{
 "S1sh3": {
  "input_fingerprint": "0f1e3212051f1753",
  "geometry": {
   "camera": {
    "look_target": {
     "x": 0.53,
     "y": 0.49
    },
    "origin": {
     "x": 0.53,
     "y": 0.54
    }
   },
   "entity_placements": [
    {
     "slot": "E1",
     "subject_en": "terrified injured running man",
     "x": 0.53,
     "y": 0.49
    }
   ],
   "rationale_ko": "남자는 A1 woodland floor area 중앙의 열린 탈출 동선에 두고 남쪽의 카메라를 향해 달리게 했다. 카메라는 남자의 정면 바로 앞에서 얼굴을 바라보며 짧은 거리를 유지해 타이트한 정면 초상을 만들고, 주변 B1 dense woodland undergrowth가 양쪽 프레임 가장자리를 압박하도록 배치했다."
  },
  "leakage": {
   "has_marker_leakage": false,
   "details_ko": "스케치에 다이어그램 기호나 텍스트 등 마커 유출이 없습니다."
  },
  "prompt": "Draw a single CAMERA-VIEW storyboard frame (16:9) of the scene staged\non the attached TOP-DOWN SITE PLAN, seen from the camera position\ngiven in the STAGING GEOMETRY data below.\n\nJurisdiction of sources — follow all three, they do not conflict:\n- The attached SITE PLAN is the SOT for what exists on the site and\n  how its zones and landmarks are arranged relative to each other.\n- The STAGING GEOMETRY data below is the SOT for XY POSITIONS and\n  horizontal heading ONLY: where each subject stands on that plan,\n  where the camera stands, and which way it looks. Distances,\n  directions and relative depths between subjects come from these\n  coordinates. Subjects outside the camera's horizontal field of\n  view are out of frame.\n- The CAMERA DIRECTION text below (when present) is the SOT for the\n  VERTICAL angle, elevation, tilt and framing (eye-level, high angle,\n  overhead crane, low angle, ...). If it asks for a high or overhead\n  angle, draw that angle — the site plan being top-down does NOT mean\n  the frame must be eye-level.\n\nSTYLE — LIGHT STORYBOARD FRAME: thin uniform dark outlines on plain\npaper. No shading, no hatching, no texture, no facial features beyond\na minimal head shape, no costume detail. Human figures are correctly\nproportioned full-body outlines in their canonical poses. Environment\nis reduced to a few placement lines (ground line, key landmark\nsilhouettes) — just enough to read positions and depth.\n\nSTAGING GEOMETRY — PLANNING DATA ONLY, NEVER RENDER IT: the lines\nbelow are normalized map coordinates on the attached site plan\n(x runs left→right 0..1, y runs top→bottom 0..1). Use them solely to\nplace the camera and subjects. Do NOT draw any of this data in the\nimage — no numbers, no coordinates, no labels, no arrows, no cones,\nno dots, no diagram symbols of any kind.\n- the camera stands at (x=0.53, y=0.54) and looks toward (x=0.53, y=0.49); subjects outside the camera's horizontal field of view are out of frame\n- terrified injured running man stands at (x=0.53, y=0.49)\n\nCAMERA DIRECTION (vertical angle / elevation / framing SOT): At eye level directly ahead of 숲속 사건의 남자 피해자, the camera tracks backward in a tight frontal portrait, keeping his wide eyes near the upper center as his running body pitches toward the lens. His injured movement produces restrained instability while dense undergrowth remains compressed at both frame edges, ending at the closest point before the hard cut.\n\nTHE LOCATION: 해 질 무렵의 숲속 바닥\n\nSHOT TEXT (authoritative, Korean): 눈을 크게 뜬 채 겁에 질린 숲속 사건의 남자 피해자의 얼굴 정면 구도\n\nCRITICAL — THE SITE PLAN IS A STAGING REFERENCE, NOT CONTENT: do NOT\ndraw the plan itself, and do NOT draw any diagram elements in the\nframe — no circles, dots, wedges, cones, arrows, callouts, codes or\nany other overlay symbols. The output is a clean storyboard drawing\nof the scene as seen by the camera only.\n\nNo text, no letters, no numbers, no captions anywhere in the image."
 },
 "S1sh4": {
  "input_fingerprint": "9942de12c384177f",
  "geometry": {
   "camera": {
    "look_target": {
     "x": 0.575,
     "y": 0.52
    },
    "origin": {
     "x": 0.79,
     "y": 0.31
    }
   },
   "entity_placements": [
    {
     "slot": "E1",
     "subject_en": "airborne silhouetted attacker",
     "x": 0.47,
     "y": 0.42
    },
    {
     "slot": "E2",
     "subject_en": "fallen injured man",
     "x": 0.68,
     "y": 0.63
    }
   ],
   "rationale_ko": "A1 woodland floor의 넓은 중앙 공터에 피해자를 우하단, 공중의 습격자를 좌상단에 두어 양팔을 뻗으며 대각선으로 덮치는 찰나를 구성했다. 카메라는 두 인물의 공격 축에 비스듬히 직교하는 우상단 쪽에 충분히 떨어져 배치해 허리 높이의 넓은 측면 구도가 되도록 했다."
  },
  "leakage": {
   "has_marker_leakage": false,
   "details_ko": "마커나 도면 기호 등의 유출 없이 장면이 정상적으로 그려졌습니다."
  },
  "prompt": "Draw a single CAMERA-VIEW storyboard frame (16:9) of the scene staged\non the attached TOP-DOWN SITE PLAN, seen from the camera position\ngiven in the STAGING GEOMETRY data below.\n\nJurisdiction of sources — follow all three, they do not conflict:\n- The attached SITE PLAN is the SOT for what exists on the site and\n  how its zones and landmarks are arranged relative to each other.\n- The STAGING GEOMETRY data below is the SOT for XY POSITIONS and\n  horizontal heading ONLY: where each subject stands on that plan,\n  where the camera stands, and which way it looks. Distances,\n  directions and relative depths between subjects come from these\n  coordinates. Subjects outside the camera's horizontal field of\n  view are out of frame.\n- The CAMERA DIRECTION text below (when present) is the SOT for the\n  VERTICAL angle, elevation, tilt and framing (eye-level, high angle,\n  overhead crane, low angle, ...). If it asks for a high or overhead\n  angle, draw that angle — the site plan being top-down does NOT mean\n  the frame must be eye-level.\n\nSTYLE — LIGHT STORYBOARD FRAME: thin uniform dark outlines on plain\npaper. No shading, no hatching, no texture, no facial features beyond\na minimal head shape, no costume detail. Human figures are correctly\nproportioned full-body outlines in their canonical poses. Environment\nis reduced to a few placement lines (ground line, key landmark\nsilhouettes) — just enough to read positions and depth.\n\nSTAGING GEOMETRY — PLANNING DATA ONLY, NEVER RENDER IT: the lines\nbelow are normalized map coordinates on the attached site plan\n(x runs left→right 0..1, y runs top→bottom 0..1). Use them solely to\nplace the camera and subjects. Do NOT draw any of this data in the\nimage — no numbers, no coordinates, no labels, no arrows, no cones,\nno dots, no diagram symbols of any kind.\n- the camera stands at (x=0.79, y=0.31) and looks toward (x=0.57, y=0.52); subjects outside the camera's horizontal field of view are out of frame\n- airborne silhouetted attacker stands at (x=0.47, y=0.42)\n- fallen injured man stands at (x=0.68, y=0.63)\n\nCAMERA DIRECTION (vertical angle / elevation / framing SOT): After the hard cut, a waist-height side view holds a wide slice of space around 숲속 사건의 남자 피해자, with his fallen body in the lower right and 정체불명의 습격자 suspended across the upper left-to-center. The frame freezes the peak of the attacker's diagonal flight, both arms reaching toward the victim, while the wider camera distance becomes the sole strong break from the preceding close portrait.\n\nTHE LOCATION: 해 질 무렵의 숲속 바닥\n\nSHOT TEXT (authoritative, Korean): 공중에 뜬 검은 실루엣의 정체불명의 습격자가 바닥에 쓰러진 숲속 사건의 남자 피해자를 향해 양팔을 뻗은 찰나\n\nCRITICAL — THE SITE PLAN IS A STAGING REFERENCE, NOT CONTENT: do NOT\ndraw the plan itself, and do NOT draw any diagram elements in the\nframe — no circles, dots, wedges, cones, arrows, callouts, codes or\nany other overlay symbols. The output is a clean storyboard drawing\nof the scene as seen by the camera only.\n\nNo text, no letters, no numbers, no captions anywhere in the image."
 }
}