Hi.

TITLE: "Hi." FORMAT: 4:3 landscape, 18 seconds, cinematic multi-shot short film. STYLE: Photorealistic fashion-film fantasy; restrained deadpan humor, warm late-summer daylight. No text, subtitles, logos or MUSIC.

1. REFERENCE MAP — STRICT SOURCE ROLES

@Image1 — GIRL / HIGHEST PRIORITY: EXACT IDENTITY AND WARDROBE. RIGHT-SIDE PORTRAIT is the only face anchor: EXACT eyes, nose, lips, cheeks, jaw, proportions and skin tone in EVERY shot and angle. Full-body: 175 cm tall, slim model build; long dark braids/locs, silver hoop earrings, light-blue short-sleeved overshirt, white T-shirt, wide black cropped trousers, white socks, black loafers. Ignore green masks on full-body views. Never redesign, widen, age or replace face; @Image1 overrides other references. Recheck likeness in low angles, side glances, close-ups, over-the-shoulder views, the overhead tree shot and the final seated scene. The face must not become a different model when speaking or smiling. GIRL'S GROOMING: No visible arm hair, peach fuzz, baby hairs, fuzzy hairline or stray facial hairs. No forehead wrinkles, brow furrows or expression creases, even when she lifts her eyebrows. Neat braids; youthful photographic skin, not waxy; no inflated cheeks. Fine hair texture in the main braids is fine, but never synthesize tiny hairs on arms, cheeks, forehead, hairline or neck.

@Image2 — GOLDFISH BALLOON / PROP LOCK. ONE orange goldfish in ONE clear, spherical, water-filled floating balloon. Lock fish anatomy, fins, scales, waterline, knot, thin continuous string and realistic refraction.

@Image3 — CHILD KING / IDENTITY + CLOTHES. Exact same small boy, face, short natural hair, handmade golden paper crown, oversized cream knit and realistic proportions in every shot.

@Image4 — BEAT 1 SET + COMPOSITION, NOT IDENTITY. Round fishbowl large at screen LEFT on a reflective tabletop; girl at screen RIGHT, CHIN DIRECTLY ON TABLE, looking SIDEWAYS at goldfish. Low table-height camera, green sunlit bokeh and water caustics. Take girl's facial appearance ONLY from @Image1.

@Image5 — BEAT 2 ESCALATOR SET. TWO distinct parallel metal escalators: moving black ribbed steps, yellow step edges, continuous central metal divider. Preserve lane geometry and opposite directions.

@Image6 — BEATS 3 + 4 TREE LOCATION AND FINAL SHOT. Same real giant tree, broad canopy, bark, grass-level angle, dappled light, blue sky. The tree has NO face, smile, eyes, mouth or speech.

2. AUDIO MAP — DO NOT CONFUSE CLIPS

@Audio2 = GIRL'S SHY "...Hi" TO THE KING, BEAT 2 ONLY. Match original delivery; lip-sync.
@Audio1 = GIRL'S WARMER "Hi!" TO THE TREE'S WIND, BEAT 3 ONLY. Lip-sync; same voice.
Beat 1: girl whispers a quiet "Hi" to fish, matching that same voice. KING: one consistent realistic 5–6-year-old boy voice for BOTH his "Hi!" lines—bright, sweet, sincere, childlike articulation, NEVER cartoonish. TREE: WIND AND LEAVES ONLY; no voice, laughter or words.

3. CINEMATOGRAPHY AND PERFORMANCE

Real-location LIVE ACTION, not CGI. ARRI ALEXA LF / Signature Prime-inspired natural optics, 24 fps motion blur, fine grain, plausible depth of field and real reflections. Environmental frames: approx. 35–40 mm-equivalent; portrait faces: 50–75 mm-equivalent at a flattering distance. Real fabric/skin texture; no emphasized pores, fuzz, plastic sheen or face distortion. Light: warm-neutral end-of-summer daylight, muted natural green, pale blue sky, orange fish as the key vivid accent. Match sun/exposure/grade across cuts. Motivate every camera move. Use measured tracking, a deliberate fish-to-escalator pullback, readable shared two-shots, low grass-level reveal and the tree's overhead POV. Avoid random orbiting and abrupt scene morphs.

GIRL: shy, calm, observant. Small eye movements and natural breathing. No overacting, large gasps, exaggerated brows or smiles. No forehead/brow lines from any expression. Lock @Image1 face in EVERY beat.

KING: happy and friendly but physically understated. The contrast is his easy cheerful greeting versus her socially awkward, hesitant reply. Both become quieter beneath the tree.

4. ANATOMY, SIDES AND REAL PHYSICS

GIRL holds string in anatomical LEFT hand throughout. At tree KING is on her anatomical RIGHT, his LEFT hand holding her RIGHT. Balloon opposite king. Never switch. When BOTH face AWAY from camera: girl screen LEFT, king screen RIGHT, balloon far LEFT. When BOTH face TOWARD camera: king screen LEFT, girl screen RIGHT, balloon far RIGHT. Anatomical sides override screen placement.

ESCALATOR EXCEPTION IN SCREEN PLACEMENT: girl is facing DOWN toward camera. Her anatomical LEFT hand appears on the RIGHT side of her body IN THE FRAME, beside the central divider. Balloon stays in LEFT LANE, never crosses divider.

Balloon: single intact string from knot to hand; tension, inertia, wind-drift and delayed settling. Match lighting/focus/grain/refraction. Waterline stays gravity-level with slight slosh. Fish swims with fluid, self-powered tail and fin movements, no jerks, rigid rotation, teleporting or sticker overlay. No duplication or detached string.

5. EXACT SHOT TIMELINE

BEAT 1 — FISH / 0.0–4.5s

SHOT 01 / 0.0–1.2s — OPENING TABLEAU. @Image4: bowl BIG screen LEFT, girl RIGHT, CHIN ON TABLE (not hands/arms), face at fish-height. Keep a recognisable 3/4 face angle matching @Image1 while her EYES glance SIDEWAYS to the goldfish. Quietly curious, nearly deadpan, relaxed lips, no smile. Goldfish looks back. One blink. 50 mm table-height camera; tiny lateral creep; real sun-caustics. Do not broaden or replace her face.

SHOT 02 / 1.2–1.9s — FIRST "HI". Chin stays on table. She whispers "Hi" with minimal lip movement. Same fish slowly turns toward her. Small quiet pause, no music sting.

SHOT 03 / 1.9–3.2s — TRANSFORM. Brief white vapor transforms bowl into @Image2 balloon; SAME fish, no debris/duplicates/leaks. Balloon rises. Camera tilts UP slightly, physically following it. GIRL: tiny eye/brow lift WITHOUT CREASES, soft inhale, barely parted lips. No big gasp, head jerk, shock or face drift.

SHOT 04 / 3.2–4.5s — CATCH. She reaches naturally, catches dangling string with LEFT HAND. String taut, sphere sways. Portrait: tiny relieved exhale and glance up. End ON A CLOSE FISH-IN-BALLOON IMAGE that visually anchors the next shot.

BEAT 2 — KING / 4.5–10.0s

ESCALATOR AXIS: Camera at the BOTTOM landing of @Image5 looking UP. SCREEN-LEFT escalator carries GIRL DOWN from TOP/BACKGROUND to BOTTOM/FOREGROUND, toward camera. SCREEN-RIGHT escalator carries KING UP from BOTTOM/FOREGROUND to TOP/BACKGROUND, AWAY from camera. Both run ~0.5 m/s, with visible OPPOSITE step/rail movement. Central solid divider remains visible. Both stand still on THEIR OWN steps; NEVER walk or change lanes. Never reverse camera axis.

SHOT 05 / 4.5–6.2s — ESSENTIAL TRANSITION. Close on fish inside balloon. Straight smooth PULLBACK from lower-landing axis; fish anchors frame as two lanes appear. Immediately make clear that girl is ALREADY ON THE UPPER PART OF SCREEN-LEFT DOWN ESCALATOR, moving toward camera, NOT ascending. Show tread direction and both handrails. The screen-right UP lane is also established. Keep camera looking uphill; show parallel rails converging toward the upper landing, making descent legible rather than implying she is rising. Balloon/string stay within left lane, off divider. Girl smiles privately AT HER FISH, not lens or king. Preserve this fish-close-up-to-escalator reveal; correct its spatial composition.

SHOT 06 / 6.2–7.6s — OPPOSING BOY REVEAL. Widen on SAME camera axis: king emerges on screen-RIGHT lane from the near lower foreground, carried UP toward distant upper landing. Show his back/rear 3/4 profile, golden crown and cream knit. His torso faces UPHILL, away from lens. Girl continues DOWN on screen LEFT. Same speed brings them level. Keep divider visible; slight lateral drift only.

SHOT 07 / 7.6–9.5s — GREETING. As they reach same elevation across divider, king keeps feet, hips and shoulders facing UPHILL; he turns HEAD to HIS LEFT to look at GIRL, NOT camera. With small real joy he says a cheerful "Hi!" Girl stops smiling at fish, glances at unexpected stranger, looks aside a moment, then back. After awkward beat she softly replies "...Hi" using @Audio2. Girl shy but not frightened; boy happy. Correct lip-sync/eye lines. Neither moves toward, leans over or reaches across divider. Minimal facial changes; no wrinkles on girl.

SHOT 08 / 9.5–10.0s — SEPARATE PASS. Keep riding in their OWN lanes: she goes DOWN toward camera, he goes UP away from it. No crossing/straddling/reaching/overlap across divider by bodies, clothing or balloon. Fish makes ONE smooth interested turn toward boy. CUT before awkward crossing/overlap. Keep the two moving stairways mechanically independent even during dialogue. Go to hill with a clear edit, NEVER morph escalators into grass.

BEAT 3 — TREE / 10.0–14.5s

SHOT 09 / 10.0–11.0s — ARRIVAL. @Image6 tree only. Grass-level camera ~35 cm high BEHIND pair facing tree. From behind: GIRL screen LEFT, KING screen RIGHT; his LEFT hand holds her RIGHT, and balloon in her LEFT hand floats far screen LEFT. Girl's tall 175 cm silhouette and child's small stature are readable. Gentle light through canopy. Slight upward camera rise to reveal the giant tree, real bark and moving leaves.

SHOT 10 / 11.0–12.0s — WIND GREETING. A sudden, brief real gust moves through the enormous canopy in a coherent wave. Leaves flutter, thicker branches respond later; grass, braids, shirt and knit respond in same wind direction. Balloon swings with tension and inertia, fish remains swimming in water. Camera TILTS from trunk UP into sunlit foliage and blue sky, motivated by sound. ONLY rush of real wind and leaves; tree NEVER speaks or has a face.

SHOT 11 / 12.0–13.8s — TREE'S POV. Camera HIGH IN LOWER BRANCHES, facing DOWN at pair about 60 degrees. Foreground leaves frame image; behind pair lie grassy hill and distant pale sky. From this FRONT high angle: KING screen LEFT, GIRL screen RIGHT, their hands joined. Balloon far screen RIGHT in girl's LEFT hand. They exchange ONE brief surprised glance, then look UP DIRECTLY into THIS lens (their gaze at the tree). Reply TOGETHER "Hi!": girl uses @Audio1, boy the SAME child voice. Tiny genuine smiles only; NO bouncing, big grin, shouting or brow lines. Goldfish turns smoothly and happily.

SHOT 12 / 13.8–14.5s — NATURAL LEAF WIPE. A softer second breeze rustles leaves and plays natural dappled light over them. A handful of foreground leaves crosses lens, covering it briefly to motivate transition to later rest. Neither tree speech nor music.

BEAT 4 — REST AT SAME TREE / 14.5–18.0s

SHOT 13 / 14.5–18.0s — CALM SINGLE TAKE, NO WIDE PULLBACK. After leaf wipe both ALREADY SIT supported by SAME TREE TRUNK from @Image6, in a restrained medium-wide shot. From front: KING screen LEFT, GIRL screen RIGHT; boy rests on HER RIGHT side. Girl's LEFT hand holds balloon at far frame RIGHT, fish drifting gently within it. Keep both faces clear and stable; no sudden pop into position. Boy lets out a SMALL, GENUINE sleepy yawn, blinking slowly. His shoulders lift then soften. Gradually he moves pelvis and TORSO slightly toward her; neck, chest, shoulders and spine follow as ONE gravity-led movement. His cheek gently meets HER RIGHT SHOULDER; his weight settles with subtle compression of her shirt and a small natural secondary motion. Her shoulder yields slightly and she shifts comfortably to support him. No robotic neck pivot/drop. She looks fondly at him with a tiny smile, slowly exhales, then looks at the balloon while feeling the gentle breeze rustling through the leaves. She quietly closes her eyes. Natural breeze/light and subtle balloon/fish movement. Hold frame to 18.0s. Reserve the final beat for stillness and a faint natural breeze; show neither character changing positions abruptly. End quietly. No flying vehicles, no extreme crane-out, no fade, and absolutely no music of any kind.

6. AUDIO OUTPUT — STRICTLY DIALOGUE + NATURAL SFX

NO MUSIC OF ANY KIND AT ANY TIME. No BGM, score, melody, pads, drones, rhythmic beds or musicalized wind. Generate DRY DIALOGUE + DIEGETIC SFX ONLY: the girl's fish "Hi", the boy's two "Hi" greetings, girl @Audio2 shy reply in Beat 2, girl @Audio1 warm reply in Beat 3, one short nonmusical transformation "poof", slight water sound, escalator motor/steps, realistic gust/rustling leaves, natural yawn and breathing, quiet outdoor ambience. Allow silence. Tree makes NO voice. Output MUST contain zero music.

7. FINAL PRIORITIES / NEGATIVE LOCK

Priority 1: MATCH @Image1 EXACTLY IN ALL FOUR BEATS, with no wrinkles/arm hair/baby hair/facial drift. Priority 2: LEFT escalator GIRL DOWN / RIGHT escalator KING UP with equal speed, full divider, ZERO lane crossing. Priority 3: restrained believable acting; @Audio2 = shy escalator, @Audio1 = warm tree. Priority 4: real fish, water, balloon/string physics. Priority 5: naturally weighted boy-to-shoulder lean ending CLOSE UNDER @Image6 tree. NO music, extra people, face on tree, talking tree, text, unrelated props, plastic skin, abrupt morphing or artificial random camera orbits.