PUT YOURSELF
IN THE SCENE.

CHARACTER RECAST / SOURCE-ACTOR TRACKING

Use the uploaded SOURCE VIDEO as the absolute master reference for motion, choreography, timing, camera movement, framing, lens perspective, depth, occlusion, environment and shot progression.

This is a MAIN-SUBJECT IDENTITY REPLACEMENT task. Preserve the existing guard cast and scene.

SOURCE-SPECIFIC PERFORMANCE MAP — AMMAR'S SUPPLIED VIDEO

The uploaded source is already a generated scene. Use THIS uploaded video as the only motion and scene reference; no other original clip is required.

Observed sequence (approximate landmarks for orientation, never replacement keyframes or new timing):
- Opening: a tight front-facing view of the white-shirt performer with a cigarette held near the mouth in the right hand. Smoke is visible as the head lowers and turns. The guards behind are softer in focus.
- Early section: the same performer turns away from camera, showing the back of the white shirt, then moves deeper between the guard rows. Preserve the exact route, shoulder turns, arm lifts, hand gestures and changing occlusions.
- Middle section: the performer turns back toward camera within the formation, lowers the head, raises the cigarette hand toward the mouth again, then lowers the hand. Preserve the source's precise hand-to-mouth contact, pauses and head movement.
- Later section: the white-shirt performer remains comparatively upright while the surrounding guards begin their source-specific arm/lapel gestures, head rolls, leans and deep forward bows. The guards do not all occupy the same phase at once. Preserve each person's individual timing.
- Ending: the view has widened to show the performer, multiple rows of moving guards, stone steps, the large brick archway, pale stone trim and dark vertical panels. Retain the source's ending pose and frame; do not invent a new ending.

These descriptions help identify the source performance. The actual uploaded video is authoritative for every frame, movement, onset, duration, direction, camera change and detail. Do not substitute an approximation based on this text.

CIGARETTE / HAND / SMOKE CONTINUITY:
The cigarette is a performance prop, not part of the person's identity. Preserve it through the identity replacement.
Match the source hand (right hand where visible), finger grip, cigarette position, angle, scale, hand-to-mouth movement, mouth contact, lowering of the hand and subsequent visible holding position.
Preserve visible smoke at the source moments, with the same location, direction, density and dissipation as closely as possible.
Do not omit the cigarette, swap it between hands, turn it into another object, add a second cigarette, add extra smoking gestures or make it float independently of the hand.
When the hand or cigarette is occluded or outside the frame, preserve that visibility state; do not expose it for clarity or invent hidden action.
Maintain anatomically coherent fingers, lips and contact throughout.

CAMERA / ENVIRONMENT / LIGHTING:
Preserve the continuous widening from the close opening to the broad view of the formation and architecture. Match the actual source trajectory, rate of pull-back, perspective, focus changes and subject scale; do not invent a dolly, orbit, cut or zoom beyond what the source shows.
Keep the stair geometry, rows, brick arch, pale trim, dark panels, lighting direction, muted color palette, contact shadows and source depth of field. Do not replace the setting with a studio or the white image-reference background.



==================================================
REFERENCE MAP — DO NOT MIX THESE ROLES
==================================================

@image_1 = MAIN SUBJECT — front identity reference
@image_2 = SAME MAIN SUBJECT — left-profile identity reference
@image_3 = SAME MAIN SUBJECT — right-profile identity reference
@image_4 = SURROUNDING CROWD STYLE reference ONLY

@image_1, @image_2 and @image_3 show the SAME PERSON.

Use those three images together to maintain one consistent main identity across front-facing, profile, rear-facing, moving and partially occluded views.

@image_4 does NOT represent the main subject.
Never transfer facial identity from @image_4 onto the main subject.

==================================================
1. IDENTIFY AND TRACK SOURCE ACTOR A
==================================================

In the ORIGINAL SOURCE VIDEO, identify exactly ONE performer as SOURCE ACTOR A.

SOURCE ACTOR A is the original performer who:
- wears a white dress shirt and black necktie
- wears black trousers
- has dark wavy hair, facial hair and clear metal-framed eyeglasses
- holds a cigarette in the right hand and brings it toward the mouth
- wears rolled shirt sleeves and a wristwatch
- appears in a tight front-facing opening view, then turns their back toward camera
- moves through the surrounding crowd of black-suited guards wearing dark sunglasses

IMPORTANT:
These traits describe SOURCE ACTOR A ONLY so the correct original performer can be tracked.

THEY ARE NOT APPEARANCE INSTRUCTIONS FOR THE REPLACEMENT.

Do NOT copy SOURCE ACTOR A's:
- face
- hairstyle
- clothing
- body appearance
- identity

Track this exact ORIGINAL performer continuously throughout the entire source video.

Maintain the track when SOURCE ACTOR A:
- moves away from screen center
- turns sideways
- turns their back
- walks through the formation
- passes behind another performer
- becomes partially occluded
- becomes temporarily difficult to see
- changes depth within the crowd
- returns toward camera

NEVER identify the main character by whichever performer happens to occupy the center of the frame.

SOURCE ACTOR A is a continuous source-video identity track.

==================================================
2. REPLACE SOURCE ACTOR A WITH THE MAIN SUBJECT
==================================================

Replace SOURCE ACTOR A, and ONLY SOURCE ACTOR A, with the person represented by @image_1 + @image_2 + @image_3.

ABSOLUTE ONE-PERSON CONSTRAINT:

There must be EXACTLY ONE main-subject identity mapped to SOURCE ACTOR A.

The identity from @image_1–@image_3 is forbidden from appearing on any other performer.

Never:
- duplicate the main subject
- generate a second white-shirt version of the main subject
- place the main identity in both foreground and background
- assign the main identity to another performer after an occlusion
- create another main subject because a different performer enters screen center
- allow the main identity to leak into the surrounding crowd

If SOURCE ACTOR A is hidden, the replacement must also be hidden.

If SOURCE ACTOR A leaves a visible area, do not invent another copy elsewhere.

After every occlusion, reconnect the replacement to the SAME tracked SOURCE ACTOR A.

MAIN SUBJECT APPEARANCE:

Use ONLY @image_1, @image_2 and @image_3 to determine the main subject's identity.

Preserve:
- facial geometry
- facial proportions
- skin tone
- hairstyle
- facial hair if present
- body proportions
- apparent age
- recognizable likeness

Preserve the wardrobe shown in the generated subject references:

- white dress shirt
- black necktie
- tailored black trousers
- black belt
- black dress shoes

The white-shirt / black-tie outfit belongs ONLY to the main subject.

Keep the replacement’s accessories consistent with @image_1, @image_2 and @image_3. Do not inherit the source performer’s glasses unless they are also present in the replacement references.

Do NOT make the replacement resemble SOURCE ACTOR A.

Do NOT add sunglasses to the main subject unless sunglasses are clearly and consistently present in all three main-subject references.

MOTION LOCK:

The replacement must inherit SOURCE ACTOR A's exact:
- body position
- walking path
- body rotation
- head direction
- cigarette interaction, finger grip and hand-to-mouth contact
- visible smoke timing and motion
- gestures
- arm movement
- leg movement
- posture
- timing
- screen position
- scale
- depth
- occlusion

Do not alter the original performance simply to show the replacement's face.

If SOURCE ACTOR A faces away, the replacement must face away.

==================================================
3. PRESERVE EVERY OTHER HUMAN AS SOURCE GROUP B
==================================================

EVERY other human in the uploaded video belongs to SOURCE GROUP B: the existing black-suited guard crowd.

Preserve each guard as the same distinct source individual, with stable face, hair, skin tone, body build, clothing and accessories throughout. Do not unnecessarily regenerate or redesign their identities.

This applies to foreground performers, left and right edges, middle rows, rear rows, distant performers and partially visible performers.

@image_4 is a secondary appearance reference for the shared guard styling ONLY. It must not replace the source choreography, source arrangement or each guard's individual identity. Do not clone a face from @image_4 across the crowd.

If the video model reconstructs any surrounding performer, anchor the reconstruction to that SAME source guard, maintaining source likeness, position, depth, motion and continuity.

Never transfer the replacement main identity onto any guard.

==================================================
4. SURROUNDING CROWD APPEARANCE
==================================================

Maintain SOURCE GROUP B as the coordinated black-suited guard unit already visible in the uploaded video. Preserve the existing variation in their clothing and accessories; use @image_4 to reinforce the shared style, not to overwrite individual details.

Wardrobe and styling:
- fitted matte-black or charcoal suits
- black dress shirts
- black neckties
- black dress shoes
- dark sunglasses
- discreet clear communication earpieces
- polished, controlled, intimidating professional styling

NO weapons.
NO gun holsters.
NO firearms.
NO tactical weapons.

CRITICAL CROWD IDENTITY RULE:

Every surrounding performer must remain a DIFFERENT individual person.

They belong to the same organization but must NOT look like clones.

Preserve the existing natural variation between surrounding performers in:
- facial proportions
- jaw shape
- nose shape
- brows
- face width
- hairstyles
- hairlines
- skin tones
- body builds
- facial hair where appropriate

The shared uniform creates visual unity.
Their faces must remain individually distinct.

==================================================
5. PRESERVE EACH CROWD PERFORMER'S MOTION
==================================================

Track every guard as ONE specific original SOURCE GROUP B performer.

Each guard, including any reconstructed guard, must preserve that source performer's exact:
- position
- pose
- orientation
- head direction
- bow
- lean
- turn
- arm movement
- leg movement
- timing
- spacing
- scale
- depth
- occlusion

@image_4 defines APPEARANCE ONLY.

THE SOURCE VIDEO defines MOTION.

Do NOT freeze the agents into the standing pose visible in @image_4.

Do NOT force the agents to face camera when their corresponding source performers are turning, leaning or bowing.

Do NOT synchronize performers whose source movements are not synchronized.

==================================================
6. SOURCE VIDEO LOCK
==================================================

Preserve the source video's exact:
- camera trajectory
- camera movement
- zoom / pull-back progression
- framing progression
- shot duration
- playback speed
- lens perspective
- choreography
- performer count
- performer positions
- visible architecture
- visible floor and background structures
- background
- lighting direction
- scene layout

Do not redesign the environment.

Do not add new performer positions.

Do not remove existing performer positions.

Replace ONLY SOURCE ACTOR A’s identity. Preserve SOURCE GROUP B and the scene.

==================================================
7. DEPTH, OCCLUSION AND TEMPORAL CONSISTENCY
==================================================

Preserve the original depth ordering frame by frame.

When one source performer passes in front of another, preserve the same occlusion relationship.

Maintain stable casting before, during and after every occlusion.

Do NOT allow:
- protagonist duplication
- protagonist identity switching
- protagonist identity leaking onto crowd performers
- crowd performers turning into the protagonist
- repeated clone faces across the crowd
- characters jumping between rows
- sudden scale changes
- merged bodies
- duplicated limbs
- missing limbs
- floating feet
- face morphing
- clothing flicker
- sunglasses flicker
- earpiece flicker

Match the source lighting and contact shadows naturally.

==================================================
FINAL CAST REQUIREMENT
==================================================

The finished video must look as though THIS uploaded performance was filmed again with the reference subject in place of the original white-shirt performer, while the same guards perform the same choreography.

There is ONE and only ONE main subject from:

@image_1 + @image_2 + @image_3

wearing the white dress shirt / black tie / black trousers outfit.

Every other human is a distinct black-suited covert agent styled according to @image_4.

The original:
- performance
- choreography
- camera
- timing
- depth
- architecture
- environment

must remain recognizable as the same source video.

PRIORITY ORDER:

1. NEVER duplicate the main subject.
2. Track the original white-shirt, black-tie cigarette performer as SOURCE ACTOR A. Source hair, face, eyeglasses and wristwatch identify the track ONLY; @image_1–@image_3 define replacement identity and accessories. Keep the cigarette, smoking gestures and smoke as source performance details.
3. Use @image_1–@image_3 ONLY for the replacement's physical appearance.
4. Use @image_4 ONLY for surrounding-crowd styling.
5. Keep every surrounding performer a distinct person.
6. Preserve all source choreography, timing, depth and occlusion.
7. Preserve camera movement and environment exactly.

OUTPUT:
Match the source video's 16:9 aspect ratio, approximately 30-second duration (file duration 30.04 seconds), 24 fps and framing. The supplied file is 1280 × 720. Use the exact source timeline, with no slow motion, speed changes, interpolated extra action or reframing.

Preserve the original audio track if the workflow supports source-audio retention. Otherwise keep the source audio for the final edit and align it to the unchanged video timeline; do not invent speech or replace the soundtrack.

Photorealistic live-action result.

No text.
No watermark.
No cartoon styling.
No scene redesign.

HARD CONSTRAINT: @image_1, @image_2 and @image_3 may map to SOURCE ACTOR A only. Zero other performer may use this identity in any frame. Never infer protagonist identity from screen position.