{
 "task_id": "20260928_0206_竖屏MV_S11",
 "requirement": "S11 竖屏镜头",
 "params": {
  "width": 768,
  "height": 1344,
  "duration": 5,
  "seed": 883051,
  "steps": 8,
  "fps": 24.0
 },
 "vision_description": "（外部传入提示词，未识图）",
 "prompt": "subject_definitions:\n<Subject 1> is the young swordswoman in <Picture 1>, with long wavy black hair, wearing a sleeveless long purple vest with gold piping over a white short-sleeved top, a yellow-orange sash, purple wrist guards, white trousers and purple boots. She carries a pinkish-red long whip.\n<Subject 2> is the young swordsman in <Picture 2>, with long dark hair in a high topknot bound with a white band, wearing a brown crossed-collar robe over a white inner garment, with a long sword carried on his back.\n<Picture 1>: The reference sheet shows the SAME single person from three angles (front, side, back) — the three figures are ONE person, not three different people.\n<Picture 2>: The reference sheet shows the SAME single person from three angles (front, side, back) — the three figures are ONE person, not three different people.\n\nsummary:\n[reference generation] A 5-second vertical 9:16 cinematic shot for a Chinese Paladin music video. <Subject 1> supplies the identity of the L/X characters. <Subject 2> supplies the identity of the L/X characters.\n\nretention_analysis:\n<Subject 1> (appears in [Shot 1]): fully_preserved - face, hairstyle and costume kept exactly as in <Picture 1>.\n<Subject 2> (appears in [Shot 1]): fully_preserved - face, hairstyle and costume kept exactly as in <Picture 2>.\n<Picture 1> ([Shot 1] identity anchor): partially_preserved - the three views are the same character; only identity and costume details are taken, never a group of people.\n<Picture 2> ([Shot 1] identity anchor): partially_preserved - the three views are the same character; only identity and costume details are taken, never a group of people.\n\ndetailed_description:\n[Shot 1] A rain-damp stone quay at golden hour. The swordswoman and the swordsman stand very close, blades just lowered, locking eyes; a single warm key light rakes across both faces while petals and water spray drift slowly. Camera slowly orbits them. Extremely restrained motion, almost a frozen moment. The shot lasts 5 seconds. Vertical 9:16 framing. Style: semi-realistic Chinese fantasy cinematic CGI, volumetric light, glowing mist, soft bloom, cobalt blue and teal palette with honey-gold accents. No text, no watermark, no extra people.\n\noverall_soundscape:\nDistant water lapping, a faint metallic ring fading, soft wind.\n\nnon_diegetic_music:\nA soft, restrained orchestral pad in a minor key, very low in the mix, leaving room for the song.",
 "video_url": "http://192.168.31.76:8899/10-视频生成流水线/20260928_0206_竖屏MV_S11/20260928_0206_竖屏MV_S11_h3v1.mp4",
 "time": "2026-09-28 02:14:44"
}