metres away, right hand lightly holding its edge. She watches him; he looks at the shelf. At 1.3s he places the keys on wood with his right hand, quietly and deliberately. Neither speaks. Hold the distance.

[2.8-5.4] YEON-SU MEDIUM CLOSE, 75mm, three-quarter view, mid-chest to head. She is frame-right with look-room screen-left. Begin neutral, without prepared sadness. One ordinary shallow breath. Between 3.7s and 4.6s she asks, "또 갈 거야?" Low, plain, almost practical. She wants him to say no. Brows stay relaxed; mouth closes after the line. She listens.

[5.4-10.3] JAE-HYUN MEDIUM CLOSE, 75mm, three-quarter view, mid-chest to head. He is frame-left with look-room screen-right. Mouth fully closed during the silence. Eyes down; one controlled breath and one swallow. This is active decision, not blank posing. He raises his eyes only when ready to admit it. Between 7.9s and 9.6s: "한 번쯤은... 가지 말라고 할 줄 알았어." Quiet, warm, tired. The pause after "한 번쯤은" is a failed attempt to make the sentence easier. No blame, emphasis or sigh. He keeps his eyeline toward her afterward.

[10.3-15.0] YEON-SU CLOSE, 85mm, collarbones to crown, three-quarter view, frame-right with look-room screen-left. One continuous take with no reframing or hidden cut. She is already looking down. Begin without a hurt expression. Recognition forms through stopped breath, a released jaw and eyes focusing on nothing. No tears. Between 12.7s and 13.9s she says to the floor, "안 갈 줄 알았어." Nearly matter-of-fact at first; the final word softens when she hears her own assumption. Only after the last syllable does she lift her eyes toward him. Mouth closed. Hold her listening face through 15.0s.

ACTING
Lived-in screen acting, never presented emotion. Each actor actively listens to the absent partner. Emotion appears through attention, breath and thought, not facial display. Allow imperfect human timing inside the dialogue windows. No added theatrical pauses or mirrored performances.

LIGHT AND IMAGE
One low warm living-room practical, about 2900K. Weak cold night spill from the dark entry, not saturated cyan. Faces partly in shadow; natural Korean skin tone, pores and under-eye texture. Desaturated wood, grey knit and blue-black exterior; restrained grain. The lamp motivates the warm face light. No commercial orange-teal grade, bloom, halation, glow or flare.

CAMERA
Locked and eye-level. Maintain the 180-degree line. Singles are three-quarter views, never centred frontal portraits. Shallow depth of field with both eyes readable. No drift, push-in, handheld sway, arc, rack focus, dutch angle, motion blur or lens-scale change within a shot.

PRODUCTION SOUND
Natural Korean production dialogue recorded in the apartment, not narration, dubbing or synthetic TTS. Intimate conversational level, slight room reflection, matched boom-and-lav perspective. Continuous stereo room tone beneath every line and silence: refrigerator hum, low HVAC and distant traffic through closed glass. Never hard-gate the room. Keys on wood at 1.3s are the only featured effect. No phone buzz or designed silence drop.
No music, score, drone or tonal swell. Dialogue centred; ambience has subtle width. No breath boosting, over-compression or loud social-video mastering. If controllable: 48 kHz stereo, true peak at or below -6 dBFS.

NEGATIVES
No English voice, subtitles, captions, text, logos or watermarks. No lens gaze, centred portrait composition, silent mouthing, speech outside the three lines or bad lip sync. No shouting, sarcasm, crying, tears, sobbing, gasping, widened eyes, knitted brows, trembling lips, clenched fists, pointing, arm crossing or hand gestures. No extra people or reflections, wardrobe drift, side mirroring, beauty filter, face warp or identity change. No extra cuts, jump cuts, punch-ins or framing resets, especially from 10.3s to 15.0s.

・深夜都市高速のタイムラプス
・車のライトの残像が残るタイムラプス

grim reaper in lush garden

Use Image 1 for the character, Image 2 for the UI style, and use the uploaded weapon selection image as the reference for all paint weapons. The selectable weapons should closely match the colorful ink guns shown in the reference image, including compact paint shooters, dual paint pistols, heavy paint cannons, bucket/slosher weapons, large paint brushes, paint rollers, sniper-style paint rifles, and other stylized colorful ink weapons. Match their playful proportions, vibrant colors, chunky plastic construction, and game-like appearance while avoiding exact logos or branding.

[0–2 seconds]
High-angle overhead shot. The character stands on a vivid, highly saturated purple floor, looking up toward the camera, matching Image 1 exactly. A colorful game menu matching Image 2 appears on the right: START NEW GAME, CONTINUE (highlighted), SETTINGS, EXIT GAME. Player profile MINIMAX appears in the top left. The cursor selects CONTINUE.

[2–4 seconds]
The camera pushes toward her right hand. A PAINT WEAPON SELECT panel slides in, using the uploaded weapon selection screen as its layout and visual reference. Multiple colorful paint weapons are displayed exactly like the reference image in a grid with rarity and level indicators. The cursor moves between different weapons. Every time a weapon is highlighted, the paint gun in her hands instantly transforms into the selected model with colorful ink particles, playful mechanical animation, spinning parts, and glossy paint effects. She remains standing throughout.

[4–7 seconds]
The camera smoothly arcs around her left side. The selector rapidly cycles through many weapons from the reference image: compact paint shooters, oversized paint brushes, dual paint pistols, heavy paint cannons, bucket-style splash weapons, sniper paint rifles, giant paint rollers, and experimental ink launchers. With each selection, the weapon in her hands changes immediately to match the highlighted weapon while she naturally adjusts her grip. Every swap is accompanied by bursts of colorful ink, animated UI effects, splashing paint particles, and satisfying transition animations.

[7–8.5 seconds]
Pull back to a medium shot. CONFIRM LOADOUT flashes. The cursor clicks it. All UI panels collapse and disappear. She confidently spins the selected paint weapon once before resting it on her shoulder, remaining in a relaxed standing pose.

[8.5–10 seconds]
A colorful LOADING bar fills from 0% to 100%. The purple testing room dissolves beneath spreading ink splashes that transition into the game world.

[10–15 seconds]
The world loads into a bright, colorful ink-covered city filled with graffiti, paint-coated streets, animated billboards, moving NPCs, and playful architecture. The camera settles into a third-person view behind the character. HUD elements fade in: minimap, health, paint gauge, current weapon icon matching the selected weapon, ammo counter, and objective marker. She runs forward while firing the selected colorful paint weapon, coating walls, streets, obstacles, and scenery with thick, glossy paint that matches the weapon's color.

A stunning, perfectly manicured female hand rests elegantly on a smooth, black marble surface, her fingers slightly spread to showcase each intricately designed, living fingernail. The camera remains perfectly still, ensuring an uninterrupted, high resolution full frame view of the hand and its enchanting animated nail art. The soft, warm ambient lighting casts subtle shadows across the textured marble, enhancing the hyper-realistic depth of her flawless, porcelain smooth skin.

Each fingernail is a miniature portal of magic, alive with independent yet synchronised micro scenes:

Thumb Nail  A tiny koi pond, where goldfish gracefully swim beneath a translucent, glass like surface, their scales shimmering as they dart between delicate lily pads swaying gently in the water’s micro currents.

Index Finger: A blooming cherry blossom tree, where soft pink petals flutter in an invisible breeze, some occasionally detaching and dissolving into golden light.

Middle Finger: A miniature celestial night sky, where tiny constellations shift and twinkle, meteors streaking across the deep, cosmic expanse, casting faint, dancing reflections on the nail’s glossy surface.

Ring Finger: A tiny, enchanted forest, where glowing fireflies hover and flicker, their gentle luminescence casting soft pulses of golden light across the leaves, which sway subtly as if responding to the heartbeat of the world within.

Pinky Finger: A coiled, ethereal dragon, its microscopic, iridescent scales shifting in colour from deep crimson to molten gold. It blinks once, then exhales a swirling wisp of blue fire, which vanishes the moment it drifts beyond the nail’s edge.

Her fingers remain completely still, yet the nail animations play simultaneously, creating a mesmerising symphony of motion within a perfectly composed frame. The lighting subtly shifts, casting gentle reflections off the glossy nails, reinforcing their three dimensional realism.

Style: Hyper-realistic dark fantasy tavern cinematic, grounded physical combat, realistic body momentum, medieval atmosphere, gritty lighting, practical effects, handheld cinematic camera, physically accurate movement, ultra-detailed textures, realistic collisions and environmental interaction.

Duration: 15 seconds

Aspect Ratio: 16:9

IMPORTANT:

Keep the SAME characters fully consistent with the reference: 📷Image1

— Nyssa: athletic woman, dark curly hair tied back, bronze leather armor, cloth wraps, sword on hip

— Grok: massive muscular green orc, scarred skin, fur armor, heavy axes, large physical weight

Combat MUST obey realistic physics:

— no floating

— no anime flips

— realistic inertia and recovery

— believable weight transfer

— realistic impacts and exhaustion

— grounded footwork

— environment reacts physically to impacts

[00:00-00:02]

Wide cinematic shot inside a crowded medieval tavern filled with drunk mercenaries, candles, smoke, spilled ale, and loud cheering. Nyssa walks through the tavern cautiously while patrons stare. Grok sits at a massive wooden table drinking heavily. Deep bass-heavy tavern music and crowd ambience.

[00:02-00:04]

Close-up tension sequence:

— Nyssa locks eyes with Grok

— Grok slowly stands, towering over everyone

— benches scrape across the floor

— mugs shake from his weight

— crowd forms a fighting circle cheering loudly

Handheld camera movement feels realistic and grounded.

[00:04-00:07]

The fight erupts violently. Grok swings a heavy punch with believable momentum. Nyssa narrowly dodges while stumbling realistically into a wooden table. Wood cracks and splinters physically on impact. She counters with fast grounded strikes to Grok’s ribs and legs. Crowd roars and throws coins and mugs into the air.

[00:07-00:10]

Combat intensifies through the tavern:

— Grok grabs Nyssa and throws her across a table

— table explodes realistically beneath her weight

— Nyssa rolls across the floor recovering naturally

— Grok charges heavily, smashing furniture

— patrons jump away realistically

No exaggerated movement, only physically believable combat.

[00:10-00:12]

Nyssa uses speed and positioning intelligently. She dodges another heavy attack causing Grok to crash into a support beam. Dust and debris fall from the ceiling. She climbs briefly onto the bar counter and leaps down with controlled realistic momentum, locking Grok’s arm and using leverage to throw him off balance.

[00:12-00:15]

Final cinematic climax. Nyssa tackles Grok to the ground hard onto broken wooden debris. The floor shakes from the impact. She pins him down on top of him while holding a blade near his throat. Grok struggles realistically beneath her massive weight difference and exhaustion. Tavern crowd erupts cheering wildly, slamming mugs against tables. Camera slowly pushes inward on Nyssa breathing heavily while sweat, dirt, and candlelight flicker across both fighters.

Audio:

Heavy cinematic tavern battle soundtrack with deep drums, loud cheering crowds, breaking wood, metal impacts, realistic body hits, roaring fire ambience, mugs crashing, heavy footsteps, and gritty medieval atmosphere.

Negative prompts:

No floating, no anime combat, no superhero physics, no unrealistic flips, no weightless motion, no blurry characters, no low detail, no cartoon style, no slow reactions, no clipping through objects.

Cinematic 15-second commercial shot, fast-paced editing, dark humor, creepy comedy style. Ultra-realistic, 8k resolution. A surreal hamburger restaurant with moody neon lighting (red and green). Close-ups of distorted, wide-eyed hungry people smiling creepily while eating massive juicy cheeseburgers. Fast cuts: sizzling bacon in flames, towering stacks of burger trays, a mysterious retro waiter with a unnerving wide smile. Hyper-detailed melted cheese, glossy sauce drips, chaotic and fun atmosphere, 90s thriller aesthetic crossed with a fast-food ad.

Continuous 15-second cinematic shot, extreme slow-motion: a futuristic cyberpunk samurai standing in heavy neon rain, reflective wet armor. The samurai slowly draws a glowing katana, lens flare, volumetric neon fog, dynamic camera panning around the character, continuous fluid motion, photorealistic 8k

This is an anime scene in the style of the 2000s. A girl is walking along her usual route to school. As she passes by, she looks toward the viewer and says, "おはよう"

Creative Direction:
Use a high-impact opening, six fast location changes, and a final hero push-in to turn a simple streetwear walk into a complete visual journey through Chicago.

Open with a bold establishing shot, then move through a rapid sequence of distinct Chicago locations while keeping the character, wardrobe, and attitude completely consistent.

Art Direction:
Downtown streets, elevated trains, storefronts, concrete architecture, hard cuts, dynamic tracking shots, modern streetwear, natural city light, and a polished urban-editorial finish.

Train platforms, wide-angle movement, whip-fast transitions, layered street styling, realistic motion, and crisp cinematic pacing.

The Office — Michael admires Trump. 0:00–0:04 — Medium shot, Dunder Mifflin bullpen, documentary handheld. Michael stands by Jim and Pam’s desks, hands animated, big proud smile. Jim pretends to work. Pam holds a mug. Soft office chatter, phones, AC hum. No music. 0:04–0:09 — Michael turns toward camera slightly and says brightly: “I’m just saying — Trump? Total winner. Very stable genius. Like me, but with better hair.” 0:09–0:12 — Jim slowly looks up from his monitor, deadpan, eyebrows raised. Pam freezes mid-sip, stares at Michael with a tight, confused smile. Awkward silence. A phone rings once in the background. 0:12–0:15 — Michael keeps nodding, totally oblivious, double thumbs-up. Jim and Pam exchange one weird glance. Hard cut on their shared “what did he just say” faces. Style: photoreal The Office mockumentary, subtle handheld shake, short clear dialogue, office room tone only, no music, no subtitles, no on-screen text, no logos.