Genjutsu Trend Video

Template prompt untuk digunakan semula

{
  "task": "Generate a photorealistic video using the supplied video as the exact motion and camera reference, with character and environment replacements defined by the three reference images.",
  "output": {
    "duration": "Match the full duration of @Video1.",
    "frame_rate": null,
    "framing": "Match the aspect ratio and crop of @Video1.",
    "visual_style": "Photorealistic live-action with natural movement and consistent character identities."
  },
  "references": {
    "@Video1": "Exact reference for camera movement, shot progression, composition, choreography, timing, walking routes, subject positions and occlusion.",
    "@Image1": "Character sheet for the main character. Exact reference for facial identity, hairstyle, physique, clothing, footwear and any visible markings or accessories.",
    "@Image2": "Exact identity, appearance and clothing references for the surrounding people.",
    "@Image3": "Exact reference for the replacement location, architecture, surfaces, materials and environmental details."
  },
  "reference_rules": {
    "motion_and_composition": "Follow the video frame by frame.",
    "character_appearance": "Follow @Image1 for the main character and @Image2 for surrounding people.",
    "environment": "Follow @Image3.",
    "action_exception": "Replace smoking-related actions with thumbs-up gestures while preserving all other source movement.",
    "still_image_usage": "Use the images as appearance references. Do not reproduce their static poses, reference-sheet layouts or framing."
  },
  "main_character": {
    "source_character": "The foreground man who starts with his back toward the camera, walks into the formation, turns forward and remains comparatively upright while the surrounding men perform their choreography.",
    "tracking": "Identify the source character through their continuous movement trajectory. Replace them with one consistent main character from @Image1 throughout the video.",
    "appearance": {
      "face": "Match the close-up portrait in @Image1, including facial structure, eyes, eyebrows, nose, mouth, jaw, ears and skin tone.",
      "hair": "Match the hairstyle, hairline and hair color in @Image1.",
      "physique": "Match the physique and body proportions in @Image1.",
      "markings": "Preserve any visible tattoos or distinctive markings in @Image1, including their anatomical placement during movement. Do not invent markings.",
      "clothing": "Match the complete outfit in @Image1, including colors, materials, fit and any visible logos, names or numbers. Do not invent text or numbers.",
      "footwear": "Match the footwear shown in @Image1.",
      "accessories": "Use only accessories visible in @Image1.",
      "sheet_interpretation": "Use full-body panels for clothing and body reference and the close-up portrait for facial identity. If a panel has its face removed, do not reproduce that omission. Render one complete, anatomically correct person."
    },
    "performance": {
      "movement": "Match the source character's walking route, turns, head orientation, pauses, body trajectory and timing, adapted naturally to the main character’s physique.",
      "position": "Preserve the main character’s screen position, depth, scale, overlaps and final standing position throughout.",
      "expression": "Retarget the source expressions naturally to the main character’s face, except smoking-related mouth movements.",
      "group_relationship": "The main character occupies the source character’s place among the surrounding people and follows the source character’s comparatively upright performance during the group choreography.",
      "uniqueness": "Exactly one main character. Do not duplicate them among the surrounding people."
    },
    "uniform_continuity": "Maintain the same outfit throughout. Keep any reference lettering or numbers correct, unmirrored and attached naturally to the moving fabric. Show back details only when the back is visible."
  },
  "thumbs_up_action": {
    "instruction": "Replace every smoking-related action with a natural one-handed thumbs-up.",
    "gesture": "Raise an empty hand to chest or shoulder height, curl four fingers into the palm and extend the thumb upward. Keep the hand away from the mouth and the face unobstructed.",
    "timing": "Begin, hold and release each thumbs-up within the corresponding smoking-action interval in the reference video.",
    "hand_selection": "Use the source character's active hand for the corresponding gesture. Preserve anatomical handedness through turns.",
    "transition": "Raise the hand smoothly, briefly hold the thumbs-up and lower it naturally as the source action finishes.",
    "other_hand": "Follow the source's non-smoking movement. If the other hand participates in smoking, replace that movement with a relaxed natural position.",
    "remove_completely": [
      "Cigarettes",
      "Lighters",
      "Flames",
      "Smoke",
      "Ash",
      "Cigarette-pinching gestures",
      "Smoking-related hand-to-mouth contact",
      "Inhaling or exhaling smoke"
    ],
    "motion_boundary": "Change only the hand, arm and mouth motion needed for this replacement. Preserve the remaining body movement and camera timing."
  },
  "surrounding_people": {
    "source_characters": "Every surrounding person in @Video1 except the tracked main character.",
    "replacement_identities": "Use the identities shown in @Image2.",
    "identity_details": "Replace the complete faces, head shapes, hairstyles, facial hair, skin tones and physiques. Do not retain the source crowd's faces while merely changing their clothing.",
    "mapping": "Assign a reference person to each source person track and maintain that identity throughout. Use all available identities before repeating any for additional tracks.",
    "uniform": "Match the clothing shown in @Image2. Keep newly visible clothing details consistent with those references.",
    "numbering": "Preserve any names, numbers or markings actually shown in @Image2. Do not apply the main character’s unique details to surrounding people.",
    "formation": "Match the video's exact person count, rows, spacing, relative scale, depth, edge crops and overlaps.",
    "choreography": "Transfer each person's individual forward bends, sideways leans, rises, upward head tilts, bent-arm gestures and subtle movements at the same speed and timing as the video.",
    "timing_variation": "Preserve the timing differences between individuals and rows rather than replacing them with a generic synchronized loop.",
    "continuity": "Maintain stable identities and clothing through bending, occlusion and reappearance."
  },
  "environment": {
    "location": "The location shown in @Image3.",
    "features": [
      "Match the architecture, surfaces, materials, colors and distinctive environmental features in @Image3."
    ],
    "replacement_scope": "Replace the source environment with the location in @Image3 while preserving the performers’ positions and routes.",
    "performer_integration": "Align the new standing surfaces and walking spaces with the source performers’ positions and routes. Provide sufficient clearance for their choreography without relocating them.",
    "physical_contact": "Keep feet grounded on the appropriate surfaces. Prevent bodies, limbs and footwear from intersecting environmental objects.",
    "perspective": "Maintain coherent environmental depth and parallax throughout the reference camera movement.",
    "framing": "Reveal the replacement environment only within the reference video’s framing. Do not reposition the camera to display additional scenery.",
    "lighting": "Use lighting appropriate to @Image3 with consistent skin tones, fabric shading, contact shadows and environmental illumination."
  },
  "camera": {
    "opening": "Match the rear-facing foreground view, crop and focus relationship in the reference video's first frame.",
    "progression": "Match the progressive widening as the main character moves into the formation and more of the group becomes visible.",
    "motion_lock": "Reproduce the reference camera's translation, rotation, zoom, framing drift, acceleration and any existing shot transitions at the same moments.",
    "optical_behaviour": "Match the focus progression, depth of field and motion blur.",
    "restrictions": "Do not add cuts, close-ups, orbit shots, slow motion, speed ramps, pans or alternative ending shots."
  },
  "sequence_guide": {
    "timing_priority": "Follow the actual frame timing in @Video1. Do not impose fixed timestamps.",
    "opening": "The main character starts with their back toward the camera and follows the source route toward the formation as the framing widens.",
    "entry": "The main character moves between the surrounding people, becomes partially obscured and turns toward the camera according to the reference.",
    "settling": "The main character settles into the group. Replace any smoking-related hand-to-mouth actions with thumbs-up gestures at the corresponding moments.",
    "group_performance": "Preserve individual subtle movements, repeated bends, sideways leans, rises and upward head movements. The main character follows the tracked source character’s comparatively upright performance while the camera continues its source progression."
  },
  "audio": {
    "instruction": "Preserve the reference soundtrack and timing if supported. Remove identifiable smoking-related effects. Do not add dialogue or narration."
  },
  "negative_prompt": [
    "Source crowd faces retained",
    "Clothing recoloring without identity replacement",
    "Clothing that does not match the relevant reference image",
    "Smoking, cigarettes, lighters, flames or smoke",
    "Smoking-related hand-to-mouth gestures",
    "Incorrect, invented or mirrored reference lettering or numbers",
    "Duplicated main character",
    "Identity changes between frames",
    "Changed person count or rearranged rows",
    "Frozen poses from the reference images",
    "Generic replacement choreography",
    "Additional camera movement or cuts",
    "Original environment retained instead of @Image3",
    "Flat sliding background",
    "Floating feet or intersections with environmental objects",
    "Distorted hands or extra fingers",
    "Reference-sheet panels",
    "Added captions or watermarks"
  ]
}