{"file_name": "videos/sample_001.mp4", "sample_id": "#1", "prompt": "subject_definitions:\n is Cheng Xing, a male character whose appearance and clothing are from .\n is Wang Shouming, a male character whose appearance and clothing are from .\n defines the interior setting of Xingchen Mobile Phone Store.\n\nsummary:\n[reference generation] A 14-second video set inside the Xingchen Mobile Phone Store, featuring Cheng Xing and Wang Shouming in a tense repair scene. The video consists of four shots with specific camera movements and lighting, maintaining a cold white tone and high-end visual quality.\n\nretention_analysis:\n (Cheng Xing) is fully preserved as the main character appearing in shots 1, 2, and 4, with his appearance grounded in . (Wang Shouming) is fully preserved as the second character appearing in shots 1 and 3, with his appearance grounded in . The scene setting from is fully preserved as the background for all shots.\n\ndetailed_description:\nThroughout all shots, strictly observe the 180-degree axis rule and eyeline matching; prevent axis-crossing jump cuts.\n\n[Shot 1] A full shot at eye level with a slow push in, lasting 0-3s. The scene is set inside the Xingchen Mobile Phone Store, with cluttered tools scattered on the counters. The lighting is bright incandescent with a cold white tone, creating a slightly anxious atmosphere. (Cheng Xing) sits at the glass repair desk on the left side of the frame, his hands fiddling with scattered parts on the glass desk. (Wang Shouming) stands outside the right repair counter, leaning his body forward and staring at the desk. The space is filled with cold white overhead light. The image is clean and high-end, with clear foreground, midground, and background progression, exhibiting extreme emotional tension and physical detail realism. The frame must contain no watermarks, no subtitles, and no logos. No dialogue; faint metal-part friction is heard.\n\n[Shot 2] At 00:03.000, a close-up shot from a high angle looking down with a slight handheld shake, lasting 3-5s. The lighting remains bright incandescent with a cold white tone. (Cheng Xing) is still sitting at the counter. The camera focuses on his hands; his knuckles are slightly white from tension. He holds a metal screwdriver and scrapes it hard across the motherboard, leaving a glaring white mark. Fine beads of sweat secrete on 's forehead. No dialogue; a harsh metallic scraping sound is heard.\n\n[Shot 3] At 00:05.000, a medium close-up at eye level with a fixed camera, lasting 5-10s. The lighting is bright incandescent with a cold white tone. (Wang Shouming) is still standing outside the right counter. He irritably raises his left hand, frequently looking at his watch. His eyebrows are tightly furrowed into a deep frown with three distinct vertical lines, the corners of his mouth are pressed down, and his facial muscles are slightly tense from impatience. He shoots dissatisfied glances towards on the left side of the frame. (S1) says [Chinese]我这手机,林师傅以前5分钟搞定,你都修仨小时了! in an impatient and complaining voice. Clothing friction is heard.\n\n[Shot 4] At 00:10.000, a medium close-up at eye level with a fixed camera, lasting 10-14s. The lighting is bright incandescent with a cold white tone. (Cheng Xing) is at the left counter. He suddenly looks up towards the right off-screen, sweat sliding down his cheek. He forcibly straightens his back, trying to maintain his confidence, his chest heaving. His eyes glare wide in a bluff, and he loudly defends himself. (S2) says [Chinese]急什么!专业维修,讲究精细! in a forced, raised-volume voice. No additional specific SFX is requested in this last shot.\n\noverall_soundscape:\nThe soundscape includes slight metal parts friction in the first shot, a harsh metal scraping sound in the second shot, clothing friction in the third shot, and no specific sound effects in the fourth shot.\n\nnon_diegetic_music:\nN/A", "original_prompt": "人物站位: @程星 坐在左侧的玻璃维修台前低头忙碌, @王守明 站在右侧柜台外等待,场景内杂乱的工具散落在柜台上。\n要求:画面干净又高级;画面里前中远景递进关系分明。极具情绪张力与物理细节真实感。\n画面:无水印,无字幕,无logo 。\n镜号:1 | 场次:日 内 | @星辰手机店-内部 | 0-3s | 全景,平视 | 缓慢推进 | 明亮的白炽灯光,冷白调, @程星 坐在画面左侧的柜台前,双手在玻璃台上捣鼓着散落的零件, @王守明 站在右侧的维修柜台外,身体前倾盯着桌面。整个空间充斥着冷白色的顶光,气氛略显焦躁。 | 台词 & 音效:Dialogue:无 SFX:轻微的金属零件摩擦声 | 备注:严格遵守180度轴线与视线匹配原则,严防越轴跳切\n镜号:2 | 场次:日 内 | @星辰手机店-内部 | 3-5s | 特写,俯视 | 手持轻微晃动 | 明亮的白炽灯光,冷白调, ( @程星 依然坐在柜台前)镜头聚焦在 @程星 手中,他的指节因为紧张而微微泛白,握着的金属螺丝刀在主板上狠狠刮过,划出一道刺眼的白痕, @程星 的额头渗出细密的汗珠。 | 台词 & 音效:Dialogue:无 SFX:刺耳的金属刮擦声 | 备注:严格遵守180度轴线与视线匹配原则,严防越轴跳切\n镜号:3 | 场次:日 内 | @星辰手机店-内部 | 5-10s | 近景,平视 | 固定镜头 | 明亮的白炽灯光,冷白调, ( @王守明 仍站在右侧柜台外) @王守明 烦躁地抬起左手,频频看着腕表。他的眉头紧紧皱成一个川字,嘴角下压,面部肌肉因不耐烦而微微紧绷,带着不满的目光射向画面左侧的 @程星 。 | 台词 & 音效:Dialogue:[用不耐烦的声音抱怨] @王守明 (不耐烦/抱怨):\"我这手机,林师傅以前5分钟搞定,你都修仨小时了!\" SFX:衣物摩擦声 | 备注:严格遵守180度轴线与视线匹配原则,严防越轴跳切\n镜号:4 | 场次:日 内 | @星辰手机店-内部 | 10-14s | 近景,平视 | 固定镜头 | 明亮的白炽灯光,冷白调, ( @程星 在左侧柜台前) @程星 突然抬起头看向右侧画外,额头上的汗水顺着脸颊滑落。他强行挺直脊背试图维持底气,胸口起伏不定,眼神虚张声势地瞪大,大声为自己辩解。 | 台词 & 音效:Dialogue:[用硬撑的声音喊] @程星 (硬撑/拔高音量):\"急什么!专业维修,讲究精细!\" SFX:无 | 备注:严格遵守180度轴线与视线匹配原则,严防越轴跳切\n\n参考图片说明:\n图1: 角色: 程星\n图2: 角色: 王守明\n图3: 场景: 星辰手机店-内部", "reference_images": ["references/2d77c4e30b91ffc41d2add18908941fa46008c29ec40dabf8f7488e9632224e4.jpg", "references/d99e209bbcf8194dc2a19158e6af21465e02ec6dd266d83605f0380ebe3e77e7.jpg", "references/a2bea38d15a2caab0857d5bb864423979b18f9289bdaab5e12a5d61f6888b45b.jpg"], "reference_sha256": ["2d77c4e30b91ffc41d2add18908941fa46008c29ec40dabf8f7488e9632224e4", "d99e209bbcf8194dc2a19158e6af21465e02ec6dd266d83605f0380ebe3e77e7", "a2bea38d15a2caab0857d5bb864423979b18f9289bdaab5e12a5d61f6888b45b"], "workflow": "ref2va", "prompt_sha256": "90e381777c230731e2f6458a2aaff9bf5f4113cd53c784b01dd2ac501efa91ca", "original_prompt_sha256": "44502408aefdb1a896ace802c0ce276321f335ef086658863683c6c1a1f5bf3c", "video_sha256": "28fb21bf358c4da9e572bc19e5f21d08e06f6139edf84c4d305cab3aab65f675", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_001.json", "verification": "verification/sample_001.json"} {"file_name": "videos/sample_002.mp4", "sample_id": "#2", "prompt": "subject_definitions:\n: Fina, appearance and clothing from .\n: image (Helena), appearance and clothing from .\n\nsummary:\n[reference generation] A three-shot sequence in Helena's office (). Shot 1 features an extreme macro close-up of 's fingers sliding up 's collarbone to pinch her chin, with 's pose referencing . Shot 2 is a close-up of gently kissing 's lips (reference ) before retreats. Shot 3 is a close-up where aggressively pulls 's silk tie to initiate a forceful kiss, with 's pose, action, and final frame referencing .\n\nretention_analysis:\n (Fina) is referenced via for identity, for pose in Shot 1, and for pose, action, styling, and final frame in Shot 3. (image/Helena) is referenced via for identity. The scene is referenced via . is referenced for the kiss style in Shot 2. All subject identities, scene, and specific pose/action references are fully preserved as requested.\n\ndetailed_description:\n[Shot 1] At 00:00.000, extreme macro close-up, level shot, slow rising camera (pedestal/crane up). Daytime natural side lighting, clear high-key warm white tone. 's slender, pen-holding fingers, like cool jade, slide upward from 's sunken collarbone. The delicate skin shivers with the touch of the fingertips, finally the thumb and index finger forcefully pinch 's chin. 's pose references .\n\n[Shot 2] At 00:03.000, close-up, low angle, fixed camera. Backlit sunlight, clear high-key warm white tone. 's hand lifts 's chin, her cold eyes revealing a rare hint of indulgence. Her red lips, like a dragonfly touching water, leave an extremely gentle kiss on the edge of 's open lips, referencing . slowly retreats.\n\n[Shot 3] At 00:06.000, close-up, level shot, fast push. Backlit sunlight, high saturation warm white tone, golden hair rim effect.\nInitial state (1 second before action): sits upright in a black leather boss chair, body straight, black silk tie hanging naturally in the center of her chest. leans on the edge of the desk, upper body leaning forward, about 30 centimeters from , eyes burning with fanatical desire, staring fixedly at 's lips.\nCore action sequence (must be fully generated, no skipping):\n1. Seconds 1-2: 's right hand quickly reaches out, precisely grabbing the top of the black silk tie hanging in front of 's chest.\n2. Seconds 2-4: uses arm strength to fiercely pull the tie towards herself. 's upper body is pulled forward, the black silk fabric showing obvious pulling wrinkles.\n3. Seconds 4-6: Using the pulling force, fiercely presses her lips against 's lips, dominantly and fanatically blocking all breath.\nSpatial constraints (absolutely forbidden to change): maintains her seated posture in the boss chair throughout, only her upper body slightly tilting forward from the pull. maintains her leaning posture on the desk throughout, only her left arm and upper body moving (reference ), final frame reference .\nCore expression: Through the highly aggressive action of pulling the tie, showing the instantaneous flip of power dynamics, changing from passive to active, full of controlling desire.\nTechnical requirements: Fully generate the entire dynamic process of pulling the tie, forbidden to jump directly to the kissing state; capture the fabric deformation when the tie is pulled and 's instantaneous physical reaction; reference for character styling and scene setup. Character proportions match .\n\noverall_soundscape:\nShot 1: Subtle friction sound of skin touching skin.\nShot 2: Extremely soft kissing sound.\nShot 3: Clear swish sound of fabric being forcefully pulled, followed by intense lip-lock sounds.\n\nnon_diegetic_music:\nN/A", "original_prompt": "镜号:43 | 场次:日/ @Helena办公室 | 0-3s | 极微距局部特写,平拍 | 缓慢升镜 | 白天自然侧光,通透高权暖白调,@image 那修长、常年握笔的手指犹如冰凉的玉石,从 @Fina 凹陷的锁骨一路向上滑动,细腻的肌肤随着指尖的抚摸泛起战栗,最终大拇指与食指强势地捏住 @Fina 的下巴。 | 台词 & 音效:Dialogue:无 SFX:肌肤相亲的细微摩擦声 | 备注:特写肢体接触,诱惑拉满,@Fina 的姿势参考 @图片4 \n镜号:44 | 场次:日/ @Helena办公室 | 3-6s | 特写,仰拍 | 固定镜头 | 逆光阳光,通透高权暖白调,@image 的手抬着 @Fina 的下巴,清冷的眼眸中透出一丝难得的宠溺,她的红唇犹如蜻蜓点水般,在 @Fina 张开的唇瓣边烙下一个极其轻柔的吻参考 @图片6,@Fina 慢慢后退| 台词 & 音效:Dialogue:无 SFX:极轻柔的亲吻声 | 备注: 主导的片刻温存\n镜号:45 | 场次:日/ @Helena办公室 | 6-9s | 近景,平拍 | 快推镜头 | 逆光阳光,通透高饱和暖白调,发丝金边效果\n【人物初始状态(动作前1秒)】:@image 端正坐在黑色真皮老板椅上,身体挺直,黑色真丝领带自然垂在胸前正中位置;@Fina 俯身靠在办公桌边缘,上半身前倾,距离 @image 约30厘米,眼底燃烧着狂热的欲望,眼神死死盯着 @image 的嘴唇\n【核心动作时序(强制完整生成,禁止跳过任何一步)】:\n1. 第1-2秒:@Fina 的右手快速伸出,精准攥住 @image 垂在胸前的黑色真丝领带上端\n2. 第2-4秒:@Fina 手臂用力,猛地将领带向自己身前狠狠一扯 @image 的上半身被拉力带得向前倾斜,黑色真丝布料产生明显的拉扯褶皱\n3. 第4-6秒:@Fina 借着拉扯的力道,将自己的嘴唇狠狠贴在 @image 的嘴唇上,霸道且狂热地堵住了所有呼吸\n【空间约束(绝对禁止改变)】:@image 全程保持坐在老板椅上的姿势,仅上半身因拉力轻微前倾;@Fina 全程保持俯身靠在桌子上的姿势,仅左臂和上半身有动作参考 @图片5 ,尾帧参考 @图片5 \n【核心表达】:通过拉领带这个极具侵略性的动作,展现权力关系的瞬间翻转, @Fina 从被动变为主动,充满掌控欲\n【强制技术要求】:完整生成拉领带的全部动态过程,禁止直接跳到接吻状态;捕捉领带被拉扯时的布料形变和 @image 身体的瞬间反应;参考 @图片5 的人物造型和场景布置\n台词 & 音效:Dialogue:无 SFX:布料被用力拉扯的清晰\"唰\"声,随后是激烈的唇齿交缠声\n备注:站位固定,人物比例与 @图片5 一致\n\n参考图片说明:\n图1: 角色: Fina\n图2: 角色: image\n图3: 场景: Helena办公室", "reference_images": ["references/4499dba4359317c283fb9c15821a58b5d65f16068e09739a6e9555fe7a45fba5.jpg", "references/cf40778a8e3e670e1ddfd08834c14b51c4ae5167edf168e13ce8245688e84b2f.jpg", "references/7740bc65e74e2d8a64503f1f940ce71838908939335322327f09e91a4ee196e2.jpg", "references/41428bb7ceda5924865d7447899a843bb8ba265a726537daf9eb9d1e8ee42e96.jpg", "references/ae4c1e574f5fcd8d80a7e9fcc6bfbed75bdadfbf779b1b7ce99999a1385a6db3.jpg", "references/c0beb2a34d69017ca8d043f2023cb12005ac07a8123504df95136c92b11d235b.jpg"], "reference_sha256": ["4499dba4359317c283fb9c15821a58b5d65f16068e09739a6e9555fe7a45fba5", "cf40778a8e3e670e1ddfd08834c14b51c4ae5167edf168e13ce8245688e84b2f", "7740bc65e74e2d8a64503f1f940ce71838908939335322327f09e91a4ee196e2", "41428bb7ceda5924865d7447899a843bb8ba265a726537daf9eb9d1e8ee42e96", "ae4c1e574f5fcd8d80a7e9fcc6bfbed75bdadfbf779b1b7ce99999a1385a6db3", "c0beb2a34d69017ca8d043f2023cb12005ac07a8123504df95136c92b11d235b"], "workflow": "ref2va", "prompt_sha256": "67b5f4b6d5126d66e402688bdcbfafb3ccc541fe9a91a1182b18776fb13b26f1", "original_prompt_sha256": "ae86514e425809f6df1823fbcafc8a12953d6376f3e4b543f475c0930a7ea097", "video_sha256": "245c859d13861965b36ed6d370635ac7a51df494141658f0ede152c9e4d45848", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 226, "duration_seconds": 9.416666666666666, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_002.json", "verification": "verification/sample_002.json"} {"file_name": "videos/sample_003.mp4", "sample_id": "#3", "prompt": "subject_definitions:\n is the stick (Golden Cudgel) from , emitting golden light like particle smoke. is the white-armored actor, wearing white armor and wielding a weapon that emits white-gold particle light.\n\nsummary:\n[reference generation] A cinematic sequence in max 70mm film style with low saturation and realistic grain. A glowing stick attacks from the sky, colliding with a white-armored actor's weapon, creating shockwaves and sparks.\n\nretention_analysis:\nfully_preserved: (the stick from ) is the central object in all three shots, maintaining its identity as the attacking and colliding weapon. (the white-armored actor) is consistently depicted in white armor blocking and struggling against the stick.\n\ndetailed_description:\n[Shot 1] A stick () emitting golden light resembling particle smoke, occupying one-eighth of the frame, extends diagonally from the top-left sky into the frame, attacking the crowd on the right side of the frame. [Shot 2] In the center of the frame, the white-armored actor () emits a brilliant white-gold light, raising a weapon that radiates white-gold particle light to block the stick. The stick and the white-armored actor's weapon collide, sending shockwaves scattering in all directions, with golden particles splashing outward from the point of collision. [Shot 3] The stick suppresses the white-armored actor on the right side of the frame. The white-armored actor's entire body radiates intense light, struggling to resist the stick with the weapon. Sparks fly intensely from the collision point between the weapon and the stick. The sequence maintains 4K high definition, cinematic texture, smooth and natural character movements, with no distortion or collapse, conforming to realistic physical logic, with no subtitles, in a max 70mm film texture with grainy realism, epic scale, and low saturation.\n\noverall_soundscape:\nSound effects include weapon collision sounds and the sound of sparks flying. No music is present.\n\nnon_diegetic_music:\nN/A", "original_prompt": "[Reference Aliases]\n@图片1 = 金箍棒\n\n【全局基础设定】\n使用@图片1 作为棒子参考。\n@视频1 为原始拍摄素材(包含基础演员表演与构图)\n风格要求:max70mm胶片质感,颗粒写实,史诗规模低饱和度\n音频要求:无音乐、无字幕、加入特效声音,兵器对撞声,火星四溅声。\n【时间片分镜脚本】\n镜头1,完全保留@视频1 (原始素材)的原有构图与演员表演动作,一根棒子通体冒着粒子烟雾般的金光(比例为占画面八分之一)从画面左上角的天空中斜着伸长入画,袭向画右的众人。运镜与@视频1 原有运镜保持完全一致。\n镜头2,完全保留@视频1 (原始素材)的原有构图与演员表演动作,画面中心穿白色盔甲的演员全身白金光芒大盛,举起武器(散发粒子白金色光芒)格挡棒子,棒子与白衣演员的武器相撞,冲击波相四面八方散去,金色颗粒从碰撞的位置向着四周飞溅。运镜与@视频1 原有运镜保持完全一致。\n镜头3,完全保留@视频1 (原始素材)的原有构图与演员表演动作,棒子压制着画右的白衣演员,白衣演员全身光芒大盛,努力用兵器抗衡棒子,兵器和棒子对撞的位置火星四溅。运镜与@视频1 原有运镜保持完全一致。\n【画质与防崩坏约束】\n4K高清,电影级质感,人物动作流畅自然,画面无畸变、无崩坏,符合现实物理逻辑", "reference_images": ["references/d7c7e25344a83f0af315dcdf1e820759763839576fb180451ea5f3c39e882bf6.jpg"], "reference_sha256": ["d7c7e25344a83f0af315dcdf1e820759763839576fb180451ea5f3c39e882bf6"], "workflow": "ref2va", "prompt_sha256": "3f53ef9a4ec19f35f09d01b9446c4b5617d4d73440ac6b412561fd38f426fedc", "original_prompt_sha256": "874a1d2d2b6d011e1a407236c5022a67401cc888be859a97d2954bb42ace2d0d", "video_sha256": "10b4077a244ff8f82eb82ddf395d5ca7f0a4c01dc2bd5a9bf4ae7a17a1248e06", "task_adapted": true, "width": 1344, "height": 768, "fps": 24, "num_frames": 175, "duration_seconds": 7.291666666666667, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_003.json", "verification": "verification/sample_003.json"} {"file_name": "videos/sample_004.mp4", "sample_id": "#4", "prompt": "subject_definitions:\n: The virtual female model, with facial features and appearance from , standard body proportions, wet-look wavy hairstyle, vacation-style bright makeup.\n: The clothing outfit, strictly matching the color, shape, and details from , made of polyester-spandex fabric with high elasticity, featuring a strapless design, V-type high slit cutting, and waist print.\n: The background setting, directly from , establishing a tropical vacation atmosphere with sunlight and oblique light/shadow.\n\nsummary:\n[reference generation] A 12-second women's clothing e-commerce marketing video in 9:16 vertical format, featuring a virtual model wearing the exact outfit from against the background from , with the model's face based on . The video showcases the clothing's details through three distinct shots with smooth camera movements, accompanied by bright electronic pop instrumental music. The model's movements must be natural and conform to basic physical laws; no clipping is allowed, the model must not stand on tables or chairs, and hands must not be in pockets. The clothing from must be used strictly with no changes or color differences in color, shape, or details.\n\nretention_analysis:\n (Model): fully_preserved. The facial features, hairstyle (wet-look wavy hair), and makeup (vacation-style bright makeup) must be strictly retained from .\n (Clothing): fully_preserved. The exact color, shape, details, and fabric texture (polyester-spandex with high elasticity, strapless design, V-type high slit cutting, waist print, semi-matte texture) must be strictly retained from with no alterations or color differences.\n (Background): fully_preserved. The background setting must directly use to establish the tropical vacation atmosphere.\n\ndetailed_description:\n[Shot 1] 0-4s: Full shot. walks briskly into the frame under sunlight with confident steps. The background establishes a tropical vacation atmosphere. Oblique light and shadow highlight the high-elastic wrapping of the polyester-spandex fabric of .\n\n[Shot 2] At 00:04.000, 4-8s: Medium shot with smooth rotating camera movement. turns sideways and looks back, gently brushing her hair with one hand, showcasing the strapless design and back curves of . Her movements are casual and elegant, highlighting the lengthening effect of the V-type high slit cutting on her legs.\n\n[Shot 3] At 00:08.000, 8-12s: Close-up. The camera pushes in to focus on the waist print and material details of . slightly stretches her body, and the tension rhythm of the fabric reveals a smooth semi-matte texture. The image quality is realistic and delicate.\n\noverall_soundscape:\nThe ambient soundscape is not explicitly specified in the source, but should be consistent with the tropical vacation atmosphere established by the visuals.\n\nnon_diegetic_music:\nBright electronic pop instrumental music with an upbeat rhythm, no vocals, enhancing the energetic and vacation-like mood of the video.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。服装严格使用图二,颜色/版型/细节不允许有任何改变或色差;背景直接使用图三。虚拟模特人脸参考图一,标准身材,发型为湿发感波浪卷,搭配度假风明亮妆容。模特动作自然,符合基本物理规律,禁止穿模或模特站在桌椅上,禁止插兜。 0-4s:全景。模特从阳光下轻快步入画面,步伐自信。利用斜射光影展现涤纶氨纶面料的高弹包裹性,初步建立热带度假氛围。 4-8s:中景环绕。运镜平滑旋转,模特侧身回眸,手轻拨发丝,展示抹胸设计及背部曲线,动作随性优雅,突显V型高开叉剪裁对腿部的拉长效果。 8-12s:近景。镜头推近聚焦腰间印花与材质细节,模特微微伸展身体,通过面料的张力律动展现平滑的半哑光质感,画质真实细腻。 配乐:纯音乐,无人声,明快电子流行节奏。\\n\\n素材说明:图1=模特参考(asset-20260224194849-qt54s);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552.jpg", "references/b05d4f21a3fe1a974009eb64463417dba56c502a4a0e35d8001ca98af25a284a.jpg", "references/beddc71c0e4a9e2389cc34dc9d3215ffc1e6b6249f8dfb91ac30be60981c18f4.jpg"], "reference_sha256": ["96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552", "b05d4f21a3fe1a974009eb64463417dba56c502a4a0e35d8001ca98af25a284a", "beddc71c0e4a9e2389cc34dc9d3215ffc1e6b6249f8dfb91ac30be60981c18f4"], "workflow": "ref2va", "prompt_sha256": "48d275e2dd013b8aec89ec2b3c27e6e282b8ea0e640044c6475c335aa82b950b", "original_prompt_sha256": "151f90ef51a920fb8069a31db3fe96f9665903944139f1ff3f456c36541b53a9", "video_sha256": "34faa1f865e10284df335fc460f0942e020b7b3cbe6ce52afc62a2281e0487eb", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_004.json", "verification": "verification/sample_004.json"} {"file_name": "videos/sample_005.mp4", "sample_id": "#5", "prompt": "subject_definitions:\n is the human hand. is the DJI Nano thumb camera, featuring a translucent frosted texture shell, an exquisite circular lens module, and a 'dji' LOGO with a green status indicator light, as shown in . is the DJI Nano main unit with a screen, as shown in . is the Golden Retriever dog. is the frisbee.\n\nsummary:\n[reference generation] A product showcase video for the DJI Nano thumb camera and its main unit. The video demonstrates the camera's extreme portability, magnetic attachment to the main unit and chest, and its 12mm ultra-wide first-person POV shooting capabilities, ending with the modular design's instant playback feature. The video must not contain any subtitles or watermarks.\n\nretention_analysis:\nfully_preserved: The source explicitly requires the use of for the camera's flat details, for the main unit, and for the magnetic interaction between the camera and the main unit. All visual and functional details of the product, including the translucent frosted shell, circular lens module, 'dji' LOGO, green indicator light, and the main unit's screen UI, must be strictly retained and matched to the provided reference images. The magnetic attachment action shown in must be accurately depicted when the camera module is attached to the main unit slot.\n\ndetailed_description:\n[Shot 1] At 00:00.000, a sunny city park lawn with soft golden afternoon light. Close-up shot. gently pinches the extremely small out of a light-colored sports hoodie pocket using only fingertips. The camera slightly pushes in, accompanying the hand movement. The shot showcases the translucent frosted texture shell and exquisite circular lens module of , emphasizing its 'ultimate portability' being smaller than a knuckle. The product orientation maintains a front-facing level view with a stable structure. Shallow depth of field, the background lawn is blurred, and the focus is locked on the “dji” LOGO and the green status indicator light on the body. The camera quickly pushes into the center of the camera lens, transitioning through a natural black screen.\n\n[Shot 2] At 00:02.000, a park trail with dynamic, bright natural light and shadow. The camera quickly tracks laterally following the human's running footsteps. A human and (Golden Retriever) are the subjects. The shot shows the human magnetically attaching to their chest with a light and quick movement, embodying the 'snap anytime, imperceptible wear' feature. Medium shot captures the dynamic moment of the human and the dog interacting, with light shooting from the side to create a transparent feel. Quick hard cut to first-person visual.\n\n[Shot 3] At 00:04.000, 12mm ultra-wide angle first-person POV. 12mm wide field of view with no fisheye distortion. and are the subjects. is very close to the lens, preparing to play and focused on . at the bottom of the frame is petting the dog's head, then throws . happily runs following the direction of . The shot displays the rich spatial sense brought by the 12mm ultra-wide angle, with straight and natural edge lines, and delicate, clear image quality, emphasizing the 'small camera, big field of view' capability. Low-angle looking up; the sky, lawn, and form rich color layers, with HDR effect retaining cloud details. Action match cut transition.\n\n[Shot 4] At 00:08.000, 12mm ultra-wide angle first-person POV. 12mm wide field of view with no fisheye distortion, shooting angle changes from high to low. and are the subjects. happily runs around the lawn, finally running towards the lens direction (gradually turning to a low-angle POV), happily pouncing towards the lens, and sticking out its tongue trying to lick the lens, creating a very warm and cute scene. The shot displays the rich spatial sense brought by the 12mm ultra-wide angle, with straight and natural edge lines, and delicate, clear image quality. Low angle; the sky, lawn, and form rich color layers, with HDR effect retaining cloud details. After licks the lens, a quick transition to a black screen occurs.\n\n[Shot 5] At 00:11.000, a park bench rest area with side lighting illuminating device details. Close-up shot. is holding . The human magnetically attaches the camera module to the slot of . The screen instantly lights up, displaying the wonderful footage of shot earlier. The shot focuses on the “4K/60/RS/W” UI interface displayed on the screen. The red time box in the top left corner of the screen plays sequentially from 00:00 to 00:01 to 00:02, reflecting the convenient connection and instant playback of the modular design. is in a handheld state, not fixed on a tripod. Side-rear 30° angle clearly shows the button layout of and the consistency of the translucent body. The camera pulls back from the screen details.\n\n[Shot 6] At 00:14.000, the park exit with sunset afterglow. Wide shot. The human's back silhouette is the subject. The human puts away the camera, which is stored inside and placed in a pocket, emphasizing 'burden-free storage after creation', and leads walking towards the sunset. The camera smoothly pulls back. Backlit silhouette composition with the center of the frame left blank, reflecting a relaxed and free lifestyle tone. Fades to a black screen.\n\noverall_soundscape:\nShot 1 features a crisp electronic device startup beep, with slight park bird chirps and wind in the background. Shot 2 includes cheerful footsteps. Shot 3 features the dog's excited breathing and happy barks, along with the whooshing sound of the flying frisbee. Shot 4 includes the dog's running sounds, light barks, and breathing. Shot 5 features the 'click' mechanical sound of the module magnetically locking. Shot 6 ends with the sound of natural wind.\n\nnon_diegetic_music:\nShot 2 introduces an upbeat Indie-Pop track. Shot 5 transitions the BGM to a soothing style. Shot 6 features the BGM reaching a climax before gradually fading out.", "original_prompt": "视频1\n是该视频的主要产品,保证产品细节与结构不变。\n图片1\n是产品的平面细节图。\n图片2\n是产品的主机,\n图片3\n是相机与主机的磁吸交互,磁吸后,会展示产品镜头对准发拍摄画面。 视频不要出现字幕与水印。 00:00 - 00:02 【场景/氛围】 阳光明媚的城市公园草坪,午后柔和的金灿灿光线。 【主体】 人物手部与 DJI Nano 拇指相机。 【动作描述/镜头运动】 特写镜头。人物从浅色运动卫衣口袋里,仅用指尖轻轻捏出极其小巧的 DJI Nano。镜头伴随手部动作轻微推近。 【产品使用方式与卖点细节】 展现相机半透明磨砂质感外壳与精致的圆形镜头模组,强调其比指关节还小的“极致便携”卖点。产品朝向保持正面平视,结构稳固。 【剪辑/转场方式】 镜头向相机镜头中心快速推近,通过自然黑屏转场。 【音效/BGM要求】 清脆的电子设备开启提示音,背景伴有轻微的公园鸟鸣与风声。 【构图/细节说明】 浅景深,背景草坪虚化,焦点锁定在机身的 “dji” LOGO 与绿色状态指示灯。 00:02 - 00:04 【场景/氛围】 公园步道,动感、明亮的自然光影。 【主体】 人物与萌宠(金毛寻回犬)。 【动作描述/镜头运动】 镜头快速跟随人物奔跑的脚步平移。 【产品使用方式与卖点细节】 展现人物将 DJI Nano 磁吸附于胸前,动作轻快。体现“随手即拍、无感佩戴”的特性。 【剪辑/转场方式】 快速硬切,切换至第一人称视觉。 【音效/BGM要求】 节奏轻快的独立流行乐(Indie-Pop)切入,伴有欢快的脚步声。 【构图/细节说明】 采用中景,捕捉人物与狗狗互动的动态瞬间,光线从侧方射入,营造通透感。 00:04 - 00:08 【场景/氛围】 12mm 超广角 第一人称视角(POV)。 【主体】 萌宠狗狗与人物伸出的手。 【动作描述/镜头运动】 12mm 广阔视野,非鱼眼畸变。狗狗十分靠近镜头,做出准备玩耍,专注得盯着飞盘,人物的手在画面下方正抚摸狗狗的头,随后抛出飞盘,狗狗十分开心得跟随飞盘方向奔跑。 【产品使用方式与卖点细节】 展示 12mm 超广角 带来的丰富空间感,边缘线条平直、自然,画质细腻清晰。强调“小相机,大视野”的拍摄实力。 【剪辑/转场方式】 动作匹配转场。 【音效/BGM要求】 狗狗兴奋的呼吸声与欢快叫声,飞盘飞出的破空声。 【构图/细节说明】 低角度仰视,天空、草坪与狗狗构成丰富的色彩层次,HDR 效果保留云层细节。 00:08 - 00:11 【场景/氛围】 12mm 超广角 第一人称视角(POV)。 【主体】 萌宠狗狗与人物伸出的手。 【动作描述/镜头运动】 12mm 广阔视野,非鱼眼畸变,从高角度变低角度拍摄。狗狗开心得环绕着草坪奔跑,最后奔向镜头方向(逐渐转为低角度视角),开心地扑向镜头,还伸出舌头想要舔镜头,画面十分温馨可爱。 【产品使用方式与卖点细节】 展示 12mm 超广角 带来的丰富空间感,边缘线条平直、自然,画质细腻清晰。 【剪辑/转场方式】 狗狗舔到镜头后黑场迅速转场。 【音效/BGM要求】狗狗的奔跑声和轻叫声呼吸声。 【构图/细节说明】 jiao,天空、草坪与狗狗构成丰富的色彩层次,HDR 效果保留云层细节。 00:11 - 00:14 【场景/氛围】 公园长椅休息区,侧光照亮设备细节。 【主体】 一只手拿着DJI Nano 主机(带屏幕部分)。 【动作描述/镜头运动】 近景镜头。人物将 DJI Nano 相机模块磁吸到主机槽位,屏幕瞬间亮起,显示刚才拍摄的狗狗精彩画面。 【产品使用方式与卖点细节】 重点展示主机屏幕显示的“4K/60/RS/W” UI 界面,屏幕的左上角红色时间框从00:00~00:01~00:02依次播放,体现模块化设计的便捷连接与即时回看功能。 (主机是手持状态,不是固定在三脚架上的)【剪辑/转场方式】 画面从屏幕细节拉远。 【音效/BGM要求】 模块磁吸锁定的“咔嗒”机械声,BGM 转为舒缓。 【构图/细节说明】 侧后方 30° 视角,清晰展示主机的按键布局与半透明机身的一致性。 00:14 - 00:15 【场景/氛围】 公园出口,日落余晖。 【主体】 人物背影。 【动作描述/镜头运动】 全景镜头。人物收起相机,牵着狗狗走向夕阳。镜头平稳拉远。 【产品使用方式与卖点细节】 相机已收纳在主机内并放入口袋,强调“创作之后,无负担收纳”。 【剪辑/转场方式】 渐隐至黑场。 【音效/BGM要求】 BGM 达到高潮后渐弱,收尾于大自然的风声。 【构图/细节说明】 逆光剪影构图,画面中央留白,体现休闲、自由的生活方式调性。", "reference_images": ["references/9d49efe0bb8f378f59d16ce71dcadbafc76ba999a1f0cc76825a6671e8440459.png", "references/c82698f5522ee424dfe64a5f7c4fd99daed3f9bb0586113fbb0463371fa328a8.png", "references/ac7fe051740344ddef4ecb40e43b2aedda9f9e6bd2f23f5a198eeae673b94bb6.png"], "reference_sha256": ["9d49efe0bb8f378f59d16ce71dcadbafc76ba999a1f0cc76825a6671e8440459", "c82698f5522ee424dfe64a5f7c4fd99daed3f9bb0586113fbb0463371fa328a8", "ac7fe051740344ddef4ecb40e43b2aedda9f9e6bd2f23f5a198eeae673b94bb6"], "workflow": "ref2va", "prompt_sha256": "74957c27a46c6429b09e13a90dc33b70031671c439dc51346386b066136702fa", "original_prompt_sha256": "04643bc5319d9d9cc66b9f56395618bec1e3c1de274178cb27ba1f15cefdca55", "video_sha256": "a3fc2c668dc4f6ff762f1b5b93670bcea5af913f40f1094784d0bf92f31c1bb0", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_005.json", "verification": "verification/sample_005.json"} {"file_name": "videos/sample_006.mp4", "sample_id": "#6", "prompt": "subject_definitions:\nSubject 1 is the virtual model with a standard figure and Mediterranean vacation style makeup and hair, matching the appearance from . Subject 2 is the clothing, strictly matching the color, shape, and details from , featuring a bamboo knot texture and knit fabric quality. Subject 3 is the Mediterranean terrace background, matching the environment from .\n\nsummary:\n[reference generation] A 12-second e-commerce marketing video for women's clothing to promote purchase conversion. The virtual model from wears the clothing from and walks through the Mediterranean terrace background from . The video features a long shot, medium shot, and close-up, showcasing the clothing's details, fabric texture, and overall vacation style. The video complies with physical laws, with no clipping or the model standing on tables/chairs. The clothing coverage is complete, and the outline is smooth without bulges, complying with e-commerce platform norms. The clothing color is strictly based on , and the background lighting serves the true color of the clothing, with no graying, darkening, or color shifting due to the tone of . Accompanied by lazy Bohemian guitar music.\n\nretention_analysis:\nfully_preserved: The model's appearance, standard figure, and Mediterranean vacation style makeup and hair from are retained. The clothing's color, shape, details, and bamboo knot knit fabric texture from are strictly retained. The Mediterranean terrace background from is retained. The lighting is adjusted to serve the true color of the clothing without graying, darkening, or color shifting.\n\ndetailed_description:\n[Shot 1] A long shot. The virtual model wearing the clothing from walks into the Mediterranean terrace background from . Her steps are light, driving the rhythm of the skirt. While walking, she naturally turns to connect with a horizontal stroll, showing the overall vacation proportion. The model flows continuously, with strictly no hands in pockets.\n\n[Shot 2] At 0:04.000, a medium shot. The camera translates laterally. The model turns sideways and looks back while walking. Her right hand naturally and lightly brushes the hair near her temple. She then slightly tilts her shoulder to show the side silhouette. The action is smooth, capturing the drape of the fabric.\n\n[Shot 3] At 0:08.000, a close-up. The camera pushes in to the upper body. The model sways gently with a natural breeze. Multiple angles present the neckline embroidery details and the bamboo knot fabric texture. Finally, she looks confidently at the camera and naturally walks out of the frame.\n\noverall_soundscape:\nFootsteps on the terrace, gentle breeze, and fabric rustling.\n\nnon_diegetic_music:\nPure instrumental music with no vocals, featuring a lazy Bohemian guitar.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。图一虚拟模特,标准身材,地中海度假风妆发。服装严格使用图二,颜色/版型/细节不许改变;面料呈竹节纹理与针织质感。背景直接使用图三。符合物理规律,禁止穿模/模特站桌椅等;服装覆盖完整、轮廓平整无凸点;符合电商平台规范。服装颜色以图二为准,背景光照服务服装真实颜色,禁止因图三色调致服装变灰/暗化/偏色。配乐:纯音乐,无人声,慵懒波西米亚吉他。全程模特持续流动,严禁插兜。 【0-4s】远景。穿着图二服装的模特步入地中海露台,步伐轻盈带动摆部律动,行进中自然转身衔接横向漫步,展现整体度假感比例。 【4-8s】中景。镜头平移,模特在行进中侧身回眸,右手自然轻掠过鬓角发丝,随后微微倾斜肩膀展示侧面廓形,动作衔接流畅捕捉面料垂感。 【8-12s】近景。镜头推进至上半身,模特配合自然微风轻轻摆动,多角度呈现领口刺绣细节与竹节面料肌理,最后眼神自信看向镜头并自然步出画面。\\n\\n素材说明:图1=模特参考(asset-20260224194849-qt54s);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552.jpg", "references/7589e410cd343938b9999d8010d286766cb497650f7c5e2664c72cbd68c90764.jpg", "references/3b1011496e441c7a5cfe58d9e391ee3c0aba27fd04ec99b3c494b14d7e4afb51.jpg"], "reference_sha256": ["96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552", "7589e410cd343938b9999d8010d286766cb497650f7c5e2664c72cbd68c90764", "3b1011496e441c7a5cfe58d9e391ee3c0aba27fd04ec99b3c494b14d7e4afb51"], "workflow": "ref2va", "prompt_sha256": "f33f29acb56a86fc10d9acc0472d9ee526c7cd83f09bee8797d15409eadd4582", "original_prompt_sha256": "083d05f69f036ad27de62afe0070222219b368f90a03f42bcd3b159263941017", "video_sha256": "c9631b9388bd34e11f2dc2e2f2f7541536b30d63a30a693a8cb15a0dce034327", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_006.json", "verification": "verification/sample_006.json"} {"file_name": "videos/sample_007.mp4", "sample_id": "#7", "prompt": "subject_definitions:\n (Yin Fuyao): The emperor, appearance and clothing from .\n (Pei Xingjian): A close minister, appearance and clothing from .\n (Shen Yanci): A close minister, appearance and clothing from .\n (Xie Changyuan): A close minister, appearance and clothing from .\n (Dragon desk): The imperial desk, appearance from .\n (Sword): A sword, appearance from .\n\nsummary:\n[reference generation] The scene is set in the Taiji Palace of the Li State, featuring a film photography texture, cinematic lighting, translucent textures, a matte premium texture, dynamic soft light projection, hazy glowing layers, soft side backlighting, ultimate hair light, and a darkened, blurred background with foreground obstructions. The shot continues from the previous scene where Yin Fuyao stands up to issue an edict for a personal expedition, shocking the court. Xie Changyuan quickly steps forward to kneel and intercept him, breaking the civil and military formation and creating a core conflict. Yin Fuyao is in a furious state, Pei Xingjian is high-spirited, Shen Yanci is calmly calculating, and Xie Changyuan is vigilant. Xie Changyuan suddenly kneels directly in front of Yin Fuyao on the central axis, blocking his path and forming a one-on-one confrontation, while other ministers remain in their original positions.\n\nretention_analysis:\nfully_preserved: (Yin Fuyao), (Pei Xingjian), (Shen Yanci), (Xie Changyuan), (Dragon desk), (Sword). All referenced subjects and props from to are explicitly required to appear in the scene with their identities and roles maintained.\n\ndetailed_description:\n[Shot 1] Close-up, eye-level angle, fast tilt down. The scene is set in the Taiji Palace of the Li State, featuring foreground obstructions, a film photography texture, cinematic lighting, translucent textures, a matte premium texture, dynamic soft light projection, hazy glowing layers, soft side backlighting, ultimate hair light, and a darkened and blurred background. No subtitles on screen. (Yin Fuyao) stands in front of (the dragon desk), maintaining an imposing and furious aura, not advancing or shifting position. (Pei Xingjian) and (Shen Yanci) remain in their original positions at the front of the civil and military columns, with Pei Xingjian in a high-spirited state and Shen Yanci in a calm, controlled state. (Xie Changyuan), in a vigilant state, quickly steps forward from the first position of the civil officials, thrusting straight into the central axis of the main hall. He suddenly kneels at a low position directly in front of , his body forming a straight line on the central axis, with the messy bangs on his forehead slightly disheveled. He physically blocks the emperor's path, forming a one-on-one confrontation between monarch and minister, while the rest of the court officials remain completely motionless in their original positions. (S1) says urgently, [Chinese]陛下不可!. The sound effects include the dull thud of knees hitting the ground, the simultaneous rustling of court clothes, and the sudden cessation of breathing in the hall.\n\noverall_soundscape:\nThe sound effects include the dull thud of knees hitting the ground, the simultaneous rustling of court clothes, and the sudden cessation of breathing in the hall. No background music is allowed.\n\nnon_diegetic_music:\nN/A", "original_prompt": "画面无字幕\n禁止配音乐\n【场景分析】\n场景:(黎国-太极殿)环境前景遮挡,胶片摄影质感,电影光线,半透明纹理,哑光高级质感,动态柔光投影,朦胧发光层,柔和侧逆光,极致发丝光,背景压暗虚化\n承接:上镜殷扶摇起身下诏,决意亲征,全场震动\n过渡:近臣中位拦截,打破文武分列阵型,形成核心冲突\n【人物表演底层逻辑总纲】【殷扶摇处于被背叛与权威受挑衅后的暴怒状态;裴行简处于求战与立功的昂扬状态;沈砚辞处于冷静算计战局后勤的控制状态;谢长渊处于即将出手制衡的警戒状态】\n【分镜承接】\n前一分镜末尾画面:殷扶摇立身龙案前,气场慑人,谢长渊前倾待命\n本分镜开场画面:谢长渊快步下跪,当面阻拦\n承接站位强化:承接上镜所有人物站位不变,仅核心人物动态移动。殷扶摇立于龙案前未前进、未移位;裴行简、沈砚辞保持文武分列前位,原地驻足;谢长渊从文臣首位快步上前,直插大殿正中轴线,在殷扶摇正前方低位跪地,正面阻断帝王前路,形成一对一君臣对峙,其余群臣全程原位不动。\n【镜头1】\n镜头:近景/平视/快速下摇\n画面描述: 谢长渊猛地跪地,身形正挡在殷扶摇前方,中轴线笔直,额前碎发微乱。\n核心动作: 突然下跪,抢在御令坐实前阻断。\n对手戏设计: 他以身体位置直接横在君王行动路径上,构成正面拦截。\n人物底层逻辑: 必须立刻打断帝王冲动,否则朝局与都城都将暴露风险。\n主体动作: 谢长渊猛地跪地,挡在殷扶摇面前,急声道:“陛下不可!”\n音效: 膝盖触地闷响、朝堂衣料齐响、殿内呼吸声骤停\n资产对照表(视频中涉及的角色、场景与物品如下(保持一致性)):\n视频场景对照表: 黎国-太极殿=【@黎国-太极殿/日内庄严肃杀】\n视频角色对照表: 殷扶摇=【图1】;裴行简=【图2】;沈砚辞=【图3】;谢长渊=【图4】\n视频道具对照表: 龙案=【图5】;佩剑=【图6】", "reference_images": ["references/08cc07f485bd7078883dc4cd50a1b6d75115c4d4b821a1e0a21e94dc7e512b39.jpg", "references/ce1b6ddfe0ef6e43f93cb8b327490e6d8a8c7b9d19f27ed8931999dbc34c17af.jpg", "references/405d8740e8f9141c7ea5c4b04782a3a17525bde2a237994e3499e02fe828e13e.jpg", "references/017a928928647099b2222452ed161dccd11664bfd095aef8ee96478d690f6154.jpg", "references/6a51e83196ee93ce789171a2c5262e3b157691f4d11d3134f8f236af5682ec8a.jpg", "references/d113912129bbba9bef3b60110764ce9b397b7d9c7283e02dd9b2b02493109a37.jpg"], "reference_sha256": ["08cc07f485bd7078883dc4cd50a1b6d75115c4d4b821a1e0a21e94dc7e512b39", "ce1b6ddfe0ef6e43f93cb8b327490e6d8a8c7b9d19f27ed8931999dbc34c17af", "405d8740e8f9141c7ea5c4b04782a3a17525bde2a237994e3499e02fe828e13e", "017a928928647099b2222452ed161dccd11664bfd095aef8ee96478d690f6154", "6a51e83196ee93ce789171a2c5262e3b157691f4d11d3134f8f236af5682ec8a", "d113912129bbba9bef3b60110764ce9b397b7d9c7283e02dd9b2b02493109a37"], "workflow": "ref2va", "prompt_sha256": "c22609311c7ca1a49ca344efb4caaa27992015941a64821843d350704dc0ba87", "original_prompt_sha256": "e89d5fad8317a4c0a3f9855faa68ba170c43a13216642e521a7eb6ccb130cfa3", "video_sha256": "8fd56673787d78487707afc13bf721a17c539b35f896a3d115623a7112866c4c", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 209, "duration_seconds": 8.708333333333334, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_007.json", "verification": "verification/sample_007.json"} {"file_name": "videos/sample_009.mp4", "sample_id": "#9", "prompt": "subject_definitions:\nSubject 1 is the male model from . Subject 2 is the mid-wash blue denim shirt product from .\n\nsummary:\n[reference generation] The video features the model from wearing the mid-wash blue denim shirt from , showcasing the garment in a minimalist urban industrial space through a full shot, a medium shot, and a close-up shot.\n\nretention_analysis:\nSubject 1 (the model) and Subject 2 (the denim shirt) are fully preserved as the primary subjects. The model's identity from and the product's design from must be strictly retained throughout the video, with the product worn by the model as explicitly requested.\n\ndetailed_description:\n[Shot 1] Full shot. A wide shot set in a minimalist urban industrial space with raw concrete walls and soft, diffused daylight. The model enters the frame with a light, rhythmic 3-step stride, then performs a fluid 180-degree pivot to showcase the back yoke and the curved hem of the mid-wash blue denim shirt. He smoothly turns back to face the camera with a calm, confident gaze and a naturally relaxed expression. The outfit is styled with a layered white tee, dark grey cargo pants, and clean white canvas shoes, embodying a modern streetwear aesthetic.\n\n[Shot 2] At 0:04.000, medium shot focusing on the upper body and garment construction. The model shifts to a 45-degree angle, naturally reaching up to adjust the collar and check the symmetry of the twin chest pockets, before turning back to a frontal stance. He maintains a composed and poised expression with effortless charisma. The soft lighting emphasizes the heavy twill texture of the denim and the sharp contrast of the orange topstitching against the mid-wash blue fabric, highlighting the vintage copper buttons.\n\n[Shot 3] At 0:08.000, steady close-up shot focusing on the tactile details of the chest pocket and fabric. The model's hand actively interacts with the garment, gently pinching the thick denim material and tracing the double-layered orange stitching to demonstrate the premium craftsmanship. The camera follows the hand's firm yet graceful movements, capturing the rugged texture of the denim and the clean finish of the interior white tee, emphasizing a high-quality, utilitarian feel.\n\noverall_soundscape:\nUnspecified.\n\nnon_diegetic_music:\nAn upbeat, rhythmic urban lo-fi track with a steady bassline and minimalist electronic beats, creating a cool, confident, and modern street-style atmosphere.", "original_prompt": "【图片 1】【图片 2】\n图1是模特,模特ID:asset-20260224203236-b5ttl\n图2是商品图正反面\n请将图2的商品穿在图1模特的身上,按以下要求生成视频:\n0-4S Full Shot\nA wide shot set in a minimalist urban industrial space with raw concrete walls and soft, diffused daylight. The model enters the frame with a light, rhythmic 3-step stride, then performs a fluid 180-degree pivot to showcase the back yoke and the curved hem of the mid-wash blue denim shirt. He smoothly turns back to face the camera with a calm, confident gaze and a naturally relaxed expression. The outfit is styled with a layered white tee, dark grey cargo pants, and clean white canvas shoes, embodying a modern streetwear aesthetic.\n\n4-8S Medium Shot\nMedium shot focusing on the upper body and garment construction. The model shifts to a 45-degree angle, naturally reaching up to adjust the collar and check the symmetry of the twin chest pockets, before turning back to a frontal stance. He maintains a composed and poised expression with effortless charisma. The soft lighting emphasizes the heavy twill texture of the denim and the sharp contrast of the orange topstitching against the mid-wash blue fabric, highlighting the vintage copper buttons.\n\n8-12S Close-up\nA steady close-up shot focusing on the tactile details of the chest pocket and fabric. The model's hand actively interacts with the garment, gently pinching the thick denim material and tracing the double-layered orange stitching to demonstrate the premium craftsmanship. The camera follows the hand's firm yet graceful movements, capturing the rugged texture of the denim and the clean finish of the interior white tee, emphasizing a high-quality, utilitarian feel.\n\nMusic\nAn upbeat, rhythmic urban lo-fi track with a steady bassline and minimalist electronic beats, creating a cool, confident, and modern street-style atmosphere. --ratio 3:4 --rs 480p --dur -1 --seed -1 --wm true", "reference_images": ["references/76f0272662cacbe4b46d80b1448672ac6db6756bff72dda09c24bc861c0c4b7d.jpg", "references/43d52b4f510be0307b6ced8179f67d18a7ed0241b311ea9472873a19fd46f94a.jpg"], "reference_sha256": ["76f0272662cacbe4b46d80b1448672ac6db6756bff72dda09c24bc861c0c4b7d", "43d52b4f510be0307b6ced8179f67d18a7ed0241b311ea9472873a19fd46f94a"], "workflow": "ref2va", "prompt_sha256": "5224cdb4117736eb2d9b4c6aed745beac403da0adc2dc336a95c10c67ec94545", "original_prompt_sha256": "f018593591d2d4f924a901a35315a8c2fa8a7c8a64bc50695695dcbf95563ea3", "video_sha256": "abf2a210e8cca4afa84c29065926a036dfdedcce8b12d0504eb237c760bd5fe9", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_009.json", "verification": "verification/sample_009.json"} {"file_name": "videos/sample_010.mp4", "sample_id": "#10", "prompt": "subject_definitions:\n: The bride, appearance from .\n: The groom, appearance from .\n: The groom's brother, appearance from .\n: The best female friend, appearance from .\n: The other female friend, appearance from .\n: The wedding banquet private room in a restaurant, appearance from .\n: Xifeng liquor, appearance from .\n: The groom's wine glass, appearance from .\n\nsummary:\n[reference generation] A cinematic, highly detailed, and emotionally tense wedding banquet scene in a vintage-style restaurant private room. The sequence follows the groom and bride entering, interacting with the guests, and the groom making a toast, strictly adhering to the 180-degree rule and eyeline matching. The visual style mimics a Korean drama aesthetic shot on ARRI Alexa Mini LF with shallow depth of field, teal and orange color grading, and 8K resolution.\n\nretention_analysis:\nThe user requests to retain the identities and appearances of the bride (), groom (), groom's brother (), best female friend (), and other female friend (). The setting (), Xifeng liquor (), and the groom's wine glass () are also required to be retained as specific props and locations. All 8 references are fully preserved in the detailed description to maintain character consistency, setting, and prop accuracy throughout the 4-shot sequence.\n\ndetailed_description:\n[Shot 1] (Timing unspecified in source) Medium shot, eye-level, with a surrounding/orbiting camera movement. Nighttime in the wedding banquet private room. Warm yellow lighting, vintage festive atmosphere, full of life. Guests are chatting and laughing, creating a lively and vivid scene. The center features a large round table covered with a red tablecloth, with faintly visible guests in the background, creating a vintage wedding atmosphere. The image is clean and high-end; the foreground, midground, and background have clear progressive relationships. Extreme emotional tension and physical detail realism, real human skin texture, immersive effect. No watermarks, no subtitles, no logos. SFX: ambient chatter and laughter fading in gradually and building. Strictly observe the 180-degree axis and eyeline matching rules, strictly preventing axis crossing and jump cuts.\n\n[Shot 2] (Timing unspecified in source) Close-up of hands gradually pulling back to a medium shot of the couple, three-quarter side angle, eye-level, camera follows character movement. Nighttime, warm yellow lighting, vintage festive. holds in one hand, while his other arm is linked with . They smile and naturally walk towards a table and stop. Cut to: looks at with a happy face, while shyly lowers her head and leans on 's shoulder, full of the shyness of newlyweds. All characters are strictly forbidden from lip-syncing to off-screen voices throughout. An off-screen young, lively female voice (S1) says in an off-screen voiceover with joy and surprise: [Chinese]快看是 @新娘 子来了! SFX: Footsteps and natural ambient noise of the wedding banquet. Strictly observe the 180-degree axis and eyeline matching rules.\n\n[Shot 3] (Timing unspecified in source) Wide shot, eye-level, slow push-in camera. Warm yellow lighting, vintage festive. Reveals the guests seated at the main table where and stopped—, , and . Seeing 's shy expression, the three of them get more excited, naturally stand up, and happily clap their hands to cheer them on, creating a peaceful and lively atmosphere. (S2), (S2), and (S2) shout happily and enthusiastically in a teasing, joking voice: [Chinese]亲一个!亲一个! SFX: None. Strictly observe the 180-degree axis and eyeline matching rules.\n\n[Shot 4] (Timing unspecified in source) Close-up, reaction shot, slightly low-angle, static fixed camera. Nighttime, warm yellow lighting, vintage festive. smiles and looks at , with a shy but unstoppable smile on her lips. , holding in one hand, uses his other hand to pick up the empty on the table and pours the liquor until the glass is full. He then smiles and says a few words to everyone, drinking it all in one gulp amidst cheers, creating a lively and festive scene. (S3) says in a sincere, slightly shy, gentle, and grateful voice: [Chinese]今天好酒招待,感谢大家的祝福,这杯我干了! SFX: Increasingly loud cheers and laughter. Strictly observe the 180-degree axis and eyeline matching rules.\n\nVisual Style / Cinematography: Cinematic Korean drama aesthetic style. Shot with ARRI Alexa Mini LF camera, paired with Zeiss Supreme Prime 85mm T1.5 lens, aperture f/1.5, extremely shallow depth of field, creamy smooth bokeh. Dual-tone color grading with light cool-toned shadows and warm amber highlights (teal and orange LUT). Soft diffused natural ambient light, subtle wrap-around fill light on faces, idealized skin tones with subsurface scattering, controlled fine film grain overlay, slightly crushed blacks, elevated airy highlights, presenting a prestigious television visual style, creating a melancholic yet intimate atmosphere. 8K resolution, extremely realistic.\n\noverall_soundscape:\nAmbient noise of a lively wedding banquet, including guest chatter, laughter, footsteps, and increasingly loud cheers. No specific background music is requested.\n\nnon_diegetic_music:\nN/A", "original_prompt": "人物站位:饭店包间中央摆着覆着红桌布的大圆桌,背景是隐约的宾客,环境充满年代感的婚宴现场。\n要求:画面干净又高级;画面里前中远景递进关系分明。极具情绪张力与物理细节真实感,真人皮肤质感,身临其境效果。\n画面:无水印,无字幕,无logo 。\n镜号:1 | 场次:夜晚 | @饭店婚宴包间 | 中景,平视 | 环绕镜头 | 夜晚,暖黄灯光,复古喜庆,烟火气,婚宴上宾客们谈笑着,场面热闹生动。 | 台词 & 音效:Dialogue:无 SFX:缓入渐推环境嘈杂声与欢声笑语 | 备注:严格遵守180度轴线与视线匹配原则,严防越轴跳切\n镜号:2 | 场次:夜晚 | @饭店婚宴包间 | 手部特写逐渐后退至双人中景镜头,四分之三侧,平视 | 镜头跟随人物运动 | 夜晚,暖黄灯光,复古喜庆,烟火气,@新郎 一只手拿着 @西凤酒 另一只手被 @新娘 挽着,二人笑着自然走向一张桌前停下。切换镜头, @新郎 一脸幸福的看着身边的 @新娘 ,@新娘 害羞的低下头倚靠在 @新郎 肩膀,充满新人结婚时的娇羞感;所有角色全程禁止对着画外音的口型| 台词 & 音效:Dialogue: 画外音:“快看是 @新娘 子来了!” 活泼女声,青年,喜悦惊喜。SFX:脚步声与婚宴中自然的环境喧杂声 | 备注:严格遵守180度轴线与视线匹配原则,严防越轴跳切\n镜号:3 | 场次:夜晚 | @饭店婚宴包间 | 全景,平视 | 镜头缓慢推进 | 暖黄灯光,复古喜庆,烟火气,揭示出落座在 @新娘 和 @新郎 站定的主桌的宾客们——@男主的兄弟 、@大闺蜜 、@另一位闺蜜 ,三人看着 @新娘 害羞的模样于是更来劲自然站起身开心的拍着手起哄,气氛一片祥和和热闹| 台词 & 音效:Dialogue:[用起哄开玩笑的声音大喊] @男主的兄弟 、@大闺蜜 、@另一位闺蜜 (开心/大喊/热烈):\"亲一个!亲一个!\" SFX:无 | 备注:严格遵守180度轴线与视线匹配原则,严防越轴跳切\n镜号:4 | 场次:夜晚 | @饭店婚宴包间 | 近景,反应镜头,微仰视 | 固定镜头静止拍摄 | 夜晚,暖黄灯光,复古喜庆,烟火气,@新娘 笑着看向 @新郎 嘴角挂着腼腆又止不住的笑意,@新郎 一只手拿着 @西凤酒 另一只手手上前拿起桌面上空的 @新郎的酒杯 给酒杯中倒满了酒,随后笑着对各位说了一段话,在欢呼声中一饮而尽,场面热闹喜庆。| 台词 & 音效:Dialogue:[用真诚又有点腼腆的声音说] @新郎 (真诚/温和/感激):\"今天好酒招待,感谢大家的祝福,这杯我干了!\" SFX:越来越大声的欢呼和笑声 | 备注:严格遵守180度轴线与视线匹配原则,严防越轴\n\n参考图片说明:\n图1: 角色: 新娘\n图2: 角色: 新郎\n图3: 角色: 男主的兄弟\n图4: 角色: 大闺蜜\n图5: 角色: 另一位闺蜜\n图6: 场景: 饭店婚宴包间\n图7: 道具: 西凤酒\n图8: 道具: 新郎的酒杯\n影调:具有电影质感韩剧的审美风格,采用ARRIAlexa MiniLF相机拍摄,搭配蔡司SupremePrime85mmT1.5镜头,光圈f/1.5,具有极浅的景深效果,奶油般顺滑的虚化效果,采用双色调色彩分级技术,带有淡雅冷色调阴影与温暖琥珀色亮部的处理(青绿与橙色LUIT),柔和散射的自然环境光,面部有细微的包裹式补光,理想化的肤色呈现带有次表面散射效果,控制得当的精细胶片颗粒叠加,轻微压暗黑色调,提升飘逸的亮部质感,呈现具有声望感的电视视觉风格,营造出忧郁而亲密的氛围,8K分辨率,极度逼真", "reference_images": ["references/ed0a40e1070c0a4a9479989e6205a3eb4edb3b3a699a51d1e90ad59e83c97916.jpg", "references/70ac66deb99da144fee426b99934bc7794edd140dd83e3b24897fec8d2851c97.jpg", "references/08071d32e922f097cf84a3c5c9c145d1880436955c5f8840cd13114a601e1eaa.jpg", "references/11be8db9616bdec73f3fe66eccfc5fafd75282dfe2a20947e33ea2a2f0a87716.jpg", "references/cced81a48321eaa132e4fa50d9386a2ef4085e8513c87f336bc3aec2c13003f0.jpg", "references/cedd1e369d3c1fed6140968b92c4885ffc02f044413f2076a6b5081fafa6b8a9.jpg", "references/8c33ebc397e11b3fa4b53b8a5b3a75a828aac3d1a4201dd21fcfdd50dc829496.jpg", "references/1f5d902f882be0967d3e69438ba1b962b7f36e2374617e83731884ac7bb50cde.jpg"], "reference_sha256": ["ed0a40e1070c0a4a9479989e6205a3eb4edb3b3a699a51d1e90ad59e83c97916", "70ac66deb99da144fee426b99934bc7794edd140dd83e3b24897fec8d2851c97", "08071d32e922f097cf84a3c5c9c145d1880436955c5f8840cd13114a601e1eaa", "11be8db9616bdec73f3fe66eccfc5fafd75282dfe2a20947e33ea2a2f0a87716", "cced81a48321eaa132e4fa50d9386a2ef4085e8513c87f336bc3aec2c13003f0", "cedd1e369d3c1fed6140968b92c4885ffc02f044413f2076a6b5081fafa6b8a9", "8c33ebc397e11b3fa4b53b8a5b3a75a828aac3d1a4201dd21fcfdd50dc829496", "1f5d902f882be0967d3e69438ba1b962b7f36e2374617e83731884ac7bb50cde"], "workflow": "ref2va", "prompt_sha256": "875f948d07cbb3742be930ed843f3108c20d9c929cb58a654102720963bc413f", "original_prompt_sha256": "b9218bbe63e3293ed318ff38eb0a4eacdb63d22e05b4dc2bfb08223249a75504", "video_sha256": "3814ae7ea62b4528224abfbf01613ec742d0dd2d18bcced9509a67f6dbf26d31", "task_adapted": false, "width": 1344, "height": 768, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_010.json", "verification": "verification/sample_010.json"} {"file_name": "videos/sample_011.mp4", "sample_id": "#11", "prompt": "subject_definitions:\n (Military uniform woman): Appearance and clothing from .\n (Lion Beastman): Humanoid state with lion ears and humanoid outline, appearance from .\n (Beast Lion Beastman): Giant berserk form with a huge lion head, exploded mane, thick chest and shoulders, giant claws, and a bipedal humanoid beast posture, appearance from .\n (Combat Soldier 2): Appearance and clothing from .\n: Empire Battlefield at dusk, wet mud, broken guns, broken armor, half-buried helmets, rubble, ashes, and tattered flags on the ground. Distant burning castle, orange-red firelight, thick smoke, and tattered flags pulled by shockwaves, environment from .\n\nsummary:\n[reference generation] A realistic, cinematic, low-saturation, high-contrast dusk scene at the Empire Battlefield. The sequence establishes a high-angle oppression by the Beast Lion Beastman over a fallen Combat Soldier 2, interrupted by the urgent arrival of the Military uniform woman who cuts in from the right to block the attack, forming a clear three-person spatial relationship without breaking the 180-degree axis.\n\nretention_analysis:\nThe source requires strict retention of character identities and spatial positioning from the supplied references. (Military uniform woman) must match . (Lion Beastman) must retain the humanoid state and lion ears from . (Beast Lion Beastman) must match the giant berserk form from . (Combat Soldier 2) must match . The environment must match the Empire Battlefield from . is explicitly requested to lock the specific spatial composition (Combat Soldier 2 at lower left, Beast Lion Beastman at upper middle left, Military uniform woman cutting from the right). All character identities, physical proportions, and the strict left-right axis locking are fully preserved requirements.\n\ndetailed_description:\n【Tone / Style Baseline】\nDusk, realistic live-action film style, real photography texture, low saturation, high contrast. The ground is wet mud, broken guns, broken armor, half-buried helmets, rubble, ashes, and tattered flags. In the distance is a burning castle, orange-red firelight, thick smoke, and tattered flags pulled by shockwaves. The overall atmosphere is an epic fantasy battlefield with strong visual oppression and cinematic feel, while maintaining a real photography texture. The actual generated picture is a realistic Empire Battlefield at dusk, not inheriting any white background, silhouette, or schematic illustration style.\n\n【Axis Locking】\nLeft side of the frame = (Combat Soldier 2) low position area / oppressed side.\nRight side of the frame = (Military uniform woman) cutting-in side.\nUpper middle left of the frame = (Beast Lion Beastman) high position oppression side.\nThe 180-degree axis is not broken throughout the entire sequence.\n\n[Shot 1] (0.0s - 1.5s)\nCamera: Low-angle medium shot, fixed camera, slight slow push.\nAction: The camera looks up from a low position to capture the oppression relationship. (Combat Soldier 2) is lying in the mud and water at the lower left of the frame, body leaning back, legs curled up, hands instinctively raised to protect the chest and face, expression terrified, breathing rapidly, shoulders hunched back, eyes staring fixedly upward. (Beast Lion Beastman) is positioned at the upper middle left high position, its huge body blocking out the sky, showing only partial lower limbs, chest, abdomen, and one giant front-reaching claw. The beast claw is explicitly continuing to grab down toward , claw tips pointing forward, aiming at the soldier's upper body; the danger has genuinely occurred. Mud, broken armor, and wet mud water in the foreground create a low-position oppression, while distant firelight and smoke are blurred. Core action: The beast claw is genuinely grabbing, the danger has initiated, it is not just a posing oppression. End state: The giant claw of has clearly pressed down to a lower position, approaching the upper body of ; remains in the lower left low position in terrified defense.\n\n[Shot 2] At 00:01.500 (1.5s - 2.4s)\nCamera: Ultra-low-angle close-up, fixed camera, short forward push.\nAction: The camera remains close to the mud and the soldier's hand. Just as the claw of continues to grab down, one black military boot of (Military uniform woman) suddenly rushes into the frame from the right side, heavily stepping into the mud and water right in front of . The mud and water splash only once. This step is not ordinary standing, but the final step of an urgent arrival with obvious rescue force. The landing point of the military boot cuts exactly between and the beast claw, forming a clear block. Visible in the frame are the panicked raised hand of the soldier in partial view, and the oppression shadow of the approaching giant claw above. Core action: cuts in urgently from the right, using the final step to establish the rescue. Key control: Do not generate continuous running and stepping in water; only retain the final step rushing in and heavy step stopping. End state: The military boot of has nailed into the mud, steadily cutting between and the giant claw of , the attack route is interrupted.\n\n[Shot 3] At 00:02.400 (2.4s - 4.0s)\nCamera: Low-angle medium shot, slight slow push.\nAction: The camera continues to look at the three-person relationship from a low position. (Military uniform woman) completely cuts into the frame from the right side, stopping half a step in front of , her body facing at the upper left. Behind her is still lying in the mud at the lower left. She has one foot forward and one foot back, center of gravity lowered, knees slightly bent, shoulders tense, clearly a protective posture stabilized immediately after stopping, not standing stiffly. (Beast Lion Beastman) is at the upper middle left high position, its giant beast claw forced to stop in front of ; the oppression remains, but the attack is cut off by her. The three-person relationship must be very clear: is at the lower left foreground low position, fallen and leaning back, still looking up in terror; is at the middle ground slightly right protection position, blocking in front of the soldier; is at the upper middle left high position oppressing, giant claw stopped in front of her. Core action: truly inserts between the two, the protection position is officially established. End state: remains at the lower left low position; has steadily blocked in front; the high-position oppression of is blocked.\n\noverall_soundscape:\nShot 1: Wind pressure sound of the beast claw pressing down, low beast roar, rapid breathing of the soldier, light mud and water sounds, distant burning and battlefield wind. Shot 2: Urgent running footsteps, military boot heavily stepping into mud and water with a splash, mud splashing, the sound of the beast claw pressing down is interrupted. Shot 3: Soldier's breathing, mud ground bearing force, military uniform hem blown by the wind, low beast roar, battlefield wind and distant fire sounds.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【影调 / 风格基准】\n黄昏,写实真人影视剧风格,真实摄影质感,低饱和,高反差。地面是湿泥、断枪、破甲、半埋头盔、碎石、灰烬和残旗。远处是燃烧城堡、橙红火光、厚重烟尘和被冲击波扯动的残旗。整体保持史诗奇幻战场氛围,画面压迫感强,电影感强,但保持真实摄影质感。\n 为上一段终帧参考:用于锁定【战斗士兵2】左下低位、【 @兽化狮兽人 】右侧高位压迫、双方体型差和压迫方向。实际生成画面为写实 @帝国战场 黄昏,不继承白底、剪影、示意图画风。\n【 @兽化狮兽人 】:巨型狂化形态,巨大狮头、炸开鬃毛、厚重胸肩、巨爪、双腿直立的人形猛兽姿态。\n【 @狮兽人 】:被精神力压回后的人形状态,参考 @狮兽人 人形资产图,保留狮耳和人形轮廓。\n@图片6 用于锁定:\n【战斗士兵2】位于左下方低位倒地;\n@兽化狮兽人 位于中上偏左高位压下;\n军装女性将从右侧切入两者之间。\n实际生成画面为写实电影感帝国战场,不继承白底、剪影、示意图风格。\n【轴线锁定】\n画面左侧 = 【战斗士兵2】低位区域 / 被压迫侧。\n画面右侧 = 军装女性切入侧。\n画面中上偏左 = 【兽化狮兽人】高位压迫侧。\n全段不跳轴。\n【0s-1.5s|镜头1|兽爪压下 / 士兵低位压迫关系镜头】\n景别 / 运镜:\n低机位中景,固定机位,轻微慢推。\n画面执行:\n镜头从低位看向压迫关系。\n【战斗士兵2】倒在画面左下方泥水里,身体后仰,双腿蜷起,双手本能抬起护在胸前和脸前,表情惊恐,呼吸急促,肩膀后缩,视线死盯上方。\n【兽化狮兽人】位于画面中上偏左高位,巨大身体压住天空,只露出部分下肢、胸腹和一只前探巨爪。\n那只兽爪是明确继续朝【战斗士兵2】抓下来,爪尖前探,抓向士兵上半身,危险已经真正发生。\n泥地、碎甲和湿泥水在前景形成低位压迫感,远处火光和烟尘虚化。\n核心动作:\n先明确“兽爪真的要抓人”,危险已经启动,不是摆姿势压迫。\n结束状态:\n【兽化狮兽人】的巨爪已经明显压下到更低位置,逼近【战斗士兵2】上半身;【战斗士兵2】仍在左下方低位惊恐防御。\n物理音效:\n兽爪压下的风压声,低沉兽吼声,士兵急促喘息声,泥水轻响,远处燃烧与战场风声。\n【1.5s-2.4s|镜头2|军靴切入 / 右侧急停特写】\n景别 / 运镜:\n超低机位近景,固定机位,短促前推。\n画面执行:\n镜头仍贴近泥地与士兵手边。\n就在【兽化狮兽人】的爪子继续往下抓的瞬间,军装女性一只黑色军靴从画面右侧猛地冲入,重重踩进【战斗士兵2】身前泥水里。\n泥水只炸开一次,。\n这一脚不是普通站立,而是急切赶到后的最后一步急停,带着明显救场力度。\n军靴落点刚好切进【战斗士兵2】和兽爪之间,形成清楚阻断。\n画面里可见士兵惊慌抬起的手部局部,以及上方逼近的巨爪压迫阴影。\n\n核心动作:\n军装女性从右侧急切切入,用最后一步急停建立救场。\n\n重点控制:\n不要生成连续跑步踩水。\n只保留“最后一步冲入 + 重踩急停”。\n\n结束状态:\n军装女性军靴已经钉进泥地,稳稳切在【战斗士兵2】与【兽化狮兽人】巨爪之间,攻击路线被打断。\n\n物理音效:\n急冲脚步声,军靴重踩泥水炸开声,泥点飞溅声,兽爪压下声被打断。\n\n【2.4s-4.0s|镜头3|保护位成立 / 三人关系中景】\n\n景别 / 运镜:\n低机位中景,轻微慢推。\n\n画面执行:\n镜头从低位继续看三人关系。\n军装女性从画面右侧完整切入,停在【战斗士兵2】前方半步,身体朝向左上方的【兽化狮兽人】。\n她背后是仍倒在左下方泥地里的【战斗士兵2】。\n她一脚前一脚后,重心压低,膝部微屈,肩膀绷住,明显是急停后立刻稳住的保护姿态,不是木站。\n【兽化狮兽人】位于画面中上偏左高位,巨大兽爪被迫停在军装女性前方,压迫仍在,但攻击被她硬生生切断。\n三者关系必须非常清楚:\n【战斗士兵2】在左下前景低位,倒地后仰,仍惊恐看向上方;\n军装女性在中景偏右保护位,挡在士兵前方;\n【兽化狮兽人】在中上偏左高位压迫,巨爪停在她前方。\n\n核心动作:\n军装女性真正插进两者之间,保护位正式成立。\n\n结束状态:\n【战斗士兵2】仍在左下低位;军装女性已稳稳挡在前方;【兽化狮兽人】高位压迫被拦下。\n下一镜可以直接接:军装女性抬手释放精神力。\n\n物理音效:\n士兵喘息声,泥地受力声,军装衣摆被风带起声,兽类低吼声,战场风声与远处火焰声。\n\n参考图片说明:\n图1: 角色: 军装女性\n图2: 角色: 狮兽人\n图3: 角色: 兽化狮兽人\n图4: 角色: 战场士兵2\n图5: 场景: 帝国战场", "reference_images": ["references/c69d4154747a436d3724dd5daa0155b3a349780cac88b9168076be5673dfe05d.jpg", "references/114d4ebfa77dd06ec4046d1061039c5391c1976a6c3c9d4c53690946332a72fb.jpg", "references/e8a8d7f2534423221a5a7f2688235a688db3a73996026fb1d7994d082a865350.jpg", "references/173668b3d7110607ecd02f5f925134f9ab1472be3a00cd09ecd305eb3aae5e47.jpg", "references/2cb6a550cabcbb35dd495ecab7aa0c2ec9e4f59ab97f1b4f381ec58f762db154.jpg", "references/c7c387b1a6af7be803bebb02b48c51f8e722e2bb67bd55cfcb8aee4c8b15d512.jpg"], "reference_sha256": ["c69d4154747a436d3724dd5daa0155b3a349780cac88b9168076be5673dfe05d", "114d4ebfa77dd06ec4046d1061039c5391c1976a6c3c9d4c53690946332a72fb", "e8a8d7f2534423221a5a7f2688235a688db3a73996026fb1d7994d082a865350", "173668b3d7110607ecd02f5f925134f9ab1472be3a00cd09ecd305eb3aae5e47", "2cb6a550cabcbb35dd495ecab7aa0c2ec9e4f59ab97f1b4f381ec58f762db154", "c7c387b1a6af7be803bebb02b48c51f8e722e2bb67bd55cfcb8aee4c8b15d512"], "workflow": "ref2va", "prompt_sha256": "d6b30726726d46c9961bb07a479294fa92f0c8f8820b80474f64e116cde17a21", "original_prompt_sha256": "c60e601c653060474133e17f4f82e226d829dcf9e55589ef3996158f4d6be7a4", "video_sha256": "2f8ab6e72b683d607ff7aaf54dd53a84c369fa29eb35ea67d6af4e2e4b09cdc3", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 158, "duration_seconds": 6.583333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_011.json", "verification": "verification/sample_011.json"} {"file_name": "videos/sample_012.mp4", "sample_id": "#12", "prompt": "subject_definitions:\n (Chen Ling'er): appearance and clothing from . : appearance and clothing from . : appearance and clothing from . : appearance and clothing from .\n\nsummary:\n[reference generation] A tense classroom scene at night where Chen Ling'er is confronted by three figures. She reacts in shock and retreats until blocked by a desk, demanding to know what they want.\n\nretention_analysis:\nChen Ling'er ( from ) is fully preserved as the main character in both shots. The three approaching figures ( from , from , from ) are fully preserved in the second shot. , , and are not referenced in the source prompt, so their retention is unspecified.\n\ndetailed_description:\n[Shot 1] At 0s, medium shot, level angle, fixed camera. Late night classroom illuminated by cold white fluorescent light. (Chen Ling'er) is seated carrying a schoolbag. Her expression shows shock: eyes wide open, mouth corners tight, conveying a tense and suffocating atmosphere. She stands up, gripping both sides of her schoolbag with her hands, facing , , and .\n[Shot 2] At 4s, medium shot, level angle, fixed camera. , , and walk slowly towards in a vertical line. keeps retreating until she is blocked by a desk. (S1) says [Chinese]你们想干什么?你们想干什么?. No subtitles. Quality: 4K ultra-clear, cinematic color grading, smooth character motion.\n\noverall_soundscape:\nTense and suffocating atmosphere in a late-night classroom. Footsteps as the three figures approach, and the sound of retreating and bumping into the desk.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【全局基础设定】陈灵儿站在教室内,深夜冷白日光灯光线。\n[0s-4s](中景平拍固定)图片1 -情绪要点:震惊-氛围:紧张窒息-对话语气:无-微表情细节:眼睛瞪大,嘴角绷紧-背着书包起身,双手攥住书包两侧,面对图片2 图片3 图片4 固定机位中景拍摄。\n[4s-9s](中景平拍固定)图片2 图片3 图片4 竖排缓步走向图片1 ,图片1 连连后退(当后退到桌子旁边时被桌子抵住),固定机位中景拍摄。 图片1:\"你们想干什么?你们想干什么?\"\n【画质约束】4K超清,电影级色彩,人物动作流畅,无字幕。", "reference_images": ["references/69c247c1dea6d34338fa8a99efe000a5133c66b924b3ec712b56b1ad836efb6f.jpg", "references/f7e992bd60179ec24e7af8b673ce721e86e89c7897ea450019c1c13f36189dbd.jpg", "references/0b9623e45170d9eb6fa27b6e12a312c1aa21fe233eef5ced61006efdcbc600a6.jpg", "references/26040d69dd9509691c55913953d40ab0358795b4144168fd1a56e0556157d04a.jpg", "references/c457d65332aa263a8484ee26ef01609f70f039ebda4ce45f6f8f8e394bfd5874.jpg", "references/c0db915c0067226a275b82fa869717c545fbade81e676eb64b7eb6ee3d771e4e.jpg", "references/d430267e53d7e8dc0d38beb723ab2613e9c7795bd82bda5947da3267fe07e1d7.jpg"], "reference_sha256": ["69c247c1dea6d34338fa8a99efe000a5133c66b924b3ec712b56b1ad836efb6f", "f7e992bd60179ec24e7af8b673ce721e86e89c7897ea450019c1c13f36189dbd", "0b9623e45170d9eb6fa27b6e12a312c1aa21fe233eef5ced61006efdcbc600a6", "26040d69dd9509691c55913953d40ab0358795b4144168fd1a56e0556157d04a", "c457d65332aa263a8484ee26ef01609f70f039ebda4ce45f6f8f8e394bfd5874", "c0db915c0067226a275b82fa869717c545fbade81e676eb64b7eb6ee3d771e4e", "d430267e53d7e8dc0d38beb723ab2613e9c7795bd82bda5947da3267fe07e1d7"], "workflow": "ref2va", "prompt_sha256": "107d8b822ae220dd373411ab791e274087bece6a5421cfefee237214a9ac5d56", "original_prompt_sha256": "c843cddb80f0e73ef170802f2f137aff298a9a0433e0c018a8f8706f14ac1f5e", "video_sha256": "df088f5b02aef35ac5444094da4bafd950a6a50d1224e75e7648982c6bec4bcd", "task_adapted": false, "width": 1344, "height": 768, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_012.json", "verification": "verification/sample_012.json"} {"file_name": "videos/sample_014.mp4", "sample_id": "#14", "prompt": "subject_definitions:\n (Little boy Will) is defined by the appearance, clothing, and art style from . He has soft clay-like skin with subtle freckles, rounded friendly facial features, specific hair strand detail, hair color, and skin tone matching . His clothing exactly matches the color, fabric, fit, and style from . The art style is a bright, warm, kid-friendly 3D cartoon animation style with vivid saturated colors, warm even lighting with no harsh shadows, and consistent color grading, saturation, and brightness matching .\n\nsummary:\n[reference generation] Generate a video of (Little boy Will) sitting at a colorful round wooden table, facing the camera, and gesturing excitedly toward a stack of colorful picture-only books on the table next to a sparkly purple rock. speaks directly to the camera. The entire video strictly maintains the character identity, facial features, clothing, and 3D cartoon art style from with zero text visible anywhere.\n\nretention_analysis:\nThe user requests full retention of (Little boy Will) based on , strictly matching facial features, skin freckles, hair strand detail, hair color, skin tone, clothing fabric texture, and 3D render shading in every frame. is supplied but not explicitly referenced in the prompt, so its retention role is unspecified.\n\ndetailed_description:\n[Shot 1] Medium bright shot. (S1) is sitting at a colorful round wooden table, facing the camera. He is gesturing excitedly toward a stack of colorful picture-only books on the table, which are placed next to a sparkly purple rock. All background surfaces, including walls, posters, signs, books, whiteboards, chalkboards, clothing, objects, and any background surface, are completely clean and text-free, with absolutely zero text of any kind visible anywhere in the scene, including no subtitles, captions, text overlays, lesson titles, labels, or watermarks. (S1) says directly to the camera in the same male child voice as the reference, with a fixed timbre and stable tone, speaking slowly and clearly at approximately 100 words per minute with natural brief pauses between phrases, [English]My current favorite is Ryan's World, but I also love watching all of Mr. Beast's crazy videos!. The rendering strictly avoids photorealism, dark rendering, or overly detailed textures, maintaining consistent color grading, saturation, and brightness matching .\n\noverall_soundscape:\nUnspecified\n\nnon_diegetic_music:\nN/A", "original_prompt": "ABSOLUTE RULE: the video must contain ZERO text of any kind, anywhere — no subtitles, no captions, no text overlays, no lesson titles, no labels, no watermarks, and NO text visible inside the scene itself: no words on walls, posters, signs, books, whiteboards, chalkboards, clothing, objects, or any background surface. All background surfaces must be clean and text-free. Pure clean video only. [Image 1] is the character and outfit reference for this ENTIRE video. CRITICAL: Match [Image 1]'s exact clothing, colors, and fabrics in every frame. Scene and action only: use the visual prompt for background, camera, and motion. Do NOT use it to change character identity or clothing; [Image 1] outfit always wins. Scene: Medium bright shot Will from reference image sitting at a colorful round wooden table facing camera gesturing excitedly toward a stack of colorful picture only books on the table next to a sparkly purple rock Little boy Will speaks directly to camera: \"My current favorite is Ryan's World, but I also love watching all of Mr. Beast's crazy videos!\" Character lock — [Image 1] is the SOLE reference for all visual quality. Little boy Will: strictly match facial features, skin freckles, hair strand detail, hair color, skin tone, clothing fabric texture, and 3D render shading to [Image 1] in EVERY frame. CRITICAL: Match the EXACT clothing in [Image 1] — color, fabric, fit, style, every detail. If any text prompt mentions a different outfit, color, hat, hoodie, jacket, or accessory, IGNORE that text and keep [Image 1]. No character drift, no facial distortion, no outfit change, no texture loss, no clothing color change across any frame. Art style: match [Image 1] exactly — bright, warm, kid-friendly 3D cartoon animation style. Soft clay-like skin with subtle freckles, rounded friendly facial features, vivid saturated colors, warm even lighting with no harsh shadows. Do NOT drift toward photorealism, dark rendering, or overly detailed textures. Consistent color grading, saturation, and brightness matching [Image 1] in every frame. Voice: same male child voice as the reference, fixed timbre, stable tone throughout — no pitch drift. Speak slowly and clearly at a pace comfortable for young children learning English — approximately 100 words per minute, with a natural brief pause between phrases.", "reference_images": ["references/dfec28b57e47810e3a9279c936050a10b03e6f9d72b78335fe89e293856e03d9.png", "references/703058cd9cc78a6ad72cef6fab82f8eb1d8abf0b138b3ab485014b417387977f.jpg"], "reference_sha256": ["dfec28b57e47810e3a9279c936050a10b03e6f9d72b78335fe89e293856e03d9", "703058cd9cc78a6ad72cef6fab82f8eb1d8abf0b138b3ab485014b417387977f"], "workflow": "ref2va", "prompt_sha256": "766070f650c1de00234b74130c9b38e80ca8198b24fc748cf44b296a855549f1", "original_prompt_sha256": "33cb57a97df9f94e7b1884cbb88191d43a9de54fa82f78bb4ddaa56071228745", "video_sha256": "d77aab14e7db973d86d7fd62b041855df0af9e065069642018899fff369f3887", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 243, "duration_seconds": 10.125, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_014.json", "verification": "verification/sample_014.json"} {"file_name": "videos/sample_015.mp4", "sample_id": "#15", "prompt": "subject_definitions:\n (Helper A): A helper wearing rough cloth short clothes.\n (Lin Zhi): Appearance from .\n (Helper B): A helper wearing rough cloth short clothes.\n (Other helpers): Helpers wearing rough cloth short clothes.\n (Xue Qin): Appearance from .\n\nsummary:\n[reference generation] The video depicts a tense confrontation at the Dayan Dynasty riverbank. Helper A and Helper B aggressively corner and attempt to bind Lin Zhi, who struggles fiercely. The physical struggle escalates until Lin Zhi is pushed into the mud. Suddenly, Xue Qin's off-screen voice halts the attack. The helpers step back, creating a path, and Lin Zhi looks up towards the newcomer.\n\nretention_analysis:\nfully_preserved for as the Dayan Dynasty riverbank location, for Lin Zhi's appearance, for Xue Qin's appearance, for the thick rope prop, and for the carrying pole prop. The helpers' appearance is defined by the text description 'rough cloth short clothes' as no specific image references are provided for them.\n\ndetailed_description:\nLighting and style: Film photography texture, cinematic lighting, semi-transparent texture, matte premium texture, dynamic soft light projection, hazy glowing layer, soft side backlight, ultimate hair light, darkened and blurred background.\n\n[Shot 1] Close-up, eye-level, fast push-in. Helper A steps forward from the right side of the frame, reaching in to grab Lin Zhi's forearm on the left side, forming a diagonal pulling line. Helper A's arm muscles tense up, expression tough and unquestionable. Helper A grabs Lin Zhi's arm and pulls hard towards himself. Helper A (S1) says sternly [Chinese]不是奸细能穿成这样?. SFX: Sound of fabric being pulled tight, rapid footsteps.\n\n[Shot 2] At 00:02.600, hard cut. Close-up, eye-level, handheld intense slight shake. Helper B approaches from the side with a thick rope, the rope loop opening quickly. Helper B is at Lin Zhi's side and rear, while Helper A still controls one side, forming a pincer attack. Helper B raises the thick rope to loop onto Lin Zhi's hands. Lin Zhi struggles desperately, shoulders bumping randomly. Lin Zhi (S2) shouts anxiously [Chinese]别碰我!放开!. SFX: Sound of hemp rope swinging, anxious shouting, chaotic footsteps.\n\n[Shot 3] At 00:05.700, hard cut. Close-up, high-angle looking down, tracking push down. Amidst the pushing, Lin Zhi loses balance and is fiercely thrown onto the wet muddy ground. Mud splatters, hands heavily support the ground, trouser legs quickly get stained. The helpers form a semi-circle looking down, Lin Zhi is in a low position on the ground. Lin Zhi is pushed down, shoulders and back smashing into the mud, grass and mud splashing onto sleeves and the side of the face. The helpers continue to step forward half a step. SFX: Dull thud of body falling into mud, mud and water splashing, rapid panting.\n\n[Shot 4] At 00:08.700, hard cut. Long shot, eye-level, static camera. A rapid female voice suddenly comes from outside the frame. All helpers freeze in the same instant, the rope suspended in mid-air, the wind still blowing the reeds. The helpers maintain their original poses frozen. Xue Qin (S3) says in an off-screen voiceover anxiously [Chinese]住手!. SFX: Female voice piercing the wind sound, surrounding action sounds suddenly stop, only the sound of wind blowing reeds remains.\n\n[Shot 5] At 00:10.800, hard cut. Medium shot, eye-level, slow pull-back. The helpers look at each other and then make way, a gap appears in the encirclement. Lin Zhi, lying in the mud, looks up. The helpers turn from facing Lin Zhi to parting sideways, the visual channel pointing to the newcomer outside the frame. The helpers retract their carrying poles and ropes, parting left and right. Lin Zhi supports himself on the mud and raises his head, his line of sight looking forward along the cleared path. SFX: Sound of cloth shoes stepping back, rope lightly dropping, lowered surprised and doubtful gasps.\n\noverall_soundscape:\nAmbient wind blowing reeds, chaotic and rapid footsteps, sounds of fabric pulling and hemp rope swinging, dull thuds and splashes of mud and water, rapid panting, cloth shoes stepping back, and lowered surprised gasps.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【场景分析】\n场景:(大燕朝河岸边)\n承接:伙计们已将林止逼至退无可退,包围圈正在收紧\n过渡:直接承接,情绪意图为把言语怀疑升级成身体控制,再用一声呼喊强行切断动作高潮\n人物表演底层逻辑总纲:【林止处于被侵犯边界后的激烈反抗;伙计甲处于粗暴执行的控制状态;伙计乙处于协同束缚的跟进状态;薛琴处于极度焦急、尚未看清全貌前的本能阻止状态】\n【分镜承接】\n前一分镜末尾画面:林止后退辩解,众伙计继续逼近\n本分镜开场画面:伙计甲手臂猛地探出,抓向林止胳膊\n承接逻辑:动作延续,压迫关系直接升级\n灯光:胶片摄影质感,电影光线,半透明纹理,哑光高级质感,动态柔光投影,朦胧发光层,柔和侧逆光,极致发丝光,背景压暗虚化;\n \n**【镜头1】0.0-2.6秒\n转场: 无\n镜头:近景/平视/快速推近\n画面描述:伙计甲跨前一步,手掌狠狠扣住林止胳膊,臂上肌肉绷起,神情强硬不容置疑。\n核心动作: 强行拽住林止。\n对手戏设计:伙计甲从画面右侧探手入画,抓住画面左侧林止前臂,形成斜向拉扯线。\n人物底层逻辑:通过身体控制把可疑之人彻底压入被支配位置。\n主体动作:伙计甲一把拽住林止的胳膊,向自己方向猛拉,厉声道:“不是奸细能穿成这样?”\n音效: 衣料被拽紧的拉扯声,脚步急踏声\n【镜头2】2.6-5.7秒\n转场: 硬切\n镜头: 特写/平视/手持剧烈微晃\n画面描述:伙计乙提着粗绳从侧面逼近,绳圈快速张开,林止身体后仰挣脱,双臂用力甩动,场面混乱。\n核心动作: 套绳与挣扎同时发生。\n对手戏设计:伙计乙位于林止侧后方,伙计甲仍控制一侧,形成双向夹击。\n人物底层逻辑: 一方求快速制服,一方本能拒绝被束缚。\n主体动作:伙计乙抬起粗绳往林止手上套去,林止拼命挣扎,肩膀乱撞,急喊:“别碰我!放开!”\n音效: 麻绳甩动声,急喊声,混乱脚步声\n【镜头3】5.7-8.7秒\n转场: 硬切\n镜头: 特写/俯视/跟拍下压\n画面描述: 推搡间林止失去重心,被猛地带倒在湿泥地里,泥点飞溅,手掌重重撑地,裤腿迅速沾污。\n核心动作: 林止倒地,围捕达到物理结果。\n对手戏设计: 众伙计围成半圈向下压视,林止位于地面低位,形成明显权力落差。\n人物底层逻辑: 群体暴力的惯性让场面脱离理性判断。\n主体动作: 林止在推搡中被推倒,肩背砸进泥地,草叶与泥浆溅到衣袖和脸侧,伙计们继续逼近半步。\n音效: 身体跌入泥地的闷响,泥水飞溅声,急促喘息声\n【镜头4】8.7-10.8秒\n转场: 硬切\n镜头:远景/平视/固定镜头\n画面描述:画面外侧忽然传来急促女声,所有伙计动作在同一瞬间停住,绳索悬在半空,风仍吹动芦苇。\n核心动作:画外音打断围攻节奏。\n对手戏设计: 无,使用声先入画建立新角色权重。\n人物底层逻辑: 突如其来的熟悉权威或强烈情绪令众人短暂停摆。\n主体动作: 众伙计维持原动作僵住,画外传来薛琴急促高喊:“住手!”\n音效: 女声穿透风声,周围动作声骤停,只余风吹芦苇声\n【镜头5】**10.8-15秒\n转场: 硬切\n镜头: 中景/平视/缓慢后拉\n画面描述:伙计们面面相觑后纷纷让开一条路,包围圈出现缺口,倒在泥地中的林止抬头望去。\n核心动作: 人群退让,新的关系入口被打开。\n对手戏设计: 众伙计由面向林止转为侧身分开,视觉通道指向画外来者。\n人物底层逻辑: 群体从攻击姿态切换为等待解释的悬停状态。\n主体动作:伙计们收回扁担和绳索,左右分开,林止撑着泥地抬起头,视线沿着被让开的路看向前方。\n音效: 布鞋后撤声,绳索轻落声,人物压低的惊疑吸气声\n资产对照表(视频中涉及的角色、场景与物品如下(保持一致性)):\n\n 视频场景对照表: 大燕朝河岸边=【图1】\n视频角色对照表: 林止=【图2】;伙计甲=【@伙计甲/粗布短打】;伙计乙=【@伙计乙/粗布短打】;伙计若干=【@伙计若干/粗布短打】,薛琴=【图3】\n视频道具对照表: 粗绳=【图4】;扁担=【图5】", "reference_images": ["references/443fcbb76e93fbbf6bf0acf4e8aedf38a372b1d8a2bc693956f67474ec829239.jpg", "references/0a1ca2acfef8f916e5fdf87d426bfa7b499a0d7032008b8367859f489ea39240.jpg", "references/2240651497f2fe34add88f6ab5606fc7156b6739dc63fe161bc31b423edb3007.jpg", "references/6385b934b869947bbcffdd92732da934119da38475c8aea5c1ae6c0abee4bcdb.jpg", "references/a78559a1a70c161b8f9a021a3f758dbde67cdd87d95a8a4dec144e8c8a719f90.jpg"], "reference_sha256": ["443fcbb76e93fbbf6bf0acf4e8aedf38a372b1d8a2bc693956f67474ec829239", "0a1ca2acfef8f916e5fdf87d426bfa7b499a0d7032008b8367859f489ea39240", "2240651497f2fe34add88f6ab5606fc7156b6739dc63fe161bc31b423edb3007", "6385b934b869947bbcffdd92732da934119da38475c8aea5c1ae6c0abee4bcdb", "a78559a1a70c161b8f9a021a3f758dbde67cdd87d95a8a4dec144e8c8a719f90"], "workflow": "ref2va", "prompt_sha256": "64c58f448c71e975490d8168a19ab81184e73fddf131ae6d13935a30dd213d3b", "original_prompt_sha256": "6ddf545d5f95abaad35f0bca2f90c3e54475b975d0c562c93365bed73784c029", "video_sha256": "e8d12e0343c4a2a1a6b2cbd802764379d5a6ba8247ef04dde5bffb5be5943c42", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_015.json", "verification": "verification/sample_015.json"} {"file_name": "videos/sample_016.mp4", "sample_id": "#16", "prompt": "subject_definitions:\nSubject 1: The female model featuring elegant large wavy curly hair and exquisite evening makeup, appearance from .\nSubject 2: The garment, a knitted fabric dress with side slits, three-dimensional pleats at the waist forming a slim silhouette, high elasticity, and heavy-duty luster, clothing from .\n\nsummary:\n[reference generation] A female e-commerce marketing video for a women's clothing product detail page to promote purchase conversion. The model, featuring elegant large wavy curly hair and exquisite evening makeup, wears the exact garment from against the background from . The video showcases the garment's details, fabric texture, and fit through various camera movements, adhering to physical laws and e-commerce content norms, with accurate color reproduction and lazy urban jazz background music.\n\nretention_analysis:\nThe model's identity and appearance from are fully preserved. The exact garment design, color, and details from are fully preserved, with strict instructions to avoid any color shift or deformation. The background from is fully preserved, with lighting adjusted to serve the true color of the garment.\n\ndetailed_description:\nConforms to physical laws; prohibits clipping or unreasonable scenes such as the model standing on tables or chairs; clothing coverage must be complete, and the wearing outline must be flat without bulges; conforms to e-commerce platform content norms. The garment color must be strictly based on with no changes or color differences allowed, and the background lighting must serve the true color of the garment, prohibiting the garment from turning gray, darkening, or color-shifting due to the tone of .\n\n[Shot 1] (0.000s - 4.000s) Long shot. The model, wearing the garment from , walks gracefully from the light and shadow in the background into the center of the frame. The knitted fabric captures a flowing metallic luster in the sunset as she walks, showcasing the dynamic beauty of the side slits. The camera follows and pushes in.\n\n[Shot 2] At 0:04.000 (4.000s - 8.000s), medium shot, orbiting. The model continues walking and naturally looks back, then elegantly turns sideways, her hands hanging down and swinging naturally. The camera captures the light and shadow of the three-dimensional pleats formed by the waist craftsmanship, comprehensively showcasing the slim silhouette and fabric elasticity.\n\n[Shot 3] At 0:08.000 (8.000s - 12.000s), medium close-up, focusing on the upper body and waist. The model slightly adjusts her center of gravity and arranges her hair, displaying the details of the neckline and shoulders. The light sweeps across the knitted texture, highlighting the high-elastic texture and heavy-duty luster, and finally walks towards the camera to conclude.\n\noverall_soundscape:\nNo specific ambient sound effects are specified.\n\nnon_diegetic_music:\nLazy urban jazz style pure instrumental music, no human vocals.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。模特配以优雅大波浪卷发与精致晚宴妆。服装严格使用图二,颜色/版型/细节不允许有任何改变或色差;背景直接使用图三。符合物理规律,禁止穿模/模特站桌椅等不合理画面;服装覆盖完整、穿着轮廓平整无凸点;符合电商平台内容规范。服装颜色以图二为准,背景光照服务服装真实颜色,禁止因图三色调致服装变灰/暗化/偏色。配乐:纯音乐,无人声,慵懒都市爵士风格。 [0-4s]远景,模特穿着图二服装由背景光影处步入画面中心,步伐从容。针织面料随走动在夕阳下捕捉流动金属光泽,展现侧开叉的灵动美感,镜头跟随推近。 [4-8s]中景环绕。模特持续行进并自然回眸,随后优雅侧身,双手垂落自然摆动,镜头捕捉腰部工艺形成的立体褶皱光影,全方位展现修身廓形与面料弹性。 [8-12s]中近景,聚焦上半身与腰部。模特微微调整站姿重心并整理发丝,展示领口与肩部细节。光线掠过针织纹理,凸显高弹质感与重工光泽,最后向镜头走近收尾。\\n\\n素材说明:图1=模特参考(asset-20260224194849-qt54s);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552.jpg", "references/dfe28665dd2a0e73b0ab55de4e48f6c5d0ffeeb69e06f7dbd9e681421b343543.jpg", "references/316dd4ff61d55e516bed652710b3d57ef4166b3ce8d804ce0ca73219bf86db42.jpg"], "reference_sha256": ["96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552", "dfe28665dd2a0e73b0ab55de4e48f6c5d0ffeeb69e06f7dbd9e681421b343543", "316dd4ff61d55e516bed652710b3d57ef4166b3ce8d804ce0ca73219bf86db42"], "workflow": "ref2va", "prompt_sha256": "8b180a5435c96360f4d202f64c30f5a8b3f23979e1747e775cee8015e203303e", "original_prompt_sha256": "14820de8c6ceab10d0f049574229eedd7994166d4279e201a0d62b207dc66009", "video_sha256": "b10b88b9b3774deef2b25eef810d40c86c865f7d0ed4743319d414a22c53fdbe", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_016.json", "verification": "verification/sample_016.json"} {"file_name": "videos/sample_017.mp4", "sample_id": "#17", "prompt": "subject_definitions:\n is a female model whose appearance, featuring an elegant retro updo hairstyle and exquisite makeup, strictly matches . The clothing worn by strictly matches the color, style, and details from , with no alterations or color differences allowed. The background strictly matches .\n\nsummary:\n[reference generation] A women's clothing e-commerce product detail page marketing video designed to promote purchase conversion. The video showcases the model wearing the specified clothing in the specified background, adhering to basic physical laws without any clipping or unrealistic poses such as standing on furniture. The model's hands remain out of pockets throughout, with natural movements.\n\nretention_analysis:\nThe model's appearance from is fully preserved, including the elegant retro updo and exquisite makeup. The clothing from is fully preserved with strict adherence to color, style, and details. The background from is fully preserved. The specific camera movements and shot compositions across the three timed shots are fully preserved.\n\ndetailed_description:\n[Shot 1] 0-4 seconds: Medium-long shot. confidently walks into the frame from the ambient light and stands elegantly. She demonstrates the excellent elasticity and fitted style of the knitted fabric top through subtle weight shifts. The camera smoothly pushes in. [Shot 2] At 0:04.000: Medium shot. displays the front and back views of the clothing from . She turns her head to look back over her shoulder and lightly brushes her hair. The environmental side backlight captures the diamond geometric texture and semi-transparent feel of the fabric. The camera performs a lateral translation. [Shot 3] At 0:08.000: Close-up shot focusing on the upper body. lightly adjusts the high collar and the lace ruffles at the cuffs, fully showcasing the delicate texture and craft details of the lace mesh. Her movements have natural ups and downs. The camera slowly sinks and pushes in to finish. The overall color tone serves the clothing, with light and shadow flowing naturally and realistically.\n\noverall_soundscape:\nSubtle ambient sounds of a fashion studio or indoor environment, supporting the natural movements and fabric interactions.\n\nnon_diegetic_music:\nLazy retro jazz, pure instrumental music with no vocals.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。服装严格使用图二,颜色、版型、细节不允许有任何改变或色差;背景直接使用图三。符合基本物理规律,禁止出现穿模或模特站在桌椅上等不合理画面。模特全程禁止双手插兜,动作自然。图一模特搭配优雅复古盘发与精致妆容。 0-4秒:中远景,模特从环境光中自信步入并优雅站立,通过重心微移展现上装针织面料的极佳弹力与贴身版型,镜头平稳推近。 4-8秒:中景切换,模特展示图二中的正背面视角,做侧身回眸与轻拨发丝动作,利用环境侧逆光捕捉面料的菱形几何纹理与半透明透视感,镜头横移。 8-12秒:近景聚焦上半身,模特轻理高领及袖口木耳边,充分展现蕾丝网纱的细腻质感与工艺细节,动作有起有落,镜头慢速下沉推近收尾。 整体色调为服装服务,光影流转真实自然。配乐:纯音乐,无人声,慵懒复古爵士乐。\\n\\n素材说明:图1=模特参考(asset-20260224194849-qt54s);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552.jpg", "references/845d8283a7cc116a4325e9943e6585cbe12caae6357e71add08bcd7f50c0f21c.jpg", "references/b463526dae8796106fb1e7d2f12ab853ef550d25d9fdf4e4e36cb4351f9b0644.jpg"], "reference_sha256": ["96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552", "845d8283a7cc116a4325e9943e6585cbe12caae6357e71add08bcd7f50c0f21c", "b463526dae8796106fb1e7d2f12ab853ef550d25d9fdf4e4e36cb4351f9b0644"], "workflow": "ref2va", "prompt_sha256": "5ecd2b479093201df9c5a85e214be2531804803291f4628ad51e372067272c33", "original_prompt_sha256": "70caae1be36314cb2bd770ccdbef708e97a2be8400ce23029679a7993c5b55c8", "video_sha256": "17fdb10790d641f03c37d195c746bed590ef474b3475c0e3c74daec9a6fcfdab", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_017.json", "verification": "verification/sample_017.json"} {"file_name": "videos/sample_019.mp4", "sample_id": "#19", "prompt": "subject_definitions:\nSubject 1 is the person whose appearance and identity are derived from . Subject 2 is the product whose appearance and details are derived from .\n\nsummary:\n[reference generation] Generate a commercial product promotion video using the person from and the product from . Two bags of the product appear in the scene: one held in the person's hand and one placed on the table. Both products are replaced with the product from .\n\nretention_analysis:\nSubject 1 (person from ) is fully_preserved: the person's face must remain stable, undeformed, and highly consistent throughout the video. Subject 2 (product from ) is fully_preserved: the product details must be clear, naturally fit the hand when held, and appear accurately on the table.\n\ndetailed_description:\n[Shot 1] A commercial product promotion video featuring the person from . The person holds one bag of the product from in their hand, with clear product details that naturally fit the hand and are naturally gripped. Another bag of the product from is placed on the table. The person's face is stable, undeformed, and highly consistent. Product details are clear. 4K high definition, enhanced image quality, commercial product promotion texture. No stickers, logos, watermarks, shopping carts, or other irrelevant screen text appear. No self-created text is generated.\n\noverall_soundscape:\nThe audio is primarily driven by a clear, stable, and prominent voiceover human voice located in the foreground. If background music is present, its volume is less than 20% of the voiceover volume, ensuring it does not cover, interfere with, or affect the clarity of the voiceover. Every spoken sentence is clearly discernible, natural, and comfortable.\n\nnon_diegetic_music:\nBackground music, if generated, must not exceed 20% of the voiceover volume and must not interfere with voiceover clarity.", "original_prompt": "将@视频1 中的人物替换为@图片1 中的人物,将@视频1 手中和桌面展示的产品替换为@图2 中的产品;注意:@视频1 中的产品有两袋,分别为人物手中持有的产品和桌上摆放的产品,都需要替换成@图2 中的产品;严格保留原视频的场景、背景、灯光、运镜、镜头角度、节奏、时长、转场和手部动作;注意手持的产品细节,确保产品清晰;人物面部稳定、不变形、一致性强;产品细节清晰、贴合手部、自然握持;仅替换人物和产品,其余元素完全不变;4K高清、画质增强、商业带货质感。\n禁止出现贴纸、LOGO、水印、小黄车或其他无关屏幕文字,禁止生成自创的文字。\n生成音频时,请始终以口播人声为主,确保人声清晰、稳定、突出,位于前景。严禁保留原视频音乐;如使用上传音频或必须生成背景音乐,音量必须小于口播人声,且不得超过口播人声音量的20%,不得盖过、干扰或影响口播清晰度,确保整段音频中的每一句话都清楚可辨、自然舒适。", "reference_images": ["references/d0fd95c586b2033caa05789407dc9a4a547b22517c13d3eaf74f87e9249476fc.jpg", "references/acf72c9e7f81069404f4b968ee40ca54be7951ef93b10daf3cd93ac344527caf.jpg"], "reference_sha256": ["d0fd95c586b2033caa05789407dc9a4a547b22517c13d3eaf74f87e9249476fc", "acf72c9e7f81069404f4b968ee40ca54be7951ef93b10daf3cd93ac344527caf"], "workflow": "ref2va", "prompt_sha256": "708073d576934dbb59b7649e702d04285a8d77f07b03cd415418adb67b886089", "original_prompt_sha256": "22c129f06ab11fc3aa83d028af3c8b84153dfa0a5076a1c5b0583873a6c6f75b", "video_sha256": "b1c36422c9ea6373229a4897b9540b2f492b2d167ffdc576d514029ac2fe9f47", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 209, "duration_seconds": 8.708333333333334, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_019.json", "verification": "verification/sample_019.json"} {"file_name": "videos/sample_020.mp4", "sample_id": "#20", "prompt": "subject_definitions:\n: The virtual model, featuring the face, standard figure, French elegant updo, and red lip makeup from .\n: The clothing, strictly matching the color, style, and details from , with no alterations or color differences allowed.\n: The background, directly matching .\n\nsummary:\n[reference generation] A women's clothing e-commerce product detail page marketing video designed to promote purchase conversion. The video features a virtual model wearing a specific outfit against a specific background, showcasing the clothing's details through a series of shots.\n\nretention_analysis:\n is fully preserved as the identity and appearance of the virtual model (). is fully preserved as the exact clothing worn by the model (), with strict adherence to its color, style, and details. is fully preserved as the background environment ().\n\ndetailed_description:\n[Shot 1] (0-4s) Full shot. elegantly walks in from the deep background of . The walking motion captures the elastic recovery of the knitted fabric and the structured silhouette of the A-line skirt of . The camera moves smoothly, following the model.\n[Shot 2] At 00:04.000, (4-8s) Medium shot. elegantly turns sideways and looks back, hands naturally hanging or lightly touching the clothing. The shot focuses on the three-dimensionality of the cream-colored large bow on the chest and the satin micro-gloss of , reflecting the high-end black and white color contrast. The camera slowly translates laterally in an orbiting motion.\n[Shot 3] At 00:08.000, (8-12s) Close-up. The camera focuses on the upper body of , who slightly adjusts her hair to show the one-shoulder off-shoulder design and shoulder-neck lines of . The 5% spandex brings natural fitting tension in movement. The lighting is soft. The camera slowly pulls back and freezes.\nThe background and environmental lighting serve the clothing color tone, ensuring pure black and white contrast and realistic picture texture. The overall scene conforms to basic physical laws; no clipping (such as the character penetrating tables or chairs) or the model standing on tables/chairs is allowed. The model's actions must not include hands in pockets.\n\noverall_soundscape:\nNo specific ambient sounds or sound effects are specified beyond the music.\n\nnon_diegetic_music:\nPure instrumental music, no vocals, bright and elegant French jazz piano.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。图一虚拟模特人脸,标准身材,搭配法式优雅盘发与红唇妆容。服装严格使用图二,颜色、版型、细节不允许有任何改变或色差。背景直接使用图三。整体符合基本物理规律,禁止出现穿模(人物穿透桌椅等物体)或模特站在桌椅上等不合理画面,模特动作禁止插兜。 分镜设计: 1.(0-4s)全景。模特从背景深处优雅步入,利用行走动作捕捉针织面料的弹性回复力与A字裙摆的挺括廓形。运镜平滑跟进。 2.(4-8s)中景。模特优雅侧身回眸,双手自然垂放或轻抚衣身,重点展示胸前奶油色大蝴蝶结的立体感与缎面微光,体现黑白撞色高级感。运镜慢速横移环绕。 3.(8-12s)近景特写。镜头聚焦上半身,模特微调发丝展示一字领露肩设计与肩颈线条。5%氨纶带来的合体张力在动作中自然呈现,光影柔和。运镜缓慢拉远定格。 背景与环境光为服装色调服务,确保黑白对比纯正,画面质感真实。配乐:纯音乐,无人声,明快优雅的法式爵士钢琴。\\n\\n素材说明:图1=模特参考(asset-20260224194849-qt54s);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552.jpg", "references/c735a4c064723482a662afa4f9120391bd4e38d02e4d34ad6af34751769cf3d0.jpg", "references/026ed98e09c8f8eafc487848f0d5a3bff59a45b1b5c51ebf04fbda91d0b6aa89.jpg"], "reference_sha256": ["96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552", "c735a4c064723482a662afa4f9120391bd4e38d02e4d34ad6af34751769cf3d0", "026ed98e09c8f8eafc487848f0d5a3bff59a45b1b5c51ebf04fbda91d0b6aa89"], "workflow": "ref2va", "prompt_sha256": "90a2eb0b2ab60ef69937e7eb2ee523e1971563009d45baf89b0678595d153372", "original_prompt_sha256": "8c3877181b3ba89e51bc821304f88b4ce4d94f49eace264a63cc951d30f0aeed", "video_sha256": "bb0c8d1484e6b3f7dd77278d6e2a39908d14ee8872c5be3fd43598997789ec36", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_020.json", "verification": "verification/sample_020.json"} {"file_name": "videos/sample_022.mp4", "sample_id": "#22", "prompt": "subject_definitions:\n (Chen Ping) is a male character wearing a suit, with appearance and clothing from . (Su Yuqi) is a female character wearing a white dress, with appearance and clothing from .\n\nsummary:\n[reference generation] A cinematic 3D CGI scene showing two characters exiting a parked taxi. Close-up shot captures Chen Ping exiting from the front door and Su Yuqi exiting from the rear door. Both characters stand in place, cough twice while frowning from being choked, and Su Yuqi waves her hand in front of her nose.\n\nretention_analysis:\nfully_preserved: (Chen Ping in suit from ) and (Su Yuqi in white dress from ) are the main characters whose identities and appearances must be retained throughout the scene. and serve as positional and environmental references for the taxi location and seating arrangement.\n\ndetailed_description:\n[Shot 1] Close-up shot. A taxi is parked in place. (Chen Ping) opens the front door and gets out of the car from the front seat. (Su Yuqi) opens the rear door and gets out of the car from the rear seat. The two stand in place, choked and unable to help but frown and cough twice. (Su Yuqi) waves her hand in front of her nose. The entire video uses high-end 3D CGI character rendering style. The scene maintains its original cinematic realistic environment style, focusing on the contrast between digital characters and the realistic environment. Characters have idealized facial features, presenting a slight digital luster through perfect subsurface scattering (SSS) skin effects, with clear outlines. Hair is sculpted and clean. Characters move in the original realistic environment, where reflections and lighting match the real location. Masterpiece quality, clean shadows, cinematic volumetric light. Throughout the video, it is forbidden to have two people with exactly the same appearance, clothing, and accessories, eliminating twin-like characters, no repeated characters, clones, or twin effects. UE5 rendering, high-end 3D CGI texture. Avoid facial distortion. Face stable without deformation, clear features, normal human body structure, natural and smooth movements, not stiff, no stuttering, no flickering. No red lines. No subtitles. High completion.\n\noverall_soundscape:\nNatural ambient sounds and on-site sound effects. Atmospheric sound effects to increase scene sense. Coughing sounds from the characters.\n\nnon_diegetic_music:\nN/A", "original_prompt": "生成电影级视频,UE5渲染,高端3d cgi质感,避免人物面部畸变,加入自然环境音和现场音效,氛围音效增加场景感。不要添加音乐。生成的视频不要有字幕,高完成度。面部稳定不变形、五官清晰、人体结构正常、动作自然流畅、不僵硬、画面无卡顿、无闪烁。不要出现红线。【整个视频中的所有人物角色都应用高端 3D CGI 角色渲染风格。场景保持其原始的电影写实环境风格,重点展示数字角色与现实环境之间的对比。人物应具有**理想化的五官特征,通过完美的次表面散射 (SSS) 皮肤效果呈现出微小的数字化光泽,具有清晰的轮廓。发丝必须具有雕琢感和干净。人物在原始写实环境中移动,其中的反射和照明与现实位置相符。杰作质量,干净的阴影,电影感的体积光。视频全程禁止出现外形、着装、配饰完全相同的两个人,杜绝双胞胎同款人物,不出现重复人物、分身或双胞胎效果;】具体情节如下:\n@图1为出租车停靠的位置参考,@图2为前排乘客陈平,@图3为后排乘客苏雨琪,@图4为他们在车上时的座位参考,\n近景拍摄,出租车停在原地,陈平打开前排门从车前排下车,苏雨琪打开后排车门从车后排下车,两人站在原地被呛得忍不住皱眉咳嗽两声,苏雨琪拿手在鼻子前扇着风。\n\n图片1是【车辆位置参考】,图片2是【陈平(西装)】,图片3是【苏雨琪(白色衣裙)3】,图片4是【出租车参考】。", "reference_images": ["references/e1614661dbe88590fe94872cc2086a7770d6c70618403ca95073acc31f23499c.jpg", "references/3f97a47c4176dcffe7e6cd2ca0932831ddcf4282e96f1faa6de4f5fba027b714.jpg", "references/d232733b838ca364d0867ac250a76a45b3ee828947a223794a68c7b65891a403.jpg", "references/937ab8702a9e8b2f1c3a974780d5687a680dcc47f00b43ca2ebda4df96bd1ec6.jpg"], "reference_sha256": ["e1614661dbe88590fe94872cc2086a7770d6c70618403ca95073acc31f23499c", "3f97a47c4176dcffe7e6cd2ca0932831ddcf4282e96f1faa6de4f5fba027b714", "d232733b838ca364d0867ac250a76a45b3ee828947a223794a68c7b65891a403", "937ab8702a9e8b2f1c3a974780d5687a680dcc47f00b43ca2ebda4df96bd1ec6"], "workflow": "ref2va", "prompt_sha256": "3313cc22368df6b7d886f3967eaaf2a6a93ee8504f23c5b3898fba450467682f", "original_prompt_sha256": "b504209b7949a950318daf9445f73456d17a608d6d098c0df6c72e3160f2c8c0", "video_sha256": "7b67b0567ba9d3ee6335f4b80e7006975669c59e90df275909b80190f6ae9de6", "task_adapted": false, "width": 1344, "height": 768, "fps": 24, "num_frames": 124, "duration_seconds": 5.166666666666667, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_022.json", "verification": "verification/sample_022.json"} {"file_name": "videos/sample_023.mp4", "sample_id": "#23", "prompt": "subject_definitions:\n is the light blue polo and skirt from .\n\nsummary:\n[reference generation] A static shot of a light blue polo and skirt neatly folded on a white bedsheet, with a pair of sunglasses and a tennis ball next to them, under natural afternoon light, with no subtitles, no captions, no on-screen text, and no text overlay.\n\nretention_analysis:\nThe light blue polo and skirt from are fully preserved as the main subject of the scene.\n\ndetailed_description:\n[Shot 1] A static shot of neatly folded on a white bedsheet. A pair of sunglasses and a tennis ball are placed next to it. The scene is very calm, illuminated by natural afternoon light. There are no people in the frame. The footage is realistic and cinematic, featuring natural movement, a clean frame, no subtitles, no captions, no on-screen text, no CC, no text overlay, no text, no watermark, and no social media UI. No AI artifacts, no uncanny valley, no plastic skin, no distorted hands or faces.\n\noverall_soundscape:\nUnspecified.\n\nnon_diegetic_music:\nN/A", "original_prompt": "No subtitles, no captions, no on-screen text,No CC. 禁止出现字幕:屏幕不出现任何字幕!避免生成字幕,保持无字幕! The light blue polo and skirt @Image1 neatly folded on a white bedsheet. A pair of sunglasses and a tennis ball next to it. Static shot, very calm, natural afternoon light. no people. no subtitles, no text overlay, no captions. realistic cinematic footage, natural movement, clean frame, no subtitles, no captions, no text, no watermark, no social media UI. No AI artifacts, no uncanny valley, no plastic skin, no distorted hands or faces. 禁止出现字幕:屏幕不出现任何字幕!No CC!", "reference_images": ["references/e2233dd3af06ecb75dc3556fc9e9e52ece3e130b7861bb2457bd06e84719c371.jpg"], "reference_sha256": ["e2233dd3af06ecb75dc3556fc9e9e52ece3e130b7861bb2457bd06e84719c371"], "workflow": "ref2va", "prompt_sha256": "72cdde05ce4f10dd52ed62ee156ea088507e22b8606f16635fed3e2a884d0baf", "original_prompt_sha256": "32b9c71854ee8b7e27a183b9ef2b209a71df46506eac8f2da5c376fd38ec1417", "video_sha256": "da6ecf39c81fa38a2233d37070184ce03bcb7d3c79ffadc330fd0318d038befe", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 107, "duration_seconds": 4.458333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_023.json", "verification": "verification/sample_023.json"} {"file_name": "videos/sample_024.mp4", "sample_id": "#24", "prompt": "subject_definitions:\n is the male lead, appearance and clothing from . is the female lead, appearance and clothing from .\n\nsummary:\n[reference generation] A cinematic 3D Chinese fantasy epic scene set in an empty cave at night. The injured male lead desperately grabs the female lead's dress from the ground. The female lead turns to look down at him with cold calmness. The male lead lies dying on the ground, weakly begging for help.\n\nretention_analysis:\nThe male lead's identity and appearance are fully preserved from . The female lead's identity and appearance are fully preserved from . Both subjects are actively engaged in the scene with specific actions and dialogue as requested.\n\ndetailed_description:\nNo subtitles, subtitle transparency 100%. Rich picture layers, three-dimensional and transparent lighting, fine 3D modeling, Chinese fantasy epic texture, smooth and coherent shot transitions, strong visual impact, 8K ultra-high definition, extreme details, realistic cinematic lens texture, cinematic composition, global ray tracing, volumetric lighting, realistic environmental texture, modern cinematic color grading, depth of field effect, ultra-clear resolution, transparent and natural picture, cinematic blockbuster atmosphere, flawless composition, realistic lighting layers. Gorgeous and aesthetic special effects, scene is an empty cave at night.\n\n[Shot 1] Extreme close-up, extreme low angle slow push, ground perspective. (S1)'s slender, pale hand (only the hand) suddenly enters the frame, quickly and fiercely grabbing the corner of (S2)'s dress dragged on the ground. Ultra-low angle eye-level, ground perspective. The action is fast and sudden, knuckles white from force. Eerie and oppressive atmosphere, cold tone lighting. S1 says in an off-screen voiceover with no strength [Chinese]求你, highlighting his will to survive after being injured.\n\n[Shot 2] Medium to close shot, back-side low angle, medium shot, urgent push shot. S2 suddenly turns her head sharply to look down. The foreground left shows S1's head back to the camera, foreground blurred. The camera urgently pushes towards S2, only S2 in the frame. S2 has a cold and calm expression, not affected by S1 at all. Her hair slightly flies with the movement, side backlight outlines her silhouette, creating a strong visual impact. S2 calmly looks down at S1 and says [Chinese]是你唤醒了吾?.\n\n[Shot 3] High angle shot of S1, dying vulnerability and broken feeling maxed out. S1 weakly lies on the ground with blood on his face, eyes scattered and slightly dazed, full of pain, yet with a desire to survive in the depths of his eyes. He is slightly trembling, breathing rapidly, hands weakly hanging, cold light on his face, highlighting his weakness and showing his handsomeness. S1 very weakly says [Chinese]救……我……, full of atmosphere even when injured.\n\noverall_soundscape:\nEerie and oppressive cave ambience, cold atmospheric sounds, subtle rustling of fabric, weak breathing, and faint echoing cave acoustics.\n\nnon_diegetic_music:\nN/A", "original_prompt": "无字幕!字幕透明度为100%,画面层次饱满,光影立体通透,3D 建模精细,国风玄幻史诗质感,镜头衔接流畅连贯,视觉冲击力强,8K超高清,极致细节,影视剧写实镜头质感,电影级构图,全局光影追踪,体积光,真实环境质感,现代影视调色,景深效果,超清分辨率,画面通透自然,影视大片氛围感,无瑕疵构图,写实光影层次, @图片2 为女主,@图片1 为男主,特效华丽唯美,场景是夜晚空旷的山洞里\n镜头一:大特写,极限低角度缓推,地面视角,男主 一只纤细苍白的手(只要手部),突然入画,快速的狠狠抓住女主拖到地上的裙角,超低角度平视,地面视角,动作很快,很突然,指节因用力泛白,氛围阴森压抑,冷色调打光,有气无力的男主的画外音:求你,突出男主在受伤后的求生欲 镜头二:(中景—近景)背侧仰拍,中景,推镜头,急推镜头,女主突然猛地回头向下看,画面前景左侧是男主的头背对镜头,前景虚化。镜头向女主急推过去,画面中只有女主,女主眼神冷漠的平静的表情。半点都没有受到男主的影响,表情平静,发丝随动作轻扬,侧逆光勾勒出轮廓,形成强烈视觉冲击,女主表情平静的低头看着男主说到:“是你唤醒了吾?”镜头三:俯拍男主 ,濒死脆弱感破碎感拉满,男主虚弱的躺在地上脸上带血,眼神涣散微微迷离满是痛苦,眼底却带着求生的渴望,微微颤抖、呼吸急促、无力垂手、冷光打在脸上,突出虚弱感,展现男主的帅气,非常虚弱的说:“救……我……”,即使受伤了也满满的氛围感", "reference_images": ["references/d278fdb7596198393b9f0bb386e089ba163166c3d10b0d2a27bbe8918532911f.jpg", "references/75294961f854566cc4ba3082fab4efa3311178ea1f371bac5df35b5dc946171d.jpg"], "reference_sha256": ["d278fdb7596198393b9f0bb386e089ba163166c3d10b0d2a27bbe8918532911f", "75294961f854566cc4ba3082fab4efa3311178ea1f371bac5df35b5dc946171d"], "workflow": "ref2va", "prompt_sha256": "0d66e95967927a46080cc75c98b6123f0a8e28266c85120b01efd9c6857af8f8", "original_prompt_sha256": "1942021047eb5b8cd624eb52512cdeab9c17608f37a56fb6bde3b78e9da6cbc7", "video_sha256": "9ea599b7f3129744179855d0ee1468ecd44fb08ccf9e1acd97e0efa37268918a", "task_adapted": false, "width": 1344, "height": 768, "fps": 24, "num_frames": 243, "duration_seconds": 10.125, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_024.json", "verification": "verification/sample_024.json"} {"file_name": "videos/sample_025.mp4", "sample_id": "#25", "prompt": "subject_definitions:\nSubject 1: Xue Niannian, appearance and clothing from .\nSubject 2: Xiao Zining, appearance and clothing from .\n\nsummary:\n[reference generation] A close-up, high-angle shot of Subject 1 (Xue Niannian) standing in a hospital ward, holding a baby and delivering a hypocritically pitying line, styled with a cinematic Korean drama aesthetic using specific camera and lighting setups.\n\nretention_analysis:\nSubject 1 (Xue Niannian) is fully preserved as the primary character performing the action and speaking in the shot, utilizing for identity. Subject 2 (Xiao Zining) is weakly referenced; while defined via , she is not explicitly placed in the visual action of this specific shot. The hospital ward setting from is fully preserved as the scene background.\n\ndetailed_description:\n[Shot 1] A close-up, high-angle looking down shot with a fixed camera. The scene is set in an afternoon hospital ward illuminated by cold white light. (S1), matching the appearance from , is holding a baby and standing straight on the left side of the hospital bed. She puts on a hypocritically pitying expression, her eyebrows slightly furrowed, and gently shakes her head. The visual style features a cinematic Korean drama aesthetic, shot with an ARRI Alexa Mini LF camera and a Zeiss Supreme Prime 85mm T1.5 lens at f/1.5 aperture, creating an extremely shallow depth of field with creamy smooth bokeh. The color grading uses a dual-tone technique with elegant cool-toned shadows and warm amber highlights (teal and orange LUT). Soft diffused natural ambient light is complemented by subtle wrap-around fill light on the face, presenting idealized skin tones with subsurface scattering. A controlled fine film grain overlay, slightly crushed blacks, and elevated airy highlights create a prestigious TV visual style, fostering a melancholic and intimate atmosphere. The image is requested in 8K resolution and hyper-realistic quality. (S1) says in a hypocritically pitying voice [Chinese]就算他现在不和你离婚,以后也总会和你离婚的。\n\noverall_soundscape:\nNo sound effects are specified.\n\nnon_diegetic_music:\nN/A", "original_prompt": "镜号:14 | 场次:午后 | 医院病房 | 0-4s | 近景,俯视 | 固定镜头 | 午后病房冷白光, @薛念念 抱着婴儿直直站在病床左侧,装出一副怜悯的虚伪模样,眉头微蹙,轻轻摇了摇头。 | 台词 & 音效:Dialogue:[用假惺惺怜悯的声音说] @薛念念 :\"就算他现在不和你离婚,以后也总会和你离婚的。\" SFX:无 | 备注:无\n\n参考图片说明:\n图1: 角色: 薛念念\n图2: 角色: 肖紫宁\n图3: 场景: 医院-病房\n影调:具有电影质感韩剧的审美风格,采用ARRIAlexa MiniLF相机拍摄,搭配蔡司SupremePrime85mmT1.5镜头,光圈f/1.5,具有极浅的景深效果,奶油般顺滑的虚化效果,采用双色调色彩分级技术,带有淡雅冷色调阴影与温暖琥珀色亮部的处理(青绿与橙色LUIT),柔和散射的自然环境光,面部有细微的包裹式补光,理想化的肤色呈现带有次表面散射效果,控制得当的精细胶片颗粒叠加,轻微压暗黑色调,提升飘逸的亮部质感,呈现具有声望感的电视视觉风格,营造出忧郁而亲密的氛围,8K分辨率,极度逼真", "reference_images": ["references/2671311d5615f7724e99be0bd40428b8bd772c0c27fef12db095925b11246559.jpg", "references/c58afbf2fb250b2e42944e2a436c0c1c13abc78b83df4d08e083d72dd232b2dc.jpg", "references/edbca9cb366d7aca0b2c877f1376216ecaac69fe3dd1dc7b8511653acf16707b.jpg"], "reference_sha256": ["2671311d5615f7724e99be0bd40428b8bd772c0c27fef12db095925b11246559", "c58afbf2fb250b2e42944e2a436c0c1c13abc78b83df4d08e083d72dd232b2dc", "edbca9cb366d7aca0b2c877f1376216ecaac69fe3dd1dc7b8511653acf16707b"], "workflow": "ref2va", "prompt_sha256": "b2be3ee7cc0ce3f2bfec6d439e9a4b7783638effa9fb4132418193ea3889c351", "original_prompt_sha256": "22b4f3cff619027cee9fd0d8f8ed0d70ba46fd6ccb12c6020645f2ced254909c", "video_sha256": "bd438890371fd0ab6835ec06d3857dbd44cce460bb582e6ae0deddc4563db95c", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 107, "duration_seconds": 4.458333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_025.json", "verification": "verification/sample_025.json"} {"file_name": "videos/sample_026.mp4", "sample_id": "#26", "prompt": "subject_definitions:\nSubject 1: Ariel, appearance and clothing from .\nSubject 2: Jack, appearance and clothing from .\nSubject 3: Living Cabin, environment from .\nSubject 4: Rest Cabin, environment from .\n\nsummary:\n[reference generation] A cinematic 3D animation sequence in a hardcore sci-fi wasteland and brutalist industrial style. Jack strides from the warm-lit Rest Cabin to the dark-toned Living Cabin during a violent vehicle shaking, followed by a close-up of his face and a shot of him passing Ariel, ending with an extreme close-up of Ariel's face.\n\nretention_analysis:\nThe prompt requires retaining the identities and appearances of Ariel () and Jack (), as well as the environments of the Living Cabin () and Rest Cabin (). Jack is the main subject in shots 1, 2, and 3, while Ariel is the main subject in shot 4 and appears in the background of shot 3. The environments are used as the settings for the respective shots. All four references are fully preserved as the core subjects and settings of the scene.\n\ndetailed_description:\nStyle base: Top-tier 3D game cinematic animation style, high-end aesthetic quality production, Western hardcore sci-fi wasteland and brutalist industrial aesthetics, UE5 cinematic masterpiece rendering. Characters must absolutely avoid realistic skin appearance, with flat facial lighting, clean non-greasy modeling texture, and no visible pores, wrinkles, or skin flaws to the naked eye. Environment and clothing retain extreme physical realistic PBR materials (matte gunmetal, rough concrete, mechanical battle damage), absolutely rejecting cheap plastic feel.\nLighting and effects: Cinematic color grading, extreme chiaroscuro, realistic physical light cutting and extreme cold-warm contrast. Rich dynamic particle effects, dense Tyndall volumetric light strongly penetrating fog. Extremely fine splash texture, ash, cold rain threads, and mud splashes presenting macro-level physical feedback. Mechanical and mud-water surfaces presenting sharp highlight diffuse reflection, 8K ultra-clear, cinematic deep depth of field, strong giant object oppression.\nAbsolute instructions: No watermarks, no text, no subtitles.\nOutside environment is night.\n\n[Shot 1] Third-person over-the-shoulder following shot. The front of the frame shows the protagonist (Jack)'s extremely broad shoulders and back. Jack suddenly turns and strides from the warm-lit (Rest Cabin) towards the dark-toned (Living Cabin) directly ahead. His steps are extremely steady. The hem of his black tactical coat, due to his rapid pace and the vehicle's violent bumping, produces extremely heavy and realistic cloth swinging and draping physical simulation effects. The dim blue LED lights on both walls of the corridor continuously sweep across his leather coat shoulders as he advances, forming dynamic realistic highlight flow.\n\n[Shot 2] Close-up low-angle following shot (camera synchronously retreating in front of the character). Jack is striding forward. The carriage is experiencing crazy earthquake-level shaking, and the camera accompanies the high-frequency tremor. The camera tightly locks on Jack's cold and handsome face. His brows are slightly furrowed, and his eyes reveal absolute oppression and fearlessness. As he walks into the (Living Cabin) corridor, the cold white and dim blue alternating environmental light strips on the cabin ceiling continuously sweep across his jawline and high nose bridge, forming an oppressive and tense flickering light cutting effect. Cinematic lighting.\n (S1) says [English]How big?.\nSound effect: Heavy metal vibration sound transmitted from the chassis.\n\n[Shot 3] Close-up fixed shot with extremely strong spatial depth. The camera is fixed above the metal coffee table in the (Living Cabin). During the violent carriage shaking, Jack's tall and mighty black silhouette strides briskly from the left side of the frame to the right side, bringing a strong wind sound from his coat hem. In the right rear of the frame (blurred in depth of field / Bokeh), (Ariel), wearing a white tactical suit, is sitting steadily on a black leather sofa. When Jack crosses the foreground, the frame focus produces a momentary physical breathing feel, reflecting top-tier 3D animation camera scheduling and spatial 3D feel. Wasteland dark tech-wear environment.\n\n[Shot 4] Extreme close-up facial shot (Macro Shot). The frame is extremely stable, contrasting with the shaking of the surrounding environment. Close-up of (Ariel)'s flawless face. Her silver-white bangs slightly flutter with the shaking. Her eyes slowly pan to the right following Jack's movement trajectory. The environmental light forms sharp reflection spots on the edges of her pupils, showing extremely high-precision eyeball material rendering.\n\noverall_soundscape:\nHeavy metal vibration sounds from the chassis, strong wind sounds from the coat hem, and general ambient sounds of a violently shaking vehicle carriage.\n\nnon_diegetic_music:\nN/A", "original_prompt": "@图片3 是生活舱,@图片1 是爱丽儿,@图片4 是休息舱,@图片2 是杰克 \n【风格基底】 顶级3D游戏影视动画风格,高端美学质感制作,欧美硬核科幻废土与重工业粗野主义美学 (Brutalist Industrial),UE5院线级大作渲染。人物绝对避免写实皮肤观感,面部平铺打光,干净无油腻建模质感,肉眼不可见任何毛孔、皱纹或肌肤瑕疵。环境与服饰保留极致物理写实PBR材质 (哑光炮铜、粗糙混凝土、机械战损),绝对摒弃廉价塑料感。 【光影与特效】 电影级色调 (Cinematic color grading),极高明暗对比 (Chiaroscuro),写实物理光源切割与极致冷暖反差。丰富的动态粒子特效,浓郁的丁达尔体积光强力穿透迷雾。飞溅质感极其细腻 ,灰烬、冰冷雨丝与泥浆的飞溅呈现微距级物理反馈。机械与泥水表面呈现锐利的高光漫反射,8K超清,电影级深景深,强烈的巨物压迫感。【绝对指令】禁止生成水印,禁止生成文字,禁止生成字幕,只能生成音效;\n车外环境为黑天。\n【镜头 01d:】 [提示词]:第三人称越肩跟随镜头。画面正前方是主角杰克极其宽阔的肩背。杰克猛然转身,大步从暖光源的休息舱走向正前方暗黑色调的生活舱。他步履极其沉稳,黑色战术大衣的下摆因急促的步伐和车体的狂暴颠簸,产生极其厚重、真实的布料甩动与垂坠物理演算效果(Heavy cloth physics)。走廊两壁的幽蓝LED灯光随着他的前进,不断在他的皮衣肩膀上扫过,形成动态的写实高光流转。\n镜头 01e:】 [提示词]:特写仰视跟随镜头(摄影机在人物正前方同步后退)。杰克正大步向前行走,车厢正在经历疯狂的地震级摇晃,摄影机伴随高频震颤。画面死死锁定杰克冷酷、帅气的脸庞,他眉头微皱,眼神中透着绝对的压迫感与无畏。随着他走进生活舱走廊,舱顶的冷白色与幽蓝色交替的环境光带不断划过他的下颌线与高挺的鼻梁,形成压抑且紧张的“明灭”光影切割效果。电影级打光。\n[台词与音效]:台词:JACK: How big? (尺寸多大?) / 音效:底盘传来的沉重金属震荡声。\n【镜头 01f:】 [提示词]:近景固定镜头,带有极强的空间纵深感。摄影机固定在生活舱的金属茶几上方。在剧烈的车厢震动中,杰克高大威猛的黑色身影从画面左侧大步流星地穿过画面走向右侧,带起一阵强烈的衣摆风声。在画面的右后方(景深处被虚化处理/Bokeh),身穿白色战术服的爱丽儿正稳稳地坐在黑色皮质沙发上。当杰克穿过前景时,画面焦点产生瞬间的物理呼吸感,体现出顶级的3D动画镜头调度与空间立体感。废土暗黑机能风环境。\n【镜头 01g:】 [提示词]:极近脸部特写镜头(Macro Shot),画面极度平稳,与周围环境的震动形成反差。特写爱丽儿完美无瑕的面部。她的银白色刘海微微随震动飘动。她的眼神跟随杰克的移动轨迹向右侧缓慢平移。环境光在她瞳孔边缘形成锐利的反光斑,展现出极高精度的眼球材质渲染。", "reference_images": ["references/6e5aeab6db6ddff295192f522df3a0f8b064b2412846da44a84f769bd9f2a63c.jpg", "references/96944d298d9a7ec4da80ba793ac93e4ee038b9e440eb108858ad0684fb823e96.jpg", "references/9c31bbd7c7f7e0c9356ff0b8709f196aa2d4b47e2fe4e5f26edfcb07bfb21f94.jpg", "references/46be4b00f805e54c5e20c01bfea183add0adcf9bd652deeb92ca16a96f37fad0.jpg"], "reference_sha256": ["6e5aeab6db6ddff295192f522df3a0f8b064b2412846da44a84f769bd9f2a63c", "96944d298d9a7ec4da80ba793ac93e4ee038b9e440eb108858ad0684fb823e96", "9c31bbd7c7f7e0c9356ff0b8709f196aa2d4b47e2fe4e5f26edfcb07bfb21f94", "46be4b00f805e54c5e20c01bfea183add0adcf9bd652deeb92ca16a96f37fad0"], "workflow": "ref2va", "prompt_sha256": "6cf8e991d34dd40cd3fd067b4d76b31912950549056601c8d7f12854e8adf045", "original_prompt_sha256": "7be4c57c948abc8a9ca1ddeb110d35b20383d7f499650c5ab79b97203dc28136", "video_sha256": "ea9c891004ff0c09b78ba3ec8462665486b3a880890083f0cc4d19a92375deea", "task_adapted": false, "width": 1344, "height": 768, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_026.json", "verification": "verification/sample_026.json"} {"file_name": "videos/sample_027.mp4", "sample_id": "#27", "prompt": "subject_definitions:\n is Xu Yizhou, whose facial features, hairstyle, clothing, posture, and skin tone must strictly match .\n is Jiang Chen, whose basic characteristics must strictly match .\n is Wang Jia, whose basic characteristics must strictly match .\nThe scene is a medical single patient ward, whose spatial layout, decoration style, lighting atmosphere, and color tone must strictly match .\n\nsummary:\n[reference generation] The video is set in a medical single patient ward. (Xu Yizhou) stands by the hospital bed, interacting with (Jiang Chen) and (Wang Jia) across the bed. The sequence features a realistic style with precise lighting, transitioning through various close-ups, over-the-shoulder shots, and tracking shots to capture the characters' facial expressions, subtle gestures, and interactions, strictly maintaining their appearances and the environment without any text, watermarks, or background music.\n\nretention_analysis:\nfully_preserved: is strictly retained as the exact scene environment (medical single patient ward) throughout all shots. is strictly retained as the exact identity and appearance of (Xu Yizhou). is strictly retained as the exact identity and appearance of (Jiang Chen). is strictly retained as the exact identity and appearance of (Wang Jia). All character features, clothing, and the room's layout must remain completely consistent without deformation or alteration.\n\ndetailed_description:\nGlobal constraints: Strictly no subtitles, no text of any kind, no watermarks, no logos, no sign text, no screen text, no title cards, and no bottom captions. The first frame must directly show the complete scene with all characters in their initial positions; no white/blank frames, transition frames, or progressive loading. Do not duplicate the same character within a single frame; no multiple people sharing the same face throughout the video. No exaggerated camera movement, no jump cuts, no redecoration or changes to the interior design. Do not change clothing or clothing color, hairstyle, face, or body type; do not add or remove characters; do not generate strangers.\n\n[Shot 1] The scene is a medical single patient ward, strictly maintaining the spatial layout, decoration style, lighting atmosphere, and color tone of . The style is photorealistic, highly detailed, with realistic lighting, natural colors, accurate perspective, rich texture, rigorous composition, and simple but key-detail-focused art style. The lighting is LED panel light, top light, cool light, soft light. (Xu Yizhou) stands by the hospital bed, facing the camera, opposite (Jiang Chen) across the bed. Close-up, over-the-shoulder shot. LED panel light, top light, cool light, soft light. The camera captures 's facial expression over 's shoulder. hooks a hypocritical smile at the corner of his mouth, his eyes showing obvious provocation as he looks straight ahead, slightly raising his chin. (S1) says in a yandere, gloomy, and crazy voice, with a hypocritical and provocative tone, [Chinese]姜辰,你该不会是舍不得吧?\n\n[Shot 2] The scene remains the medical single patient ward. stands by the hospital bed, facing the camera, opposite across the bed. Close-up, push-in shot. LED panel light, top light, cool light, soft light. The camera slowly pushes in towards . Pretending to be helpless, he spreads his hands, his eyebrows slightly raised, a trace of contempt flashing in his eyes, his body leaning slightly forward, striking a preaching posture. (S1) says in a yandere, gloomy, and crazy voice, with a frivolous and mocking tone, [Chinese]哎,看一个男人爱不爱老婆,就看他愿不愿意花钱就完事了。\n\n[Shot 3] The scene remains the medical single patient ward. stands by the hospital bed, slightly tilting his head to the left, standing side by side with (Wang Jia). Close-up, slight pan shot. LED panel light, side light, cool light, soft light. turns his head to look at beside him, his eyes instantly becoming deeply affectionate, a gentle smile on his lips, with an assured, resolute manner. (S1) says in a yandere, gloomy, and crazy voice, shifting to a deep and affectionate tone, deliberately raising his volume so the other person can hear, [Chinese]佳佳跟我要是夫妻,我愿意为了她倾家荡产。\n\n[Shot 4] The scene remains the medical single patient ward. (Jiang Chen) stands on the other side of the hospital bed, facing , opposite across the bed. Reaction shot, close-up. LED panel light, top light, cool light, soft light. looks expressionlessly at the two people performing a dual act opposite him, his eyes cold as frost, his jawline tight, his eyes full of disdain and mockery.\n\n[Shot 5] The scene remains the medical single patient ward. (Wang Jia) stands on the left side of , turning her body to the right, her body close to ; stands in place. Medium shot, lateral camera translation. LED panel light, side light, warm light, soft light. An extremely moved expression instantly appears on 's face, her eyes sparkling as she looks at , her body naturally leaning closer, her hands intimately holding his arm, a sweet smile on her face. (S2) says in a cold queenly voice, with a sweet and coquettish tone, shifting to a bossy tone towards in the second half, [Chinese]亦舟,你对我可真好,姜辰,你以后多跟亦舟学着点。\n\n[Shot 6] The scene remains the medical single patient ward. (Jiang Chen) stands on the other side of the hospital bed, facing , opposite across the bed. Reaction shot, close-up. LED panel light, top light, cool light, soft light. looks at 's intimate gesture, slightly squinting his eyes, his mouth pulling into an extremely mocking cold sneer, his eyes completely cooling down.\n\noverall_soundscape:\nNatural environment sounds or mute. No background music, no musical score, no musical audio tracks.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【全局基础设定】\n[图片1](医疗单人病房)为场景参考图,严格保持空间布局、装修风格、光影氛围、色调一致,不可偏离。\n[图片2](徐亦舟 - 基础形态)为主角参考图,严格保持人物五官、发型、服饰、体态、肤色完全一致,全程不可变形。\n[图片3](姜辰 - 基础形态)为配角参考图,保持人物基本特征一致。\n[图片4](王佳 - 基础形态)为配角参考图,保持人物基本特征一致。\n\n【全局强制约束】\n禁止生成字幕、禁止生成任何文字、禁止出现水印、Logo、标牌文字、屏幕文字。\n禁止生成背景音乐,禁止生成背景音乐,只保留自然环境声或静音。\n视频全程不要在同一画面中复制相同人物,不要多人同脸。\n首帧必须直接是完整场景 + 全部角色初始站位(禁止白底/过渡帧/逐帧加载)\n\n【负面约束】\n不要生成新场景,不要改变室内设计。\n不要换衣服,不要改变衣服颜色,不要改变发型,不要改变脸,不要改变体型,不要增加或减少人物,不要生成陌生人。\n不要字幕,不要文字,不要水印,不要Logo,不要海报文字,不要屏幕文字。\n不要背景音乐,不要夸张运镜,不要镜头跳切,不要重新装修。\n\n场景:[图片1](医疗单人病房)\n角色:[图片2](徐亦舟 - 基础形态)\n风格:真人写实风格,逼真精细,光影真实,色彩自然,透视准确,质感丰富,构图严谨,细节丰富,画风简洁但关键细节突出\n光线:LED平板灯,顶光,冷光,柔光\n位置:[图片2](徐亦舟 - 基础形态)站在病床旁,面向镜头方向,与[图片3](姜辰 - 基础形态)隔着病床相对\n描述:近景·正反打过肩|冷光,镜头越过[图片3](姜辰 - 基础形态)的肩膀捕捉[图片2](徐亦舟 - 基础形态)的面部表情,[图片2](徐亦舟 - 基础形态)嘴角勾起一抹假惺惺的笑意,眼神却带着明显的挑衅直视前方,微微扬起下巴。\n对话:[图片2](徐亦舟 - 基础形态):姜辰,你该不会是舍不得吧?\n配音指令:病娇阴郁疯批音,语气假惺惺,带着挑衅的意味\n\n场景:[图片1](医疗单人病房)\n角色:[图片2](徐亦舟 - 基础形态)\n风格:真人写实风格,逼真精细,光影真实,色彩自然,透视准确,质感丰富,构图严谨,细节丰富,画风简洁但关键细节突出\n光线:LED平板灯,顶光,冷光,柔光\n位置:[图片2](徐亦舟 - 基础形态)站在病床旁,面向镜头方向,与[图片3](姜辰 - 基础形态)隔着病床相对\n描述:近景·推镜|冷光,镜头缓慢向[图片2](徐亦舟 - 基础形态)推进,他故作无奈地摊开双手,眉头微挑,眼神中闪过一丝轻蔑,身体微微前倾,摆出一副说教的姿态。\n对话:[图片2](徐亦舟 - 基础形态):哎,看一个男人爱不爱老婆,就看他愿不愿意花钱就完事了。\n配音指令:病娇阴郁疯批音,语气轻浮,带着嘲讽\n\n场景:[图片1](医疗单人病房)\n角色:[图片2](徐亦舟 - 基础形态)\n风格:真人写实风格,逼真精细,光影真实,色彩自然,透视准确,质感丰富,构图严谨,细节丰富,画风简洁但关键细节突出\n光线:LED平板灯,侧光,冷光,柔光\n位置:[图片2](徐亦舟 - 基础形态)站在病床旁,微微侧头向左,与[图片4](王佳 - 基础形态)呈并肩站立关系\n描述:近景·微摇镜|冷光,[图片2](徐亦舟 - 基础形态)转头看向身旁的[图片4](王佳 - 基础形态),眼神瞬间变得深情款款,嘴角带着温柔的笑意,语气笃定。\n对话:[图片2](徐亦舟 - 基础形态):佳佳跟我要是夫妻,我愿意为了她倾家荡产。\n配音指令:病娇阴郁疯批音,语气转为深情,刻意拔高音量让对方听见\n\n场景:[图片1](医疗单人病房)\n角色:[图片3](姜辰 - 基础形态)\n风格:真人写实风格,逼真精细,光影真实,色彩自然,透视准确,质感丰富,构图严谨,细节丰富,画风简洁但关键细节突出\n光线:LED平板灯,顶光,冷光,柔光\n位置:[图片3](姜辰 - 基础形态)站在病床另一侧,面向[图片2](徐亦舟 - 基础形态),与[图片2](徐亦舟 - 基础形态)隔着病床相对\n描述:反应镜头·特写|冷光,[图片3](姜辰 - 基础形态)面无表情地看着对面唱双簧的两人,眼神冰冷如霜,下颌线紧绷,眼底满是不屑与嘲弄。\n\n场景:[图片1](医疗单人病房)\n角色:[图片4](王佳 - 基础形态)·[图片2](徐亦舟 - 基础形态)\n风格:真人写实风格,逼真精细,光影真实,色彩自然,透视准确,质感丰富,构图严谨,细节丰富,画风简洁但关键细节突出\n光线:LED平板灯,侧光,暖光,柔光\n位置:[图片4](王佳 - 基础形态)站在[图片2](徐亦舟 - 基础形态)左侧,侧身向右,与[图片2](徐亦舟 - 基础形态)身体贴近;[图片2](徐亦舟 - 基础形态)站在原地\n描述:中景·横移|暖光,[图片4](王佳 - 基础形态)脸上瞬间浮现出极度感动的神色,双眼亮晶晶地看着[图片2](徐亦舟 - 基础形态),身体顺势贴近,双手亲昵地挽住他的胳膊,笑容甜美。\n对话:[图片4](王佳 - 基础形态):亦舟,你对我可真好,姜辰,你以后多跟亦舟学着点。\n配音指令:冷酷御姐音,语气甜美娇嗔,后半句转为对姜辰的颐指气使\n\n场景:[图片1](医疗单人病房)\n角色:[图片3](姜辰 - 基础形态)\n风格:真人写实风格,逼真精细,光影真实,色彩自然,透视准确,质感丰富,构图严谨,细节丰富,画风简洁但关键细节突出\n光线:LED平板灯,顶光,冷光,柔光\n位置:[图片3](姜辰 - 基础形态)站在病床另一侧,面向[图片4](王佳 - 基础形态),与[图片4](王佳 - 基础形态)隔着病床相对\n描述:反应镜头·特写|冷光,[图片3](姜辰 - 基础形态)看着[图片4](王佳 - 基础形态)亲昵的举动,微微眯起眼睛,嘴角扯出一抹极度嘲讽的冷笑,眼神彻底冷了下来。\n\n附加限制要求:\n1. 画面中不要出现任何字幕、台词字卡、屏幕文字说明或底部文案。\n2. 不要添加背景音乐、配乐或音乐性音轨。\n3. 不要出现任何Logo、水印、品牌标识、角标或平台标记。", "reference_images": ["references/1cbc01174e5d1e134de1d9760502d8a589185e163c91459faaf1e41b6cd15250.jpg", "references/82add03805ccf97c0930e7fd4b26e93e594924154d6791dde2e47faef8403bb4.jpg", "references/1da0d94c52cf6291011a6aa741f95e439f759c539601be436e0e53a0d24e033f.jpg", "references/f8f758dad2319113ce86443e5c44432243745b8247cf989dbcac64792fe8de38.jpg"], "reference_sha256": ["1cbc01174e5d1e134de1d9760502d8a589185e163c91459faaf1e41b6cd15250", "82add03805ccf97c0930e7fd4b26e93e594924154d6791dde2e47faef8403bb4", "1da0d94c52cf6291011a6aa741f95e439f759c539601be436e0e53a0d24e033f", "f8f758dad2319113ce86443e5c44432243745b8247cf989dbcac64792fe8de38"], "workflow": "ref2va", "prompt_sha256": "f836b0703e85931e68599304c1415630b75b16de969c73403f058a6cbf6e0e0d", "original_prompt_sha256": "bdbbdfe19e2e4da46881b46388dc281d83ac03257794192d4d21b896a555d85b", "video_sha256": "9cdb7d32d62dfd8af7223570ffa131a4840efd1843b50e0810df815a195dc9e1", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_027.json", "verification": "verification/sample_027.json"} {"file_name": "videos/sample_028.mp4", "sample_id": "#28", "prompt": "subject_definitions:\n (Zhou Yutong) matches the appearance and clothing from . (Gu Chusheng) matches the appearance and clothing from . (Meng Xiaoxi) matches the appearance and clothing from . (Fang Nian) matches the appearance and clothing from . represents the photo area location.\n\nsummary:\n[reference generation] A scene in Korean webtoon sweet romance style with a soft light filter and pink-tender color tones. (Gu Chusheng) stands opposite the photo area (), looking directly at (Zhou Yutong). (Zhou Yutong) stands in front of the center position, looking coldly at (Gu Chusheng). (Fang Nian) stands behind and to the side of (Zhou Yutong). (Meng Xiaoxi) retreats to the edge of the photo area (), frozen with tears. Bright, even warm white front lighting illuminates (Gu Chusheng)'s face, while the background light is soft and blurred.\n\nretention_analysis:\nfully_preserved for (Zhou Yutong), (Gu Chusheng), (Meng Xiaoxi), and (Fang Nian) as they are explicitly positioned and described in the scene. is fully preserved as the photo area location reference. All spatial relationships, character identities, and lighting constraints from the source are retained.\n\ndetailed_description:\n[Shot 1] (0-2s): [Close-up, quick cut in]. (Gu Chusheng)'s face instantly turns ashen, his eyes widening in disbelief. Opposite him, (Zhou Yutong) looks coldly at (Gu Chusheng). (Fang Nian) stands behind and to the side of (Zhou Yutong). (Meng Xiaoxi) is frozen at the edge. [Oppressive murmuring sound on site].\n[Shot 2] At 0:02.000 (2-4s): [Extreme close-up, sudden freeze frame]. (Gu Chusheng)'s ashen, twisted face fills the lens. His widened eyes are mixed with fury and disbelief, his gentle facade completely shattered, his expression frozen at the most embarrassing moment. [Freeze frame accent, background sound instantly cut off]. No subtitles, no dialogue, no background music.\n\noverall_soundscape:\nOppressive murmuring sound on site during the first shot, followed by an instant cut to silence with a freeze frame accent during the second shot.\n\nnon_diegetic_music:\nN/A", "original_prompt": "韩漫甜宠风韩式甜漫风格,柔光滤镜,粉嫩色调 [位置关系:@图2顾楚生站在@图5合影区对面直视@图1周雨桐,@图1周雨桐站在C位前方冷冷看着@图2顾楚生,@图4方念站在@图1周雨桐身后侧,@图3孟小溪退在@图5合影区边缘含泪僵住] [全局光影:明亮均匀的暖白色正面光照亮@图2顾楚生面部,背景光线柔和虚化]。镜头1(0-2s):[近景,快速切入]。@图2顾楚生脸色瞬间变得铁青,眼睛难以置信地瞪大,@图1周雨桐在对面冷冷看着@图2顾楚生,@图4方念站在@图1周雨桐身后侧,@图3孟小溪在边缘僵住。[现场压抑哗然声]。镜头2(2-4s):[大特写,骤停定格]。@图2顾楚生铁青扭曲的脸占据镜头,他瞪大的眼睛里混着震怒与不可置信,温润伪装彻底碎裂,表情凝固在最难堪的一瞬。[定格重音,背景声瞬间切断]。保持无字幕,无对话,无背景音乐。", "reference_images": ["references/4efae96e2260347e6b6fb884991f46173122902e6a3a59550cdd9cf1e3204ab3.jpg", "references/48802a35c98395ebef733f7eba779204b978566085228ce5c88df01ac01d3b27.jpg", "references/b0af6acdf592326d4ff07c1072ea9b3219f5eb83dd6e9e8cc3b1eab6c6e6b2c3.jpg", "references/81448900faa69994032e472c40205bb314b990fc9392ef13aa3e81a4edb5f048.jpg", "references/5d4148027f06e14c258c38a4d13177080b3c89fcf762d63c92a21192f29317c4.jpg"], "reference_sha256": ["4efae96e2260347e6b6fb884991f46173122902e6a3a59550cdd9cf1e3204ab3", "48802a35c98395ebef733f7eba779204b978566085228ce5c88df01ac01d3b27", "b0af6acdf592326d4ff07c1072ea9b3219f5eb83dd6e9e8cc3b1eab6c6e6b2c3", "81448900faa69994032e472c40205bb314b990fc9392ef13aa3e81a4edb5f048", "5d4148027f06e14c258c38a4d13177080b3c89fcf762d63c92a21192f29317c4"], "workflow": "ref2va", "prompt_sha256": "4396608e418458d43fa123627edad9801e6cecc2558ca9e877e0f78603689717", "original_prompt_sha256": "50fcfa80c3b78dfead67fac665da41736286061b3dd054206940e3dd0ea091f5", "video_sha256": "303e338d1e04952c1ea104e9dc12589bafb61a362c13d1865a8f8f869cae1273", "task_adapted": false, "width": 1344, "height": 768, "fps": 24, "num_frames": 107, "duration_seconds": 4.458333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_028.json", "verification": "verification/sample_028.json"} {"file_name": "videos/sample_029.mp4", "sample_id": "#29", "prompt": "subject_definitions:\n represents the visual content and appearance from . represents the visual content and appearance from .\n\nsummary:\n[reference generation] The video is a continuous, smooth drone fly-through shot moving forward at 0.5 meters per second along the ground's center line. The first frame must have no distortion. There is no BGM or voiceover.\n\nretention_analysis:\nThe source requests the first frame to have no distortion, implying the use of a reference image as the first frame, but does not explicitly specify which of the two provided images to use. Both and are provided but their specific roles are not explicitly defined in the text. The first-frame distortion constraint is fully preserved.\n\ndetailed_description:\n[Shot 1] The video begins with a first frame that has no distortion. The camera executes a smooth track movement from a drone perspective (not a handheld camera walking perspective). It is a fly-through shot where the camera moves forward at a constant, even, and silky-smooth speed along the center line of the ground, with no stuttering. The camera remains perfectly stable with absolutely no up-and-down fluctuations, no vertical bumping, no vertical movement, and no shaking. The drone advances at a constant slow speed of 0.5 meters per second in a smooth, stable, continuous one-take shot.\n\noverall_soundscape:\nNo BGM and no voiceover are present. No other specific sounds are requested.\n\nnon_diegetic_music:\nN/A", "original_prompt": "首帧画面无变形,平稳轨道运镜,无人机视角(非手持相机走路视角),穿梭镜头,镜头沿着地面中线匀速前进,速度均匀丝滑,无卡顿,镜头平稳无丝毫上下起伏,无上下颠簸,镜头无垂直移动,无晃动,丝滑稳定前进,无人机以0.5米每秒的速度均速慢速前进,一镜到底,无BGM,无配音", "reference_images": ["references/3d2cede73092ecdea022b026b38fbe929285081f8f2352d142a85d6a254d694a.jpg", "references/88549d95cc4a7630216b8014a56f631ea3ee0906beed5d3d6ec5c6bf76bc8140.jpg"], "reference_sha256": ["3d2cede73092ecdea022b026b38fbe929285081f8f2352d142a85d6a254d694a", "88549d95cc4a7630216b8014a56f631ea3ee0906beed5d3d6ec5c6bf76bc8140"], "workflow": "ref2va", "prompt_sha256": "e5e9e6a966b3aff2fc5a93e231acd3deb212ff47358cff13bf41fb6584d97049", "original_prompt_sha256": "c392ca710db6042f6f30f6ad2db2eaa8a77039900771b6833ff61e46ebfb1a49", "video_sha256": "1fa8b175a97147f506f24cfc6aaa47f990c865b76bd5f2a776741be4a073800c", "task_adapted": false, "width": 1344, "height": 768, "fps": 24, "num_frames": 209, "duration_seconds": 8.708333333333334, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_029.json", "verification": "verification/sample_029.json"} {"file_name": "videos/sample_030.mp4", "sample_id": "#30", "prompt": "subject_definitions:\n (Della): Appearance and clothing from .\n (Maid 1), (Maid 2), (Maid 3): Three maids.\n (Hermès suitcase): Appearance from .\nLocation: Scene from .\n\nsummary:\n[reference generation] A 4-second shot in a vertical portrait mode. Della and three maids are inside a main door. The three maids rush forward to roughly grab a Hermès suitcase on the ground, while Della gets scared and dodges to the side. The scene features a cold, high-grade gray and low-saturation color palette, realistic skin details, and a tense, high-pressure atmosphere.\n\nretention_analysis:\n (Della) from is fully preserved as the main character reacting to the maids. , , and (the three maids) are fully preserved as the aggressors. (Hermès suitcase) from is fully preserved as the central prop. The location from is fully preserved as the setting. All referenced subjects and props are actively involved in the described actions and must be retained throughout the generation.\n\ndetailed_description:\n[Shot 1] The video is in a vertical portrait mode with a cold, high-grade gray and low-saturation color palette, featuring stable dark shadows and clear bright parts. The camera uses hard cut transitions, steady tracking, and a slight narrative breathing shake, maintaining a stable 180° axis without crossing the line. The shot size is specified as a medium shot, but the composition section requests a centered full shot to show the group action; this conflict is retained. The framing highlights a bullying dynamic. (Della), whose appearance matches , stands inside the main door. Next to her on the ground is (the Hermès suitcase), matching . , , and (the three maids) immediately take action, rushing forward with rough and impatient movements. They reach out to pull and grab the Hermès suitcase with rough actions. Startled by the rushing maids, (Della) dodges to the side. The group action is both neat and chaotic. The camera captures the intense, high-pressure confrontation and overwhelming aura. Facial features are accurate and consistent with the references, with clear skin pores and textures, distinct hair strands, and realistic clothing textures. There is no clipping, distortion, or facial deformation. Human body structures are normal, and clothing and appearance remain completely consistent. No subtitles are present.\n\noverall_soundscape:\nUrgent footsteps and clothing friction sounds.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【风格】:固定写实风格,真实人像摄影,现代豪门写实商战剧,竖屏构图(Portrait Mode),适配短视频平台,人物占比突出、压迫感强,冷色调高级灰 + 低饱和质感,暗部沉稳、亮部通透,无过度磨皮,保留真实皮肤细节,大量人物近景 / 面部特写、关键冲突慢动作、稳定运镜、局部特写(首饰 / 伤口 / 枪械 / 文件),镜头特点为极致微表情面部特写、快切硬切转场、慢动作氛围感镜头、第一视角代入镜头、手持微晃纪实镜头,谁说话切谁近景,人物互动用中景,环境氛围用全景,关键表情 / 道具 / 手部动作用特写;镜头带叙事呼吸感微晃,景别切换自然顺滑,180° 轴线稳定不越轴;角色样貌全程统一、情绪连贯、动作自然、人体结构精准,无穿模无扭曲;情绪基调:张力拉满、冷静克制、爽感强烈、高压对峙、气场碾压\n【时长】4秒​\n【景别】中景​\n【摄法】硬切转场,平稳跟拍,轻微呼吸微晃​\n【场景】图1\n【出镜人物】图2是黛拉,三位佣人​\n【道具】图3是爱马仕行李箱​\n【画面描述】三位佣人立刻行动,扑向爱马仕行李箱,动作粗鲁急躁​,黛拉被冲过来的佣人吓到,往一旁躲开\n1. 构图:全景居中构图,群像动作整齐且杂乱,凸显仗势欺人​\n2. 人物站位:黛拉站在大门内侧,爱马仕行李箱在旁边地上\n3. 人物动作:核心动作:三位佣人快步扑上、伸手拉扯行李、动作粗暴​\n【对白】无​\n【音效】急促脚步声、衣物摩擦声\n【约束】人物五官精准不崩坏,人物五官与参考保持一致,脸部毛孔皮肤纹理清晰,发丝根根分明,衣物纹理写实,无穿模无扭曲,人物面部稳定不变形,人体结构正常,服装样貌与前文完全一致,无字幕,无背景音乐", "reference_images": ["references/52a869cfcfc9695db093d2269bf0af0e3ab6e7faf37a13a8e6e5b4558e054124.jpg", "references/f142c3a1a4d4e0af7023fcc00b282de7b0a8864173fe7ec6279f4c91a0d19fce.jpg", "references/94f183f9ff727d6e41187354331b2782a32d22f860f344a3d461949facc9be0f.jpg"], "reference_sha256": ["52a869cfcfc9695db093d2269bf0af0e3ab6e7faf37a13a8e6e5b4558e054124", "f142c3a1a4d4e0af7023fcc00b282de7b0a8864173fe7ec6279f4c91a0d19fce", "94f183f9ff727d6e41187354331b2782a32d22f860f344a3d461949facc9be0f"], "workflow": "ref2va", "prompt_sha256": "54bfb334406b3cecb2fb931c9989bdcedf6ad15727d9e4a7fa8f42dee19b358d", "original_prompt_sha256": "8d66baab22e8dfed6debbd520a2bdeeb5c3d1e4f00a8a8d55861e1cf654c62f1", "video_sha256": "efacc2e54191535fd04c0d751822796331a7f5c3534a5853e1a775243b92c3e7", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 107, "duration_seconds": 4.458333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_030.json", "verification": "verification/sample_030.json"} {"file_name": "videos/sample_031.mp4", "sample_id": "#31", "prompt": "subject_definitions:\n: Franka Panda robotic arm, appearance from .\n\nsummary:\n[reference generation] A continuous first-person view in a clean modern kitchen where a Franka Panda robotic arm picks up and pours soy sauce.\n\nretention_analysis:\nThe Franka Panda robotic arm from is fully preserved as the primary subject, positioned at the right-hand location in the first-person perspective throughout the video. Its mechanical appearance and structure from the reference image must be retained consistently without adding extra arms.\n\ndetailed_description:\n[Shot 1] The scene is set in a clean and tidy modern kitchen environment with bright and soft indoor kitchen lighting. The countertop is clean and the background is stable, with no cluttered objects added. The entire video maintains a first-person perspective. At the right-hand position in the first-person view is the Franka Panda robotic arm (), whose appearance matches ; the Franka robotic arm is not an additionally added robotic arm but the sole mechanical arm in the scene. At 00:04.000, the Franka robotic arm picks up a bottle of soy sauce. At 00:09.000, the Franka robotic arm pours the soy sauce into a pot. The pot is stably placed on the kitchen countertop. The soy sauce bottle and the pot remain stable and consistent throughout the entire video, without flickering, deforming, drifting, or suddenly disappearing. The object interaction is natural, the hand-eye perspective is realistic, and the image is clear.\n\noverall_soundscape:\nUnspecified.\n\nnon_diegetic_music:\nN/A.", "original_prompt": "使用视频1作为唯一动作、时序、相机运动、第一视角构图、相机距离和镜头轨迹参考。把视频1中杂乱的机器人操作间替换为干净整洁的现代厨房环境,明亮柔和的室内厨房光照,台面干净,背景稳定,不要新增杂乱物体。保持全程第一视角,相机视角、相机距离、相机运镜、视频分辨率、帧率和时长与视频1一致。把画面里第一视角的右手替换成图片1中的 Franka Panda 机械臂外观,Franka 机械臂必须占据原右手的位置、大小、方向和运动轨迹,不是额外新增一条机械臂。第4秒原本拿起方块的动作,改为 Franka 机械臂拿起一瓶酱油;酱油瓶的位置、运动轨迹、遮挡关系和接触时序沿用原方块。第9秒原本将方块放入篮子的动作,改为将酱油倒入锅中;锅位于原篮子所在区域,稳定放在厨房台面上。酱油瓶和锅在全程保持稳定一致,不要闪烁、变形、漂移或突然消失。保持物体交互自然,手眼视角真实,画面清晰。", "reference_images": ["references/6e050759b750c049e65b17c0703c4d01c5ff78976628c4572b8bc6d3382c87ca.png"], "reference_sha256": ["6e050759b750c049e65b17c0703c4d01c5ff78976628c4572b8bc6d3382c87ca"], "workflow": "ref2va", "prompt_sha256": "d87275ec1b93e1165d59acf7e0ec0934550338021a5285d055de31baaa647dec", "original_prompt_sha256": "12306b20fc5b654934da133e83e0f182f7587166bf5217c7ec89cd53a9cc9b4e", "video_sha256": "e801737a98cb9721429a2caca91d6f7984e1d936fb52ce029133edd1ecb9b6d3", "task_adapted": true, "width": 1344, "height": 768, "fps": 24, "num_frames": 328, "duration_seconds": 13.666666666666666, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_031.json", "verification": "verification/sample_031.json"} {"file_name": "videos/sample_033.mp4", "sample_id": "#33", "prompt": "subject_definitions:\n is Lu Chen, a male character with appearance and clothing from . is Bai Tiantian, a female character with appearance and clothing from .\n\nsummary:\n[reference generation] A 15-second scene set in a bar corridor at night featuring (Lu Chen) and (Bai Tiantian), with four timed shots depicting an intimate, passionate encounter under flickering neon and ceiling lights. The sequence progresses from a whip-pan transition to close-ups and an extreme close-up of a kiss, with handheld and fixed camera work, ending on a frontal angle emphasizing emotional tension. The sequence is purely visual content with no dialogue or narration, and all content must conclude within 14.8 seconds.\n\nretention_analysis:\nAll three supplied images are fully preserved in the generation. defines the appearance of (Lu Chen) as the male lead. defines the bar corridor night setting as the scene location. defines the appearance of (Bai Tiantian) as the female lead.\n\ndetailed_description:\n[Shot 1] A fast whip pan transitions from 's flushed cheeks to 's deep eyes in the bar corridor at night. The corridor lights flicker bright and dim in chaotic disorder, with neon lights and ceiling lights alternating, casting changing light and shadow on the two characters.\n[Shot 2] At 00:01.000, a close-up with handheld shaky camera movement shows 's lips slightly parted from . He glances at the flickering light above, and the light instantly returns to normal as if obeying his command. 's cheeks are flushed, her eyes hazy. 's side profile is sharp like carved by a knife under the light, his eyes deep and unfathomable. Only the heavy breathing of both characters and clothing rustling sounds are heard.\n[Shot 3] At 00:04.000, a medium shot with fixed camera shows hazily reaching to grab 's neck, wanting more. grabs 's hand with one hand and pins it behind her back, while his other hand cups her face, controlling the pace. Only 's soft whimpers and 's low breathing are heard. The restored normal lighting softly illuminates the two, forming a warm, ambiguous atmosphere.\n[Shot 4] At 00:08.000, a close-up with slow push-in camera movement shows cupping 's face, his thumb caressing her hot cheeks, his eyes locked on her lips as he slowly lowers his head. 's eyelashes tremble lightly, her lips slightly parted in waiting. Only the heavy breathing of both characters is heard. The image is high-definition and delicate, with realistic skin texture and clearly visible pores, no skin-smoothing effect.\n[Shot 5] At 00:12.000, an extreme close-up with fixed camera shows kissing down, a close-up of their lips with tongues intertwined. goes soft all over from the kiss, her fingers tightly grabbing 's collar. Only lingering kissing sounds and 's suppressed breathing are heard. The angle is frontal, emphasizing emotional tension.\nNegative prompt: low resolution, deformed human body structure, deformed hands, text, image errors, missing fingers, extra fingers, incomplete fingers, image cropping, worst quality, low quality, normal quality, compression noise, signature, watermark, username, blurry image, ugly appearance, repeated image, pathological deformity, missing limbs, out of frame, distorted body, ruined appearance, mutated body, extra limbs, missing limbs, floating limbs, severed limbs, overly long neck, disproportionate, severe proportion deformity, face cloning, multiple faces, extra faces, deformed limbs, fused fingers, too many fingers, deformed feet, missing feet, too many people. No text or subtitles appear on the screen.\n\noverall_soundscape:\nHeavy breathing from both characters, clothing rustling sounds, soft whimpers from the female character, low breathing from the male character, and lingering kissing sounds.\n\nnon_diegetic_music:\nN/A", "original_prompt": "1-4第四镜:意乱情迷(00:00-00:15)\n故事发生场景:(酒吧走廊 - 夜 @图2 )\n出演人物:陆沉 @图1 、白甜甜 @图3 \n 请仔细理解人物位置关系与镜头之间的合理连贯性\n场景光影:走廊灯光忽而明亮忽而昏暗,错乱无序,霓虹灯与顶灯交替闪烁,在两人身上投下变幻的光影\n分镜过渡(00:00-00:01):快速甩镜头,从白甜甜绯红的脸颊甩到陆沉深邃的眼睛\n镜头 1(00:01-00:04):\n景别:近景\n运镜:手持晃动\n画面内容:陆沉的唇微微分开白甜甜,眼风瞥了一眼头顶闪烁的灯光,那灯光仿佛听从指令般瞬间恢复正常;白甜甜脸颊绯红,眼神迷离。\n对话:(无台词,仅两人粗重喘息声和衣物摩擦声)\n动作描述:陆沉侧脸轮廓在灯光下如同刀削,眼神深邃不见底\n镜头 2(00:04-00:08):\n景别:中景\n运镜:固定\n画面内容:白甜甜迷离着去抓陆沉的脖子,想要更多;陆沉一手捞过白甜甜的手,反扣在她身后,另一手捧着她的脸,掌控着节奏。\n对话:(无台词,仅白甜甜细微的呜咽声和陆沉低沉的呼吸)\n氛围光感:恢复正常的光线柔和地照亮两人,形成温暖的暧昧氛围\n镜头 3(00:08-00:12):\n景别:特写\n运镜:缓慢推进\n画面内容:陆沉捧着白甜甜的脸,拇指摩挲她滚烫的脸颊,眼神锁定她的唇,缓缓低头;白甜甜睫毛轻颤,嘴唇微张等待。\n对话:(无台词,仅两人粗重喘息声)\n画质要求:高清细腻,皮肤纹理真实,毛孔清晰可见,无磨皮感\n镜头 4(00:12-00:15):\n景别:大特写\n运镜:固定\n画面内容:陆沉吻下,两人唇部特写,舌尖交缠,白甜甜被吻得浑身发软,手指紧紧抓住陆沉的衣领。\n对话:(无台词,仅缠绵亲吻声和白甜甜压抑的喘息)\n视角:正面角度,突出情感张力\n(视频中的所有台词,对话,旁白等,必须合理完整的在 14.8 秒之前说完,纯视觉内容,不要出现背景音乐与 BGM),注意:反向提示词(低分辨率、人体结构畸形、手部畸形、文字、画面错误、手指缺失、多余手指、手指残缺、画面裁切、最差画质、低画质、普通画质、压缩噪点、签名、水印、用户名、画面模糊、颜值丑陋、画面重复、病态畸形、肢体残缺、画面出框、形体扭曲、容貌损毁、形体变异、多余肢体、肢体缺失、悬浮肢体、肢体断裂、脖颈过长、比例失调、严重比例畸形、脸部克隆、多张人脸、多余面部、肢体畸形、手指粘连、手指过多、脚部畸形、脚部缺失,人物过多)No text or subtitles appear on the screen", "reference_images": ["references/a510fb3c53ed9e15fa148c87161021a7b0dfaf52ca555932c372161f65538778.jpg", "references/dab6f7f499503028bdf2c45604afc2494b2fb261380c00cfd4d1c5a9c68885e8.jpg", "references/31edcc6100f58e0bf13b03a5c6d6a04a2284c2e0cc9d17c6bf1191e32a1cf48f.jpg"], "reference_sha256": ["a510fb3c53ed9e15fa148c87161021a7b0dfaf52ca555932c372161f65538778", "dab6f7f499503028bdf2c45604afc2494b2fb261380c00cfd4d1c5a9c68885e8", "31edcc6100f58e0bf13b03a5c6d6a04a2284c2e0cc9d17c6bf1191e32a1cf48f"], "workflow": "ref2va", "prompt_sha256": "d89ad528bbfcd8e477c8e6f6fba33e758e350b8b24c59252cf904cd15f27e943", "original_prompt_sha256": "8feb7d61a4df242495166698fdf2257f76a7d8289428ae8db040c6c40495729d", "video_sha256": "4ee4d4f9db7d0dee769cbb99b0e724b47baf6076c25dc77765173d285f8bde89", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_033.json", "verification": "verification/sample_033.json"} {"file_name": "videos/sample_034.mp4", "sample_id": "#34", "prompt": "subject_definitions:\n is the supplied visual reference; its particular subject or environmental role is not specified.\n\nsummary:\n[reference generation] Recreational shrimp fishing, followed by suddenly catching a shark.\n\nretention_analysis:\nThe source supplies as visual reference but does not specify particular attributes or identities to preserve.\n\ndetailed_description:\n[Shot 1] Depict recreational shrimp fishing. Suddenly, a shark is caught.\n\noverall_soundscape:\nUnspecified\n\nnon_diegetic_music:\nN/A", "original_prompt": "钓虾娱乐 突然钓到一条鲨鱼", "reference_images": ["references/0c5373d0507dddc3751761bd593ee1f839f4c45a50ce56cb89b9e7f3da11a601.jpg"], "reference_sha256": ["0c5373d0507dddc3751761bd593ee1f839f4c45a50ce56cb89b9e7f3da11a601"], "workflow": "ref2va", "prompt_sha256": "580d53755065c7fa79dea4a82de7582194863ef22d9851183382d25c605b22eb", "original_prompt_sha256": "cc9a094f76d5622cc167a7b6dac6db08f806b9d17047ba6a71827a09d05786fb", "video_sha256": "b518216b54508daf579f3db0d7e5a3fd50ae1942f959409fea685ff8f1410521", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 107, "duration_seconds": 4.458333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_034.json", "verification": "verification/sample_034.json"} {"file_name": "videos/sample_035.mp4", "sample_id": "#35", "prompt": "subject_definitions:\n is the female model, whose appearance, standard figure, lazy French updo hairstyle, and light makeup are derived from . is the women's clothing, whose exact color, shape, and details are derived from . provides the background environment.\n\nsummary:\n[reference generation] A women's clothing e-commerce product detail page marketing video designed to promote purchase conversion. The video strictly features the clothing from without any alterations in color, shape, or details, set against the background from . The model from showcases the outfit with a standard figure, styled in a lazy French updo and light makeup, adhering to physical laws and e-commerce platform content norms.\n\nretention_analysis:\n is fully_preserved as the model's identity, requiring her standard figure, lazy French updo, and light makeup to remain consistent. is fully_preserved as the clothing's identity, demanding strict adherence to its exact color, shape, and details without any color shift, graying, or darkening caused by the background lighting. is fully_preserved as the background, with its lighting adjusted to serve the true color of the clothing rather than altering it.\n\ndetailed_description:\n[Shot 1] At 0:00.000, a full body long shot shows walking confidently from far to near, showcasing the elasticity of the knitted fabric of and its overall slim silhouette. During her movement, she performs light side turns while the camera smoothly pushes forward. 's hands remain out of pockets throughout the entire video, and the clothing coverage is complete with a smooth wearing outline without any bulges. [Shot 2] At 0:04.000, a medium shot tracks as she continues walking and naturally touches her hair, revealing the neckline and side button craftsmanship details of . She naturally turns to display the drape of the fabric, with smooth action transitions, strictly adhering to physical laws without any clipping or unreasonable scenes such as standing on tables or chairs. [Shot 3] At 0:08.000, a close-up focuses on the upper body, where 's center of gravity naturally sways, delicately presenting the true texture of the fabric under soft oblique light. Subsequently, turns sideways, looks back with a smile, and walks away into the distance while the camera slowly pulls back. The background lighting from is carefully adjusted to serve the true color of , ensuring no graying, darkening, or color shift occurs due to the background tone.\n\noverall_soundscape:\nUnspecified ambient sounds, with the focus on the non-diegetic music.\n\nnon_diegetic_music:\nUpbeat and rhythmic lazy jazz guitar, pure instrumental music with no vocals.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。服装严格使用图二,颜色/版型/细节不允许有任何改变或色差;背景直接使用图三。模特使用图一,标准身材,搭配慵懒法式盘发与淡雅妆容。符合物理规律,禁止穿模/模特站桌椅等不合理画面;服装覆盖完整、穿着轮廓平整无凸点;符合电商平台内容规范。服装颜色以图二为准,背景光照服务服装真实颜色,禁止因图三色调致服装变灰/暗化/偏色。模特全程禁止插兜。 分镜1(0-4s):全身远景,模特迈着自信步伐由远及近,展示针织面料的随身弹力与整体修身廓形,行进中伴随轻盈侧身,镜头平稳推进。 分镜2(4-8s):中景跟踪,模特持续行进并自然拨动发丝,展现服装领口及侧边纽扣工艺细节,通过自然转身展示面料的垂坠感,动作衔接流畅。 分镜3(8-12s):近景聚焦上半身,模特重心自然摆动,细腻展现面料在柔和斜射光下的真实质感,随后模特侧身回头微笑着向远方走去,镜头缓慢拉远。 配乐:纯音乐,无人声,明快且富有节奏感的慵懒爵士吉他。\\n\\n素材说明:图1=模特参考(asset-20260224194849-qt54s);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552.jpg", "references/1714b625b0bfd9b395ca8d5a16c7d6fd95ca5f3834bf882ede151862f4ccdb95.jpg", "references/20772ecdd334d6ac88fa7fcb7d4bbc49a1804560bbb399a1414f018dcde994fe.jpg"], "reference_sha256": ["96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552", "1714b625b0bfd9b395ca8d5a16c7d6fd95ca5f3834bf882ede151862f4ccdb95", "20772ecdd334d6ac88fa7fcb7d4bbc49a1804560bbb399a1414f018dcde994fe"], "workflow": "ref2va", "prompt_sha256": "335e9824d4c5e1e81efa6f13a26b604d298ada4d48c9dcaf36fbf8a568e12f40", "original_prompt_sha256": "c832f5a8877c7187379371c123a527c233d47c20618552c41c3bf850134a51b1", "video_sha256": "3cfae3891623819c7086d3aa21055e5bd2dacb0778cba94049107e6bb65e30c4", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_035.json", "verification": "verification/sample_035.json"} {"file_name": "videos/sample_036.mp4", "sample_id": "#36", "prompt": "subject_definitions:\n: Maodou, the POV character (from ).\n: The female lead (from ).\n: The product (from ).\n: The product (from ).\n: The light green paste-like product (from ).\n\nsummary:\n[reference generation] A POV shot from Maodou's perspective in a hotel bathroom. The female lead squeezes a product onto her palm, rubs it into a light green paste, applies it to her face, and washes it off, showcasing her clean and moisturized skin. They interact playfully, with Maodou gifting the product to the female lead.\n\nretention_analysis:\n (Maodou, ): Fully preserved as the POV character and speaker.\n (Female lead, ): Fully preserved as the main on-screen actor and speaker.\n (Product, ): Fully preserved as the product held by the female lead.\n (Product, ): Fully preserved as the product squeezed onto the palm.\n (Light green paste product, ): Fully preserved as the product applied to the face.\n\ndetailed_description:\n[Shot 1] POV perspective from (Maodou). Inside a hotel bathroom, (the female lead) squeezes (the product) onto her palm. She rubs her hands together to spread (the light green paste-like product), then applies the light green paste to her face. The camera pulls in for a close-up of 's face as she uses her fingertips to rub the sides of her nose. (S1) says [Chinese]哇,味道很清晰耶。 (S2) says [Chinese]用一次能管三天呢,而且还搭配了多重护肤级植萃,洗的干净还温和不紧绷,怎么样?是不是还不错?\n[Shot 2] cups her hands to splash water and wash the product off her face. The camera is positioned at a side profile angle, capturing the water splashing down with a flowing water effect on her face. She then uses a facial cleansing towel to wipe the water off her face, showing her skin condition to the camera, and reaches out to pinch her cheek. (S1) says [Chinese]嗯嗯,确实很润欸,而且感觉毛孔都透气了呢哎呀! (leaning close to the camera with a surprised expression) [Chinese]我那熬夜熬出来的陈年黑头,真的洗掉了欸!\n[Shot 3] (S2) says [Chinese]那是,本豆甄选,能有错吗? in a proud tone. giggles, holding (the product) and asks tentatively, (S1) says [Chinese]那……这支归我啦? gently scrapes 's nose. (S2) says [Chinese]真滑溜啊,看在戏里欺负你的份儿上,这支送你了。\n(Note: No subtitles are to be rendered in the video. Characters speak only the specified lines; no additional dialogue or improvised speech. All dialogue is delivered in a crisp, gentle, and slightly louder voice. Rendering constraints: stable facial features without deformation, clear facial features, normal human anatomy, natural and fluid movements without stiffness, no stuttering or flickering in the video, and highly realistic and smooth character performance.)\n\noverall_soundscape:\nThe sound of water splashing and washing, the rustling of a facial cleansing towel, and natural indoor bathroom ambience.\n\nnon_diegetic_music:\nN/A", "original_prompt": "毛豆图片1的pov视角:酒店洗手间内,女主图片2将产品图片4挤到手心,双手搓开产品图片5,图片2将淡绿色的糊状产品图片5搓到脸上,镜头拉给女主图片2面部特写,女主图片2用指尖揉搓鼻翼:“哇,味道很清晰耶。”毛豆图片1道:“用一次能管三天呢,而且还搭配了多重护肤级植萃,洗的干净还温和不紧绷,怎么样?是不是还不错?”女主图片2捧起水冲掉脸上的产品(机位在侧脸,水泼下来,脸上有水流感。),图片2接着用洗脸巾擦掉脸上的水向镜头展示肌肤状态并伸手捏了捏脸,图片2惊喜道:“嗯嗯,确实很润欸,而且感觉毛孔都透气了呢哎呀!(凑近镜头,表情惊喜)我那熬夜熬出来的陈年黑头,真的洗掉了欸!”毛豆图片1傲娇道:“那是,本豆甄选,能有错吗?”女主图片2嘻嘻一笑,拿着产品图片3试探道:“那……这支归我啦?”毛豆图片1刮了刮女主的鼻子:真滑溜啊,看在戏里欺负你的份儿上,这支送你了。”(注意:要求人物演绎真实流畅不卡顿,不要说设定外的台词,不要字幕,不要背景音乐,想象你就是真实的导演,而她就是真实的演员,说台词时嗓音清脆温柔(声音大一些)面部稳定不变形,五官清晰,人体结构正常,动作自然流畅,不僵硬,画面无卡顿,无闪烁。)", "reference_images": ["references/ac4f8d8ac99204cf0518ea900d4daa54281b78e3578e8fd9fdca7fa63547b117.jpg", "references/25ac723609f95bd15531cac79eb89c82def3e1721a7c0cdc0aa0186f309a373b.jpg", "references/d21a9d19be5514ca6bfbeb59e12686d8227a4756af55c24aea5e788217c6b441.jpg", "references/051a1362ccabd424cd49a5190d7ee68ff326be0046fba3c047acc5b657e60678.jpg", "references/cb54fc8a0fe942b2c66a0a77c81ea71cc9c0d93c3cbcecd0adb36611b195a19f.jpg"], "reference_sha256": ["ac4f8d8ac99204cf0518ea900d4daa54281b78e3578e8fd9fdca7fa63547b117", "25ac723609f95bd15531cac79eb89c82def3e1721a7c0cdc0aa0186f309a373b", "d21a9d19be5514ca6bfbeb59e12686d8227a4756af55c24aea5e788217c6b441", "051a1362ccabd424cd49a5190d7ee68ff326be0046fba3c047acc5b657e60678", "cb54fc8a0fe942b2c66a0a77c81ea71cc9c0d93c3cbcecd0adb36611b195a19f"], "workflow": "ref2va", "prompt_sha256": "ab160edef118b5009b851581d417555bf5365fedb9193773514ef16315589495", "original_prompt_sha256": "8303db071f3a3f29ca42040c4ef73e530178d6e95da48dab1ad99eb8e211f0db", "video_sha256": "c3ae878fe2128c4896c42876df8fb8d7467e8e92458b5c29b41eb6cd1bfbd82c", "task_adapted": false, "width": 1344, "height": 768, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_036.json", "verification": "verification/sample_036.json"} {"file_name": "videos/sample_037.mp4", "sample_id": "#37", "prompt": "subject_definitions:\n is the black-haired woman whose appearance and identity are from .\n\nsummary:\n[reference generation] A close-up static shot of in the location from . Her eye sockets quickly turn red, tears fill her eyes but do not fall. While crying, she slowly grits her teeth as anger spreads across her face.\n\nretention_analysis:\nfully_preserved: (the black-haired woman from ) and the location from are fully retained as the primary subject and setting.\n\ndetailed_description:\n[Shot 1] Close-up, static shot. The scene takes place in the location from . (S1), a black-haired woman whose appearance is from , is shown in a close-up. Her eye sockets quickly turn red, and tears fill her eyes but do not fall. While crying, she slowly grits her teeth, and anger gradually spreads across her face. [cut]\n\noverall_soundscape:\nNo background music. No specific ambient sounds or sound effects are specified.\n\nnon_diegetic_music:\nN/A", "original_prompt": "Location:图片1\n\n1.特写 | 静止镜头 | 黑发,图片2眼眶迅速泛红,泪水积满却没有落下,一边流泪一边慢慢咬牙切齿愤怒爬上了她的脸庞。 [cut]\nRules:不要背景音乐, 不要有字幕 , 不要引入不存在的角色", "reference_images": ["references/ec6a290668ae9e9ea881cf7c757bdb0bc909f51d71492fd3a043351d63a6f7f4.jpg", "references/06568c787e8c5d8c73f88d22200a06378476ad957fbaea8f23d374765887e20f.jpg"], "reference_sha256": ["ec6a290668ae9e9ea881cf7c757bdb0bc909f51d71492fd3a043351d63a6f7f4", "06568c787e8c5d8c73f88d22200a06378476ad957fbaea8f23d374765887e20f"], "workflow": "ref2va", "prompt_sha256": "2b87ea04bda30514f4768da89ac37af79c22aa893759b9b0bc3543a549963b55", "original_prompt_sha256": "b92537cc6be2a398589f2fcd3f2133693ed1d2e717f08f52069ee1649e696dba", "video_sha256": "c3a8584ee35eeaea7d11bf2680f10e4013a6a39b8d2701d30ae66af9fd739389", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 107, "duration_seconds": 4.458333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_037.json", "verification": "verification/sample_037.json"} {"file_name": "videos/sample_038.mp4", "sample_id": "#38", "prompt": "subject_definitions:\n (Little boy Will): Appearance, clothing, fabric texture, and 3D render shading strictly match . Bright, warm, kid-friendly 3D cartoon animation style. Soft clay-like skin with subtle freckles, rounded friendly facial features.\n\nsummary:\n[reference generation] A medium bright shot of Little boy Will from the reference image, leaning forward slightly with open arms and palms up on a colorful table, speaking directly to the camera in a bright, warm, kid-friendly 3D cartoon animation style.\n\nretention_analysis:\n (Little boy Will) is fully preserved, with strict requirements to match the facial features, skin freckles, hair details, skin tone, clothing fabric texture, and 3D render shading from in every frame. The exact clothing, colors, and fabrics from must be matched without any character drift, facial distortion, outfit change, texture loss, or clothing color change. is not explicitly referenced in the text prompt, so its retention is unspecified.\n\ndetailed_description:\n[Shot 1] Medium bright shot of (Little boy Will) leaning forward slightly, looking excited with open arms and palms up, resting on a colorful table. The scene is rendered in a bright, warm, kid-friendly 3D cartoon animation style, matching exactly, with soft clay-like skin, subtle freckles, rounded friendly facial features, vivid saturated colors, and warm even lighting with no harsh shadows. Do not drift toward photorealism, dark rendering, or overly detailed textures. Consistent color grading, saturation, and brightness are maintained. The video must contain absolutely zero text of any kind anywhere—no subtitles, captions, text overlays, lesson titles, labels, watermarks, or text visible inside the scene on walls, posters, signs, books, whiteboards, chalkboards, clothing, objects, or any background surface; all background surfaces must be clean and text-free. (S1) says [English]I think it’s fascinating! And how about playing games? Is that exciting? Yes! I love it too! directly to the camera, using the same male child voice as the reference with a fixed timbre and stable tone throughout, speaking slowly and clearly at a pace comfortable for young children learning English (approximately 100 words per minute) with a natural brief pause between phrases.\n\noverall_soundscape:\nN/A\n\nnon_diegetic_music:\nN/A", "original_prompt": "ABSOLUTE RULE: the video must contain ZERO text of any kind, anywhere — no subtitles, no captions, no text overlays, no lesson titles, no labels, no watermarks, and NO text visible inside the scene itself: no words on walls, posters, signs, books, whiteboards, chalkboards, clothing, objects, or any background surface. All background surfaces must be clean and text-free. Pure clean video only. [Image 1] is the character and outfit reference for this ENTIRE video. CRITICAL: Match [Image 1]'s exact clothing, colors, and fabrics in every frame. Scene and action only: use the visual prompt for background, camera, and motion. Do NOT use it to change character identity or clothing; [Image 1] outfit always wins. Scene: Medium bright shot Will from reference image leaning forward slightly excited with open arms palms up on the colorful table Little boy Will speaks directly to camera: \"I think it’s fascinating! And how about playing games? Is that exciting? Yes! I love it too!\" Character lock — [Image 1] is the SOLE reference for all visual quality. Little boy Will: strictly match facial features, skin freckles, hair strand detail, hair color, skin tone, clothing fabric texture, and 3D render shading to [Image 1] in EVERY frame. CRITICAL: Match the EXACT clothing in [Image 1] — color, fabric, fit, style, every detail. If any text prompt mentions a different outfit, color, hat, hoodie, jacket, or accessory, IGNORE that text and keep [Image 1]. No character drift, no facial distortion, no outfit change, no texture loss, no clothing color change across any frame. Art style: match [Image 1] exactly — bright, warm, kid-friendly 3D cartoon animation style. Soft clay-like skin with subtle freckles, rounded friendly facial features, vivid saturated colors, warm even lighting with no harsh shadows. Do NOT drift toward photorealism, dark rendering, or overly detailed textures. Consistent color grading, saturation, and brightness matching [Image 1] in every frame. Voice: same male child voice as the reference, fixed timbre, stable tone throughout — no pitch drift. Speak slowly and clearly at a pace comfortable for young children learning English — approximately 100 words per minute, with a natural brief pause between phrases.", "reference_images": ["references/7076c109492027650170cea448cef7f43477992ef672425d846c777aa7faf9e3.png", "references/1c0f0c09cc4c8fa4dab89c6b4a24813ed280809fe35cface21aa9b6e90097bea.jpg"], "reference_sha256": ["7076c109492027650170cea448cef7f43477992ef672425d846c777aa7faf9e3", "1c0f0c09cc4c8fa4dab89c6b4a24813ed280809fe35cface21aa9b6e90097bea"], "workflow": "ref2va", "prompt_sha256": "caf1d05e46b74038b881cd7df5a8053232960e9ae307b4dcea68ae5a7559a48a", "original_prompt_sha256": "16ab7fdb2f6c58a0a8b91ba40439bcbd9a58adfc1dbef74f3f6c53a60d1b377c", "video_sha256": "3e7f6369de9831a639d7d0b95bf8a07483861379465541d5d2ea26b9da6ce9dd", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 243, "duration_seconds": 10.125, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_038.json", "verification": "verification/sample_038.json"} {"file_name": "videos/sample_040.mp4", "sample_id": "#40", "prompt": "subject_definitions:\n is Pei Yingying, a young woman with tears on her face, her expression twisted in anger and panic, wearing clothing appropriate for a wealthy family dining setting. Her appearance and clothing are derived from .\n\nsummary:\n[reference generation] A 7-second close-up shot of Pei Yingying () in the Pei family dining room. She is being yelled at, her face showing a mix of anger, panic, and embarrassment, with tears still on her cheeks. The scene features warm golden morning sunlight, ultra-realistic texture, a cinematic shallow depth of field, and no subtitles. Pei Yingying delivers an inner voiceover expressing her frustration.\n\nretention_analysis:\nfully_preserved: The identity and appearance of Pei Yingying from are fully preserved as the main subject of the close-up shot. The reference is directly tied to the character's facial expressions and emotional state.\n\ndetailed_description:\n[Shot 1] The shot is a 7-second close-up (chest and above) with a fixed camera and a slight breathing feel. The scene is set in the Pei family dining room, bathed in warm golden morning sunlight pouring through massive floor-to-ceiling windows. The lighting is unified and realistic, with a cinematic large aperture creating a natural, soft shallow depth of field. (Pei Yingying, appearance and clothing from ) is framed in the foreground. Her face is illuminated by the warm golden sunlight, revealing ultra-realistic skin texture, visible pores, nasolabial folds, and hair layers. Her brows are tightly furrowed, and her eyes are filled with anger and panic. Tears still hang on her face, and her expression is twisted, conveying a mix of fury and anxiety. Her mouth is slightly open as if she wants to speak but is stunned into silence by the yelling. Her body slightly twists as if trying to break free, but she is powerless. Her previous pitiful demeanor is completely gone, replaced by an angry and embarrassed state. In the blurred background, the out-of-focus figure of Zhen Wanwan is visible. The camera maintains a strict 180-degree axis without crossing. The overall aesthetic is warm gold and passionate, with off-white and light brown low-saturation tones. The texture is 100% ultra-realistic live-action, with no AI feel, no plastic feel, and no flat look. No subtitles are to be displayed on screen. The source requests 8K resolution and 60fps; the generation pipeline's native settings are 768x1344 at 24fps, so these cannot be fully honored — render at maximum available fidelity. (S1) says in an off-screen voiceover [Chinese]这个暴发户到底在发什么疯?!我只是想装个可怜而已。\n\noverall_soundscape:\nThe soundscape includes the rapid, panicked breathing of Pei Yingying and the environmental acoustics of the grand dining room. A symphony sound effect is specified to play in the environment.\n\nnon_diegetic_music:\nN/A", "original_prompt": "全剧核心统一设定8K 超写实真人电影质感,60 帧,现代豪门宅斗爽剧题材,脸部毛孔、皮肤纹理、法令纹、颈纹、发丝层次、衣物布料编织纹理、褶皱垂感、皮肤光泽度 100% 还原真人;无背景 BGM、无字幕,仅保留台词与环境声场,无多余现代元素。电影级大光圈浅景深,焦外虚化自然柔和;光影统一、光质真实,;谁说话切谁近景,关键调度用中景 / 近景,重要动作 / 物件必加特写;镜头带轻微呼吸感,运镜丝滑稳定,180° 轴线不越轴;人物动作自然松弛、无僵硬感、无 AI 感、无塑料感;情绪真实、微表情细腻、肢体舒展有张力,完全还原真人表演状态。第 11 集单集核心限定\n暖金激昂餐厅美学,米白 + 浅棕低饱和色调,清晨暖金色自然光从落地窗倾泻而入;运镜以低角度仰拍、快推特写、定点定格、肩扛呼吸感为主,情绪递进从诡异凝滞→慷慨激昂→错愕混乱→阴阳暴击;人物动作松弛、肢体舒展、无僵硬、无 AI 感;面部微表情细腻,反差情绪转折真实;景深清晰,焦外虚化自然,画面层次丰富;全程超写实真人质感,皮肤肌理、发丝、布料细节拉满。11-1 裴家餐厅 日 内场景详细描述\n延续第 10 集的裴家餐厅,挑高四米的空间被清晨暖金色的阳光彻底笼罩,巨大的落地窗透进明亮的光线,大理石地面光洁如镜,长条形大理石餐桌摆在房间中央,银质餐具、骨瓷餐盘依旧整齐摆放,地上散落着玻璃碎片与水渍,与温馨的早餐氛围形成强烈反差。餐桌旁的真皮餐椅纹理清晰,一旁的餐边柜上精致的花艺微微晃动。陈设摆放符合顶级豪门餐厅标准,木质家具的纹理、金属装饰的光泽、陶瓷的质感真实清晰,无塑料感。氛围从最初的极度诡异凝滞,逐渐被甄绾绾的演讲点燃为激昂热血,最后转为裴盈盈的错愕与愤怒,画面电影感拉满,完全无 AI 感、无平面感、无假质感。\n超写实真人质感,毛孔清晰,具有符合相应年龄段的肌理感。去掉人物 AI 感,增强人物皮肤质感,高清写实人物质感。8k 分辨率。\n镜号:11-1-09\n秒数:7 秒\n景别:近景(胸部以上)\n运镜:固定镜头,轻微呼吸感\n画面提示词:8K 超写实真人质感,裴盈盈图1被吼得耳膜嗡嗡作响,眉头紧锁,眼神中充满了愤怒与慌乱,她的眼泪还挂在脸上,表情扭曲,透露出又气又急的情绪。暖金色的阳光打在她的脸上,皮肤纹理、毛孔清晰可见,大光圈浅景深,背景虚化的甄绾绾身影模糊。\n面部情绪:眉头紧锁,眼神愤怒又慌乱,嘴巴微张,想要说话却被甄绾绾的吼声震得说不出来,身体微微扭动想要挣脱,却发现根本无能为力,整体神态又气又急,狼狈不堪,之前的楚楚可怜荡然无存。\n台词(裴盈盈 OS):这个暴发户到底在发什么疯?!我只是想装个可怜而已。\n音效:交响乐、裴盈盈急促的呼吸声\n特效:无\n备注:无字幕,镜头流畅,超写实真人风格", "reference_images": ["references/0c4b513b7bd54dcc9e16e37d62d1034402eb738c8cb9b34f4547ee02e76c748a.jpg"], "reference_sha256": ["0c4b513b7bd54dcc9e16e37d62d1034402eb738c8cb9b34f4547ee02e76c748a"], "workflow": "ref2va", "prompt_sha256": "0a48045b054df4a5a2bc8324376c740e551adc1ed0be07ad2ba6334f52d12e5d", "original_prompt_sha256": "1c2f35d26a04de1b764b9308dbbd728298f8998f376c41297dc7a11aa9957a35", "video_sha256": "7b13af8c64b7c16e5b9a6d92e1647be3c4de2681da96039e9970172097ca4a17", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 124, "duration_seconds": 5.166666666666667, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_040.json", "verification": "verification/sample_040.json"} {"file_name": "videos/sample_041.mp4", "sample_id": "#41", "prompt": "integrated_multimodal_description:\nMax 70mm film texture, realistic grain, epic scale, low saturation. 4K high definition, cinematic texture, smooth and natural character movements, no distortion or deformation, conforming to realistic physical logic. No subtitles. [Shot 1] (0-1s) The actress in black raises her hand, and thick black-red particle demonic energy bursts from her hand, occupying one-tenth of the frame. [Shot 2] At 0:01.000, the thick black-red particle demonic energy from the actress in black's hand, occupying one-tenth of the frame, shoots out of frame at a 45-degree angle to the left. [Shot 3] At 0:03.000, the black-red demonic energy wraps around the actor in yellow on the ground. [Shot 4] At 0:04.000, the black-red demonic energy wraps around the woman in yellow and seeps into the ground. [Shot 5] At 0:05.000, the actor in black withdraws her hand, and the black-red demonic energy dissipates.\n\noverall_soundscape:\nOnly natural environmental sounds and on-set sound effects are retained.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【全局基础设定】\n@视频1 为原始拍摄素材(包含基础演员表演与构图)\n风格要求:max70mm胶片质感,颗粒写实,史诗规模低饱和度\n音频要求:无音乐、无字幕、仅保留自然环境音和现场音效\n【时间片分镜脚本】\n0-1s:完全保留@视频1(原始素材)黑衣女演员抬手,手中冒出浓稠的黑红粒子魔气(画面占比十分之一),运镜与@视频1 原有运镜保持完全一致。1-3s:完全保留@视频1 (原始素材 )原有构图与演员表演动作,黑衣女演员手中的浓稠黑红粒子魔气(画面占比十分之一),射向画左45度的画外。运镜与@视频1 原有运镜保持完全一致。3-4s:完全保留@视频1 (原始素材)原有构图与演员表演动作,黑红魔气将地上的黄衣演员包裹住。运镜与@视频1 原有运镜保持完全一致。4-5s:完全保留@视频1 (原始素材)原有构图与演员表演动作,黑红魔气包裹住黄衣女子,渗入地下。运镜与@视频1 原有运镜保持完全一致。\n5-7s:完全保留@视频1 (原始素材)原有构图与演员表演动作,黑衣演员收回手,黑红魔气消散。运镜与@视频1 原有运镜保持完全一致。\n【画质与防崩坏约束】\n4K高清,电影级质感,人物动作流畅自然,画面无畸变、无崩坏,符合现实物理逻辑", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "28b8b81579ae26d0cf05ade567ea2718b78a7924c2963d943c374b06ee7c2ec2", "original_prompt_sha256": "7881a0455c0f48a8e3303bb0630d7a6d9e5a043c88e9260e413f458ff22ef5de", "video_sha256": "91cff538120d1f1d49997c57621f4eb894a396f71835d2753ca7453e591fc3a0", "task_adapted": true, "width": 1344, "height": 768, "fps": 24, "num_frames": 175, "duration_seconds": 7.291666666666667, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_041.json", "verification": "verification/sample_041.json"} {"file_name": "videos/sample_042.mp4", "sample_id": "#42", "prompt": "subject_definitions:\n is the female model, with facial appearance and standard figure from , featuring clean and flawless skin texture, natural slightly wavy wet hair, and vacation-style sheer makeup. is the clothing, strictly using the exact color, shape, and details from , featuring knitted hollow fabric, geometric hollow texture, and a wavy hem. is the background, directly using the architectural light and shadow scene from .\n\nsummary:\n[reference generation] A 12-second women's clothing e-commerce marketing video designed to promote purchase conversion. The video features a female model wearing specific clothing in a specific architectural background, showcasing the fabric's texture and drape through three distinct shots with specific camera movements.\n\nretention_analysis:\n is fully preserved as the facial and physical reference for the model, requiring her standard figure and flawless skin to match the image. is fully preserved as the strict clothing reference, requiring the exact color, shape, details, knitted hollow fabric, geometric hollow texture, and wavy hem to match without any alteration or color difference. is fully preserved as the direct background reference, requiring the architectural light and shadow scene to match the image.\n\ndetailed_description:\n[Shot 1] Full body long shot. confidently walks into the frame from the architectural light and shadow of . She walks lightly, showcasing the transparency and drape of the knitted hollow fabric of . The camera follows and pushes in. has natural slightly wavy wet hair and vacation-style sheer makeup. Her hands are not in her pockets. She does not stand on any tables or chairs, and no clipping occurs where she penetrates objects.\n[Shot 2] At 0:04.000, medium shot. turns sideways and looks back, lightly brushing her hair. This showcases the dynamic rhythm of the geometric hollow texture and wavy hem of . The camera slowly orbits around her. Her hands are not in her pockets.\n[Shot 3] At 0:08.000, upper body close-up. interacts vividly with the camera. Light and shadow create light and dark rhythms on the crochet fabric of , fully reflecting the fabric texture. The camera smoothly pulls back to finish. Her hands are not in her pockets.\n\noverall_soundscape:\nNo specific sound effects are requested beyond the non-diegetic music. The scene follows basic physical laws with no clipping or unnatural interactions.\n\nnon_diegetic_music:\nPure instrumental music with no vocals. The style is lazy and high-end bossa nova guitar.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。模特面部参考图一,标准身材,肤质洁净无瑕。服装严格使用图二,颜色/版型/细节不允许有任何改变或色差;背景直接使用图三。符合基本物理规律,禁止出现穿模(人物穿透物体)或模特站在桌椅上,双手禁止插兜。模特采用自然微卷湿发,搭配度假风清透妆容。第一段(4秒):全身远景。模特从建筑光影中自信步入,通过轻盈行走展现针织镂空面料的通透感与垂坠性,镜头跟随推进。第二段(4秒):中景。模特侧身回眸并轻拨发丝,展示几何镂空纹理与波浪边下摆的灵动节奏,镜头慢速环绕。第三段(4秒):上半身近景。模特面对镜头生动互动,光影在钩编面料上产生明暗律动,充分体现面料质感,镜头平滑拉远收尾。配乐:纯音乐,无人声。风格为慵懒高级的波萨诺瓦吉他。\\n\\n素材说明:图1=模特参考(asset-20260224194849-qt54s);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552.jpg", "references/da9b841efce2b0b754c8a8cce9b4b6c46b7aec6fff9a09b8d1dd48646ffe163f.jpg", "references/6e0374d98930816df122c6a7b492451f93c5bf162d9e9290c4ba3993133e3fc1.jpg"], "reference_sha256": ["96d15ee0c7c247746ae68346d9ff44b531d8fdd47781e2fab420eccb0a9b3552", "da9b841efce2b0b754c8a8cce9b4b6c46b7aec6fff9a09b8d1dd48646ffe163f", "6e0374d98930816df122c6a7b492451f93c5bf162d9e9290c4ba3993133e3fc1"], "workflow": "ref2va", "prompt_sha256": "8933e36330eaddf62a255ddf3273b7f72f01cceee1170ff0b4c8c4284d5c4c2d", "original_prompt_sha256": "fe87bd9381485151e789fe8c2e18366a7c8da490fcc332e31b7a24bf83d7e44e", "video_sha256": "5c75dd415925b41392f8ed485bb2e777cfdde50b2e25d7cb55f1a5b524e1f342", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_042.json", "verification": "verification/sample_042.json"} {"file_name": "videos/sample_043.mp4", "sample_id": "#43", "prompt": "integrated_multimodal_description:\nGlobal constraints: no text or lettering anywhere in the image; retain environmental sounds and sound effects. Su Jinghong has a slightly plump build and must not be depicted as thin. [Shot 1] From 0.0 to 2.4 seconds, begin with a hard cut. Daytime in the Changchun Palace bedchamber. Close-up, eye level, extremely slow push-in. Xiao Che suddenly freezes, his brow locking; surprise flashes through his eyes, then quickly turns to anger. He pauses for half a beat, jaw tightening. At first, the outer parts of his eyebrows rise, the brow peaks lift, and his lips part with a 5 mm gap between the upper and lower teeth. His face then darkens: eyebrows lower, brow furrows tightly, upper eyelids lower, lips press together, and his gaze looks straight ahead, conveying authority publicly challenged. He cannot accept Su Jinghong, normally obedient, answering him defiantly. Xiao Che (S1) says through gritted teeth in anger [Chinese]你说什么? No extra characters appear.\n\noverall_soundscape:\nEnvironmental sounds and sound effects, specifically the sound of the jaw tightening and a slight sleeve tremor.\n\nnon_diegetic_music:\nN/A", "original_prompt": "画面不要出现任何文本和文字,保留环境声和音效!!!\n苏惊鸿体型稍胖,画面中不要显瘦!!!\n【场景分析】\n场景:(长春宫寝殿/日间)\n【镜头2】6.0-8.4秒\n转场: 硬切\n镜头: 特写,平视,极缓推近\n画面描述: 萧彻神情骤停,眉心锁死,眼中先闪过愕然,再迅速沉成怒色。\n人物视线镜头参考图1,但是不要出现多余的人物\n核心动作: 他整个人愣住半拍,牙关随即绷紧。\n对手戏设计: 萧彻先是眉毛外侧上扬,眉峰抬高,嘴巴微张,上下牙间距5毫米。随后面色一沉表情转变眉毛下压加眉心紧锁加上眼睑下垂加嘴唇抿紧,目光直视,权威感被当众顶回。\n人物底层逻辑: 他无法接受一向顺从的苏惊鸿当面反刺自己。\n主体动作: 萧彻面色一沉,咬牙切齿地说:“你说什么?”\n音效:牙关收紧声,衣袖轻震声\n资产对照表(视频中涉及的角色、场景与物品如下(保持一致性)):\n视频场景对照表: 长春宫寝殿=【图2/日间】\n视频角色对照表:萧彻=【图3】;", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "56e7e2259ed2865172c6396605f397af4faa1c89f54b1a483eba7694dc99f2d4", "original_prompt_sha256": "95c1f8d079f5724e9c7da90232cd05918fa698dfffb59b1cf4984f29ba4dc40c", "video_sha256": "462ebc0b2d30c92eff7a2d1fe616540762aabddd9154abc252d2f9584843da38", "task_adapted": true, "width": 768, "height": 1024, "fps": 24, "num_frames": 107, "duration_seconds": 4.458333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_043.json", "verification": "verification/sample_043.json"} {"file_name": "videos/sample_044.mp4", "sample_id": "#44", "prompt": "subject_definitions:\n: The main UI screenshot of the Government Affairs Knowledge Graph Platform (from ), featuring a blue sidebar on the left, a central radial knowledge graph node tree, and an entity card list on the right.\n: The complete dialogue window screenshot of the AI Intelligent Government Affairs Customer Service Xiao Lan (from ), featuring a dialogue history list on the left, government affairs Q&A dialogue bubbles between Mr. Li and Xiao Lan in the center, and a Related Resources and Assistants panel on the right.\n\nsummary:\n[reference generation] The video animates the provided UI screenshots of a government affairs knowledge graph platform and an AI customer service dialogue window. It features a fixed camera with a slight pull-back at the end, highlighting data flow, node activation, and UI element growth while strictly preserving all original text and layout pixel-perfectly.\n\nretention_analysis:\n (from ): fully_preserved. The entire UI layout, including the sidebar, knowledge graph tree, and entity cards, must be strictly maintained as the foreground and layout anchor throughout the video.\n (from ): fully_preserved. The dialogue window flies in and overlays as a sub-card element. All dialogue bubbles, history list, and panels must retain their exact original appearance and text.\n\ndetailed_description:\n[Shot 1] At 0:00.000, fixed camera. The central root node of lights up, and a blue halo slowly diffuses from the root node to four main branches (政策文件, 政府部门, 服务事项, 法律法规). On the left, the numbers \"实体总数 12,458\" and \"关系总数 34,192\" grow in place via scrolling accumulation. The UI text position, font size, and stroke thickness must be completely consistent with the original image. Overall style is government blue (#1A4FBA) and cold white background tech style.\n[Shot 2] At 0:03.000, the camera remains fixed. The AI customer service dialogue window of flies in smoothly from the front right with a slight 3D flip, leaving a trail of light filaments upon landing, followed by a micro-bloom pulse on its surface. The dialogue bubbles inside containing the text \"您好!我是您的专属智能政务客服小蓝...\" and \"我想了解一下,现在注册一家科技类有限责任公司...\" remain completely clear and readable, with no altered copy or garbled text.\n[Shot 3] At 0:06.000, the rounded bars in the 实体筛选 area for 政策文件, 政府部门, 服务事项, and 法律法规 grow from zero width to full width with a sweeping light effect from left to right, with yellow fill sweeping across. In , Xiao Lan's response content (the 5-item material list for enterprise opening one-netcom service) highlights and flashes line by line, each line highlighting for 0.4 seconds, sequentially from top to bottom.\n[Shot 4] At 0:09.000, the camera pulls back extremely slowly (only push/pull motion, amplitude ≤6%). The connecting lines between all knowledge graph nodes flow with blue data stream light bands from the root node to the leaf nodes. The leaf cards slightly pitch forward by 6°, reflecting cold white environmental light. The shot finally freezes on the coexisting side-by-side composition of (Knowledge Graph Panorama) and (AI Customer Service Dialogue Window). The top titles \"政务知识图谱平台\" and \"AI 智能客服\", all text inside the dialogue bubbles, and the update timestamp in the bottom left corner remain completely unchanged. The right panel displays 相关资源与助手.\nOverall visual style: 720P HD, government tech promotional documentary style, cold high-tech atmosphere, soft lighting, distinct UI element layers, conforming to government standardized visualization aesthetics. Camera motion strictly uses only fixed camera and slight pull-back at the end, prohibiting yaw, roll, fast panning, shaking, or handheld feel. The screen must have no subtitles, no watermarks, and no third-party logos. Natural and smooth motion. Government blue main color tone throughout, prohibiting high-saturation entertainment colors. Strict pixel-level reference to the uploaded screenshots for all UI text, layout, controls, and colors, ensuring no typos, garbled text, or AI-generated pseudo-characters. Text areas must not be blurry, jittery, deformed, torn, flickering, or misaligned. If any text becomes occluded during animation, the occluded portions must remain unchanged; no new/spurious text may be generated in occluded areas.\n\noverall_soundscape:\nUnspecified.\n\nnon_diegetic_music:\nN/A", "original_prompt": "@图片1 为「政务知识图谱平台」UI 截图主体(含左侧蓝色侧边栏、中央放射状知识图谱节点树、右侧实体卡片列表),全程作为画面前景与版式锚点;@图片2 为「AI 智能政务客服小蓝」的完整对话窗口截图(含左侧对话历史列表、中央李先生与小蓝的政务问答对话气泡、右侧'相关资源与助手'面板),作为画面中飞入并叠加的子卡片元素。整体为政务蓝(#1A4FBA)+冷白底科技风。\n\n0–3 秒:固定机位起手,@图片1 中央根节点亮起,蓝色光晕从根节点向四条主分支(政策文件/政府部门/服务事项/法律法规)缓慢扩散,左侧'实体总数 12,458''关系总数 34,192'数字以滚动累加方式生长到位,UI 文字位置、字号、笔画粗细与原图完全一致。\n\n3–6 秒:镜头保持固定,@图片2 的 AI 智能客服对话窗口以轻微 3D 翻面方式从右前方平稳飞入,落位时带尾迹光丝,落定后表面浮现一次微 bloom 脉动;@图片2 内的对话气泡(小蓝的'您好!我是您的专属智能政务客服小蓝...'与李先生的'我想了解一下,现在注册一家科技类有限责任公司...')文字保持完全清晰可读,不得改变文案、不得乱码。\n\n6–9 秒:'实体筛选'区域的政策文件、政府部门、服务事项、法律法规圆角条从 0 宽度向右扫光生长到满,黄色填充由左至右扫过;@图片2 中小蓝的应答内容(企业开办一网通办的 5 项材料清单)逐行高亮闪现一次,每行高亮 0.4 秒,顺序由上至下。\n\n9–12 秒:镜头极缓慢后拉(仅推拉一种运镜,幅度≤6%),全图谱节点之间连接线以蓝色数据流光带从根节点向叶子节点顺向流动,叶子卡片轻 pitch 前倾 6°,表面反射环境冷白光,最终定格在「@图片1 知识图谱全景 + @图片2 AI 客服对话窗口」并列共存的画面上,顶部'政务知识图谱平台'标题、'AI 智能客服'标题、对话气泡内全部文字、左下角更新时间戳保持完全不变。\n\n720P 高清,政务科技宣传片纪实风,冷调高科技氛围,光影柔和,UI 元素层次分明,符合政务规范可视化美学。\n\n运镜:全程仅使用'固定机位+末端轻微后拉'一种运镜,禁止 yaw/roll/快速转向/晃动/手持感;画面无字幕、无水印、无第三方 logo;动作自然流畅,不僵硬、不卡顿、不闪烁;政务蓝主色调贯穿全片,禁止高饱和娱乐化色彩。\n\n【素材保真硬约束】@图片1(政务知识图谱平台 UI 截图)与 @图片2(AI 智能政务客服小蓝对话窗口截图)为本视频的真实素材主体,全程必须严格、像素级地参考上传截图的图片素材:(a) 所有界面文字(中文汉字、英文、数字、标点)的字形、笔画、字号、字重、字距、行距、颜色、对比度、排版位置必须与原图完全一致,严禁出现错字、缺笔、多笔、合字、变体字、外文乱码、火星文、AI 伪汉字或任何无法识别的字符;(b) 节点名称、卡片标题、菜单项、按钮文案、统计数字('实体总数 12,458''关系总数 34,192'等)、对话气泡内的文字内容必须逐字与原截图一致;(c) UI 整体版式、控件位置、图标、配色、圆角、阴影、分割线必须保持原样,禁止 AI 自由发挥重新设计 UI;(d) 文字区域不得模糊、抖动、变形、撕裂、闪烁、错位;(e) 若动画过程中文字会被遮挡,被遮挡部分仍要保持原样,不得在被遮挡处生成新的虚假文字;(f) 视频任何一帧的 UI 文字都应做到清晰锐利、可读、不变形。", "reference_images": ["references/9c2ed8f075f78bb326ef9e5b9a5dffee5add6424c431437d0a87c7ae2e499cc7.png", "references/3315ec7206dadb6f4a2e001ed5897d377a78040b93308f2f6a93afb2580a515d.png"], "reference_sha256": ["9c2ed8f075f78bb326ef9e5b9a5dffee5add6424c431437d0a87c7ae2e499cc7", "3315ec7206dadb6f4a2e001ed5897d377a78040b93308f2f6a93afb2580a515d"], "workflow": "ref2va", "prompt_sha256": "24a9cfc778dd5d8a7080a3457610aa6697e2a9be03600499863acb211ed7e3cf", "original_prompt_sha256": "b0c4ad37f89dadd89ab32f0da3d13276ff07769f945a02154eb3f04ea23800ce", "video_sha256": "b8be311ed6dac0b563e9fb30861df8af228f5e47af873e308f8ef25a07b3c90f", "task_adapted": false, "width": 1344, "height": 768, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_044.json", "verification": "verification/sample_044.json"} {"file_name": "videos/sample_045.mp4", "sample_id": "#45", "prompt": "integrated_multimodal_description:\n[Shot 1] Begin with a hard cut. From 0:00.000 to 0:03.600, in the Lin family main room, Lin Jianjun stands on the right side of the frame while He Yutang sits on the left. The camera is in a medium close-up, eye-level shot with a slow push-in. Lin Jianjun looks at He Yutang at a round table, gently pushing a white porcelain bowl of porridge towards her. His movement is deliberately gentle, and a smile remains on his face. The distance between them is controlled to create an illusion of closeness without pressing. His steady tone is meant to lower her defenses, using the amnesia claim as a universal explanation. Lin Jianjun (S1) places the white porcelain bowl on the round table and says considerately: [Chinese]雪芳,医生说你是失忆了,只要在熟悉的环境,就会慢慢好起来的。\n\n[Shot 2] From 0:03.600 to 0:05.800, transitioning with a dissolve, the camera remains in a medium close-up, eye-level, fixed shot. He Yutang nods slightly, her expression still stiff, her gaze not fully moving away from the porridge bowl. She makes a minimal response out of politeness, avoiding direct eye contact with Lin Jianjun and only looking down to avoid direct emotional contact. She complies temporarily to protect herself in a strange environment. He Yutang (S2) says in a low voice: [Chinese]谢谢……\n\n[Shot 3] From 0:05.800 to 0:09.200, transitioning with a hard cut, the camera shifts to a close-up, high-angle looking down, with a slow push-in. Steam rises from the white porridge, and He Yutang's gaze rests on the surface of the porridge, the steam blurring the edges of her eyes. She stares at the white porridge in a daze, not reaching out immediately. Her hands stop at the edge of the table, forming a hesitant relationship between food and trust. She silently observes whether the mother-and-son story is credible. He Yutang stares silently at the bowl of white porridge, her fingers curling slightly at the edge of the table.\n\nNo subtitles appear in the video.\n\noverall_soundscape:\nShot 1: light sound of porcelain bowl touching the table, porridge surface swishing. Shot 2: slight swallowing sound, fabric rustling. Shot 3: extremely light sound of porridge steam, environmental static pressure increasing.\n\nnon_diegetic_music:\nN/A", "original_prompt": "场景:图1参考视频1视频中站位,不要参考视频内容和人物动作。图2站在右边,图3坐在画面左边位置【镜头1】 0-3.6秒 转场: 硬切 镜头: 近景/平视/缓慢推近 画面描述: 林建军在圆桌旁看向贺雨棠,将桌上的白粥瓷碗轻轻推到贺雨棠面前,动作刻意轻柔,脸上笑意不减。 核心动作: 放碗、靠近、安抚,一气呵成。 对手戏设计: 距离被控制在“亲近但不至逼迫”的假象范围。 人物底层逻辑: 用稳定语气削弱她的防御,让“失忆”成为解释一切的万能理由。 主体动作: 林建军把白粥瓷碗放在圆桌上,体贴地说:“雪芳,医生说你是失忆了,只要在熟悉的环境,就会慢慢好起来的。” 音效: 瓷碗触桌轻响,粥面晃动声 【镜头2】 3.6-5.8秒 转场: 叠化 镜头: 近景/平视/固定镜头 画面描述: 贺雨棠轻轻点头,表情仍僵,视线没完全从粥碗上移开。 核心动作: 她出于礼貌做出最小回应。 对手戏设计: 她不直视林建军,只以低头姿态回避正面情感接触。 人物底层逻辑: 暂时顺从是为了在陌生环境里保护自己。 主体动作: 贺雨棠低声说:“谢谢……” 音效: 轻微吞咽声,衣料摩擦声 【镜头3】 5.8-9.2秒 转场: 硬切 镜头: 特写/俯视/缓慢推近 画面描述: 热气从白粥上升起,贺雨棠的视线停在粥面,蒸汽模糊她的眼睛边缘。 核心动作: 她盯着白粥出神,没有立刻伸手。 对手戏设计: 画面中人物手部停在桌边,形成“食物—信任”的迟疑关系。 人物底层逻辑: 她在用沉默观察这对母子的说辞是否可信。 主体动作: 贺雨棠无言地盯着那碗白粥,手指在桌边轻轻蜷起。 音效: 粥的热气声极轻,环境静压感增强 资产对照表(视频中涉及的角色、场景与物品如下(保持一致性)): 视频场景对照表: 林家正屋=【图1】 视频角色对照表: 贺雨棠=【图3】;林建军=【图2】 视频道具对照表: 白粥瓷碗=【图4】视频中不要出现字幕。", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "6bfba4f86a8ee4fd177eda69cd6722588a4f3508c676336c22fe2a4611771a2f", "original_prompt_sha256": "f2c0813a13eefae0371c39cdd3aac7af86816a7d18c42a72343309cf8b5802ba", "video_sha256": "e9ac7f807ee6718080dd8a4da0b083560d305c1089514ffc3870401ac084db16", "task_adapted": true, "width": 768, "height": 1024, "fps": 24, "num_frames": 226, "duration_seconds": 9.416666666666666, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_045.json", "verification": "verification/sample_045.json"} {"file_name": "videos/sample_046.mp4", "sample_id": "#46", "prompt": "integrated_multimodal_description:\nThe entire sequence is rendered in max70mm film texture with realistic grain, epic scale, low saturation, and all shots maintain a large aperture with shallow depth of field. The visual quality is 4K high-definition with cinematic texture, ensuring smooth and natural character movements, no visual distortion or structural collapse, and strict adherence to realistic physical rules.\n\n[Shot 1] A thick, dense black-red demonic energy quickly falls and lands between the two actors in the frame.\n[Shot 2] At 0:01.000, a thick, dense black-red demonic energy quickly falls and lands between the two actors.\n[Shot 3] At 0:02.000, the demonic energy rolls and churns, transforming into a woman dressed in yellow clothing. She reaches out and grabs the hand of the actress positioned on the left side of the frame.\n[Shot 4] At 0:03.000, the scene continues with the established visual constraints, maintaining 4K high-definition cinematic quality, smooth and natural character movements, no visual distortion or structural collapse, and strict adherence to realistic physical rules. No subtitles.\n\noverall_soundscape:\nOnly natural environmental sounds are present. No music.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【全局基础设定】\n@视频1 为原始拍摄素材(包含基础演员表演与构图)\n@图片1 作为环境和位置参考。风格要求:max70mm胶片质感,颗粒写实,史诗规模低饱和度,所有镜头保持大光圈,浅景深。\n音频要求:无音乐、无字幕、仅保留自然环境音。\n【时间片分镜脚本】\n0-1秒,镜头1:完全保留@视频1 (原始素材)的原有构图与演员表演动作,一道浓稠的黑红魔气快速落到画面中两个演员中间。运镜与@视频1 原有运镜保持完全一致。1-2秒,镜头2:完全保留@视频1 (原始素材)的原有构图与演员表演动作,一道浓稠的黑红魔气快速落到两个演员中间。运镜与@视频1 原有运镜保持完全一致。2-3秒:镜头3::完全保留@视频1 (原始素材)的原有构图与演员表演动作,魔气翻滚,变化为了黄衣女子,她一把抓住了画左女演员的手。运镜与@视频1 原有运镜保持完全一致。3-4秒,镜头4:完全保留@视频1 (原始素材)的原有构图与演员表演动作,运镜与@视频1 原有运镜保持完全一致。【画质与防崩坏约束】\n4K高清,电影级质感,人物动作流畅自然,画面无畸变、无崩坏,符合现实物理规则。", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "155d59a0e0ff04bb10645024fefe136465ce533485817e223303527564eb1777", "original_prompt_sha256": "7a72370a64c4f225f20214c29f1aeac6270e967cf75f15a518f7ab6d50224d3b", "video_sha256": "555fe0b97ab03f6509759ff61458516c65a1e80962cd62a04c540d1123b0d9af", "task_adapted": true, "width": 1344, "height": 768, "fps": 24, "num_frames": 107, "duration_seconds": 4.458333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_046.json", "verification": "verification/sample_046.json"} {"file_name": "videos/sample_047.mp4", "sample_id": "#47", "prompt": "integrated_multimodal_description:\n[Shot 1] At 0.0s–2.4s, hard cut to a long shot, eye-level, static camera. The scene is the Changchun Palace bedchamber in daytime. The palace doors are violently pushed open inwards, shaking the doors and letting sunlight rush in. Xiao Che (S1) strides into the main hall with his robes surging forward, and Bai Lingwei (S2) follows closely behind. Xiao Che's eyebrows are pressed down and his brow is furrowed, occupying the main foreground position. Bai Lingwei's mouth corners are raised but her eyes are still with no wrinkles around them, positioned behind him. Su Jinghong sits steadily in the center of the hall, her eyebrows, eyes, and mouth corners slightly raised with an expression of sure victory, forming a face-to-face confrontation. Sound effects include the crash of the doors opening, rapid footsteps, and the rustling of robes.\n\n[Shot 2] At 2.4s–4.8s, hard cut to a medium close-up, slightly low-angle, slow push-in. Xiao Che's face is full of anger, his jaw clenched, and his shoulders and back press forward, his dragon robe projecting an imposing momentum under the light. His upper eyelids are slightly drooping, his eyes firm, and his gaze is locked directly onto Su Jinghong in the center of the hall, maintaining a dominant oppressive position. Xiao Che (S1) shouts sternly: [Chinese]苏惊鸿!你好大的胆子! The sound effect is his shouting echoing in the hall.\n\n[Shot 3] At 4.8s–7.0s, hard cut to a medium close-up, eye-level, slow lateral tracking (trucking) move. Bai Lingwei shrinks half a step behind Xiao Che, using his cover to reveal a grievance posture. Her eye sockets are red, her eyes slightly squinted to hide a trace of tears, her lower lip slightly pouted, and her brow furrowed, forming a front-protecting/back-hiding relationship with Xiao Che. She lightly touches the corner of her eye with a handkerchief, her body slightly trembling. Her gaze passes over Xiao Che and lands on Su Jinghong. Bai Lingwei (S2) says while sobbing: [Chinese]皇后娘娘,臣妾知道您不喜欢臣妾。 Sound effects include light sobbing and the fine sound of the handkerchief brushing her eye.\n\n[Shot 4] At 7.0s–11.8s, hard cut to a medium shot, eye-level, slow push-in. Bai Lingwei supports her abdomen with one hand and clutches her handkerchief with the other. Her eyelashes are stained with tears, her chest heaves with crying, and her lower lip is pouted. She leans diagonally into Xiao Che's protection zone, her gaze weak but accurately pointing at Su Jinghong. She touches her belly to use the unborn heir to strengthen her accusation. Bai Lingwei (S2) says with aggrieved sobbing: [Chinese]可臣妾腹中怀着龙裔,您就算再生气,也不能想着自尽,把这畏罪自尽的污名扣在臣妾头上啊…… Sound effects include choking sobs, light rustling of clothes in the hall, and an enhanced sense of air pressure.\n\n[Shot 5] At 11.8s–14.6s, hard cut to a medium close-up, eye-level, static camera. Su Jinghong remains seated in the center of the hall, her expression unchanged. She slowly swallows the last bite of osmanthus cake, her throat moving up and down in a natural swallowing motion taking 1 second, her composure almost provocative. Her blinking rate is 15 times per minute, natural and unforced (not deliberate), without staring. She does not avoid the gazes of Xiao Che and Bai Lingwei, holding the center position. She leisurely picks up a handkerchief to wipe her mouth. Sound effects include a light swallowing sound and a brief silence in the hall.\n\nGlobal constraints: Absolutely no text or characters are allowed to appear on the screen. Su Jinghong's body type is slightly plump; she must not look thin in the frame.\n\noverall_soundscape:\nThe soundscape consists entirely of diegetic ambient sounds and physical sound effects, including the loud crash of palace doors opening, rapid footsteps, rustling of robes, echoing shouts, light sobbing, the fine sound of a handkerchief, choking sobs, light rustling of clothes, and a light swallowing sound followed by brief silence. No non-diegetic music is present.\n\nnon_diegetic_music:\nN/A", "original_prompt": "画面不要出现任何文本和文字,保留环境声和音效!!!\n苏惊鸿体型稍胖,画面中不要显瘦!!!\n【场景分析】\n场景:(长春宫寝殿/日间)\n承接:上一分镜末尾门外脚步声逼近,苏惊鸿已做好迎战姿态\n过渡:硬切爆发式入门,情绪意图为将对峙正式点燃\n人物表演底层逻辑总纲:【萧彻处于被挑动后的威压审判状态;白凌薇处于借势卖惨、巩固指控的表演状态;苏惊鸿处于表面松弛、实则观察来势的控场状态】\n【分镜承接】\n前一分镜末尾画面:苏惊鸿抬眸,神态冷定\n本分镜开场画面:殿门猛然向内推开,门扇掀动光影\n承接逻辑:空间内强衔接与动势爆发\n**【镜头1】0.0-2.4秒\n转场: 硬切\n镜头:远景,平视,固定镜头\n画面描述: 长春宫殿门被猛地推开,门外日光涌入,萧彻踏入正殿,白凌薇紧跟其后。\n核心动作: 门扇震开,萧彻大步跨入,衣袍前冲。\n对手戏设计: 萧彻眉毛下压,眉心紧锁位于前景主位,白凌薇嘴角上扬但眼睛不动,眼周无纹收在其后侧,苏惊鸿:眉眼和嘴角微微上扬,一种势在必得的样子稳坐于殿心形成迎面对峙。\n人物底层逻辑: 萧彻以压迫性入场先声夺人。\n主体动作: “砰”的一声后,萧彻带怒进入,白凌薇跟着进门。\n音效: 殿门撞开的响声,急促脚步声,袍摆破风声\n【镜头2】2.4-4.8秒\n转场: 硬切\n镜头: 中近景,略仰视,缓慢推近\n画面描述: 萧彻满脸怒气,下颌绷紧,肩背前压,龙袍在光下压出威势。\n核心动作: 他盯住苏惊鸿,眉毛下压,眉心紧锁,眼神坚定,高声喝斥。\n对手戏设计: 萧彻:眉毛下压,眉心紧锁,上眼睑微垂,眼神坚定,视线直锁殿中央的苏惊鸿,保持主导压迫位。\n人物底层逻辑: 他默认苏惊鸿有罪,要用气势压垮对方。\n主体动作: 萧彻厉声大喝:“苏惊鸿!你好大的胆子!”\n音效:斥喝声在殿内形成回响\n【镜头3】4.8-7.0秒\n转场: 硬切\n镜头:近景,平视,缓慢平移\n画面描述: 白凌薇缩到萧彻身后半步,眼眶通红,手帕轻点眼角,身体微微发颤。\n核心动作: 她借着皇帝的遮挡露出委屈姿态。\n对手戏设计: 她眼睛微微眯起藏着一丝泪水,下唇微撅,眉心紧锁与萧彻形成前护后藏的关系,目光越过萧彻投向苏惊鸿。\n人物底层逻辑: 她故意弱化自己的攻击性,只保留受害感。\n主体动作: 白凌薇往萧彻身后缩了缩,抽泣着说:“皇后娘娘,臣妾知道您不喜欢臣妾。”\n音效:轻微抽泣声,手帕拂过眼角的细响\n【镜头4】7.0-11.8秒\n转场: 硬切\n镜头: 中景,平视,缓慢推近\n画面描述: 白凌薇一手扶着腹部,一手攥着手帕,睫毛沾泪,语气哀婉却句句带刺。\n核心动作: 她摸向肚子,借腹中龙裔强化指控。\n对手戏设计: 白凌薇, 胸腔随着哭泣起伏,下唇微撅,眼神斜看着苏惊鸿,身体斜靠萧彻保护范围,视线柔弱却精准指向苏惊鸿。\n人物底层逻辑: 她要把苏惊鸿置于“自尽诬人”的更深污名里。\n主体动作: 白凌薇摸着自己肚子,委屈抽泣地说:“可臣妾腹中怀着龙裔,您就算再生气,也不能想着自尽,把这畏罪自尽的污名扣在臣妾头上啊……”\n音效: 哽咽声,殿内衣料轻响,空气压迫感增强\n【镜头5】**11.8-14.6秒\n转场: 硬切\n镜头:近景,平视,固定镜头\n画面描述: 苏惊鸿仍坐在殿中,面色不变,最后一口桂花糕缓慢咽下,动作从容得近乎挑衅。\n核心动作: 她吞下糕点,为反击蓄势。\n对手戏设计: 苏惊鸿:喉结上下移动,吞咽动作,用时1秒,眨眼频率每分钟15次,自然不刻意,不瞪眼。不回避萧彻与白凌薇的目光,稳住中心位。\n人物底层逻辑: 她用极度松弛回应外部施压,反向制造失衡。\n主体动作: 苏惊鸿慢悠悠地咽下最后一口糕点,拿起手帕准备擦嘴。\n音效:轻微吞咽声,殿内短暂静场\n资产对照表(视频中涉及的角色、场景与物品如下(保持一致性)):\n视频场景对照表: 长春宫寝殿=【图1/日间】\n视频角色对照表: 苏惊鸿=【图2】;萧彻=【图3】;白凌薇=【图4】", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "afef98629e0aaa5546824b8baeab9374bb4e84386328aa3d16dc094b5ae98a76", "original_prompt_sha256": "fe16f1c001b5280fe64980a871aaeed81d3fcba518e29f7c66cd54266b595737", "video_sha256": "1f6a708ecc62d88b6c48e6b459ee283006546dfbdf9527ffab7186429b8fa25a", "task_adapted": true, "width": 768, "height": 1024, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_047.json", "verification": "verification/sample_047.json"} {"file_name": "videos/sample_048.mp4", "sample_id": "#48", "prompt": "integrated_multimodal_description:\n[Shot 1] In a dim storage room illuminated by the cold light of a flashlight, Zhou Ran (S1) appears in a single-person close shot. He looks nervous and quickly takes out his phone, aiming it at a large iron barrel printed with the visible text \"XJ-2305\". The camera focuses on the barrel's batch number, intercut with a close-up of the batch number \"XJ-2305\". He takes photos and records videos of the surrounding warehouse environment for evidence, with the camera following his shooting actions. The atmosphere is tense and urgent. Zhou Ran (S1) says [Chinese]五吨明胶,原封未动,这就是马文斌吃回扣做黑账的铁证! in an urgent, low voice with a firm tone while recording.\n\n[Shot 2] In a close shot with side lighting from the door and predominantly dark areas, Lao Zhao (S2) guards the door outside the warehouse. His body is tense as he pokes his head out, his brows furrowed, vigilantly looking around. He turns his head back urgently and worriedly to speak to the person behind him, saying [Chinese]周总监,动作快点,我总觉得心里不踏实。\n\n[Shot 3] In a single-person close shot with the micro-light of a flashlight sweeping over his body, Zhou Ran (S1) quickly stuffs his phone into his inner pocket. He stands straight with determined eyes and speaks decisively, turning to the right side of the frame to say [Chinese]取证完毕。老赵,我们撤!\n\nStyle: 8K ultra-realistic live-action texture, no plastic feel, no AI traces, flat lighting for smooth faces, ultimate picture quality, soft light, bright and transparent, airiness, clean light and shadow, delicate texture, no over-sharpening, soft edges, no noise, no gray fog, weak shadows, high definition. Compact video rhythm, single shot duration not exceeding 5 seconds. Shots include close-up, close shot, medium shot, full shot, over-the-shoulder shot-reverse shot. Characters speak naturally with vivid expressions, real micro-expressions, slight blinking and breathing, and natural movements. Subtitle transparency is 0 (no subtitles visible). Delicate internal light and shadow, 45-degree side lighting, multiple 45-degree side shots, dark background, cinematic feel. No background music.\n\noverall_soundscape:\nOnly the characters' spoken dialogue is audible; no background music and no additional sound effects.\n\nnon_diegetic_music:\nN/A", "original_prompt": "重要规则:1. 仅使用参考音频的音色,完全忽略所有参考音频的内容,严格按照以下文本内容生成语音,不要添加任何额外内容;2. 所有参考音频仅用于音色克隆,不得作为背景音乐使用,不得播放音频中的任何原始内容;3. 禁止生成任何字幕,禁止添加任何背景音乐,视频仅保留角色对话语音:[图片1] 是场景储藏室, [图片2] 是周然, [图片3] 是老赵,[图片4] 是仓库外的门,[音频1] 是周然的声音,[音频2] 是老赵的声音,【8K超写实实拍质感,无塑料感,无AI痕迹,平面光照使人面部平整,极致画质,柔和光线,明亮通透,空气感,干净光影,细腻质感,无过度锐化,边缘柔和,无噪点,无灰雾,弱阴影,高清晰度,视频节奏紧凑,单镜头时长不超5秒,分镜含特写、近景、中景、全景、过肩正反打,人物自然说话、表情生动、微表情真实、轻微眨眼与呼吸、动作自然不僵硬;【视频中的字幕透明度为0】视频节奏紧凑,分镜包括特写近景中景全景等多个镜头。内光影细腻,人物微表情真实,动作自然流畅,无AI痕迹。45度侧光,多45度侧面镜头,暗调背景,电影感,8K超写实。】故事内容如下:\n1光影:储藏室昏暗手电筒冷光,昏暗的仓库内,周然(单人近景)神色紧张,迅速掏出手机,对准印有XJ-2305批号的大铁桶,聚焦铁桶批号(穿插铁通批号的特写镜头),对仓库周边环境进行全方位拍照、录像取证,(镜头跟随拍摄动作),氛围紧张急促,周然(单人近景)(一边录像一边低声)(急促低声,语气笃定):五吨明胶,原封未动,这就是马文斌吃回扣做黑账的铁证!\n2.近景 | 光影:门口侧光,暗部为主 | 老赵守在门口(门口参考 图片4 )身体紧绷,探出头眉头紧锁、警惕张望四周,语气急促担忧的回头对身后说:周总监,动作快点,我总觉得心里不踏实。\n3.单人近景 | 光影:手电筒微光扫过周身 | 周然迅速将手机塞进贴身内兜,身体站直、眼神坚定,语气果断利落转身对着画面右侧说:取证完毕。老赵,我们撤!", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "8b063dceea56365343123dbf6fd781d6afc9f5413fa2ab280d56351b9854e67d", "original_prompt_sha256": "e3592d5f1c3aa120bcce2b5626715ecb84925a4bfa766f8a5d1b82d333c01d39", "video_sha256": "7e4084498508f9c2e7308d50b4e5ae37e12ec9eed89c94173e0ef021cdb7b7ee", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_048.json", "verification": "verification/sample_048.json"} {"file_name": "videos/sample_049.mp4", "sample_id": "#49", "prompt": "subject_definitions:\n is the right-side ALOHA robotic arm. is the single cutout tactile-glove hand from , acting as the left-side executor. The objects are a playing-card box and a basket with perforated side walls.\n\nsummary:\n[reference generation] In a first-person robot manipulation video, the right ALOHA arm puts a playing-card box into a perforated basket. Then the left tactile-glove hand lifts and moves the basket.\n\nretention_analysis:\nfully_preserved: supplies the appearance of the single cutout tactile-glove hand used as the left executor. The glove is the executor itself, rather than an additional glove prop or an extra hand.\n\ndetailed_description:\n[Shot 1] First-person robot manipulation view, with a clean white tabletop background, soft laboratory illumination, natural shadows and clear image quality. The right-side executor is an ALOHA robotic arm. The left-side executor is the single cutout tactile-glove hand from . First, the right robotic arm grasps a playing-card box and puts it into a basket with perforated side walls. Subsequently, the left tactile-glove hand lifts and moves that basket. The tactile-glove hand is the left executor itself, not a new hand or arm additionally reaching into view and not an added glove object. Keep the right robotic arm, playing-card box and basket stable; contact, occlusion and physical interactions must be natural. The tactile-glove hand enters from the left to carry out the basket-lifting and moving action. Do not additionally place a static glove in the first frame, on the tabletop or beside the basket. Do not add any small objects unrelated to the task.\n\noverall_soundscape:\nUnspecified.\n\nnon_diegetic_music:\nN/A", "original_prompt": "使用视频1作为唯一动作、时序、相机运动和构图参考。视频1是干净白色背景下的第一视角机器人操作视频,画面中有左右两个 ALOHA 机械臂。视频1的任务语义是:右机械臂先抓取一个扑克牌盒,将扑克牌盒放入带孔侧壁的筐子中;随后左机械臂把这个带孔侧壁的筐子提起来并移动。输出视频必须保持视频1的分辨率、宽高比、帧率、时长和镜头视角,不要跟随图片1的分辨率或构图。保持干净白色桌面背景、实验室柔和光照、阴影关系和画面质量与视频1一致。默认保持右机械臂、扑克牌盒、带孔侧壁筐子的位置、轨迹、接触关系、遮挡关系和物理交互不变。只替换画面中左边的 ALOHA 机械臂及其夹爪外观:这是对原左机械臂像素区域的外观替换,不是在画面外额外伸入新的手或新的机械臂。替换后的主体必须占据原左机械臂的位置、尺度、朝向、可见范围和运动轨迹,并在原左机械臂提起并移动筐子的时间段执行同样动作。不要新增任何与原任务无关的小物体。把视频1中从左侧伸进画面的 ALOHA 机械臂和夹爪替换为图片1中的单只触觉手套手。这不是新增手套物体,而是把原左机械臂本身改成触觉手套手。原左机械臂什么时候进入画面,触觉手套手就什么时候进入画面;原左机械臂不在画面中时,画面里也不要出现触觉手套。不要在首帧、桌面或篮子旁边额外放置一只静止手套。触觉手套手必须沿用原左机械臂的位置、大小、方向和运动轨迹,完成提起并移动带孔侧壁筐子的动作。最终效果应像同一段机器人操作视频,只是左侧执行器被准确替换为图片1中的单只抠图触觉手套手。", "reference_images": ["references/a0522ec16309814fec171c00dc1dbbaaf91784f2d97d643bc6d0e445565fb991.png"], "reference_sha256": ["a0522ec16309814fec171c00dc1dbbaaf91784f2d97d643bc6d0e445565fb991"], "workflow": "ref2va", "prompt_sha256": "16fddc0cf5a67eaa095108285fb98a4ae2e7e80fc1687226b234babca60c83a1", "original_prompt_sha256": "f504590136a7fb06bb4867b23c10aedfc96ce4987ad289c773becdd86f7f1f6d", "video_sha256": "792320707ad0d718b26d8bdb3e513f11aab1177e67e41d018dd11c125caca995", "task_adapted": true, "width": 1024, "height": 768, "fps": 24, "num_frames": 209, "duration_seconds": 8.708333333333334, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_049.json", "verification": "verification/sample_049.json"} {"file_name": "videos/sample_051.mp4", "sample_id": "#51", "prompt": "subject_definitions:\n (Zhao Lei): young man, appearance and clothing from .\n (Sun Yue): young woman, appearance and clothing from .\n (Liu Guifang): older woman in homewear, appearance and clothing from .\n (Lin Xiaoke): young woman in a black dress, appearance and clothing from .\n: reference image for library character positioning.\n: reference image for the university library background.\n\nsummary:\n[reference generation] Generate a realistic cinematic video of approximately 12 seconds with two internal shots. Four characters sit around a desk in a university library; the three young people keep their heads down in a somber mood while (Liu Guifang) sits at the head of the table silently watching them. (Zhao Lei) and (Sun Yue) speak in-scene dialogue. Maintain cinematic lighting, soft focus, rim light, backlight, diffuse reflection, and air dust. Reject 3D, CG, plastic, UE5, and game feels. Use for spatial arrangement and for the background environment.\n\nretention_analysis:\n (Zhao Lei): fully_preserved, appearance and clothing from .\n (Sun Yue): fully_preserved, appearance and clothing from .\n (Liu Guifang): fully_preserved, appearance and clothing from .\n (Lin Xiaoke): fully_preserved, appearance and clothing from .\n (Library character positioning): fully_preserved, used for spatial arrangement of the four characters around the desk.\n (University library): fully_preserved, used for the background environment.\n\ndetailed_description:\nThe background is the university library in , in daytime: sunlight streams through large glass windows, rows of bookshelves extend into the distance, and the reading area is quiet and bright. The arrangement in has Liu Guifang in the center, Lin Xiaoke on the left in a black dress, Zhao Lei on the inner right, and Sun Yue on the outer right in a long white dress. [Shot 1] medium close-up, stable fixed camera: The scene features bright natural daylight with clear main light direction, preserved shadows, receding background, and an exposure hierarchy prioritizing the subject. (Zhao Lei) sits at the desk, frowning, staring blankly at the black screen of his phone, thumb unconsciously rubbing the edge of the phone, phone glass case reflecting the halo from the window; he smiles bitterly, eyes showing deep embarrassment, eyelids slightly raised looking at the desk, and (S1) says [Chinese]我这个月生活费花完了。再交不上辅导班的钱,我可能得放弃了。\n[Shot 2] medium close-up, slow lateral tracking: focus shifts to (Sun Yue) sitting next to him, (Sun Yue) still with head down, shoulders slightly hunched, hands tightly clutching the hem of her clothes, cotton fabric wrinkled into folds of varying depth; she keeps her eyes down not daring to look at anyone, and (S2) says in a very small voice [Chinese]我爸妈还是不同意我考文学。说再坚持就断我生活费。 No subtitles appear on screen.\n\noverall_soundscape:\nScripted character dialogue.\n\nnon_diegetic_music:\nMusic accompanying Sun Yue; no particular musical style is prescribed.", "original_prompt": "分镜2,预估12秒\n@赵磊_默认 这是赵磊;给赵磊配音\n@孙悦_默认 这是孙悦;给孙悦配乐\n@刘桂芳_居家服 这是刘桂芳 @刘桂芳的音色 \n@林小可_默认 这是林小可 @林小可的音色 \n@图书馆人物站位 中间的是刘桂芳 左边黑裙子的是林小可,右边靠里的是男生赵磊,右边靠外穿白色长裙的是孙悦\n请使用真人写实电影质感生成视频,画面稳定连贯,拒绝3D感、CG感、塑料感、UE5感、游戏感,保持电影光线、柔光镜、轮廓光、逆光、漫反射、空气微尘等氛围承载项中至少一项作为本镜氛围基底,场景内自然光线明亮但保持曝光层级和主体分离,主光方向明确、暗部保留、背景退后、主体或主承光面优先曝光,背景是@大学图书馆 大学图书馆(白天,阳光透过大玻璃窗洒入,一排排书架延伸,阅读区安静明亮)。\n 参考图片@图书馆人物站位 中间的是刘桂芳 左边黑裙子的是林小可,右边靠里的是男生赵磊,右边靠外穿白色长裙的是孙悦\n(全局锚点:四人围坐书桌,三个年轻人都低着头,气氛沉闷,刘桂芳坐在主位默默看着他们。)\n近景,平稳固定镜头:赵磊坐在桌边,眉头紧锁,低头盯着黑屏的手机屏幕发呆,拇指无意识地摩挲着手机边缘,手机玻璃壳反射着窗外的光晕;他苦笑了一下,眼神中透着深深的窘迫,眼皮微抬看向桌面,开口\"我这个月生活费花完了。再交不上辅导班的钱,我可能得放弃了。\"——画面内对白。\n近景,缓慢侧移跟拍:视线焦点平移到坐在旁边的孙悦身上,孙悦依然低着头,双肩微微瑟缩,双手紧紧攥着衣角,棉布衣料被揉出深浅不一的褶皱;她眼帘低垂不敢直视任何人,声音很小地开口\"我爸妈还是不同意我考文学。说再坚持就断我生活费。\"——画面内对白,不要出现字幕。 图片1是赵磊_默认,图片2是孙悦_默认,图片3是刘桂芳_居家服,音频1是刘桂芳的音色,图片4是林小可_默认,音频2是林小可的音色,图片5是图书馆人物站位,图片6是大学图书馆", "reference_images": ["references/a45f4283e8b205ac2320d3360973743b429af2ba3ce5499fad229cfb17278e0a.jpg", "references/2d1a0dc0db427d3f70142a567433fdfeec74b28cf0084684aec26f29c384b015.jpg", "references/342f48533378e95160b5edbc6933e4f03f476876d056541661fb661b7122df28.jpg", "references/16122656784ee729eca256903b8772e094dc9a9baeb28754fe876d7645575cd5.jpg", "references/f048e2e08b0563344346ba84484e68a1ade587bc2ddc1228a922ac0ef307e2f4.jpg", "references/dd7f570ad747265bf85e3ec3625632ea3bd872108f33e07d317a19d4a287e116.jpg"], "reference_sha256": ["a45f4283e8b205ac2320d3360973743b429af2ba3ce5499fad229cfb17278e0a", "2d1a0dc0db427d3f70142a567433fdfeec74b28cf0084684aec26f29c384b015", "342f48533378e95160b5edbc6933e4f03f476876d056541661fb661b7122df28", "16122656784ee729eca256903b8772e094dc9a9baeb28754fe876d7645575cd5", "f048e2e08b0563344346ba84484e68a1ade587bc2ddc1228a922ac0ef307e2f4", "dd7f570ad747265bf85e3ec3625632ea3bd872108f33e07d317a19d4a287e116"], "workflow": "ref2va", "prompt_sha256": "b174ce922228ac81112a9a88d71c274905c0ab0ee184d5785c5e570a405e8c3b", "original_prompt_sha256": "b7765a0058a410521e5208ae02e68ffa7f814d2b190bf752fba4292cc6f7d0eb", "video_sha256": "810bb991c4e0dcbd6edd0db464896c26ce922c2a2e65bc4670359ff70a1ef917", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_051.json", "verification": "verification/sample_051.json"} {"file_name": "videos/sample_052.mp4", "sample_id": "#52", "prompt": "subject_definitions:\n: A young woman whose face, hairstyle, age, gender characteristics, and overall vibe must strictly match . She wears a white short-sleeve shirt.\n\nsummary:\n[reference generation] Generate a 12-second 9:16 short-form e-commerce video featuring in an office setting promoting a cheese durian pie cookie. The video consists of 4 timed shots with synchronized on-screen Mandarin dialogue, bottom subtitles, and a unified lively BGM, maintaining strict character consistency with throughout. The finished video must look directly publishable, with unified natural visuals, stable subjects and natural continuous motion. No explanatory text, prompt wording or task descriptions may appear in the finished video. Smart packaging suited to short-form e-commerce pacing may include stickers, button prompts, keyword highlights and necessary sound effects without over-stacking. Spoken delivery is Mandarin, natural and emotionally consistent; dialogue, caption timing, sound effects and the shot visuals should match.\n\nretention_analysis:\n (from ): fully_preserved. The character's facial features, hairstyle, age, gender, and overall vibe must remain strictly consistent with across all four shots. No other images override this identity constraint.\n\ndetailed_description:\n[Shot 1] From 00:00.000 to 00:02.000, medium shot, fixed camera. An office desk scene. (S1), a young woman wearing a white short-sleeve shirt, sits at the desk. She covers her mouth with both hands, making a mysterious expression. On the desk is a packaging box of cheese durian pie cookies. The background shows office ceiling lights and a desk computer. At the bottom, a white subtitle with a black border displays the exact text “偷偷给你们看我工位藏的宝”. (S1) says [Chinese]偷偷给你们看我工位藏的宝 in a mysterious tone to arouse curiosity.\n\n[Shot 2] At 00:02.000, medium close-up and hand close-up, cut. The camera first cuts to a medium close-up of picking up the cookie box with both hands, opening it, and taking out a cookie, placing it in her palm, and showing its appearance to the camera. Then it quickly cuts to a hand close-up: her two hands hold both ends of the cookie and gently break it apart, revealing the bright yellow durian cheese filling inside with a distinct stretching effect. At the bottom, subtitles sequentially display “就是这个芝士榴莲软心饼” then “夹心超浓郁”. (S1) says [Chinese]就是这个芝士榴莲软心饼,夹心超浓郁.\n\n[Shot 3] At 00:05.000, medium close-up, fixed camera. The camera cuts back to a medium close-up of 's upper body. She takes a cookie, puts it to her mouth, and takes a bite. While chewing, she closes her eyes with an enjoying expression, then raises her hand and waves it to show strong approval of the taste. At the bottom, subtitles sequentially display “外酥内软甜而不腻” then “摸鱼吃超解压”. (S1) says [Chinese]外酥内软甜而不腻,摸鱼吃超解压.\n\n[Shot 4] From 00:09.000 to 00:12.000, medium close-up, fixed camera. holds two stacked, broken cookies showing the filling in one hand. With her other hand, she first makes a '1' gesture to emphasize the discount, then points her finger downward to guide the viewer to place an order. Her expression is excited and compelling. At the bottom, subtitles sequentially display “现在拍超划算” then “喜欢的赶紧囤啦”. (S1) says [Chinese]现在拍超划算,喜欢的赶紧囤啦.\n\noverall_soundscape:\nA unified, light, lively, and fast-paced BGM plays throughout the entire video. In Shot 2, a slight crisp sound of breaking the cookie is added. In Shot 3, a crisp sound of biting the cookie is added. No other extra sound effects are used.\n\nnon_diegetic_music:\nA light, lively, and fast-paced BGM is used throughout the video, matching the e-commerce rhythm and smartly matching the storyboard pacing.", "original_prompt": "生成一条适合短视频平台传播的成片视频,时长控制在12秒以内,画幅比例9:16。\n\n已提供参考视频,请参考原视频的整体节奏、镜头推进、运镜方式、信息密度和转场逻辑进行重制。\n如用户编辑分镜与参考视频不一致,以用户最终编辑的分镜内容为准。\n\n固定素材引用规则:\n- 视频1为固定参考视频,整体节奏、镜头推进、运镜方式、信息密度和转场逻辑都要优先参考视频1;用户额外上传的视频素材从视频2开始按顺序依次编号为视频2、视频3……,如无用户明确改动,不要偏离视频1的核心表达逻辑。\n- 图片1为固定人物形象参考图;用户额外上传的图片素材从图片2开始按顺序依次编号为图片2、图片3……,其余图片用于场景、商品、字幕版式或辅助画面参考,不可覆盖人物形象一致性要求。\n- 音频1为固定声音素材。\n\n用户设置如下:\n- 比例:9:16\n- 人物素材:使用图片1作为固定人物形象参考,人物出镜时必须保持图片1中的脸部、发型、年龄感、性别特征和整体气质一致,其余图片仅作为辅助素材参考。\n- 音频素材:使用音频1作为固定声音参考,口播音色、语气、情绪表达和声音节奏都要参考音频1,保持整条视频声音风格统一。\n- 视频情绪:智能匹配。要求:视频情绪智能匹配参考视频内容和分镜节奏,优先保证表达自然、情绪统一、贴合画面。\n- 语言口音:普通话。要求:在遵守音频1声音基准的前提下,口播措辞、停顿、语气风格尽量贴近“普通话”的表达感觉,避免口播风格偏差过大。\n- 画面包装:智能匹配。要求:可根据短视频带货节奏智能补充字幕、贴纸提示、按钮引导、重点词高亮和必要音效,但不要过度堆砌。\n- 视频风格:智能匹配。要求:画面风格智能匹配参考视频和带货内容,保证成片统一、自然、可发布。\n\n分镜执行:\n镜头一:时间0-2s;景别中景;运镜固定;画面画面为办公室工位场景,穿白色短袖的年轻女生坐在桌前,双手捂住嘴巴做出神秘的表情,桌面摆放着芝士榴莲派软馅饼干的包装盒,背景可见办公室的天花板灯光、工位电脑,底部同步显示白色带黑边的字幕“偷偷给你们看我工位藏的宝”,用神秘的情绪勾起用户好奇心,吸引用户停留。;口播偷偷给你们看我工位藏的宝;音效搭配轻快活泼的快节奏轻松BGM,无额外音效\n镜头二:时间2-5s;景别近景+手部特写;运镜切镜;画面先切近景,女生双手拿起饼干盒,打开后取出一块饼干放在手心向镜头展示饼干的外观,随后快速切到手部特写,两只手捏住饼干的两端轻轻掰开,露出内部亮黄色的榴莲芝士夹心,夹心呈现明显的拉丝效果,底部字幕依次显示“就是这个芝士榴莲软心饼”“夹心超浓郁”,直观展示产品的核心卖点。;口播就是这个芝士榴莲软心饼,夹心超浓郁;音效延续同一段轻快活泼的快节奏BGM,搭配轻微的掰饼干的酥脆音效\n镜头三:时间5-9s;景别近景;运镜固定;画面画面切回女生上半身近景,女生拿起一块饼干放到嘴边咬下一口,咀嚼时闭上眼睛露出享受的表情,随后抬起手摆了摆表示对口感非常认可,底部字幕依次显示“外酥内软甜而不腻”“摸鱼吃超解压”,用真实的试吃反应增强产品的说服力。;口播外酥内软甜而不腻,摸鱼吃超解压;音效延续同一段轻快活泼的快节奏BGM,搭配咬饼干的酥脆音效\n镜头四:时间9-12s;景别近景;运镜固定;画面女生一只手举着两块叠在一起、已经掰开露出夹心的饼干,另一只手先比出1的手势强调优惠力度,随后手指向画面下方做出引导下单的动作,表情兴奋有号召力,底部字幕依次显示“现在拍超划算”“喜欢的赶紧囤啦”,直接传递优惠信息引导用户完成转化。;口播现在拍超划算,喜欢的赶紧囤啦;音效延续同一段轻快活泼的快节奏BGM,无额外音效\n\n最终成片要求:\n1. 严格围绕用户设置和最终分镜生成,不要偏离分镜主体。\n2. 画面风格统一,人物或主体保持稳定,动作自然连续。\n3. 若存在固定人物形象或固定音频素材,必须优先遵守对应参考约束。\n4. 口播节奏、字幕节奏、音效节奏与镜头画面匹配。\n5. 整体成片要像可直接发布的短视频,不要出现说明文字、提示词文字或任务描述。", "reference_images": ["references/a34436916f244075bf084116a13ed4aba22547bd238f3bc9037fb32842b82f6c.jpg"], "reference_sha256": ["a34436916f244075bf084116a13ed4aba22547bd238f3bc9037fb32842b82f6c"], "workflow": "ref2va", "prompt_sha256": "1ce9b772c2e79adffa5670d8d8c2bbc38256051f56289aef2699eeff09dd03b8", "original_prompt_sha256": "e90b1617c649061c48c034973f45e7d6db06dc54643f6d59993792f7e93de94e", "video_sha256": "91c9c6fa88f3a1feee2d90c6497a7bdeaa97c5f4b741c4078e05d30c6bba8f7c", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_052.json", "verification": "verification/sample_052.json"} {"file_name": "videos/sample_053.mp4", "sample_id": "#53", "prompt": "subject_definitions:\nSubject 1: The model with a standard figure and fresh fairy-tale style makeup, appearance from .\nSubject 2: The clothing outfit, strictly matching the color, shape, and details from , with no alterations or color differences allowed.\nSubject 3: The forest background scene, directly using .\n\nsummary:\n[reference generation] A 12-second e-commerce marketing video for a women's clothing product detail page to promote purchase conversion. The video features a model wearing the exact clothing from the reference, walking continuously in a forest background, showcasing the garment's details through various camera angles and movements, accompanied by dreamy healing piano and string instrumental music.\n\nretention_analysis:\n is fully preserved as the model's identity and makeup reference. is fully preserved as the strict clothing reference, requiring exact color, shape, and detail matching without any text description overriding it. is fully preserved as the background reference, with lighting adjusted to serve the true color of the clothing without causing graying, darkening, or color shifting.\n\ndetailed_description:\nThe entire video must comply with e-commerce platform content standards. The entire video must comply with physical laws, strictly prohibiting clipping or the model standing on tables or chairs. The clothing must cover the body completely with a smooth wearing contour and no bulging. The background lighting must serve the true color of the clothing, strictly prohibiting the clothing from turning gray, darkening, or shifting color due to the background tone. Strictly prohibit the use of text descriptions to define the color and style of the clothing; rely entirely on the visual reference. Throughout the video, the model must maintain continuous flowing movement, strictly prohibiting hands in pockets, with no pausing or freezing. [Shot 1] [0:00.000 - 0:04.000] Full-body long shot. The model (Subject 1) lightly steps into the forest scene (Subject 3), showcasing the overall silhouette of the clothing (Subject 2). While walking, capture the airy and bouncy feel of the layered lace on the lower garment, followed by a natural turn. [Shot 2] [0:04.000 - 0:08.000] Medium shot with lateral camera translation. The model walks slowly in the forest and looks back over her shoulder from the side. Her hands hang naturally or gently touch her hair, showcasing the silky luster of the jacquard fabric on the top and the structured shape of the corset waist. [Shot 3] [0:08.000 - 0:12.000] Medium close-up with a camera push-in. Focus on the layered feel of the two-piece set. The model naturally shifts her center of gravity, showcasing the elastic tension of the knitted fabric and the drape of the lace, ending with an elegant side pose.\n\noverall_soundscape:\nUnspecified\n\nnon_diegetic_music:\nPure instrumental music, no vocals, dreamy and healing piano and strings.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。服装严格使用图二,颜色/版型/细节不允许有任何改变或色差,严禁文字描述颜色款式;背景直接使用图三。符合物理规律,禁止穿模/模特站桌椅等不合理画面;服装覆盖完整、穿着轮廓平整无凸点;符合电商平台内容规范。图一模特标准身材,清新童话感妆造。服装颜色以图二为准,背景光照服务服装真实颜色,禁止因图三色调致服装变灰/暗化/偏色。全程模特持续流动,严禁插兜,无停顿定格。 [0-4s] 全身远景,模特轻盈步入森林场景,展现图二服装整体轮廓,行走中捕捉下装层叠蕾丝的空气感与跳动感,自然转身。 [4-8s] 中景横移,模特在林间缓步并侧身回眸,双手自然垂放或轻抚发丝,展现上衣提花面料的丝滑光泽与束腰挺括版型。 [8-12s] 中近景推镜,聚焦两件套层次感,模特自然切换重心,展现针织面料的弹性张力与蕾丝垂坠感,以优雅侧身收势。 配乐:纯音乐,无人声,梦幻治愈钢琴钢琴弦乐。\\n\\n素材说明:图1=模特参考(asset-20260224202629-db8dr);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/b6a086feeefbfc1368067945dee0421d9e7aa21ed0db1e841c16e809c66f2874.jpg", "references/09ff37d8da769e1e63c49b3f0a8460f757eef9384c8c73bf0491461b911d4ae7.jpg", "references/9f34552a9d4053e067e410ddd96ff6a03f06b5057d0465f14e6fa1ae32d57be4.jpg"], "reference_sha256": ["b6a086feeefbfc1368067945dee0421d9e7aa21ed0db1e841c16e809c66f2874", "09ff37d8da769e1e63c49b3f0a8460f757eef9384c8c73bf0491461b911d4ae7", "9f34552a9d4053e067e410ddd96ff6a03f06b5057d0465f14e6fa1ae32d57be4"], "workflow": "ref2va", "prompt_sha256": "7410def541926b75c855e5092e1c82906428b98b35fea633de04bb42de6c0e12", "original_prompt_sha256": "0b79473751160ee6da233b45c56997c9b5cff2f648a1210106df966c5b9aff66", "video_sha256": "270fa23657fab06a8d25d162d94522a1afc00a3933efb3601c8d71c2c90a6bfd", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_053.json", "verification": "verification/sample_053.json"} {"file_name": "videos/sample_054.mp4", "sample_id": "#54", "prompt": "integrated_multimodal_description:\n[Shot 1] 0.0-3.6s. Transition: Dissolve. Medium shot, eye-level angle, static camera. Daytime. Inside a quiet Maybach, parked in a roadside parking space outside the court. Through the front windshield, the flashing red and blue police lights at the alley entrance are visible in the distance, reflecting onto the dark interior. I sit in the car, quietly looking far away at the police lights, watching the net closing in the distance. I am in a state of calm summary after complete detachment, emotions no longer looking back, only retaining conclusions and self-rebuilding. I only confirm the result, maintaining a distance between myself and the event, no longer participating in any emotions. No interaction with others.\n\noverall_soundscape:\nFading siren sounds heard through the car windows, creating a low-frequency, quiet ambient environment inside the car.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【场景分析】\n场景:(法院外路边停车位)\n承接:上一分镜结束于巷口警灯亮起、Ryker Thorne被当场控制。\n过渡:从高压现场切回车内静观,以稳定构图完成叙事收束与命运裁定。\n人物表演底层逻辑总纲:【我处于彻底抽离后的平静总结状态,情感不再回头,只保留结论与自我重建】\n【分镜承接】\n前一分镜末尾画面:警察在巷口控制Ryker Thorne,警灯映亮墙面\n本分镜开场画面:车内视角透过前挡风玻璃远望闪烁的警灯\n承接逻辑:空间回切,视觉元素延续为同一方向的巷口警灯\n**【镜头1】0.0-3.6秒\n转场:叠化\n镜头: 中景/平视/固定镜头\n画面描述:白天,车内安静,前挡风玻璃外可见巷口闪烁的红蓝光,映在深色内饰上。\n核心动作: 我安静看着远处收网。\n对手戏设计: 无,人与事件保持距离。\n人物底层逻辑: 她只确认结果,不再参与任何情绪。\n主体动作: 我坐在车里远望警灯\n音效: 隔着车窗的警笛余音,车内低频安静环境\n资产对照表(视频中涉及的角色、场景与物品如下(保持一致性)):\n视频场景对照表: 法院外路边停车位=【图1】\n视频角色对照表: 我=【图2】\n视频道具对照表:迈巴赫=【图3】", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "c46729817990f096308eccb89b22ae41a8459bd4198146627b6196ac2fb3ad61", "original_prompt_sha256": "bbd5e661504ce65b657b8c98ee77bb3d4ad596fe5ed824def045b5f6b9fed7e2", "video_sha256": "c4da70e52e2a942633de1b9a7b0cb20001fafd9a1bf9c024bd3e216489fb2ade", "task_adapted": true, "width": 768, "height": 1024, "fps": 24, "num_frames": 107, "duration_seconds": 4.458333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_054.json", "verification": "verification/sample_054.json"} {"file_name": "videos/sample_055.mp4", "sample_id": "#55", "prompt": "integrated_multimodal_description:\n[Shot 1] First-person perspective office desk scene. The left and right mechanical arms of the Unitree G1 robot serve as the first-person left and right hands, not as additional mechanical arms placed on the desk. Subsequent actions are completed in sequence: First, the first-person perspective moves to the front right, walking to the front right of the desk. Second, the G1 mechanical arms close the laptop on the desk, and consequently, the desktop monitor connected to the laptop immediately goes black. Third, the camera perspective turns to the left, and the G1 mechanical arms push away the chair on the left seat. Fourth, the G1 mechanical arms pick up the large mineral water bottle printed with the visible text \"喝开水\" on the desk, bringing it close to the first-person perspective to take a sip. Fifth, finally, the camera perspective turns towards itself, and a single G1 mechanical hand makes a heart-shaped gesture towards the camera. Throughout the entire process, maintain first-person perspective continuity with natural camera movement. The positions of the desk, chair, laptop, desktop monitor, and mineral water bottle must remain stable and consistent. Do not let objects flicker, drift, deform, or suddenly disappear; the monitor going black must occur after the laptop is closed.\n\noverall_soundscape:\nUnspecified.\n\nnon_diegetic_music:\nN/A", "original_prompt": "使用视频1作为第一视角办公桌场景、相机初始位置、相机距离和画面构图参考。图片1用于参考宇树机器人 G1 机械臂的外观、材质、比例和运动风格。首先编辑视频1的首帧:把首帧里已经出现的一双人手替换成宇树机器人 G1 的左右双机械臂。这是对原双手像素区域的外观替换,不是在桌面上额外放置机械臂;原双手在哪里,G1 双机械臂就在哪里,大小、朝向、可见范围和后续运动轨迹都跟随原双手。基于这个编辑后的首帧继续向后生成完整视频。后续动作按顺序完成:第一,第一视角向右前方移动,走到桌面右前方;第二,G1 机械臂合上桌面上的笔记本电脑,因此与笔记本连接的桌面显示器随即黑屏;第三,相机视角转向左侧,G1 机械臂推开左边座位的椅子;第四,G1 机械臂拿起桌面上印有“喝开水”的大矿泉水瓶,靠近第一视角喝一口;第五,最后相机视角转向自己,单只 G1 机械手对着相机比一个心形手势。全程保持第一视角连贯,相机运动自然,桌面、椅子、笔记本电脑、桌面显示器、矿泉水瓶的位置稳定一致。不要让物体闪烁、漂移、变形或突然消失;显示器黑屏必须发生在笔记本合上之后。", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "f55f415ff32d6ab19e0c9061dd4de2da1d7eed81a1469ed9612c6459231f5829", "original_prompt_sha256": "bfc00e3eca8b4d7fb26f0ccf522b4264ae6822e79c63495e76c7e5abed35fb13", "video_sha256": "0f042e297b57d4a68c66997e45bb6183f01abeff6a96736294ba8c10da0743c7", "task_adapted": true, "width": 1344, "height": 768, "fps": 24, "num_frames": 328, "duration_seconds": 13.666666666666666, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_055.json", "verification": "verification/sample_055.json"} {"file_name": "videos/sample_056.mp4", "sample_id": "#56", "prompt": "subject_definitions:\nN/A\n\nsummary:\n[reference generation] A 15-second sequence depicting a tense morning confrontation in a traditional pavilion lobby, featuring a woman descending the stairs to face an official and his armed guards, culminating in a sharp verbal exchange.\n\nretention_analysis:\nfully_preserved: All 7 supplied images are explicitly bound to specific characters and the scene location in the source prompt. The visual identities of Lv Hen, Qin Guangyi, the yamen runners, the shop assistant, Li Yuan, and Chun Tao, along with the lobby setting, must be strictly retained from their respective reference pictures throughout the 5 shots.\n\ndetailed_description:\nTotal duration: 15 seconds. All shots use a fixed camera or slight handheld shake (simulating a documentary breathing effect), with no smooth gimbal movement. Natural blinking is required ( approx. 1 time per 1.5s, approx. 1 time per 1.5s). Lighting: Early morning interior, cold white scattered light filtering from high windows, no artificial fill light, strictly no warm yellow filters. Overall cold gray tone, shadows leaning towards cyan-gray, highlights cold white. Hard cuts only, no dissolves, no subtitles, no background music.\n\nScene: Early morning, first-floor lobby of Lv Jin Pavilion (from ). The left side features a wooden staircase leading to the second floor with carved handrails. The door faces the street.\n\n[Shot 1] (2 seconds) Extreme close-up of (appearance from )'s footsteps, stepping down the stairs layer by layer. Fixed camera, low angle. The camera is placed on the side of the stairs, height level with the soles, filming 's feet stepping down. 's skirt hem slightly fluttering, footsteps steady, stepping down the wooden stairs one by one. The wooden boards have slight wear marks. Cold white morning light shines diagonally from the high window onto the shoes and steps, showing a matte texture. Hard cut when the foot lands firmly on the second step.\n\n[Shot 2] At 00:02.000 (2 seconds) Close-up of (from ) walking down the stairs (waist up). Fixed camera. The camera is placed on the side of the stairs, height level with 's chest. is walking down the last few steps, looking straight ahead. wears an elegant and simple long dress, calm expression, unhurried pace. Morning light from the door hits her side profile, cold white. Her gaze is steady, but the fingertips of her hands hanging by her sides are slightly tense. Facial micro-expressions: Brow relaxed, mouth straight, eyes calm but alert. Hard cut when finishes the last step.\n\n[Shot 3] At 00:04.000 (2 seconds) Medium shot, full view of the door (centered (from ), yamen runners on both sides, (from )'s reaction). Fixed camera. The camera is placed inside the lobby, facing the door, eye-level. stands in the center of the door, with a line of (from ) (holding long sticks) on both sides behind him. leans against the side of the door, hand on the door panel, body slightly shrinking. Bright swords and weapons reflect matte cold light in the morning light. 's face is sinister, eyes scanning the lobby. is scared stiff, eyes wide, lips slightly parted. Overall cold gray tone, no warm colors. Facial micro-expressions: brows furrowed, mouth corners turned down; eyelids wide open, pupils constricted. Hard cut the moment 's gaze sweeps in.\n\n[Shot 4] At 00:06.000 (2 seconds) Medium shot ( (from )'s upper body and threshold area). Fixed camera. The camera is placed slightly left inside the lobby, chest-level, filming his walking in. steps over the threshold into the lobby, facing . takes a big step over the threshold, the hem of his dark purple official robe slightly fluttering, boots stepping on the blue brick floor with a dull thud. His face is sinister, gaze locking straight ahead (towards ). The background shows partial views of (from ) on both sides. Facial micro-expressions: Brows furrowed, eyelids half-closed, mouth corners pulled down. Hard cut the moment stands still.\n\n[Shot 5] At 00:08.000 (7 seconds) Medium shot, over-the-shoulder reverse shot (blurred (from )'s right shoulder and back of head in foreground, (from ) in center, blurred (from ) and (from ) behind ). Fixed camera, focus on 's face. The camera is placed behind 's right shoulder, using his right shoulder and back of head as a 40% blurred foreground, placing in the center-left of the frame, with and blurred in the background.\nPositioning: stands in front of the stairs, facing the door; stands inside the door, facing ; and stand about 2 meters behind , bodies slightly stiff.\n0-1s: sees the bright weapons, fingertips suddenly clenching the cuff tightly (right hand fingers clenching tightly, fabric wrinkles). She suppresses the shock in her heart, face remaining calm as water.\n1-7s: She speaks in a cold tone, moderate pace, every word clear: (S1) says [Chinese]秦大人,大清早带兵围我铺子,不知我是犯了哪条国法? (The whole line is about 5s, shot gives 7s to ensure it's finished with emotional resonance). While speaking, she looks straight at (blurred direction on the left), brow relaxed but eyes sharp, mouth straight. Morning light from the door illuminates her side profile, cold white, real skin texture visible.\nSimultaneously, blurred background and : brows furrowed, body slightly leaning forward (vigilant); hands wringing together, lips pressed tight (nervous).\nBlurred foreground : mouth corners slightly moving, not speaking.\nFacial micro-expressions: : knuckles turning white when clenching cuff; lower lip tucked in when speaking, throat not moving (calm with pressure). (blurred): brows furrowed. (blurred): lower lip tucked in.\nHard cut after the line is fully spoken, ending in the lingering look of .\n\noverall_soundscape:\nStairs creaking, footsteps, skirt rustling, metal clinking of sticks, armor rustling, boots thudding, clothing rustling, and speaking voices. No background music.\n\nnon_diegetic_music:\nN/A", "original_prompt": "总时长:15秒镜头顺序:\n分镜1:极特写绿痕脚步,一层层下楼梯(2秒) → 分镜2:近景绿痕走下楼梯(2秒) → 分镜3:中景门口全景(秦广义居中,两侧衙役,伙计甲反应)(2秒) → 分镜4:中景秦广义走进来,面向绿痕(2秒) → 分镜5:外反打(秦广义右肩后虚化前景,画面中央绿痕,绿痕身后李元和春桃虚化),绿痕说完台词(7秒)\n拍摄提示:所有镜头采用固定机位或轻微手持晃动(模拟纪录片呼吸感),无稳定器丝滑运动。自然眨眼(绿痕约1.5秒1次,秦广义约1.5秒1次)。光影环境:清晨室内,冷白散射光从高窗透入,无人工补光,杜绝暖黄滤镜。整体清冷灰调,暗部偏青灰,亮部冷白。无BGM,仅有环境音、脚步声、衣料摩擦声、说话声。硬切转场,无叠化。绿痕台词完整,语速适中,镜头给足时长。场景与站位(贯穿全镜)\n时间:清晨。\n场景:绿锦阁一楼大堂图1。左侧木质楼梯通往二楼,楼梯扶手雕花。门口朝街。\n人物站位:\n楼梯上:绿痕图2正从二楼缓步走下。\n门口正中:秦广义图3站立,身后两侧各有一列衙役图4(手持长棍)。\n门口侧边:伙计甲图5身体微微颤抖,面露惊恐望向秦广义。\n大堂内(绿痕身后):李元图6、春桃图7分镜1(2秒):极特写绿痕脚步,一层层下楼梯\n分镜编号:1\n秒数:2秒\n景别:极特写\n运镜:定点固定,低角度\n镜头位置:摄影机置于楼梯台阶侧面,高度与鞋底平齐,拍摄绿痕的脚一层层踏下。\n出场人物:绿痕图2\n画面描述:绿痕裙摆微扬,脚步沉稳,一层层踏下木质楼梯。楼梯木板有细微磨损痕迹,冷白晨光从高窗斜射,照在鞋面和台阶上,呈哑光质感。\n音效:楼梯“吱呀”声,脚步“哒、哒”声。\n备注:无字幕,无音乐。画面在脚步落定第二级台阶时硬切收束。分镜2(2秒):近景绿痕走下楼梯\n分镜编号:2\n秒数:2秒\n景别:近景(绿痕腰部以上)\n运镜:定点固定\n镜头位置:摄影机置于楼梯侧面,高度与绿痕胸部平齐,拍摄她走下楼梯的过程。\n出场人物:绿痕图2\n人物站位:绿痕正从最后几级楼梯走下,目光平视前方。\n画面描述:绿痕一身素雅长裙,神色从容,步履不疾不徐。晨光从门口透入,打在她侧脸,呈冷白。她的目光沉稳,但垂在身侧的手指尖微微绷紧。\n面部微表情:眉心舒展,嘴角平直,眼神平静中带着警觉。\n音效:脚步落地的轻响,裙摆摩擦声。\n备注:无字幕,无音乐。画面在绿痕下完最后一级台阶时硬切收束。分镜3(2秒):中景门口全景(秦广义居中,两侧衙役,伙计甲反应)\n分镜编号:3\n秒数:2秒\n景别:中景(绿锦阁门口全景)\n运镜:定点固定\n镜头位置:摄影机置于大堂内,正对门口,高度与眼睛平齐。\n出场人物:秦广义图3、衙役图4若干、伙计甲图5\n人物站位:秦广义站立门口正中,身后两侧各列一队衙役,刀棍在手。伙计甲靠在门边,手扶门板,身体微缩。\n画面描述:明晃晃的刀兵在冷白晨光中泛着哑光寒光。秦广义面色阴鸷,目光扫向大堂内。伙计甲吓得不敢动弹,眼睛瞪大,嘴唇微张。整体清冷灰调,无暖色。\n面部微表情:秦广义眉头紧锁,嘴角下撇;伙计甲眼睑瞪大,瞳孔收缩。\n音效:刀棍轻微的金属碰撞声,衣甲摩擦声。\n备注:无字幕,无音乐。画面在秦广义目光扫入的瞬间硬切收束。分镜4(2秒):中景秦广义走进来,面向绿痕\n分镜编号:4\n秒数:2秒\n景别:中景(秦广义上半身及门槛区域)\n运镜:定点固定\n镜头位置:摄影机置于大堂内偏左,高度与秦广义胸部平齐,拍摄他走进来的动作。\n出场人物:秦广义图3\n人物站位:秦广义一步跨过门槛,进入大堂,面向绿痕。\n画面描述:秦广义大步跨入门槛,暗紫色官服下摆微扬,靴子踏在青砖地面上发出闷响。他面色阴鸷,目光直直锁向前方(绿痕所在方向)。背景中可见两侧衙役的局部。\n面部微表情:眉头紧锁,眼睑半垂,嘴角下拉。\n音效:靴子落地“咚”,衣料摩擦声。\n备注:无字幕,无音乐。画面在秦广义站定的瞬间硬切收束。分镜5(7秒):外反打(秦广义右肩后虚化,画面中央绿痕,绿痕身后李元和春桃虚化),绿痕说完整台词\n分镜编号:5\n秒数:7秒\n景别:中景(绿痕上半身及身后李元、春桃虚化,前景秦广义右肩及后脑虚化)\n运镜:定点固定,焦点在绿痕面部\n镜头位置:摄影机置于秦广义右肩后方(过肩视角),以他的右肩及后脑为前景虚化40%,使绿痕位于画面中央偏左,身后李元和春桃在背景虚化。\n出场人物:绿痕图2、秦广义图3(虚化前景)、李元图6(虚化背景)、春桃图7(虚化背景)\n人物站位:绿痕站在楼梯前,面向门口;秦广义站在门口内,面向绿痕;李元和春桃站在绿痕身后约两米处,身体微僵。\n画面描述:\n0-1秒:绿痕看到明晃晃的刀兵,指尖猛地攥紧了袖口(右手手指收紧,布料褶皱)。她强压下心头的惊骇,面上依旧平静如水。\n1-7秒:她语气清冷,语速适中,字字清晰:“秦大人,大清早带兵围我铺子,不知我是犯了哪条国法?”(整句台词约5秒,镜头给足7秒,保证说完且有情绪余韵)。\n说话时,她目光直视秦广义(画面左侧虚化方向),眉心舒展但眼神锐利,嘴角平直。晨光从门口透入,照亮她的侧脸,呈冷白,皮肤真实纹理可见。\n同时,虚化背景中李元和春桃:李元眉头紧锁,身体微微前倾(戒备);春桃双手绞在一起,嘴唇抿紧(紧张)。\n秦广义虚化前景:嘴角微动,未出声。\n面部微表情:\n绿痕:指尖攥袖口时指节泛白;说话时下唇内收,喉结未动(平静中带压迫)。\n李元(虚化):眉头紧锁。\n春桃(虚化):下唇内收。\n台词:绿痕(语气清冷):“秦大人,大清早带兵围我铺子,不知我是犯了哪条国法?”\n音效:说话声(清冷),衣料摩擦声(攥袖口),背景中极轻的刀棍碰撞声。\n备注:无字幕,无音乐。台词完整说完后,画面在绿痕视线的余韵中硬切收束,全片结束。\n总时长:2+2+2+2+7=15秒,严格遵守清冷灰调、自然散射光、硬切、无BGM、无暖黄滤镜。绿痕说话镜头7秒,语速适中,台词完整。每镜落幅硬切收束。", "reference_images": ["references/d5e4b642cef460180e2dda3c91b6b00aaacd2d06e59ffc4f93fefe8f31904e55.jpg", "references/82c070b32b6a6d9f88437c4754170f8df2f73080b56ed3f3a756386d87b24c13.jpg", "references/42fecc912a4ab3382894f01203f6de86609d7629f46ad77e2fb8daf7021fd295.jpg", "references/31284faaf4cabe0440e35ae6ae5f5dae2191e6f0af8e83109b8efde850814a1b.jpg", "references/3bc9342c0173d0df962126ae8d8435688155f3c31e0f70906f4430eca44e2bbd.jpg", "references/fcc44bc4f50640843d89ee5a46271363576d355894ceb780d5f32a97a3fc3253.jpg", "references/405bf82e63a9c5bc252e4b81d907d995f097936e97323316cc18a6d7f273e994.jpg"], "reference_sha256": ["d5e4b642cef460180e2dda3c91b6b00aaacd2d06e59ffc4f93fefe8f31904e55", "82c070b32b6a6d9f88437c4754170f8df2f73080b56ed3f3a756386d87b24c13", "42fecc912a4ab3382894f01203f6de86609d7629f46ad77e2fb8daf7021fd295", "31284faaf4cabe0440e35ae6ae5f5dae2191e6f0af8e83109b8efde850814a1b", "3bc9342c0173d0df962126ae8d8435688155f3c31e0f70906f4430eca44e2bbd", "fcc44bc4f50640843d89ee5a46271363576d355894ceb780d5f32a97a3fc3253", "405bf82e63a9c5bc252e4b81d907d995f097936e97323316cc18a6d7f273e994"], "workflow": "ref2va", "prompt_sha256": "694706160df0d3bdd6ef9e2646413910309076e92d9ebbfa2bd485f3fdaaf9d3", "original_prompt_sha256": "e96d218d4c38f5457d23476f9fe1518ed7348f88141f5a9a37b5b5a0c9ba668d", "video_sha256": "6f6c79c192afcd1195233fc93f0ec2b6b5bda661a23d0cb1cf6d1c7b9f44103d", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_056.json", "verification": "verification/sample_056.json"} {"file_name": "videos/sample_057.mp4", "sample_id": "#57", "prompt": "subject_definitions:\n is Li Yanhong, a woman whose appearance and clothing match . is Wang Laoshuan, a man whose appearance and clothing match . The courtyard wall exterior night scene matches .\n\nsummary:\n[reference generation] A tense night scene outside a courtyard wall. Wang Laoshuan, furious, confronts and aggressively grabs Li Yanhong's arm, accusing her. He drags her toward the courtyard entrance while she struggles and protests her innocence. The scene conveys rough dragging and panic.\n\nretention_analysis:\nfully_preserved: (Li Yanhong) and (Wang Laoshuan) are explicitly referenced by the source for character identity and appearance. is referenced for the courtyard wall exterior night scene setting. All three references are actively used in the scene description and dialogue assignments.\n\ndetailed_description:\n[Shot 1] In the courtyard wall exterior night scene matching , Wang Laoshuan can no longer contain his anger. Wang Laoshuan , whose appearance matches , suddenly steps forward and grabs Li Yanhong 's arm, whose appearance matches . His fingers tighten like iron clamps. Wang Laoshuan (S1) says [Chinese]倒泔水能倒到墙外男人跟前去? Wang Laoshuan grits his teeth, neck veins bulging, and harshly adds: Wang Laoshuan (S1) says [Chinese]我刚才那双眼还没瞎!\n[Shot 2] Li Yanhong is pulled off balance, her heels slipping half a step on the dirt ground. The swill bucket by the wall flips over with a clang. Panicking, Li Yanhong raises her hands to pry at Wang Laoshuan's grip. Li Yanhong (S2) says [Chinese]我真没跟他咋样!\n[Shot 3] Wang Laoshuan ignores her, dragging Li Yanhong toward the courtyard entrance. His steps are urgent and heavy, as if afraid she will run if he lets go. Wang Laoshuan (S1) says [Chinese]进屋说!我今儿倒要听听,你这回还咋编! The scene conveys rough dragging and panic. No subtitles or any on-screen text overlays.\n\noverall_soundscape:\nSound of a swill bucket flipping over and falling, sound of shoe soles urgently dragging on dirt.\n\nnon_diegetic_music:\nN/A", "original_prompt": "参考image_3 <<>> 的院墙外夜景里,王老栓这口火已经压不住了。 \n镜头一:参考image_2 <<>> 的王老栓猛地上前,一把攥住参考image_1 <<>> 的李艳红胳膊,手指像铁钳子一样越收越紧:{倒泔水能倒到墙外男人跟前去?} 王老栓咬着牙,脖筋都绷起来,又狠狠干脆脆砸下一句:{我刚才那双眼还没瞎!} \n镜头二:李艳红被拽得一个趔趄,脚后跟先在土地上打滑半步,墙根边的泔水桶跟着“咣当”一下翻倒;李艳红慌得连气都散了,只顾抬手去掰王老栓的手:{我真没跟他咋样!} 使用audio_1 <<>> 的音色作为李艳红说话声。 \n镜头三:王老栓根本不听,拽着李艳红就往院里拖,脚下又急又重,像生怕一松手人就跑了:{进屋说!我今儿倒要听听,你这回还咋编!} rough dragging panic,<泔水桶翻倒声><鞋底急拖土声>,no music,no subtitles。", "reference_images": ["references/55927ff2829ccf442bd124478ef54591e492596babca062eaf6edc42d9588798.jpg", "references/166f0d62a66f1f548590a6971e0acade0517bb7301b656b79bd00501130e26fd.jpg", "references/4732fbc41f3264a2ff0ce8d4ede7c168d6d5bb21c23b5c599293d28849de1d25.jpg"], "reference_sha256": ["55927ff2829ccf442bd124478ef54591e492596babca062eaf6edc42d9588798", "166f0d62a66f1f548590a6971e0acade0517bb7301b656b79bd00501130e26fd", "4732fbc41f3264a2ff0ce8d4ede7c168d6d5bb21c23b5c599293d28849de1d25"], "workflow": "ref2va", "prompt_sha256": "b978d462054d06cb58e3403e98bc12178d67fcf588ad371957ed49d5595add84", "original_prompt_sha256": "8c5ce4ae24e15b5ae0e43e428fdf2b916cf9000caec646f77dd15132ee755a5a", "video_sha256": "9a5c4b01e7a7c8d3dc1994e6a72d360dd291668598aca71a5f803377c77381a3", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_057.json", "verification": "verification/sample_057.json"} {"file_name": "videos/sample_059.mp4", "sample_id": "#59", "prompt": "subject_definitions:\n is a 3-month pregnant woman, whose appearance and clothing are derived from .\n\nsummary:\n[reference generation] A continuous long take with a fixed camera captures a 3-month pregnant woman performing a sequence of lively and playful actions around a living room window, from walking to the window, opening it, breathing fresh air, closing it, and walking back to her starting position.\n\nretention_analysis:\nThe subject's identity as a 3-month pregnant woman is fully preserved through linked to . The specific actions, expressions, and camera constraints (fixed camera, one continuous shot) are fully preserved. is included in the reference bindings but its specific role is unspecified in the source prompt.\n\ndetailed_description:\n[Shot 1] 0:00.000 - 0:03.000: The camera remains fixed throughout this one continuous shot. , a 3-month pregnant woman, takes light, playful mincing steps with an agile and funny posture. She sidesteps lightly from the right path, bypassing the living room sofa to walk to the window. Her walking posture is lively and cute, her face showing a lazy and pleasant expression, with light and agile small body movements. She walks to the indoor window position, stands firmly, leans her body slightly forward, her eyes full of expectation, and her expression lively and playful, getting ready to push the window open.\n[Shot 2] At 0:03.000 - 0:05.000: places both hands on the window and pushes it outward lightly and quickly. Her window-pushing action is exaggerated, cute, neat, and smooth, with a playful movement amplitude full of cartoon fun. Her face reveals a joyful expression full of expectation.\n[Shot 3] At 0:05.000 - 0:10.000: leans close to the window, slightly tilting her head up to take deep breaths of the fresh air outside, her expression relaxed and comfortable. Her gaze looks at the natural scenery outside, then her eyebrows and eyes relax, revealing a sweet expression of comfort, satisfaction, relaxation, and peace of mind, creating a strong atmosphere.\n[Shot 4] At 0:10.000 - 0:12.000: Feeling soothed and pleasant, uses her left hand to pull the left window and her right hand to pull the right window, slowly closing them well. Her window-closing action is gentle and smooth, her expression leisurely and relaxed, and her movements light and quick without dragging.\n[Shot 5] At 0:12.000 - 0:15.000: again takes light and flexible steps, returning to her initial standing position along the original route from the right path, bypassing the living room sofa. She adjusts her body shape, straightens her posture, and returns to her initial pose.\n\noverall_soundscape:\nUnspecified.\n\nnon_diegetic_music:\nN/A", "original_prompt": "全程固定镜头,一镜到底。0—3秒 3月孕妈迈着轻快俏皮的小碎步,身姿灵动搞笑,侧身轻巧从右边的道路绕过客厅沙发走到窗前,行走姿态活泼可爱,面部带着慵懒愉悦的神情,肢体小动作轻盈灵动,孕妈走到室内窗边位置,稳稳站定身体,身体微微前倾,眼神充满期待,表情灵动俏皮,做好推窗准备。3—5秒 3月孕妈双手放在窗户上,把窗户向外面轻快推开,孕妈推窗动作夸张可爱、利落流畅,动作幅度俏皮充满卡通趣味,面部露出满怀期待的欢快神情。5—10 秒孕妈凑近窗边,微微仰头大口呼吸窗外新鲜空气,神情舒展放松;目光望向窗外自然风光,随即眉眼舒展,浮现惬意满足、轻松舒心的甜美神态,氛围感十足。10—12 秒 3月孕妈心情舒缓愉悦,左手拉着左边窗户,右手拉着右边窗户,把窗户缓缓关好,关窗动作轻柔流畅,神情悠然闲适,动作轻快不拖沓。12—15 秒 3月孕妈再次迈着轻快灵活的步伐,原路从右边道路绕过客厅沙发走回最初站立位置,调整身形摆正姿态,恢复初始姿势。", "reference_images": ["references/879bddee03744223e99ef3cb50e317ef96e7177649a9b46abd19669c157b1a1a.jpg", "references/879bddee03744223e99ef3cb50e317ef96e7177649a9b46abd19669c157b1a1a.jpg"], "reference_sha256": ["879bddee03744223e99ef3cb50e317ef96e7177649a9b46abd19669c157b1a1a", "879bddee03744223e99ef3cb50e317ef96e7177649a9b46abd19669c157b1a1a"], "workflow": "ref2va", "prompt_sha256": "efbbde51a2efbc390d604c9609e0f69bdecd7931f07c92e0d574727b71dd392d", "original_prompt_sha256": "25a39a98d4471e79934b89e5e52cff2b8d887d34c11d665076a5a7a99393ea34", "video_sha256": "37df200d9ff96026ca504748f842c653710fcb0ee750095338864d1a3b387efa", "task_adapted": false, "width": 1344, "height": 768, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_059.json", "verification": "verification/sample_059.json"} {"file_name": "videos/sample_060.mp4", "sample_id": "#60", "prompt": "subject_definitions:\nSubject 1 is the model, whose appearance and physical features are referenced from . She has a plus-size hourglass figure with full, round hips, a natural waist curve, a height of 165-168cm, clean skin, French lazy slightly curly hair, and vivid eyes. Subject 2 is the clothing, strictly referenced from , with no alterations allowed to its color, style, details, or color difference. Subject 3 is the background, directly referenced from .\n\nsummary:\n[reference generation] A women's clothing e-commerce product detail page marketing video designed to promote purchase conversion. The video strictly adheres to basic physical laws, avoiding any clipping or unrealistic scenes such as the model standing on tables or chairs. The model wears the exact clothing from against the background from , showcasing the garments through a three-shot sequence with specific camera movements and natural details. Target 9:16 aspect ratio, 12s duration, 720p resolution; note actual generation is 768x1024 (3:4), a conflict to resolve.\n\nretention_analysis:\nfully_preserved for Subject 1 (model appearance and physical features from ), fully_preserved for Subject 2 (exact clothing from with no alterations), and fully_preserved for Subject 3 (exact background from ).\n\ndetailed_description:\n[Shot 1] (0-4s) Full shot. The model (S1) confidently walks in from a vintage door. The camera follows her movement, utilizing her walking motion to showcase the thick and crisp texture of the knitted fabric. Oblique sunlight outlines her body contour.\n[Shot 2] (4-8s) Medium shot. The model (S1) stands sideways to display the visual slimming effect of the stripes and the retro silhouette of the micro-puff sleeves. She performs a natural hair-adjusting motion to demonstrate the slight elasticity of the fabric in a dynamic state. Her hands are strictly not in her pockets.\n[Shot 3] (8-12s) Medium close-up. The camera pushes in to focus on her upper body, highlighting the color-blocking details of the Polo collar and the relief texture of the fabric. The model (S1) gently looks back. The background light and shadow naturally interweave, creating a cinematic visual quality.\nNatural details: A gentle breeze blows her hair, and the fabric texture exhibits a realistic gravitational drape and rebound with her movements.\n\noverall_soundscape:\nUpbeat, elegant retro jazz pop pure instrumental music with no vocals.\n\nnon_diegetic_music:\nUpbeat, elegant retro jazz pop pure instrumental music.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。符合基本物理规律,禁止出现穿模或模特站在桌椅上等不合理画面。模特为大码沙漏身材,臀部丰满圆润,腰部有自然曲线,身高165-168cm,皮肤洁净,发型为法式慵懒微卷,眼神生动。服装严格使用图二,颜色、版型、细节不允许任何改变或色差;背景直接使用图三。 分镜设计: 1.(0-4s)全景。模特从复古大门前自信步入,运镜跟随,利用行走动作展现针织面料的厚实挺括感,斜射阳光勾勒身材轮廓。 2.(4-8s)中景。模特侧身站立展示条纹视觉收缩效果及微泡泡袖的复古廓形,做自然理发动作,展示面料在动态下的微弹力,严禁双手插兜。 3.(8-12s)中近景。镜头推进聚焦上半身,重点呈现Polo翻领撞色细节与面料凹凸肌理,模特温婉回眸,背景光影自然交织,营造电影感画质。 自然细节:微风拂动发丝,面料质感随动作呈现真实的重力垂感与回弹。配乐:纯音乐,无人声,节奏明快、优雅复古的爵士流行乐。\\n\\n素材说明:图1=模特参考(asset-20260330154928-zjzqh);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/2280c4f845a8065dc10ea8924ca11cb8ffd6e3af81866abf7eca7fb14b038ed2.png", "references/acc07cd6be9a78a4ca6bb41f3a817f48c4f62b951f9454bd3a90788061ea406f.jpg", "references/5456500def3e422af5c0802d851810a87f40377ef6efedfc5b9f595aea3594f5.jpg"], "reference_sha256": ["2280c4f845a8065dc10ea8924ca11cb8ffd6e3af81866abf7eca7fb14b038ed2", "acc07cd6be9a78a4ca6bb41f3a817f48c4f62b951f9454bd3a90788061ea406f", "5456500def3e422af5c0802d851810a87f40377ef6efedfc5b9f595aea3594f5"], "workflow": "ref2va", "prompt_sha256": "fb5d03262bd64ce1669e957517e3e29f0711563cb93c65401aa5c40edc0e9b51", "original_prompt_sha256": "b40db28dc9101626aa1eb0434c37481a04d2f96042f1d5df58fe0b0e14a409b1", "video_sha256": "97edd10510fc27fae663c6ddd731ef98ce3d3ce683092cacb17dbf9e5d3aa5e4", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_060.json", "verification": "verification/sample_060.json"} {"file_name": "videos/sample_061.mp4", "sample_id": "#61", "prompt": "subject_definitions:\n: Lin Zhi, appearance from .\n: Helper A, wearing coarse cloth short clothes.\n: Helper B, wearing coarse cloth short clothes.\n: Several Helpers, wearing coarse cloth short clothes.\n\nsummary:\n[reference generation] The scene takes place at the riverbank of the Da Yan Dynasty, matching . Lin Zhi, matching , is standing in a daze looking at the distant city wall. Suddenly, helpers burst out from the reeds behind him, surrounding and interrogating him, suspecting him of being a spy. Lin Zhi panics and denies the accusation.\n\nretention_analysis:\nThe scene location is retained from . Lin Zhi's identity is retained from . The props, including the thick rope from and the carrying pole from , are retained. The helpers' identities are referenced via text aliases without explicit image bindings, which requires clarification regarding their visual consistency.\n\ndetailed_description:\n[Shot 1] 0.0-2.2 seconds. Medium shot, eye-level, fast lateral tracking. The reed marsh at the riverbank (from ) behind Lin Zhi (from ) is suddenly pushed apart, and several helpers rush out from behind, their carrying poles (from ) and thick ropes (from ) swaying in the sunlight. The group of helpers suddenly closes in to form an encirclement. The helpers press in from behind and the rear side of Lin Zhi, who is positioned in the foreground near the center of the frame, forming the center of the encirclement. The group takes the initiative, establishing control through numbers and tools. Several helpers quickly rush out of the reeds, their shoe soles crushing the mud surface, carrying poles raised, thick ropes flung open, rapidly closing the distance to Lin Zhi. Lighting: film photography texture, cinematic lighting, semi-transparent texture, matte premium texture, dynamic soft light projection, hazy glowing layer, soft side backlight, ultimate hair light, background darkened and blurred.\n\n[Shot 2] At 2.2 seconds, hard cut. Close-up, eye-level, handheld slight shake. Helper A stands at the very front, brows furrowed, shoulders pressing forward, carrying pole slanted in front of his body, expression fierce. Helper A stares directly at Lin Zhi, gaze completely locked, while the helpers behind form a blurred background pressure. (S1) steps forward half a step, carrying pole raised horizontally, and shouts sternly: [Chinese]什么人!\n\n[Shot 3] At 4.2 seconds, hard cut. Medium shot, eye-level, slow orbit. The helpers instantly surround Lin Zhi completely, their gazes falling on his short hair, short sleeves, and exposed arms, the encirclement gradually tightening. Lin Zhi is at the center, multiple gazes focusing inward from all sides, maintaining the same-side axis without flipping. The helpers slowly close the circle around Lin Zhi, carrying poles pressed down, thick ropes pulled taut, Lin Zhi unconsciously steps back, his gaze shifting back and forth.\n\n[Shot 4] At 6.8 seconds, hard cut. Close-up, eye-level, slow push in. Helper B stares with wide eyes at Lin Zhi's hair and clothing, expression surprised and vigilant, thick rope in hand slightly raised. Helper B faces Lin Zhi from a diagonal side, upper body leaning forward, gaze scanning from head to toe. (S2) grips the thick rope tightly, staring at Lin Zhi and mutters with a tense voice: [Chinese]短发?还穿这种露出四肢的奇装异服……\n\n[Shot 5] At 9.2 seconds, hard cut. Close-up, slight low-angle, static camera. Helper A's expression suddenly becomes certain, arm raised pointing at Lin Zhi, surrounding helpers respond accordingly, the encirclement even tighter. Helper A is in the visually dominant position, other helpers turn their bodies and direct their gazes in obedience to his judgment. (S1) points at Lin Zhi and shouts loudly: [Chinese]肯定是北方的奸细!兄弟们,抓起来,送官查办!\n\n[Shot 6] At 11.9 seconds, hard cut. Close-up, eye-level, handheld pull back. Lin Zhi's pupils suddenly contract, feet stumbling backward in panic, behind him are swaying reeds and wet mud, face full of panic. Lin Zhi faces multiple helpers, gaze quickly switching between different people, retreat cut off by the encirclement. (S3) raises both hands halfway in a defensive posture, continuously stepping back, and shouts in panic: [Chinese]我不是!我不是奸细!\n\noverall_soundscape:\nRustling of reeds being parted, footsteps on mud, wooden pole collisions, short and powerful shouts, background wind sound lowered, fine footsteps, rope friction, clothing rustling, rope strands tightening, lowered murmuring, command voice raised, surrounding footsteps responding, wooden pole moving sounds, rapid breathing, muffled sound of shoes sinking into mud, reeds slapping pants.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【场景分析】\n场景:(大燕朝河岸边)\n承接:林止刚刚认出远处城墙,身体仍处于震惊僵立状态\n过渡:硬切进入突发围捕,情绪意图为用集体压迫感打断人物的短暂确认\n人物表演底层逻辑总纲:【林止处于骤然受惊后的慌乱自保;伙计甲处于本能防范与主导局势状态;伙计乙处于惊异观察与跟随定性状态;伙计若干处于从众式戒备状态】\n【分镜承接】\n前一分镜末尾画面:林止站在河岸空地,望着远处城墙失神\n本分镜开场画面:林止背后的芦苇突然剧烈晃动,几道人影破丛而出\n承接逻辑:同一场景强衔接,利用背后动势制造惊吓切入\n灯光:胶片摄影质感,电影光线,半透明纹理,哑光高级质感,动态柔光投影,朦胧发光层,柔和侧逆光,极致发丝光,背景压暗虚化;\n \n**【镜头1】0.0-2.2秒\n转场: 无\n镜头: 中景/平视/快速平移跟拍\n画面描述: 林止背后的芦苇丛猛地被拨开,几名伙计从后方冲出,手中扁担与绳索在阳光下晃动。\n核心动作: 众伙计突然逼近形成包围起势。\n对手戏设计:伙计群从林止后方与侧后方压入,林止处于画面前景偏中位置,形成被围中心。\n人物底层逻辑: 群体先发制人,以人数与工具建立控制权。\n主体动作:伙计若干快步冲出芦苇,鞋底踏碎泥面,扁担抬起,粗绳甩开,迅速缩小与林止的距离。\n音效: 芦苇被拨开的刷响,脚步踩泥声,木杆碰撞声\n【镜头2】2.2-4.2秒\n转场: 硬切\n镜头:近景/平视/手持微晃\n画面描述:伙计甲站在最前方,眉头紧锁,肩膀前压,扁担斜拦在身前,神情凶厉。\n核心动作:伙计甲厉声喝止。\n对手戏设计:伙计甲正面逼视林止,视线完全锁定,后方伙计形成虚焦压阵。\n人物底层逻辑:先用声音和姿态震慑对方,建立审问者地位。\n主体动作:伙计甲上前半步,扁担横起,厉声喝道:“什么人!”\n音效: 喝声短促有力,背景风声被压低\n【镜头3】4.2-6.8秒\n转场: 硬切\n镜头: 中景/平视/缓慢环绕\n画面描述: 众伙计瞬间将林止团团围住,目光纷纷落在他的短发、短袖与裸露手臂上,包围圈逐步收紧。\n核心动作: 围拢、打量、压迫。\n对手戏设计: 林止位于中心,四周多道视线向内聚焦,保持同侧轴线不翻转。\n人物底层逻辑: 对未知装束的本能排异迅速转化为共同判断。\n主体动作: 众伙计绕着林止缓慢收圈,扁担下压,粗绳被扯直,林止不自觉后退,目光来回游移。\n音效:细碎脚步声,绳索摩擦声,衣料窸窣声\n【镜头4】6.8-9.2秒\n转场: 硬切\n镜头:近景/平视/缓慢推近\n画面描述:伙计乙瞪大眼睛盯着林止的头发和衣着,神色惊异而戒备,手中粗绳微微上提。\n核心动作:伙计乙给出外观判断。\n对手戏设计:伙计乙斜侧面对林止,上半身前倾,视线从头到脚扫视。\n人物底层逻辑:通过异样外形强化对可疑身份的想象。\n主体动作:伙计乙攥紧粗绳,盯着林止喃喃又发紧地说:“短发?还穿这种露出四肢的奇装异服……”\n音效: 绳股收紧声,人物压低的议论声\n【镜头5】9.2-11.9秒\n转场: 硬切\n镜头:近景/略仰视/固定镜头\n画面描述:伙计甲神色骤然笃定,手臂抬起指向林止,周围伙计随之响应,包围圈更紧。\n核心动作:伙计甲下达抓人命令。\n对手戏设计:伙计甲处于视觉主导位,其他伙计以身体转向与目光服从他的判断。\n人物底层逻辑: 用明确结论凝聚众人行动,避免局势失控。\n主体动作:伙计甲指着林止,大声喝道:“肯定是北方的奸细!兄弟们,抓起来,送官查办!”\n音效: 命令声抬高,周围脚步立刻响应,木杆挪动声加重\n【镜头6】**11.9-15秒\n转场: 硬切\n镜头:近景/平视/手持后撤\n画面描述: 林止瞳孔猛缩,脚下慌乱后退,身后是摇晃的芦苇与湿泥,脸上惊慌失措。\n核心动作: 林止急忙辩解并本能后退。\n对手戏设计: 林止面对多名伙计,视线在不同人之间快速切换,退路被包围圈切断。\n人物底层逻辑:语言辩白是他在陌生秩序中唯一能立刻使用的自保方式。\n主体动作: 林止双手半抬作防御姿势,连续后退,慌张喊道:“我不是!我不是奸细!”\n音效: 呼吸急促声,鞋底陷进泥地的闷响,芦苇拍打裤腿声\n资产对照表(视频中涉及的角色、场景与物品如下(保持一致性)):\n视频场景对照表: 大燕朝河岸边=【图1】\n视频角色对照表: 林止=【图2】;伙计甲=【@伙计甲/粗布短打】;伙计乙=【@伙计乙/粗布短打】;伙计若干=【@伙计若干/粗布短打】\n视频道具对照表: 粗绳=【图3】;扁担=【图4】\n \n ", "reference_images": ["references/443fcbb76e93fbbf6bf0acf4e8aedf38a372b1d8a2bc693956f67474ec829239.jpg", "references/0a1ca2acfef8f916e5fdf87d426bfa7b499a0d7032008b8367859f489ea39240.jpg", "references/6385b934b869947bbcffdd92732da934119da38475c8aea5c1ae6c0abee4bcdb.jpg", "references/a78559a1a70c161b8f9a021a3f758dbde67cdd87d95a8a4dec144e8c8a719f90.jpg"], "reference_sha256": ["443fcbb76e93fbbf6bf0acf4e8aedf38a372b1d8a2bc693956f67474ec829239", "0a1ca2acfef8f916e5fdf87d426bfa7b499a0d7032008b8367859f489ea39240", "6385b934b869947bbcffdd92732da934119da38475c8aea5c1ae6c0abee4bcdb", "a78559a1a70c161b8f9a021a3f758dbde67cdd87d95a8a4dec144e8c8a719f90"], "workflow": "ref2va", "prompt_sha256": "134ea5400d7eee90eb1046870cca496da81c80b963b0f1f7b057fee729ec343e", "original_prompt_sha256": "39e9990464fab81f7cdbf605228969bf278d2bcc3be054e80cce606a766a3432", "video_sha256": "2a0722500040a981003dfd105aa5779003cd106b9e3b340b0c66db5a0586091e", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_061.json", "verification": "verification/sample_061.json"} {"file_name": "videos/sample_062.mp4", "sample_id": "#62", "prompt": "subject_definitions:\n: Host 1, matching the appearance and identity from .\n: The podium and daytime scene setting.\n\nsummary:\n[reference generation] A realistic, cinematic documentary-style scene featuring Host 1 standing behind a podium on the right side of the frame, delivering a speech with a solemn yet warm smile, with the camera transitioning from a medium shot to a close-up.\n\nretention_analysis:\nThe identity and appearance of Host 1 from are fully preserved. The podium and daytime scene environment from are fully preserved as the physical setting for the shot.\n\ndetailed_description:\n[Shot 1] At 00:00.000, 0-6s, medium shot, eye-level angle. The camera has a subtle documentary breathing feel (gentle organic micro-movements) while remaining a fixed, static setup. The visual style is live-action cinematic with authentic human skin texture, natural and restrained expressions, realistic clothing materials, and a unified realistic aesthetic. The lighting features high-level depth, creating a clean, warm, solemn, and delicate atmosphere that highlights a sense of era and heritage. Avoid any animation, AI-generated look, plastic feel, over-smoothing, exaggerated acting, or fake lighting effects. Host 1 stands behind the podium from on the right side of the frame, holding a microphone, wearing a solemn yet warm smile. He pauses slightly, takes a deep breath, and sweeps his gaze across the audience. As the dialogue reaches the second half, the shot transitions to a medium close-up of Host 1. Host 1 (S1) says [Chinese]接下来,让我们用最热烈的掌声,有请本次结对仪式最特殊的一组师徒登场—— in a loud, powerful voice filled with respect and warmth. Ensure continuous character positioning and consistent voice tone. No subtitles are allowed.\n\noverall_soundscape:\nLoud and powerful applause that begins before the dialogue finishes.\n\nnon_diegetic_music:\nN/A", "original_prompt": "真人实拍电影质感,真实人物皮肤纹理,表情自然克制,服装材质真实,统一写实画风;光影层次高级;镜头有纪录片呼吸感,画面干净温暖、庄重细腻,突出年代感与传承感;避免动画感、AI感、塑料感、过度磨皮、夸张表演、虚假光效。(no srt,no music),人物音色统一,视频场景,注意人物站位要连贯,不允许出现字幕\n参考 @视频1 ,更换主持人 @主持人1,其它场景动作都不变;\n镜号:4 | 场次:日, 为场景参考图 | 0-6s | 中景,平拍, | 固定镜头,主持人 @主持人1 站在画面右侧的演讲台上 @图片2,台词说到后半段改成主持人 近景 |主持人 @主持人1 站在画面的右侧演讲台 @图片2 的后面,手持话筒,面带庄重又温暖的微笑。他微微停顿,深吸一口气,目光扫过全场。 | 台词 & 音效:Dialogue:[声音洪亮有力,带着满满的敬意与温情] @主持人 (音色参考 ):\"接下来,让我们用最热烈的掌声,有请本次结对仪式最特殊的一组师徒登场——\" SFX:话音未落,掌声已起\n\n参考图片说明:\n图1: 角色: 主持人1", "reference_images": ["references/8143e5cbef78949482e4cc70ee85a5321f61085a112b2faa7ec48a024d097737.jpg", "references/3ac1026ff34279324ad7852b435f6f00a7d4b4b0a0ade1498de8d283dae03a74.jpg"], "reference_sha256": ["8143e5cbef78949482e4cc70ee85a5321f61085a112b2faa7ec48a024d097737", "3ac1026ff34279324ad7852b435f6f00a7d4b4b0a0ade1498de8d283dae03a74"], "workflow": "ref2va", "prompt_sha256": "31c20d2a6d5a3336d3c7447f7396080e95cf7b01215b6bd7c177745a5e46b26e", "original_prompt_sha256": "d8691145321743086a894398d05e7afa97ee03e214ff10df7c8f89d72c89ec2b", "video_sha256": "dd4c7e4fe771f066d3082d522cfa15ed3ab5b359c3b67490aabdbcb268ccd8bd", "task_adapted": true, "width": 1344, "height": 768, "fps": 24, "num_frames": 158, "duration_seconds": 6.583333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_062.json", "verification": "verification/sample_062.json"} {"file_name": "videos/sample_063.mp4", "sample_id": "#63", "prompt": "subject_definitions:\n: A female model demonstrating skincare, appearance from .\n: A bottle of Changbai Mountain Ginseng Cream, appearance from .\n\nsummary:\n[reference generation] A 15-second skincare product advertisement video for Changbai Mountain Ginseng Cream, featuring a female model demonstrating the product's effects and the product itself, with voiceover and subtitles highlighting its benefits.\n\nretention_analysis:\nThe two supplied images are retained as the primary visual subjects. (the female model from ) is featured in close-up shots showing facial skin texture and cream application. (the Changbai Mountain Ginseng Cream bottle from ) is displayed prominently in product shots. Both subjects are fully preserved as the core visual elements throughout the 15-second advertisement.\n\ndetailed_description:\n[Shot 1] A quick close-up of 's face showing dry lines and fine wrinkles, with a simple hand gesture. (S1) says in an off-screen voiceover [Chinese]皮肤干长细纹,看着显老气? The visible text \"皮肤干长细纹,看着显老气?\" appears on screen. The voiceover delivery is crisp and clear.\n\n[Shot 2] At 0:04.000, a quick display of (the Changbai Mountain Ginseng Cream bottle), followed by a close-up of scooping the cream. (S1) says in an off-screen voiceover [Chinese]长白山人参面霜,深润养护肌肤。 The visible text \"长白山人参面霜,深润养护肌肤。\" appears on screen. The voiceover delivery is clear and powerful.\n\n[Shot 3] At 0:09.000, a quick shot of applying the cream, with the camera freezing on the skin texture to show the effect. (S1) says in an off-screen voiceover [Chinese]清爽好吸收,肌肤紧致水润。 The visible text \"清爽好吸收,肌肤紧致水润。\" appears on screen. The voiceover delivery is light and rhythmic.\n\n[Shot 4] At 0:13.000, is centered and frozen in a clean frame. (S1) says in an off-screen voiceover [Chinese]护肤实力派,放心入手。 The visible text \"护肤实力派,放心入手。\" appears on screen. The voiceover delivery is capable and compelling.\n\noverall_soundscape:\nThe overall soundscape includes the off-screen voiceover delivering the script in a crisp, clear, and compelling manner, along with subtle sound effects for the hand gestures, product display, and cream application.\n\nnon_diegetic_music:\nN/A", "original_prompt": "0-3秒 痛点引入\n口播:皮肤干长细纹,看着显老气?\n画面动作:快速特写面部干纹,手势简洁\n字幕:皮肤干长细纹,看着显老气?\n配音:干脆明快\n\n4-8秒 讲解产品核心\n口播:长白山人参面霜,深润养护肌肤。\n画面动作:快速展示瓶身,舀取膏体特写\n字幕:长白山人参面霜,深润养护肌肤。\n配音:清晰有力\n\n9-12秒 效果展示\n口播:清爽好吸收,肌肤紧致水润。\n画面动作:快速涂抹,镜头定格肤感\n字幕:清爽好吸收,肌肤紧致水润。\n配音:节奏轻快\n\n13-15秒 催单引导\n口播:护肤实力派,放心入手。\n画面动作:产品居中定格,画面干净\n字幕:护肤实力派,放心入手。\n配音:干练有号召力", "reference_images": ["references/b28cdf77c4a74cc24decc3ad274506faa81421f765afba6e0e02587fc92a6aeb.jpg", "references/1b0ca38c989bc96947928a01975aa0fa64cc8a9aa3be9a545a76126da482565c.jpg"], "reference_sha256": ["b28cdf77c4a74cc24decc3ad274506faa81421f765afba6e0e02587fc92a6aeb", "1b0ca38c989bc96947928a01975aa0fa64cc8a9aa3be9a545a76126da482565c"], "workflow": "ref2va", "prompt_sha256": "30dda614dbf5af3508b2e29295e8061ed5afce7021f4d3943e33beb2006f7c5c", "original_prompt_sha256": "8008cef87eb92e4bfa512c20083d1ac0967bf76297eb815c626e619ff29b5294", "video_sha256": "693e2e633ebc412862df9b7cf62678ab3b7fd1ffccbd455810a23aedd8d027b2", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_063.json", "verification": "verification/sample_063.json"} {"file_name": "videos/sample_064.mp4", "sample_id": "#64", "prompt": "integrated_multimodal_description:\n[Shot 1] Medium shot, eye-level, static camera. Morning cold light slants down from the high sealed window, and a beam of cold light cuts into the dim room as an old wooden door opens from the outside. He Yutang sits on the bed deep in the background, while Li Lian enters in the foreground holding a white porcelain bowl of rice porridge, the thin porridge slightly swaying; the two keep a controlled distance. Li Lian bends down and places the bowl on a wooden stool. No dialogue. Sound effects: light creak of door hinges, porcelain bowl lightly tapping the wooden stool, sound of cloth shoes stepping on the floor.\n\n[Shot 2] At 00:02.200, medium close-up, eye-level, slow push-in. He Yutang sits on the edge of the bed, hands resting on her knees, shoulders deliberately relaxed. She forces a weak smile on her lips, but her eyes momentarily sharpen with hidden alertness. She looks up at Li Lian, softening her gaze, keeping her body from leaning forward to lower the other's guard. He Yutang (S1) says [Chinese]谢谢妈。. Sound effects: soft speaking voice, indoor silence.\n\n[Shot 3] At 00:04.400, medium close-up, eye-level, slight pan. Li Lian is stunned for a moment, then her eyes and brows relax. She pats her thigh, turns her whole body towards the door, and reveals a proud and surprised smile. She shouts towards the door while He Yutang stays in the background observing. Li Lian (S2) says [Chinese]建军,你快看!雪芳终于想通了,肯叫我妈了!. Sound effects: thigh patting sound, shouting voice spreading out the door.\n\n[Shot 4] At 00:06.800, medium shot, eye-level, static camera. Lin Jianjun stands at the doorway, backlit by the outside light, his figure blocking the exit. Lin Jianjun holds the doorway while He Yutang remains seated on the bed, keeping a stable 180-degree axis; the doorway stays the power position. His gaze scrutinizingly falls on He Yutang's face. After a pause, his expression slightly softens. Lin Jianjun (S3) says [Chinese]想通了就好,只要你乖乖跟我过日子,以后少不了你的福气。. Sound effects: slight wind from outside the door, low male voice, light rustle of clothing.\n\n[Shot 5] At 00:09.600, close-up, slightly high-angle looking down, slow lateral translation. Lin Jianjun and Li Lian listen from the foreground side. He Yutang picks up the white porcelain bowl and takes a sip. Her eyelashes droop to hide the emotions in her eyes; she averts her gaze to show dependence, and her line of sight follows the bowl and drops towards the door lock. He Yutang (S1) says [Chinese]我以前的事记不起来了,现在只能指望你们了。. Sound effects: subtle sound of drinking porridge, porcelain spoon lightly touching the bowl rim, quiet breathing.\n\n[Shot 6] At 00:12.400, medium close-up, eye-level, slow pull-back. Lin Jianjun smiles after hearing this, turns and leaves the doorway. Li Lian collects the empty bowl, looks back at He Yutang for a glance, and turns to walk outside, transitioning from inside the room towards the door. He Yutang stays in the background watching quietly, hinting the watcher/watched positions are about to reverse. The bright light at the door becomes the only exit again. Li Lian (S2) says [Chinese]行了,你先歇着,我得去后面菜地转转。. Sound effects: porcelain bowl lightly tapping, footsteps moving towards the door, slightly stronger wind at the door.\n\nConstraint: No subtitles or any visible text appear on the screen.\n\noverall_soundscape:\nPhysical sounds and ambience include the light creak of door hinges, porcelain bowls tapping on wood, cloth shoes on the floor, soft speaking and shouting voices, thigh patting, slight wind from outside, rustling clothes, drinking porridge, spoon clinking, quiet breathing, and footsteps. No music is specified.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【场景分析】\n场景:(林图1/白日残光)\n承接:上一分镜停在高处封死的小窗与贺雨棠求生念头上\n过渡:硬切到晨光入室,情绪意图为表面缓和、内里算计的反差推进\n人物表演底层逻辑总纲:【贺雨棠处于高度警觉下的伪装顺从状态;李莲处于放松戒心的得意状态;林建军处于半信半疑的占有与安抚状态】\n【分镜承接】\n前一分镜末尾画面:小窗木板与锈铁丝网特写\n本分镜开场画面:同一房间内,晨间冷光从高处斜照下来,木门被打开\n承接逻辑:同一场景时间推进,空间锚点延续\n**【镜头1】0.0-2.2秒\n转场: 无\n镜头: 中景/平视/固定镜头\n画面描述:旧木门从外侧打开,一束白日冷光切入灰暗室内,李莲端着白粥瓷碗走进来,碗里稀饭轻微晃动。\n核心动作: 李莲进门并把稀饭放到木凳上。\n对手戏设计: 李莲自门口进入前景,贺雨棠坐在床边深处,二人保持可控距离。\n人物底层逻辑: 李莲以施舍式照料维持控制秩序。\n主体动作: 李莲弯腰把碗放上木凳,无台词。\n音效: 门轴轻响,瓷碗轻磕木凳声,布鞋踩地声\n【镜头2】2.2-4.4秒\n转场: 硬切\n镜头:近景/平视/缓慢推近\n画面描述: 贺雨棠坐在床边,双手放在膝上,肩线刻意放松,唇角勉强牵起一抹虚弱笑意,眼神却在一瞬间收住锋芒。\n核心动作: 她主动示弱示好。\n对手戏设计: 贺雨棠目光抬向李莲,视线柔化,身体不前倾,降低对方戒心。\n人物底层逻辑:以顺从姿态换取近身机会与信息。\n主体动作: 贺雨棠轻声说道:“谢谢妈。”\n音效:轻弱说话声,室内静默\n【镜头3】4.4-6.8秒\n转场: 硬切\n镜头: 中近景/平视/小幅摇移\n画面描述: 李莲先是一愣,随即眉眼舒展开,拍了拍大腿,整个人转向门外,脸上露出得意又惊喜的笑。\n核心动作: 李莲被一句“妈”彻底安抚,兴奋地向门外招呼。\n对手戏设计: 李莲朝门外喊,贺雨棠留在后景观察,形成信息差。\n人物底层逻辑: 她相信驯服已经见效,防备开始松动。\n主体动作: 李莲高兴喊道:“建军,你快看!雪芳终于想通了,肯叫我妈了!”\n音效: 拍腿声,喊话声向门外扩散\n【镜头4】6.8-9.6秒\n转场: 硬切\n镜头: 中景/平视/固定镜头\n画面描述: 林建军站在门口逆着外面亮光,身形堵住出口,目光审视地落在贺雨棠脸上,停顿片刻后神色略缓。\n核心动作: 林建军确认贺雨棠的“顺从”,口头施压安抚并存。\n对手戏设计: 林建军守门,贺雨棠坐床边,轴线稳定,门口始终是权力高位。\n人物底层逻辑: 他希望通过软硬兼施让她彻底放弃逃跑。\n主体动作: 林建军松口气般说道:“想通了就好,只要你乖乖跟我过日子,以后少不了你的福气。”\n音效: 门外风声微入,低沉男声,衣料轻动\n【镜头5】9.6-12.4秒\n转场: 硬切\n镜头: 特写/略俯视/缓慢平移\n画面描述: 贺雨棠端起白粥瓷碗,喝了一口,睫毛低垂遮住眼底情绪;她的视线顺着碗延落向门锁方向。\n核心动作: 她一边示弱编造失忆依赖,一边借低头动作观察门锁位置。\n对手戏设计: 林建军与李莲处于前景外侧听她说,贺雨棠通过避视表现依附姿态。\n人物底层逻辑:语言上交出主动,视觉上锁定逃生关键点。\n主体动作: 贺雨棠端着碗低声说:“我以前的事记不起来了,现在只能指望你们了。”\n音效: 喝粥细响,瓷勺轻碰碗沿声,安静呼吸声\n【镜头6】**12.4-15.0秒\n转场: 硬切\n镜头: 中近景/平视/缓慢后拉\n画面描述: 林建军听后笑了笑转身离开门口;李莲收起空碗,回头看贺雨棠一眼,转身往外走。门口亮光再次成为唯一出口。\n核心动作: 两人戒心下降并准备离开房间。\n对手戏设计: 李莲由室内转向门外,贺雨棠留在后景静静注视,形成猎人与猎物位置即将反转的前奏。\n人物底层逻辑: 外部看守松动,为突发行动制造缝隙。\n主体动作: 李莲笑着说:“行了,你先歇着,我得去后面菜地转转。”\n音效: 瓷碗轻碰声,脚步声向门口移动,门口风声略强\n资产对照表(视频中涉及的角色、场景与物品如下(保持一致性)):\n视频场景对照表: 林家卧室=【图1/白日残光】\n视频角色对照表: 贺雨棠=【图2】;李莲=【图3】;林建军=【图4】\n视频道具对照表: 白粥瓷碗=【图5】;木凳=【@木凳】;门锁=【@门锁/生锈铁锁】\n视频中不要出现任何字幕。", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "4909f9edd16d856b125b22ebaedffa13c5b57149817ce90501deea157ea4b3e2", "original_prompt_sha256": "0a31c8d3c8d36445fd7d03b31e63991aef3923cd4190fcd35eb80470c23ba0b0", "video_sha256": "c1c7c06f15cb50ba1abde69848fac923d6da7b4d227a2a1727af088bc7d4797f", "task_adapted": true, "width": 768, "height": 1024, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_064.json", "verification": "verification/sample_064.json"} {"file_name": "videos/sample_065.mp4", "sample_id": "#65", "prompt": "subject_definitions:\n is Lin Dagang, whose appearance and clothing are from .\n is Wang Xiulan, whose appearance and clothing are from .\n is Xu Youtian, the village head, whose appearance and clothing are from .\n\nsummary:\n[reference generation] Generate a 15-second video sequence in the old house's main hall during the afternoon, featuring Lin Dagang, Wang Xiulan, and village head Xu Youtian. The scene strictly follows the 180-degree axis and eyeline matching rules to prevent axis jumping. The visual style is clean, high-end, with clear foreground, midground, and background depth, high emotional tension, and realistic physical details. No watermarks, no subtitles, no logos.\n\nretention_analysis:\n (Lin Dagang) is fully preserved as a main character appearing in shots 1 and 3, with appearance from .\n (Wang Xiulan) is fully preserved as a main character appearing in shot 1, with appearance from .\n (Xu Youtian) is fully preserved as the main speaking character appearing in shots 1, 2, and 4, with appearance from .\n is fully preserved as the scene setting (old house main hall).\n is fully preserved as the reference for the spatial positioning of the characters.\n\ndetailed_description:\n[Shot 1] 0:00.000 - 0:03.000 | Medium shot, eye-level, static camera. Afternoon natural light. On the left side of the frame, (Lin Dagang) and (Wang Xiulan) stiffen their expressions simultaneously, as if awakened from a daze. On the right side, village head (Xu Youtian) sits on a stool, slowly half-closing his eyes, his head swaying left and right extremely slowly, his posture as steady as Mount Tai. (S1) says [Chinese]静瑶有没有打小报告,两说。 in a low, experienced, and calm voice.\n\n[Shot 2] 0:03.000 - 0:07.000 | Close-up, eye-level, slow push-in. Afternoon natural light. Close-up of 's weathered and wrinkled face. He slowly opens his eyes, his gaze deep and unfathomable. His dry fingers tap rhythmically on the armrest of the chair, revealing a strong sense of control. (S1) says [Chinese]但信用社来,是迟早的事。你叔账户里有存款, in a rational, majestic, and unquestionable voice. SFX: The sound of fingers tapping on wood.\n\n[Shot 3] 0:07.000 - 0:11.000 | medium close-up, eye-level, static camera. Afternoon natural light. Cut back to a medium close-up of 's reaction. 's eyebrows are tightly furrowed into a 'Chuan' shape, his gaze fixed unblinkingly on the right off-screen space (towards the village head). As the off-screen voiceover continues, the tension in his shoulders slightly relaxes. (S1) says in an off-screen voiceover [Chinese]他们系统里肯定有记录,人去世了,按规定他们得来核实。 in a calm and analytical tone.\n\n[Shot 4] 0:11.000 - 0:15.000 | medium close-up, eye-level, slow pull-back. Afternoon natural light. Cut back to again. gently presses his raised palm downwards in mid-air, signaling them to calm down, his eyes full of insight into the rules of the world. (S1) says [Chinese]这事儿,怪不到别人头上。 in a firm, soothing, and majestic voice.\n\noverall_soundscape:\nScripted character dialogue, with rhythmic finger taps on the wooden armrest in Shot 2. No other sound effects.\n\nnon_diegetic_music:\nN/A", "original_prompt": "人物站位:@林大刚 @许友田 @王秀兰 参考 @图片5 位置关系。\n要求:画面干净又高级;画面里前中远景递进关系分明。极具情绪张力与物理细节真实感。\n画面:无水印,无字幕,无logo 。\n镜号:9 | 场次:午后 | @老宅堂屋 | 0-3s | 中景,平视 | 固定镜头 | 午后,自然光, 画面左侧的大刚和秀兰表情同时僵了一下,似被惊醒。右侧凳子上的村长 @许友田 则慢条斯理地半闭上眼睛,脑袋极度缓慢地左右摇晃了两下,姿态稳如泰山。 | 台词 & 音效:Dialogue:[用低沉且老练的声音说] @许友田 (低沉/老练/沉静):\"静瑶有没有打小报告,两说。\" SFX:无 | 备注:严格遵守180度轴线与视线匹配原则,严防越轴跳切\n镜号:10 | 场次:午后 | @老宅堂屋 | 3-7s | 特写,平视 | 缓慢推进 | 午后,自然光, 村长布满沧桑皱纹的面部特写。他缓缓睁开眼,眼神深邃不可测,干枯的手指在椅子扶手上有节奏地敲击着,显露出强烈的掌控感。 | 台词 & 音效:Dialogue:[用理智且不容置疑的声音说] @许友田 (理智/威严/不容置疑):\"但信用社来,是迟早的事。你叔账户里有存款,\" SFX:手指敲击木头的笃笃声 | 备注:严格遵守180度轴线与视线匹配原则,严防越轴跳切\n镜号:11 | 场次:午后 | @老宅堂屋 | 7-11s | 近景,平视 | 固定镜头 | 午后,自然光, 切回大刚的反应近景。大刚的眉头紧紧锁成一个川字,目光一动不动地死死盯着右侧画外(村长的方向),随着画外音的陈述,他肩膀的紧绷感微微有些松懈。 | 台词 & 音效:Dialogue:[用平缓解析的语调说] @许友田 (VO)(平缓/解析):\"他们系统里肯定有记录,人去世了,按规定他们得来核实。\" SFX:无 | 备注:视线匹配,展示大刚的认真倾听状态\n镜号:12 | 场次:午后 | @老宅堂屋 | 11-15s | 近景,平视 | 缓慢后退 | 午后,自然光, 再次切回村长。村长将抬起的手掌在半空中轻轻向下压了压,示意他们稍安勿躁,眼神里满是对世事规则的洞悉。 | 台词 & 音效:Dialogue:[用笃定且安抚的声音说] @许友田 (笃定/安抚/威严):\"这事儿,怪不到别人头上。\" SFX:无 | 备注:严格遵守180度轴线与视线匹配原则,严防越轴跳切\n\n@许友田(图片3) 的音色按照 @许有田(音频2) 进行演绎\n\n参考图片说明:\n图1: 角色: 林大刚\n图2: 角色: 王秀兰\n图3: 角色: 许友田\n图4: 场景: 老宅堂屋", "reference_images": ["references/6f9862bfda73412ef6c412f93b6d3d679c302e45d628df6beeb30aa55ac045d5.jpg", "references/dc350033ab7c53822030da3da8e0130404bcc4b7ccab3e7dfb41662135dd4b63.jpg", "references/198084bc25afb0e1361c09c0866c3913fee4e4eae47dea95e179fa3a8a6e207a.jpg", "references/481be503542fab6d35cde14f903788805db48a653406e8bce1bb9fa075dea226.jpg", "references/8275b6139e2e6afcbded5b37f2a65d12357e635c3acf4d9265f085ebfd7bd266.jpg"], "reference_sha256": ["6f9862bfda73412ef6c412f93b6d3d679c302e45d628df6beeb30aa55ac045d5", "dc350033ab7c53822030da3da8e0130404bcc4b7ccab3e7dfb41662135dd4b63", "198084bc25afb0e1361c09c0866c3913fee4e4eae47dea95e179fa3a8a6e207a", "481be503542fab6d35cde14f903788805db48a653406e8bce1bb9fa075dea226", "8275b6139e2e6afcbded5b37f2a65d12357e635c3acf4d9265f085ebfd7bd266"], "workflow": "ref2va", "prompt_sha256": "70a647b84f37aff2cc576759261ea54ece10ad93702206dd2efb62f996131f99", "original_prompt_sha256": "cada4cfed450a01a8f7ab9e0680a71d857271974383003e0303536109829bb02", "video_sha256": "4d86905f9f1541367d7e680c6083f575d78cc0a6c08ed8263ab26eb4bcdbe6fc", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_065.json", "verification": "verification/sample_065.json"} {"file_name": "videos/sample_066.mp4", "sample_id": "#66", "prompt": "integrated_multimodal_description:\nNo subtitles appear on the video. The scene is set at an airport pickup area. The sequence continues in the same space to reveal the truth, using hard cut shot-reverse-shot to let Shen Huaichuan puncture his own fantasy step by step. Shen Huaichuan is in self-judgment after his fantasy shatters; He Yutang remains calm and does not actively comfort; Lu Jinglin maintains a protective positioning without over-explaining.\n\n[Shot 1] At 0.0 seconds, hard cut. Medium close-up, eye-level angle, slow push-in. He raises his head to look at Lu Jinglin and He Yutang standing side by side. Shen Huaichuan stands in the center of the foreground. Behind him, the airport crowd continues to move, but he remains still as if pinned in place, his face pale, his Adam's apple bobbing. He looks toward Lu Jinglin and He Yutang, struggling to voice the old information. His gaze passes through the front of the camera, landing on the same plane as the two people opposite him; the two remain motionless, reinforcing his sense of isolation. Shen Huaichuan (S1) says [Chinese]原来上次你说的那个太太,真的就是雨棠……\n\n[Shot 2] At 3.4 seconds, hard cut. Two-shot medium close-up, eye-level angle, fixed camera. He Yutang stands beside Lu Jinglin, her arm lightly linked with his. Lu Jinglin turns slightly to shield her on the outer side, their postures stable and intimate. The two remain silent, responding to Shen Huaichuan with their positioning. He Yutang and Lu Jinglin stand close together, forming a united front to the outside; both look toward Shen Huaichuan. He Yutang links her arm with Lu Jinglin's, her expression composed, while Lu Jinglin maintains a steady, protective posture.\n\n[Shot 3] At 6.0 seconds, hard cut. medium close-up, eye-level angle, handheld with slight shake. Shen Huaichuan's gaze shifts to He Yutang. Pain rapidly accumulates in his eyes, his brow furrows tightly, the corners of his mouth drop involuntarily, his entire face looking as if instantly drained. He finally voices the guess he least wants to believe. Shot-reverse-shot with He Yutang, their gazes meet but receive no emotional response. Shen Huaichuan (S1) says, his voice trembling, [Chinese]雨棠……你嫁的人不是陆家老爷子……而是他?\n\n[Shot 4] At 9.8 seconds, no transition. Close-up, eye-level angle, extremely slow push-in. He Yutang's face is calm, her eyes clear. Lu Jinglin's shoulder line and suit collar occupy the edges of the frame, forming a stable support. The airport's cold white light illuminates her silhouette clearly. She silently watches Shen Huaichuan, waiting for the second before she speaks. Her gaze briefly meets Shen Huaichuan's, but the emotional distance is obvious; Lu Jinglin remains on her dependable side. He Yutang links her arm with Lu Jinglin's, her expression composed, the stillness before her slight nod becoming the prelude to her answer.\n\noverall_soundscape:\nAirport ambient sounds including blurred public broadcasts, low crowd noise, distant luggage wheel sounds, and airport air conditioning low-frequency hum. Character breathing sounds are slightly amplified in emotional moments, with broken breathing and fabric friction sounds. Background environmental sounds are lowered during emotional peaks. A low-frequency emotional drone bed is present during the emotional peak as an SFX ambience, distinct from background music.\n\nnon_diegetic_music:\nN/A", "original_prompt": "视频上不要出现任何字幕,可以有音效但是不能有背景音乐\n【场景分析】\n场景:图1\n承接:上一分镜落在红绳特写与沈淮川确认后的表情,继续在同一空间内推进真相揭示\n(参考图2的站位,不参考内容和镜头角度。)\n过渡:硬切正反打,情绪意图为让沈淮川一步步自我戳穿幻想\n人物表演底层逻辑总纲:【沈淮川处于幻想破灭后的自我审判;贺雨棠保持冷静,不主动安抚;陆景林维持占位性保护,不做多余解释】\n【分镜承接】\n前一分镜末尾画面:红绳在冷白灯下轻微晃动,沈淮川眼神被其牢牢钉住\n本分镜开场画面:沈淮川抬头,看向陆景林与贺雨棠并肩的姿态\n承接逻辑:视线延续与空间强衔接\n**【镜头1】0.0-3.4秒\n转场: 硬切\n镜头: 中近景/平视/缓慢推近\n画面描述: 沈淮川站在前景中央,身后机场人流持续运动,只有他像被钉住一般静止,脸色苍白,喉结上下滚动。\n核心动作: 他看向陆景林与贺雨棠,艰难把旧日信息说出口。\n对手戏设计:视线穿过镜头前方,落在对面两人同一平面;两人不动,强化他的孤立感。\n人物底层逻辑: 他试图用语言确认真相,其实是在否认自己曾错失的一切。\n主体动作: 沈淮川看向陆景林和贺雨棠,声音发涩地说:“原来上次你说的那个太太,真的就是雨棠……”\n音效: 环境广播模糊、人群低噪、呼吸声被轻微放大\n【镜头2】3.4-6.0秒\n转场: 硬切\n镜头: 双人近景/平视/固定镜头\n画面描述: 贺雨棠站在陆景林身侧,手臂轻挽着他,陆景林略侧身挡在她外侧,两人姿态稳定而亲密。\n核心动作: 两人沉默不语,用站位回应沈淮川。\n对手戏设计: 贺雨棠与陆景林彼此距离紧密,对外形成统一阵线;两人都看向沈淮川。\n人物底层逻辑: 不解释,本身就是最直接的答案。\n主体动作: 贺雨棠挽着陆景林的手,神色自若,陆景林保持沉稳保护姿态。\n音效: 布料轻碰声、远处行李轮声、机场空调低频声\n【镜头3】6.0-9.8秒\n转场: 硬切\n镜头:近景/平视/手持微晃\n画面描述: 沈淮川视线转向贺雨棠,眼底痛苦迅速累积,眉心拧紧,嘴角不受控下沉,整张脸像被瞬间抽空。\n核心动作: 他终于把最不愿相信的猜测问出口。\n对手戏设计: 与贺雨棠形成正反打,视线对准但得不到情感回应。\n人物底层逻辑: 他还想从贺雨棠口中听到否定,但内心已经知道答案。\n主体动作: 沈淮川看向贺雨棠,表情充满极致的痛苦,颤着声问:“雨棠……你嫁的人不是陆家老爷子……而是他?”\n音效: 呼吸断续声、背景环境声被压低、情绪低频铺底\n【镜头4】**9.8-14.6秒\n转场: 无\n镜头: 特写/平视/极缓推近\n画面描述: 贺雨棠面部安静,眼神清醒,陆景林肩线与西装领口占据画面边缘,构成稳定依靠,机场冷白光把她轮廓照得清晰。\n核心动作: 她沉默看着沈淮川,等待自己开口前的那一秒。\n对手戏设计: 她与沈淮川视线短暂相接,但情感距离明显;陆景林始终处于她可依靠的一侧。\n人物底层逻辑: 她不再逃避,也不再解释过往,只陈述现实。\n主体动作: 贺雨棠挽着陆景林的手,神色自若,轻微点头前的静止成为回答前奏。\n音效: 环境声稳定、轻微衣料摩擦声、情绪留白静默\n资产对照表(视频中涉及的角色、场景与物品如下(保持一致性)):\n视频场景对照表:机场接机口=图1\n视频角色对照表: 贺雨棠=图3;沈淮川=图4;陆景林=图5\n视频道具对照表: 红绳=图6", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "65af2f06f5bee1fef508e606fdb0b4d0bface9db4515cf2ecc1f7fa15c4b9b77", "original_prompt_sha256": "c1f8fb64bbb9eae97f99a52a3a66919630ba231297c89a32685ffb1358cc10b0", "video_sha256": "a08d3b958cacbd542a99a55f63fd2cc0119c683be725a57846bfc48399b9427e", "task_adapted": true, "width": 768, "height": 1024, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_066.json", "verification": "verification/sample_066.json"} {"file_name": "videos/sample_067.mp4", "sample_id": "#67", "prompt": "subject_definitions:\nSubject 1: Vivian Hawthorne, a female character whose appearance, hairstyle, facial features, clothing patterns, and accessories are strictly preserved from .\nSubject 2: The Book of Destiny, a dark brown, ruined and cracked thick book with yellowed parchment pages, whose appearance is strictly preserved from .\n\nsummary:\n[reference generation] A cinematic, highly realistic 4-shot sequence set at night inside a small canvas tent at a border medical camp (as depicted in ). Vivian Hawthorne (from ), restrained by iron shackles, interacts with the Book of Destiny (from ), reading its glowing text with morbid obsession. The visual style features 3D realistic human texture, cinematic quality, delicate and natural skin surface with fine texture, smooth 18-stop shadow transitions, and an overall steady and heavy image. 8K resolution as originally requested (note: actual generation runs at 768x1344). Extremely realistic with normal human body proportions, no deformed limbs, no exaggerated distortion, and a reasonable skeletal structure. The cinematography emulates the 2017 live-action 'Beauty and the Beast,' utilizing an ARRI Alexa Studio camera with an Angenieux Optimo Prime lens, featuring rich depth of field transitions, soft and delicate bokeh texture, and a cold-warm dual color grade (dark blue-gray-black shadows and amber-gold-brown highlights). No subtitles, no music.\n\nretention_analysis:\nSubject 1 (Vivian Hawthorne) is fully_preserved across all shots, with strict adherence to her original appearance, styling, and clothing details from . Subject 2 (The Book of Destiny) is fully_preserved from its introduction in Shot 2 onwards, maintaining its exact appearance from . The scene environment from is fully_preserved as the setting for all shots.\n\ndetailed_description:\n[Shot 1] 0-3s. Wide shot, quarter side view, slow push in. The scene is a small canvas tent at night (matching ), lit by dim side lighting and weak candlelight, creating a claustrophobic cold-warm contrast. Vivian (Subject 1, from ) sits alone on the edge of a simple wooden bed like a prisoner. Cold iron shackles on her wrists reflect glaring light. She emits a gloomy and twisted aura.\n\n[Shot 2] At 00:03.000, 3-7s. Medium shot, quarter side view, fixed camera. Dim side lighting, weak candlelight, claustrophobic cold-warm contrast. Vivian (Subject 1) twists her body to the right rear. Constrained by the handcuffs, she can only bring her hands together and struggles to reach under a deflated pillow. She forcefully pulls out the dark brown, ruined and cracked Book of Destiny (Subject 2, from ). From Vivian's subjective POV looking down, the thick Book of Destiny lies flat on her white skirt. The yellowed parchment surface suddenly glows with an eerie weak golden light, dispelling the surrounding darkness.\n\n[Shot 3] At 00:07.000, 7-11s. Close-up, low angle, slow push in. Dim side lighting, weak candlelight, claustrophobic cold-warm contrast. Golden light shines from below onto Vivian's (Subject 1) pure face, making her appear extremely twisted and fanatical. Her eyeballs reflect the golden light. The corners of her mouth pull into an extremely morbidly smug arc as she obsessively reads the text. Vivian (S1) says in an infatuated and morbidly fanatical voice [English]Sylvia is cast aside by Cyrus and loses her inheritance.\n\n[Shot 4] At 00:11.000, 11-15s. Extreme close-up, eye level, fixed camera. Dim side lighting, weak candlelight, claustrophobic cold-warm contrast. The frame focuses on the glowing text on the parchment of the Book of Destiny (Subject 2). Vivian's (Subject 1) fair fingers, with shackle chains on her wrists, slowly caress and slide over the rough paper surface as if caressing a rare treasure. Vivian (S1) says in an extremely proud and intoxicated voice [English]Vivian becomes the Empire’s greatest channeler. All four beast-men submit to her.\n\noverall_soundscape:\nShot 1: Metal handcuffs clinking. Shot 2: Fabric rustling and heavy book dragging sound, mysterious magical humming. Shot 3: No SFX. Shot 4: Paper rustling sound.\n\nnon_diegetic_music:\nN/A", "original_prompt": "立体真人质感,电影级画面。皮肤表面呈现细腻肌理,细腻自然,顺滑 18 档阴影过渡,整体画面沉稳厚重。8k分辨率,极度逼真。(Remove thenoise and high-frequency details from theimage. Keep all the lines, colors, and brightness unchanged.) 严格保留 的原有形象、造型、着装细节(发型、五官、服装纹样、配饰完全不变), 保持人脸高度相似度。备注:no srt, no music\n视频整体保持人体比例正常,无畸形四肢,无夸张变形,身形差异自然写实,不头大身小,骨架结构合理\n镜号:35 | 场次:夜/医疗营-Vivian帐篷 | 0-3s | 全景,四分之一侧视 | 缓慢推进 | 暗淡侧光,微弱烛光,幽闭压抑的冷暖对比,狭小的帆布帐篷内,Vivian如同囚徒般孤独地坐在简陋的木床边缘,手腕上冰冷的铁铐反光刺眼,整个人散发出阴郁扭曲的气息。 | 台词 & 音效:Dialogue:无 SFX:金属手铐碰撞的叮当声\n镜号:36 | 场次:夜/医疗营-Vivian帐篷 | 3-7s | 中景,四分之一侧视 | 固定镜头 | 暗淡侧光,微弱烛光,幽闭压抑的冷暖对比,Vivian身体向右后方扭转,受制于手铐她只能双手并拢,艰难地探入干瘪的枕头下方,用力抽出一本表面破败皲裂的暗褐色 @命运之书 。从Vivian的主观视角向下看,厚重的 @命运之书 平摊在她的白色裙摆上,泛黄的羊皮纸表面突然亮起诡异的微弱金光,驱散了周围的黑暗。 | 台词 & 音效:Dialogue:无 SFX:布料摩擦与厚重书本拖拽声 ,神秘的魔法嗡鸣声| 备注:新道具出场,道具特写\n镜号:37 | 场次:夜/医疗营-Vivian帐篷 | 7-11s | 特写,仰视 | 缓慢推进 | 暗淡侧光,微弱烛光,幽闭压抑的冷暖对比,金色的光芒自下而上打在Vivian清纯的脸上,显得极为扭曲和狂热,她的眼珠倒映着金光,嘴角扯起极度病态得意的弧度,痴迷地读出文字。 | 台词 & 音效:Dialogue:[用迷恋且病态狂热的声音念诵]Vivian:\"Sylvia is cast aside by Cyrus and loses her inheritance.\" SFX:无 | 备注:反派精神状态异常\n镜号:38 | 场次:夜/医疗营-Vivian帐篷 | 11-15s | 极特写,平视 | 固定镜头 | 暗淡侧光,微弱烛光,幽闭压抑的冷暖对比,画面聚焦在羊皮纸上的发光字迹,Vivian白皙的手指(手腕带着手铐链条)在粗糙的纸面上缓慢抚摸滑过,如同抚摸稀世珍宝。 | 台词 & 音效:Dialogue:[用极度自豪且沉醉的声音念诵]Vivian:\"Vivian becomes the Empire’s greatest channeler. All four beast-men submit to her.\" SFX:纸张摩擦声 | 备注:物理状态延续,保持手铐和书的存在\n\n参考图片说明:\n图1: 角色: Vivian Hawthorne\n图2: 场景: 边境医疗营-简易帐篷\n图3: 道具: 命运之书\n影调:参考电影《美女与野兽》2017 真人版的影调风格,采用 ARRI Alexa Studio 相机拍摄,搭配安琴 Optimo Prime 镜头,具备层次丰富的景深过渡,柔润细腻的虚化质感,运用冷暖双色阶调色手法,兼具清冷暗蓝色阴影与华贵鎏金暖调高光处理(幽蓝灰黑与琥珀金棕 LUT),古堡室内烛光漫射柔光与乡村通透自然柔光结合,光影层次柔和且明暗对比分明,丝绒石材金属雕花肌理呈现精致复古的哑光温润质感,暗部沉稳内敛且高光温润不刺眼。", "reference_images": ["references/ec96cd680ee8fc5f7c1ac1170c2bd81162d4181bb8e0f6480f61685c547f4b3e.jpg", "references/6a2d172efe2416f120382d4a8861962d7bc21d2b6c88b6a9b4f56f2153b7948a.jpg", "references/4a120aa212e6f26330dae77b694eb2417dbdc3e1bc32456c21a6cda06cd3fea0.jpg"], "reference_sha256": ["ec96cd680ee8fc5f7c1ac1170c2bd81162d4181bb8e0f6480f61685c547f4b3e", "6a2d172efe2416f120382d4a8861962d7bc21d2b6c88b6a9b4f56f2153b7948a", "4a120aa212e6f26330dae77b694eb2417dbdc3e1bc32456c21a6cda06cd3fea0"], "workflow": "ref2va", "prompt_sha256": "2f0c469c6814191b538981ea6038f4cef4d6ae3a87f46f377821ddd5d1975f20", "original_prompt_sha256": "9aaa273859bd969ba0822b7deda6f343cdce145b4568857043e24c8ef0a471c2", "video_sha256": "68aba0432619803d5bb2dcd81780759c04e1d173d4ee1df297b0326ff2360014", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_067.json", "verification": "verification/sample_067.json"} {"file_name": "videos/sample_072.mp4", "sample_id": "#72", "prompt": "subject_definitions:\n (Zhong Ran): appearance and clothing from .\n (Lin Wanzhu): appearance and clothing from .\n (Master / Lin Hongqiang): appearance and clothing from .\nScene environment from .\nCharacter positioning and spatial layout from .\n\nsummary:\n[reference generation] Three characters — Zhong Ran, Lin Wanzhu, and the Master (Lin Hongqiang) — are locked in a single space following the positioning from . Zhong Ran is tied high on a stone pillar at center; Lin Wanzhu floats in the air directly opposite her; the Master stands on the ground to Lin Wanzhu's right side. Lin Wanzhu approaches with a dark-qi orb and threatens Zhong Ran, who then erupts a golden shockwave that staggers Lin Wanzhu while the Master continues his spell-casting stance.\n\nretention_analysis:\nfully_preserved: all five supplied images are explicitly bound — as Zhong Ran's identity, as Lin Wanzhu's identity, as the Master's identity, as the scene environment, and as the mandatory spatial layout. The source demands 100 % strict adherence to the positioning map and the presence of all three named characters; each is retained verbatim in the detailed description.\n\ndetailed_description:\n[Shot 1] 8K ultra-realistic live-action quality. The frame contains exactly three characters — (Zhong Ran), (Lin Wanzhu), and (Master / Lin Hongqiang) — with no extra figures. Their positions strictly follow : is tied to a stone pillar at a higher position in the center of the frame; floats suspended in the air directly opposite ; stands on the ground to the right side of 's body. All three occupy the same space with clear spatial relationships. is fully visible to 's right, never cropped or omitted. Cold-white translucent skin, no plastic feel, no AI artifacts, no noise or haze. Flat soft light combined with soft, bright, transparent air-quality illumination; clean and delicate light layers, weak shadows, soft edges, no over-sharpening, natural and smooth facial surfaces. 45-degree side light plus strong back-light rim light, overlaid with Tyndall volumetric light effect; dark atmospheric background; ultimate cinematic image quality. Dynamic camera movement with rich shot-scale variation — close-up, medium close-up, medium shot, and full shot intercut smoothly and naturally; continuous camera motion, never static. Compact rhythm, each single shot no longer than 5 seconds. Vivid expressions, full emotions, realistic micro-expressions, slight blinking, breathing texture, fluid body movements, natural live-action speaking performance. Delicate and premium in-frame lighting throughout. No subtitles, no extra elements.\n\n[Shot 2] (Lin Wanzhu) hovers with both feet suspended in the air, slowly floating forward while approaching. One hand's fingernails glow with eerie blue poisonous light as she slowly reaches toward 's cheek, who is tied high on the stone pillar. 's eyes are vicious and mean. (S1) says [Chinese]钟冉,你不是凭着这张脸勾引大师兄吗?我今天就毁了它! in a triumphant, sinister-smiling tone while pressing a black-qi magic orb against 's face.\n\n[Shot 3] Close-up of (Zhong Ran) tied to the stone pillar. She clenches her fists tightly, knuckles turning white, arms taut, nails digging into her palms. She trembles all over from pain yet makes no sound. Her eyes, like cold stars, stare fixedly at floating in the air.\n\n[Shot 4] suddenly clenches her fists and erupts a pure-golden furious shockwave. Golden light explodes outward from her fists as the center, intense light sweeping forward, instantly staggering , who is floating in the air, sending her stumbling and flying backward two steps, her form unsteady and her expression utterly shocked and terrified.\n\n[Shot 5] (Master / Lin Hongqiang) stands motionless in place. His robes and long hair are blown wildly backward by 's golden shockwave, fluttering and streaming, robes snapping in the wind. His expression grows ever more cold and ferocious. His hands continue forming seals as golden magical patterns surge and expand. No subtitles.\n\noverall_soundscape:\nOnly the scripted character dialogue is audible. No background music.\n\nnon_diegetic_music:\nN/A", "original_prompt": "重要规则:1. 仅使用参考音频的音色,完全忽略所有参考音频的内容,严格按照以下文本内容生成语音,不要添加任何额外内容;2. 所有参考音频仅用于音色克隆,不得作为背景音乐使用,不得播放音频中的任何原始内容;3. 禁止生成任何字幕,禁止添加任何背景音乐,视频仅保留角色对话语音:角色与场景分配:图一为钟冉,图二为林婉珠,图三为师父,图四为场景,图五为站位图。林婉珠的音色参考音频一。\n【全局人物与站位强制锁定指令】\n1. 画面中必须且只能出现三个人物:钟冉、林婉珠、师父(林宏强),严禁缺少任何一人,严禁出现多余人物!\n2. 站位必须100%严格遵循图五(站位图)的相对关系:钟冉在中间被绑在石柱子上(处于较高位置),林婉珠悬空飘浮在钟冉的正对面,师父(林宏强)站在地面上位于林婉珠身体的右侧。三人必须在同一空间内呈现清晰的方位关系。\n3. 师父(林宏强)必须出现在画面中(林婉珠右侧),绝不能被裁切或遗漏!\n\n画面质感与风格:\n8K 超写实实拍质感,角色冷白皮肌肤透亮通透,无塑料感、无 AI 痕迹、无噪点无灰雾;平面柔光 + 柔和明亮通透空气感光影,干净细腻光影层次,弱阴影、边缘柔和、无过度锐化,面部平整自然;45 度侧光 + 强逆光轮廓光,叠加丁达尔体积光效应,暗调氛围感背景,极致电影感画质;镜头动态运镜、景别变化丰富,特写 / 近景 / 中景 / 全景多视角穿插切换流畅自然,镜头持续运动不呆板;视频节奏紧凑,单镜头时长不超 5 秒,人物表情生动情绪饱满到位,微表情真实自然,轻微眨眼、呼吸肌理、肢体动作流畅不僵硬,真人生活化自然说话演绎,画面内光影细腻高级,全程无字幕、无多余元素干扰。\n\n画面呈现:\n▲林婉珠双脚悬空,在空中缓缓飘浮逼近,单手指甲泛幽蓝毒光,慢慢伸向被绑在石柱高处的钟冉脸颊,眼神恶毒刻薄。\n【配音指令:林婉珠 | 音色必须100%完美克隆参考音频 | 状态:得意狞笑 | 强制台词:钟冉,你不是凭着这张脸勾引大师兄吗?我今天就毁了它!】\n林婉珠(在空中飘浮着,得意狞笑,单手将黑气法球按在钟冉脸部):钟冉,你不是凭着这张脸勾引大师兄吗?我今天就毁了它!\n钟冉被绑在石柱上死死攥拳,指节发白,手臂绷紧,指甲掐入掌心,痛得浑身发抖却一声不吭,特写眼眸如寒星死死瞪着飘在空中的林婉珠。\n\n下一秒 —— 她捏拳猛然爆发【纯正金色暴怒冲击波】,金光以拳头为中心轰然炸开,强光席卷前方,瞬间震得飘在空中的林婉珠在半空中踉跄倒飞出两步,身形不稳、惊骇至极!\n\n▲林宏强原地不动,袍与长发被钟冉的金色冲击波疯狂往后吹得飘逸飞扬,衣袍猎猎作响,神色越发阴冷狰狞,双手继续结印,金色法术纹路暴涨。\n【全局文本及语音控制指令】\n1. 语音内容必须与上述台词完全一字不差,绝对不能出现漏字、错字、改字或多余的语气词!\n2. 严禁AI根据参考音频的内容擅自篡改、拼接或覆盖提示词中规定的台词!", "reference_images": ["references/2c8d7d122706351f4ef297319382d3e30c398bc37db2825a076fe3fba641fdc4.jpg", "references/9300df50d19768cbd4bab0a932444c13ec8fd83cd6420bf05833e60751437c6f.jpg", "references/e65c9c1a22a28fdb7b511d315f587bed1ec54f5a9d56d0647085e71a87e92311.jpg", "references/8b579b99f319ba9fe3ca5e111a721ada5e926815ecb23e70b615f8ff2a0d2021.jpg", "references/110c3dc8d84c8e0e311f1d71ee301e19a90f76dc0869c4cdb60aefc7c56d2300.jpg"], "reference_sha256": ["2c8d7d122706351f4ef297319382d3e30c398bc37db2825a076fe3fba641fdc4", "9300df50d19768cbd4bab0a932444c13ec8fd83cd6420bf05833e60751437c6f", "e65c9c1a22a28fdb7b511d315f587bed1ec54f5a9d56d0647085e71a87e92311", "8b579b99f319ba9fe3ca5e111a721ada5e926815ecb23e70b615f8ff2a0d2021", "110c3dc8d84c8e0e311f1d71ee301e19a90f76dc0869c4cdb60aefc7c56d2300"], "workflow": "ref2va", "prompt_sha256": "21abc9e3ddaeb61fde164fc975a5ccab1f68968ba226f300164b598a5c514e4a", "original_prompt_sha256": "757e34aaf6b9e0b8216af9e51ed43f2df3493a4631f03f592ac74818a9baf8f2", "video_sha256": "23acc44bfbf6a25e7a20c0c91582556d6533595be1c3fce8e1032a45718409c6", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_072.json", "verification": "verification/sample_072.json"} {"file_name": "videos/sample_073.mp4", "sample_id": "#73", "prompt": "subject_definitions:\n is a plus-size model with an hourglass figure, full and round hips, a natural waistline, and a height of 165-168cm, with makeup and styling presenting an energetic sports academy style, appearance from .\n\nsummary:\n[reference generation] A women's clothing e-commerce product detail page marketing video to promote purchase conversion. wears the exact clothing from (color, style, and details must not change or have any color difference) against the background from .\n\nretention_analysis:\nfully_preserved: must retain the appearance, body type, and styling from ; the clothing must strictly match the color, style, and details from without any alteration or color shift; the background must directly use .\n\ndetailed_description:\n[Shot 1] [0-4s] Full body long shot. confidently walks in from the window and transitions into a natural turn. The tight elasticity of the knitted fabric expands and contracts with the steps, the high-waisted large pleated skirt hem moves rhythmically with the body's center of gravity, showing the fluffiness and drape of the fabric, camera smoothly pushes in. [Shot 2] At 00:04.000 [4-8s] Medium shot lateral translation. stands sideways to show the waist-hip curves, then naturally looks back and lightly touches hair. Light and shadow outline the square collar tailoring and chest patch details, showing the delicate matte texture of the fabric, 's movements are smooth and uninterrupted. [Shot 3] At 00:08.000 [8-12s] Close-up to full body. continues walking, camera first focuses on the upper body to show the wrapping and fit of the clothing, then pulls back to full body. lightly sways the body to show the rebound elasticity of the pleats in dynamic state, the plus-size figure proportions are coordinated, natural ending. The clothing color is based on , the background lighting serves to present the true color of the clothing, the tone of must not cause the clothing to turn gray, darken, or shift color. 's hands are prohibited from being in pockets throughout. The scene complies with physical laws, no clipping or unreasonable scenes such as the model standing on tables or chairs; strictly no nipple protrusion or areola showing through, no naked pornographic content.\n\noverall_soundscape:\nUnspecified.\n\nnon_diegetic_music:\nPure instrumental music, no vocals, bright and light electronic pop style.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。图一模特为大码沙漏身材,臀部丰满圆润,腰线自然,165-168cm,妆造呈现活力运动学院风。服装严格使用图二,颜色、版型及细节不允许有任何改变或色差;背景直接使用图三。符合物理规律,禁止穿模或模特站立桌椅等不合理画面;严禁乳头凸起或乳晕透出,无裸露色情内容,符合平台规范。模特全程双手禁插兜。服装颜色以图二为准,背景光照服务于服装真实颜色呈现,禁止因图三色调导致服装变灰、暗化或偏色。 [0-4s] 全身远景。模特从窗边自信步入并衔接一个自然转身。针织面料的紧致弹性随步伐收放,高腰大百褶裙摆随身体重心移动而律动,展现面料的蓬度与垂坠感,运镜平稳推近。 [4-8s] 中景横移。模特侧身站立展示腰臀曲线,随后自然回眸并轻撩发丝。光影勾勒出方领剪裁与胸前贴标细节,展示面料细腻的哑光纹理,模特动作流畅不间断。 [8-12s] 近景转全身。模特持续行进,镜头先聚焦上半身展现服装的包裹感与贴合度,随后拉远至全身。模特轻摆身体展示褶皱在动态下的回弹力,大码身材比例协调,自然收尾。 配乐:纯音乐,无人声,明快轻盈的电子流行风格。\\n\\n素材说明:图1=模特参考(asset-20260224202454-g6xs6);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/b1f1f77fa1ebc09ba350eab9e2e5dfcdaf181b16b11038f325a7bab1c5552909.jpg", "references/a8ac62167f082305fcfb54e79149a01dc06c9a7142798e15c78658c9a6c335ed.jpg", "references/a37be8fff07665b547ccbb86cb56c1e622ac84c33ffb2dc9b5abe8d6349eb60d.jpg"], "reference_sha256": ["b1f1f77fa1ebc09ba350eab9e2e5dfcdaf181b16b11038f325a7bab1c5552909", "a8ac62167f082305fcfb54e79149a01dc06c9a7142798e15c78658c9a6c335ed", "a37be8fff07665b547ccbb86cb56c1e622ac84c33ffb2dc9b5abe8d6349eb60d"], "workflow": "ref2va", "prompt_sha256": "bd9e98ba8a24449b6bd14dae59c2d4b6e8b6d5df0da12263402215e8d4b7b9b0", "original_prompt_sha256": "556b110bc846dfe34fb52b47811a8f1990ab448725c98c2d5b394176575e4666", "video_sha256": "c9280471a17ebfa6bb556b0dbb50919acc1bf267004766a9470bad5757cc7876", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_073.json", "verification": "verification/sample_073.json"} {"file_name": "videos/sample_074.mp4", "sample_id": "#74", "prompt": "subject_definitions:\n is a character whose appearance is defined by , identified as Tang San. is a character whose appearance is defined by , identified as Xiao Wu.\n\nsummary:\n[reference generation] A cinematic 3D Chinese animation epic scene in fantasy realistic style with Unreal Engine 5 hyper-realistic rendering, strictly with no subtitles and no watermarks. The scene is a medium shot where Xiao Wu () pushes the target character (source's undefined \"A\", most plausibly Tang San) down onto the bed to stop Tang San () from continuing to release hidden weapons. The scene environment references . The ending standing position and action of the characters reference .\n\nretention_analysis:\nScene environment from is fully preserved as the setting reference. Tang San's identity and appearance from are fully preserved. Xiao Wu's identity and appearance from are fully preserved. The ending pose and standing position of the characters from are fully preserved as the final action reference. The starting action and standing position reference is partially preserved as the source does not specify which image to reference for the starting position.\n\ndetailed_description:\n[Shot 1] A cinematic 3D Chinese animation epic scene in fantasy realistic style with Unreal Engine 5 hyper-realistic rendering, 8K ultra-high detail (Note: Source requires 8K; actual run is 1536x672 — resolution conflict, retaining source requirement), and global volumetric lighting, strictly containing no subtitles and no watermarks. The scene environment is based on . The shot is a medium shot featuring (Xiao Wu, appearance from ) and (Tang San, appearance from ). The composition uses asymmetrical framing to express the tension of the action storyboarding, with high-contrast lighting, strong light and shadow layers, and high-saturation vibrant premium colors that are not eye-straining. The entire frame is full of tension and shock, presenting large-scale epic special effects and camera movement. (Xiao Wu) pushes the target character (source's undefined \"A\", most plausibly Tang San) down onto the bed to stop (Tang San) from continuing to release hidden weapons. The starting action and standing position are based on an unspecified reference. The ending standing position and action of the characters match . The spatial logic is coherent with no sense of incongruity, and the environment transitions are seamless.\n\noverall_soundscape:\nNo specific audio requirements are provided in the source.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【生成需求】制作一个无字幕、无水印的3D顶级游戏CG动画片段,具有院线级奇幻大片质感。整体为3D国漫史诗虐心唯美风格,电影级色调,超写实3D渲染,虚幻引擎5极致画质。画面要求:高对比度光影,明暗层次强烈,高饱和、鲜艳高级但不辣眼睛的色调。构图根据情绪和动作设计,多用不对称构图以表现动作分镜的张力,整个画面需充满张力和震撼力。需要呈现大场面史诗级特效和运镜。镜头顺序需按指定呈现,并细化角色的姿势、表情、台词动作及环境衔接,确保空间逻辑连贯,无违和感。 电影级动画镜头,奇幻写实风格,UE5超写实渲染,8K超高细节,全局体积光。场景参考第二张图片,中景,小舞第四张图片为了阻止唐三第三张图片继续释放暗器,将A推倒坐到了床上,开始的动作站位参考,结束时的站位动作参考第一张图片", "reference_images": ["references/f91f7b19702acb4b0eafc90be806f5f44fb92297b413a27c8fc95ef18540008a.jpg", "references/140d0fd7c9e38e9496b024018b06f352aac60870f2ca5b36280fba6748e9c548.jpg", "references/1767b2ae87ab6eebe3f6a054cf8ee00ca64d94fab4767269f42ddfad0fcb60f1.jpg", "references/150219636022b372878842b18b66fc1e7127e44ad3c5830ede26c573da701083.jpg"], "reference_sha256": ["f91f7b19702acb4b0eafc90be806f5f44fb92297b413a27c8fc95ef18540008a", "140d0fd7c9e38e9496b024018b06f352aac60870f2ca5b36280fba6748e9c548", "1767b2ae87ab6eebe3f6a054cf8ee00ca64d94fab4767269f42ddfad0fcb60f1", "150219636022b372878842b18b66fc1e7127e44ad3c5830ede26c573da701083"], "workflow": "ref2va", "prompt_sha256": "e2c887430a39675759918ed1a42ee1cab7da1dc2e9a04fbace95ec9ce705e18a", "original_prompt_sha256": "7f05c171f4caf4e51b9e3542708b882cd2d325297cc6d5fb3f9be806ad1b52ad", "video_sha256": "c6f730333a2ca769302a8a597fb7945d646d3ff6bdf5134fedf394b3a471e300", "task_adapted": false, "width": 1536, "height": 672, "fps": 24, "num_frames": 107, "duration_seconds": 4.458333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_074.json", "verification": "verification/sample_074.json"} {"file_name": "videos/sample_075.mp4", "sample_id": "#75", "prompt": "integrated_multimodal_description:\n[Global Style & Technical Requirements] 8K ultra-realistic live-action cinematic texture, 60fps, ancient Chinese royal high-couture political drama. Clear visibility of facial pores, skin texture, hair layers, gold-woven fabric texture, jewelry cut gloss, metal hairpin texture, fox fur flow, and python robe dark patterns. Cinematic large aperture shallow depth of field, unified scene lighting logic, blending soft light, top light, side backlight, rim light, and 10%-35% Tyndall effect. Dialogue switches to close-ups, character confrontation and emotional outbursts use close-ups/extreme close-ups, key choreography uses wide/medium shots, important actions, props, design drafts, fabrics, and dresses use extreme close-ups. The camera has an emotional breathing feel, shot size switching follows the cool drama rhythm, maintaining a stable 180-degree axis. Ultra-clear sharpened texture, high-contrast ancient style color grading. Ultra-realistic live-action texture, clear pores, skin texture matching the corresponding age and identity, removing AI feel, enhancing real skin texture and high-end makeup. Characters' costumes, makeup, and props remain consistent throughout without clipping, strict royal etiquette, no cheap or discordant elements. No subtitles, no background music.\n\n[Jiangnan Aesthetics & Narrative] Jiangnan stable aesthetics, warm gold highlights + gratitude of the people contrast, high-healing national color tone, large light ratio prosperous era contrast. Narrative emotions: gratitude, reverence, gratification, stability. Soft light + top light mainly, 30% Tyndall effect, warm light fills the stable moments. Warm gold envelops to strengthen national peace, plain white commoners and vermilion official uniforms as iconic colors, warm light and shadow highlight the princess's benevolence.\n\n[Shot 1] Duration: 3.5 seconds. Scene 48, Shot 14. Frontal face close-up (above shoulders). Camera: shoulder-mounted handheld narrative breathing feel + cinematic gratification breathing shot + fixed, extremely slight gentle floating. Scene: Jiangnan small town street. Character: Shen Ruowei. Prop: horse-drawn carriage. Composition & Lighting: centered composition, character's face in the center of the frame; warm gold soft light, 30% Tyndall effect, shallow depth of field, blurred background. Background details: blurred scenery outside the carriage window. Action: Shen Ruowei looks out the carriage window, lips slightly moving, inner monologue. Texture: Shen Ruowei's skin is delicate and glossy, pores clearly visible, natural eye reflection. Layered scene: Foreground: none; Midground: Shen Ruowei's face; Background: blurred scenery outside the window. VFX: warm gold highlight effect. Sound effect: slight horse-drawn carriage driving sound. Constraints: precise facial features without collapse, clear facial pores and skin texture, distinct hair strands, realistic clothing texture, no clipping or distortion, stable face without deformation, normal human structure, costume and appearance completely consistent with previous text, no subtitles, no background music. Shen Ruowei (S1) says in an off-screen voiceover [Chinese]看着他们能重返家园,这半月奔波便值了。\n\n[Shot 2] At 00:03.500, Duration: 3.5 seconds. Scene 48, Shot 15. Side face close-up (above shoulders). Camera: shoulder-mounted handheld narrative breathing feel + cinematic nostalgic breathing shot + fixed, extremely slight gentle floating. Scene: Jiangnan small town street. Character: Shen Ruowei. Prop: horse-drawn carriage. Composition & Lighting: centered composition, character's eyes in the center of the frame; warm gold soft light, 30% Tyndall effect, shallow depth of field, blurred background. Background details: blurred scenery outside the carriage window. Action: Shen Ruowei's eyes are gentle, continuing inner monologue, corners of the mouth slightly raised. Texture: Shen Ruowei's sclera is clear, pupil reflection is clear, eyelashes are distinct. Layered scene: Foreground: none; Midground: Shen Ruowei's eyes; Background: blurred scenery outside the window. VFX: eye reflection effect. Sound effect: none, no sound effect beyond the general environment field. Constraints: precise facial features without collapse, clear facial pores and skin texture, distinct hair strands, realistic clothing texture, no clipping or distortion, stable face without deformation, normal human structure, costume and appearance completely consistent with previous text, no subtitles, no background music. Shen Ruowei (S1) says in an off-screen voiceover [Chinese]娘亲,这就是您生长的地方,真的很美。\n\noverall_soundscape:\nEnvironmental sound field including footsteps, fabric friction, jade pendant collision, palace gate opening and closing, horse hooves, armor collision, and wind and snow sounds as per key plot points. No redundant background music.\n\nnon_diegetic_music:\nN/A", "original_prompt": "8K 超写实真人电影质感,60 帧,古风王侯高定权谋爽剧,脸部毛孔、皮肤纹理、发丝层次、织金面料肌理、珠宝切割光泽、金属钗环质感、狐裘毛流感、蟒袍暗纹清晰可见;无冗余背景 BGM,关键剧情配极简氛围感音效,无遮挡字幕,仅保留台词与环境声场(脚步声、布料摩擦、玉佩碰撞、宫门开合、马蹄声、甲胄碰撞、风雪声);电影级大光圈浅景深,场景化光影逻辑统一,融合柔光、顶光、侧逆光、轮廓光与 10%-35% 丁达尔光效;对白切换近景,人物对峙与情绪爆发用近景 / 特写,关键调度采用全景 / 中景,重要动作、道具、设计稿、面料、礼服使用特写,镜头自带情绪呼吸感,景别切换遵循爽剧节奏,保持 180° 轴线稳定;超清锐化质感,高对比古风色调,虐戏低饱和冷青灰、逆袭戏冷白高锐、高光戏暖金鎏光、宫廷侯府戏华贵金红、北疆战场凛冽银白、祭墓戏清冷素白,画面暗角自然过渡,强化破碎感、爽感、高级感与宿命感;超写实真人质感,毛孔清晰,具有符合相应年龄段与身份的肌理感,去掉人物 AI 感,增强皮肤真实质感与妆面高级感,高清写实人物质感,8K 分辨率;人物服化道全程统一不穿模,王侯礼制严谨,无廉价感与违和元素。\n江南安定美学,暖金高光 + 万民感恩对冲,高治愈家国色调,大光比盛世反差;叙事情绪为感激、崇敬、欣慰、安定;调度以灾区破败与重建安定对比、百姓跪送与父女从容对比;柔光 + 顶光为主,丁达尔光效 30%,暖光铺满安定时刻;运镜百姓跪送全景、馒头干粮特写、女主扶老慢推、地方官恭敬怼脸;暖金笼罩强化国泰民安,素白百姓、朱红官服为标志性色彩,光影温暖凸显郡主仁心。\n【时长】3.5 秒\n【场次】第 48 场\n【镜号】14\n【景别】正面脸部近景(肩部以上)\n【摄法】肩扛手持叙事呼吸感 + 电影级欣慰呼吸镜头 + 固定,极微小幅度舒缓浮动\n【场景】江南小镇街道图1\n【出镜人物】沈若微图2音频1\n【道具】马车图3\n【画面描述】\n构图光影:居中构图,人物面部位于画面中央;暖金色柔光,30% 丁达尔光效,浅景深,背景虚化\n场景细节:背景是模糊的马车窗外风景\n画面动作:沈若微眼神望向马车窗外,嘴唇微动,内心独白\n画面质感:沈若微皮肤细腻有光泽,毛孔清晰可见,眼神反光自然\n分层场景:前景层:无;中景层:沈若微的面部;后景层:模糊的窗外风景\n【对白】沈若微图2音频1:(内心独白)看着他们能重返家园,这半月奔波便值了。\n【音效】轻微马车行驶声\n【特效】暖金高光特效\n【约束】人物五官精准不崩坏,人物五官与参考保持一致,脸部毛孔皮肤纹理清晰,发丝根根分明,衣物纹理写实,无穿模无扭曲,人物面部稳定不变形,人体结构正常,服装样貌与前文完全一致,无字幕,无背景音乐\n【时长】3.5 秒\n【场次】第 48 场\n【镜号】15\n【景别】侧面脸部近景(肩部以上)\n【摄法】肩扛手持叙事呼吸感 + 电影级怀念呼吸镜头 + 固定,极微小幅度舒缓浮动\n【场景】江南小镇街道图1\n【出镜人物】沈若微图2音频1\n【道具】马车图3\n【画面描述】\n构图光影:居中构图,人物眼部位于画面中央;暖金色柔光,30% 丁达尔光效,浅景深,背景虚化\n场景细节:背景是模糊的马车窗外风景\n画面动作:沈若微眼神温柔,继续内心独白,嘴角微微上扬\n画面质感:沈若微眼白清澈,瞳孔反光清晰,睫毛根根分明\n分层场景:前景层:无;中景层:沈若微的眼部;后景层:模糊的窗外风景\n【对白】沈若微图2音频1:(内心独白)娘亲,这就是您生长的地方,真的很美。\n【音效】无\n【特效】眼神反光特效\n【约束】人物五官精准不崩坏,人物五官与参考保持一致,脸部毛孔皮肤纹理清晰,发丝根根分明,衣物纹理写实,无穿模无扭曲,人物面部稳定不变形,人体结构正常,服装样貌与前文完全一致,无字幕,无背景音乐", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "c2f41a6414fbbb5669b16b7e750174d48decf306787022347ba030f9d46f586c", "original_prompt_sha256": "6c3deb0cf0a52d0614a130293f8ae83d330fedb6eada7d0df3480ddc4aa4d6a3", "video_sha256": "e7ee8740733c21b5b87613c81cf3ad3eb7b7a46b8f92d33ca5c3b6001a611e62", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 175, "duration_seconds": 7.291666666666667, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_075.json", "verification": "verification/sample_075.json"} {"file_name": "videos/sample_076.mp4", "sample_id": "#76", "prompt": "integrated_multimodal_description:\n[Shot 1] 0.0-4.8 seconds. medium close-up, eye-level, slight push-in. In a bright hospital ward during the day, He Yutang (S1) reaches out and gently places her hand over Lu Jinglin's hand. She looks at him with a playful, coquettish helplessness, teasing him about staying in the hospital. He Yutang (S1) says [Chinese]景林,你还想让我在这里住多久呀?到处都是消毒水的味道,我都觉得自己快要发霉了。 The hand-covering gesture establishes a comforting dynamic. [Shot 2] At 00:04.800, hard cut. medium close-up, eye-level, slow push-in. He Yutang's (S1) eyes soften, showing a smile mixed with longing as she mentions her daughter. Her gaze briefly pauses in the distance before returning to Lu Jinglin. He Yutang (S1) says [Chinese]我想家了,想棠棠。这么久没见,棠棠肯定都不认得妈妈了。 [Shot 3] At 00:09.200, hard cut. medium close-up, eye-level, static camera. Lu Jinglin's (S2) expression immediately softens, as if correcting a baseless worry. He looks back at He Yutang with gentle, doting eyes, keeping his hand under hers without pulling away. Lu Jinglin (S2) says [Chinese]胡说,棠棠天天在家里抱着你的照片看,怎么会不认得? No subtitles or any visible text appear on the screen throughout the video.\n\noverall_soundscape:\nLight and relaxed human voices, the soft sound of a hand gently touching another hand, soft and gentle human voices, and the ambient daytime background noise of a hospital ward.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【场景分析】\n场景:(图1/日内明亮)\n承接:上一分镜医生确认无碍后,紧张感开始下降\n过渡:以贺雨棠的玩笑与思家情绪转入家庭温度\n人物表演底层逻辑总纲:【贺雨棠处于即将回家的幸福期待;陆景林处于顺势哄她放松的宠溺状态】\n【分镜承接】\n前一分镜末尾画面:贺雨棠准备伸手覆上陆景林手背\n本分镜开场画面:她轻轻把手覆上去,带笑开口\n承接逻辑:动作与情绪延续\n**【镜头1】0.0-4.8秒\n转场: 无\n镜头:近景/平视/轻微推近\n画面描述: 贺雨棠伸出手,轻轻覆在陆景林手背上,语气带着撒娇式的无奈。\n核心动作: 她打趣自己在医院快发霉。\n对手戏设计: 手部覆盖动作建立安抚关系,视线看向陆景林。\n人物底层逻辑: 她想用轻松口吻表达“我想回家”。\n主体动作: 贺雨棠失笑地说:“景林,你还想让我在这里住多久呀?到处都是消毒水的味道,我都觉得自己快要发霉了。”\n音效:轻松人声,手掌轻触声\n【镜头2】4.8-9.2秒\n转场: 硬切\n镜头:近景/平视/缓慢推近\n画面描述: 她眼神软下来,提到女儿时笑里带着想念。\n核心动作:说出想家、想棠棠。\n对手戏设计:目光短暂停向远处,再回到陆景林。\n人物底层逻辑: 康复后的第一愿望是回到完整的家。\n主体动作: 贺雨棠幸福地说:“我想家了,想棠棠。这么久没见,棠棠肯定都不认得妈妈了。”\n音效: 柔和人声\n【镜头3】**9.2-13.4秒\n转场: 硬切\n镜头:近景/平视/固定镜头\n画面描述: 陆景林神色立刻柔下来,像在纠正她一个根本不成立的担心。\n核心动作: 他宠溺反驳。\n对手戏设计:目光温柔地回望贺雨棠,手没有抽开。\n人物底层逻辑: 他知道女儿对母亲的思念,也想让她彻底安心。\n主体动作: 陆景林宠溺地说:“胡说,棠棠天天在家里抱着你的照片看,怎么会不认得?”\n音效: 温柔人声,病房日间底噪\n资产对照表(视频中涉及的角色、场景与物品如下(保持一致性)):\n视频场景对照表: 医院-病房=【图1/日内明亮】\n视频角色对照表: 贺雨棠=【图2】;陆景林=【图3】\n视频道具对照表: 无\n视频中不要出现任何字幕。", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "8678856a4ca3bc2191d901094977035f7b2b5357054723f8ccb6b8bbeb5e1e61", "original_prompt_sha256": "db594a16dcd598b8954070a0a7313d4710332b20ea8f314d2fc255dac4655ec9", "video_sha256": "4b8ddca17707e6bd2e0a13ae09e77357e4aa5a57252768626075064b14c789f2", "task_adapted": true, "width": 768, "height": 1024, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_076.json", "verification": "verification/sample_076.json"} {"file_name": "videos/sample_077.mp4", "sample_id": "#77", "prompt": "integrated_multimodal_description:\n[Shot 1] (7 seconds) Fixed realistic style, real portrait photography, vertical portrait composition, desk lamp lighting. Character A and Character B maintain fully consistent appearance, clothing, hairstyle, and makeup throughout. Indoor bedside and desk scene. Character A and Character B sit separately. Over-the-shoulder shot over Character B's shoulder, framing Character A's upper body. Character A walks to the bedside and sits down, hands still intertwined and placed on the knees, looks at Character B with a hesitant expression, lips moving slightly before speaking. Character A (S1) says [Chinese]想跟你商量件事。 in a tentative and hesitant tone. The mattress springs make a slight sinking sound. HD cinematic quality, rich details, natural and soft lighting, stable facial features, clear facial features, normal body structure, natural and smooth movements, no stuttering, flickering, or clipping. No subtitles.\n\n[Shot 2] (4 seconds) Fixed realistic style, real portrait photography, vertical portrait composition, desk lamp lighting. Character A and Character B maintain fully consistent appearance, clothing, hairstyle, and makeup throughout. Indoor bedside and desk scene. Character A and Character B sit separately. medium close-up over-the-shoulder shot over Character A's shoulder, framing Character B. Character B puts down the pen in hand, turns around, and quietly looks at Character A with calm eyes. The pen makes a light touching sound on the desk surface. HD cinematic quality, rich details, natural and soft lighting, stable facial features, clear facial features, normal body structure, natural and smooth movements, no stuttering, flickering, or clipping. No subtitles.\n\noverall_soundscape:\nReal environmental sound effects and character action sound effects are retained, including the slight spring sound of the mattress sinking and the light sound of the pen touching the desk. No background music is present.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【风格定位】固定写实风格,真实人像摄影,竖屏构图(Portrait Mode),【台灯光影】,\n【角色锚定】图1图2,全程保持角色外观、服装、发型、妆容完全一致\n【场景环境】图3\n【人物站位】图1坐在绿色方框位置,图2坐在黄色方框位置。\n【其他要求】无字幕、无BGM音乐,保留真实环境音效与人物动作音效\n【时长配置】总时长7秒\n【分镜时序(时间轴拆解)】\n镜头15:过肩镜头,过图2的肩拍图1,镜头框住图1上半身,图1走到床边坐下,双手仍绞在一起放在膝盖上,看了图2一眼欲言又止,嘴唇动了动才开口\n【音效】床垫轻微下陷的弹簧声\n【对白/口型】图1音频1+试探、犹豫:\"想跟你商量件事。\"\n【画质与强约束】高清电影质感,细节丰富,光影自然柔和;人物面部稳定不变形、五官清晰、人体结构正常、动作自然流畅不僵硬、画面无卡顿无闪烁无穿模\n\n\n【风格定位】固定写实风格,真实人像摄影,竖屏构图(Portrait Mode),【台灯光影】\n【角色锚定】图2全程保持角色外观、服装、发型、妆容完全一致\n【场景环境】图3\n【人物站位】图1坐在绿色方框位置图2坐在黄色方框位置\n【其他要求】无字幕、无BGM音乐,保留真实环境音效与人物动作音效\n【时长配置】总时长4秒\n【分镜时序(时间轴拆解)】\n镜头14:近景,过肩镜头,过图1的肩拍图2,图2放下手中的笔,转过身来,安静地看图1目光平静\n【音效】笔放下轻触桌面的声音\n【画质与强约束】高清电影质感,细节丰富,光影自然柔和;人物面部稳定不变形、五官清晰、人体结构正常、动作自然流畅不僵硬、画面无卡顿无闪烁无穿模", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "8187e7b694631100d6e0177b9c393af11420a248bace57c5d6c997e98119371d", "original_prompt_sha256": "8dd807149add58686289e91e95d41c4bb0dca51638e59edae9dd7810c48f0de3", "video_sha256": "0c94f272d3bf891739e72a15d5370e30b6e75b7ee5da39c04ff41738d82aa8c2", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 277, "duration_seconds": 11.541666666666666, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_077.json", "verification": "verification/sample_077.json"} {"file_name": "videos/sample_078.mp4", "sample_id": "#78", "prompt": "subject_definitions:\n: The security guard, appearance and clothing from .\n: Wen Wuchen (the third young master), appearance and clothing from .\n: Wen Zaozao, appearance and clothing from .\n\nsummary:\n[reference generation] A Hollywood cinematic live-action style scene set outside the Wen family villa gate during the day. The security guard looks at Wen Wuchen with an obsequious expression and speaks to him. The composition maintains a certain distance between characters, referencing for positioning, with clear foreground, midground, and background depth.\n\nretention_analysis:\n (security guard) is fully preserved as the main speaking character, retaining appearance from .\n (Wen Wuchen) is fully preserved as the listener, retaining appearance from .\n (Wen Zaozao) is partially preserved as a referenced character, retaining appearance from , though not explicitly acting in the described shot.\n (Wen family villa gate) is fully preserved as the scene location.\n is fully preserved as the composition and character positioning reference.\n\ndetailed_description:\n[Shot 1] Hollywood cinematic live-action style, cinematic lens. Add camera movement to each shot per the source request; note the shot also specifies a side fixed lens (possible conflict). Smooth, natural action without stutter; camera movement transitions naturally. Day exterior. Cinematic atmospheric lighting with corresponding Tyndall light effects, background and character face lighting. The scene is set at the Wen family villa gate (). The composition is clean and high-end, with distinct progressive relationships in the foreground, midground, and background, exhibiting extreme emotional tension and physical detail realism. Character positioning references , maintaining a certain distance between the characters. Over-the-shoulder shot, extreme close-up | Side fixed lens: 135mm, aperture: 1.2. (S1) stares obsequiously at and says [Chinese]谢谢三少,帮我看住这娘们。哟,三少,您脸是怎么了?是不是这娘们打的?都是我管教不好,我把她带回家,一定好好收拾她. The security guard's performance shows Hollywood-level emotional tension, with detailed emotional fluctuations. Clear details of character skin, clothing, and hands, highlights do not blow out. Avoid continuity errors, crossing the line, regular errors, clipping, CG feel, plastic texture, cheap feel, blur, frame drops, low definition. No subtitles.\n\noverall_soundscape:\nDay exterior ambience at a villa gate.\n\nnon_diegetic_music:\nN/A", "original_prompt": "制作风格为好莱坞电影级实拍风格,电影镜头,每个镜头添加运镜,运镜衔接自然,动作顺滑无卡顿,人物皮肤、衣物、手部细节清晰,亮部不溢出,画面必须避免:穿帮、跃轴、规律性错误、穿模、CG感、塑料质感、廉价感、模糊、画面掉帧、不高清,无字幕。好莱坞级别演员表演情绪,配角以及主演的情绪起伏细节,演员的表演张力。电影级对应场景氛围光,相应的丁达尔光效,日外,背景和人物面光\n人物站位参考: @图片5 ,保持人物之间有一定距离\n要求:画面干净又高级;画面里前中远景递进关系分明。极具情绪张力与物理细节真实感。\n过肩镜头,大近景 | 侧面固定镜头:135,光圈:1.2, @保安 殷勤的盯着 @温无尘 说:“谢谢三少,帮我看住这娘们。哟,三少,您脸是怎么了?是不是这娘们打的?都是我管教不好,我把她带回家,一定好好收拾她”\n\n参考图片说明:\n图1: 角色: 保安\n图2: 角色: 温无尘\n图3: 角色: 温早早\n图4: 场景: 温家别墅大门口", "reference_images": ["references/25a11b4d4c25b85f971cc2ba260cf5dff9320f0645d73be9948747f263a380f5.jpg", "references/76b02f65513a6311e6764bff9c8528722e5dcef64f96041128aaf25cb6f71c9c.jpg", "references/e71bc892e19cbe99753dc891004ade9c0de14df872e57613fb43df3c223198d6.jpg", "references/a002d2366bc611a3a7d34eeb857ff71a655776c41c9a9f53da79842f61b2b98a.jpg", "references/0b36c6e9d7dc06df45337b71aa57575daab061a92b894e9b168962a58ed2d9c2.jpg"], "reference_sha256": ["25a11b4d4c25b85f971cc2ba260cf5dff9320f0645d73be9948747f263a380f5", "76b02f65513a6311e6764bff9c8528722e5dcef64f96041128aaf25cb6f71c9c", "e71bc892e19cbe99753dc891004ade9c0de14df872e57613fb43df3c223198d6", "a002d2366bc611a3a7d34eeb857ff71a655776c41c9a9f53da79842f61b2b98a", "0b36c6e9d7dc06df45337b71aa57575daab061a92b894e9b168962a58ed2d9c2"], "workflow": "ref2va", "prompt_sha256": "2c06a942836552d4e71773fe0fd9b71d01137637ba6ac1386ad3069df12c2397", "original_prompt_sha256": "60351a87208ae3a085e4a87ba6d6995b5545fa7a2677485869508666227c663d", "video_sha256": "3970f269fc3cbdbd43515a5b860d924fbb07d48524aa25480aaf68974dd81fcb", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 277, "duration_seconds": 11.541666666666666, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_078.json", "verification": "verification/sample_078.json"} {"file_name": "videos/sample_079.mp4", "sample_id": "#79", "prompt": "subject_definitions:\n (Lucian): appearance and clothing from , a male character in meticulously tailored 18th-century noble period costume.\n (Aila): appearance and clothing from , a female character in meticulously tailored 18th-century noble period costume.\n: scene/environment from .\n: a carriage stopped on the road with horses moving naturally, from .\n: a coachman pulling the carriage, from .\n\nsummary:\n[reference generation] Epic cinematic live-action video in 18th-century noble aesthetics with a cold, dark Gothic tone. Strong shadow contrast, medieval European style, extremely highly detailed, 32k resolution, hyper-realistic. High-end warm soft lighting with deep, rich colors. Romantic and artistic composition, poetic atmosphere, high psychological tension. Deep depth of field, artistic facial expressions, highly detailed clean and sharp faces, extremely clear character outlines. Four sequential shots depicting Lucian carrying Aila through a doorway and down steps, ending with an over-the-shoulder exchange.\n\nretention_analysis:\n (Lucian, ) is fully preserved as the primary male character appearing in Shot 1, Shot 2, and Shot 4 with detailed actions and dialogue. (Aila, ) is fully preserved as the primary female character appearing in Shot 2, Shot 3, and Shot 4 with detailed actions and dialogue. (scene, ) is partially preserved as the environmental setting implied by the metal door and stone steps in Shot 1. (carriage with horses, ) is a weak reference: explicitly defined in the source but not depicted in any of the four shot descriptions. (coachman, ) is a weak reference: explicitly defined in the source but not depicted in any of the four shot descriptions.\n\ndetailed_description:\nAbsolute directives: Strictly no watermarks, no on-screen text, no music (only sound effects are allowed), and no duplicated/identical persons. [Shot 1] Empty wide shot, low-angle upward camera. Epic cinematic live-action style, 18th-century noble aesthetics, cold dark Gothic tone with strong shadow contrast, medieval European style, extremely highly detailed, 32k resolution, hyper-realistic, high-end warm soft lighting, deep rich colors, deep depth of field. No characters visible. Large water droplets fall densely in linear streams. The main subject is the bottom steps and a tightly closed metal door above. At the 1st second, the metal door receives a physical impact from the inside bottom, and the door panel swings outward rapidly at a 90-degree angle. A single leather boot belonging to (Lucian) remains at the bottom position where the door panel originally stood. Dense rain impact sounds transition into the loud crash of the metal door hinge breaking and the door panel striking the wall. The camera remains fixed at a low angle.\n\n[Shot 2] Medium close-up, frontal flat shot. (Lucian) enters the frame from the front. His arms are held horizontally to support (Aila). Aila's torso is suspended in the air, her limbs swinging violently and irregularly. Lucian's facial muscles remain completely still with no contraction or deformation. Lucian's legs alternately move downward along the Z-axis as he walks down the steps. Heavy boot-sole impacts on stone steps transition into high-frequency clothing friction sounds. The camera tracks downward along the Y-axis in sync with the rhythm of the characters descending the steps.\n\n[Shot 3] Over-the-shoulder shot, high-angle downward camera. The upper left of the frame shows the shoulder outline of (Lucian). The focus is on the face of (Aila) in the lower part of the frame. Aila's body maintains high-frequency micro-tremors, and her gaze is fixed upward toward Lucian's face. (Aila) (S1) says [English]Where are you taking me? with her voice accompanied by physical tremors. The camera is at a high angle with fixed focus.\n\n[Shot 4] Over-the-shoulder shot, low-angle upward camera. The lower foreground shows the back and back-of-head outline of (Aila). The focus is on the close-up of (Lucian) above. Lucian's eyeballs deflect downward by 15 degrees, hold for 0.5 seconds, then return to looking straight ahead. His face shows no speaking motion. Lucian's body then moves toward the edge of the frame until he completely exits the frame. (Lucian) (S2) says in an off-screen voiceover with internal reverb [English]A reunion only I remember… The camera is at a low angle and remains completely fixed.\n\noverall_soundscape:\nDense rain impact sounds, loud crash of a metal door hinge breaking and door panel striking a wall, heavy boot-sole impacts on stone steps, high-frequency clothing friction sounds. Strictly no music; only diegetic sound effects are generated.\n\nnon_diegetic_music:\nN/A. The source explicitly prohibits music generation; only sound effects are permitted.", "original_prompt": "@图片2 是艾拉。@图片1 是卢锡安。 @图片3 是场景。 @图片4 是停在路上的马车,马儿自然的动。 @图片5 拉着马车的车夫\n【整体风格:史诗级电影真人风格,18世纪贵族美学,整体色调偏向冷暗的哥特式风格,需要强调强烈的阴影对比,属于中世纪欧美风格,极其高度精细,32k分辨率,超写实。高级的暖光柔和色调,色彩深沉浓郁。人物身着精心剪裁的时代服饰。浪漫而富有艺术感的构图,诗意的氛围,高度的心理张力,堪称杰作,【光影细节】 深景深,人物表情刻画文艺,高度精细的面部,干净利落,人物轮廓极致清晰,32k高清画质。\n 【绝对指令】禁止生成水印,禁止生成文字,禁止生成音乐,只能生成音效;禁止生成相同的人】。\n\n\n 镜号 01 | 破门 (空境 WS | 低机位仰拍)\n\n画面内容: 画面无人物。大体积水滴呈线状密集下落。画面主体为底部台阶与上方紧闭的金属大门。第1秒,金属门受到来自内部底端的物理撞击,门板呈 90 度向外快速弹开。卢锡安的单只皮靴停留在门板原本所在的底部位置。\n\n台词 / 音效: 密集的雨水撞击声 ➔ 金属门铰链断裂与门板撞击墙壁的巨响。\n\n镜头运动: 固定低机位(Low Angle)。\n\n镜号 02 | 降临 (近景 MCU | 正面平拍)\n\n画面内容: 卢锡安正面入画。其双臂呈水平横向托举姿态,承托艾拉。艾拉躯干悬空,四肢处于无规律的剧烈摆动状态。卢锡安面部肌肉维持静止,无收缩形变。卢锡安双腿交替发生 Z 轴向下的位移,走下台阶。\n\n台词 / 音效: 沉重的靴底与石台阶接触的撞击声 ➔ 衣物高频摩擦声。\n\n镜头运动: 镜头沿 Y 轴(垂直方向)随人物下台阶的节奏同步产生梯级下沉跟拍。\n\n镜号 03 | 质问 (过肩镜头 OTS | 俯拍)\n\n画面内容: 画面左上方为卢锡安肩部轮廓。焦点位于画面下方的艾拉面部。艾拉躯体维持高频率的微小震颤,其视线向上方(卢锡安面部位置)固定。\n\n台词 / 音效: 艾拉(声带伴随物理震颤):“Where are you taking me?”\n\n镜头运动: 俯拍机位(High Angle),固定焦点。\n\n镜号 04 | 无视 (过肩镜头 OTS | 仰拍)\n\n画面内容: 画面下方前景为艾拉背部与后脑轮廓。焦点位于上方的卢锡安近景特写。卢锡安眼球向下产生 15 度偏转,视线停留 0.5 秒后,眼球回正视向前方。面部无发音动作。随后卢锡安身体整体向画面边缘移动,直至完全移出画框(Exit frame)。\n\n台词 / 音效: 卢锡安(OS 内部混响):“A reunion only I remember…”\n\n镜头运动: 仰拍机位(Low Angle),机位完全固定。", "reference_images": ["references/572430c34e8d80e69b327ec28eefbe1e270e97a73030572c67847f0c328a43b9.jpg", "references/ac85e19328940113dd0ba5ca14f9cb1f75b9784ad16804150f9dd74def1cc309.jpg", "references/46fc4c7116b30005bd1b31499776f0d14a9371239671ad2d1c1c1ec35a81a7d2.jpg", "references/b346290fb74f26fa597f663c7d70b5cdba6581691c4e8dcd09eb85369e93f167.jpg", "references/64ac59ce55ecc9520c86d7309ab466c3b22576c02c0ce6ad4c33044c6d69ba52.jpg"], "reference_sha256": ["572430c34e8d80e69b327ec28eefbe1e270e97a73030572c67847f0c328a43b9", "ac85e19328940113dd0ba5ca14f9cb1f75b9784ad16804150f9dd74def1cc309", "46fc4c7116b30005bd1b31499776f0d14a9371239671ad2d1c1c1ec35a81a7d2", "b346290fb74f26fa597f663c7d70b5cdba6581691c4e8dcd09eb85369e93f167", "64ac59ce55ecc9520c86d7309ab466c3b22576c02c0ce6ad4c33044c6d69ba52"], "workflow": "ref2va", "prompt_sha256": "580598292f05b4c013753d0cb787a8338fa3079b124c997aecbfdaf374dde888", "original_prompt_sha256": "2e1f3e629bce88e917221dc67a9df15326ec7017c186281ab164f49567f8440d", "video_sha256": "0006f6d577e8cf4015795534d3209ddd19c9897ffad314e91f1a9a461074f3ef", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_079.json", "verification": "verification/sample_079.json"} {"file_name": "videos/sample_080.mp4", "sample_id": "#80", "prompt": "subject_definitions:\n: Xiao Ran, appearance and clothing from .\n: Shen Mo, appearance and clothing from .\n: Gu Nian, appearance and clothing from , wearing a gas station uniform.\n: Zhan Yuan, appearance and clothing from , wearing casual clothes.\n: Beauty 1 and Beauty 2, appearance and clothing from .\n\nsummary:\n[reference generation] Generate a 15-second over-the-shoulder close-up shot at the entrance of a mountaintop restaurant. (Zhan Yuan) walks forward two steps to look (Xiao Ran) up and down, then delivers two lines of dialogue with a smug expression, focusing on subtle facial and breathing changes. The scene features strong sunlight, road reflections, and a background with parked vehicles including a small scooter, a red Harley, and a blue Harley. No on-screen text, captions, or subtitles of any kind. Do not reduce image quality.\n\nretention_analysis:\nfully_preserved: (Xiao Ran), (Shen Mo), (Gu Nian), (Zhan Yuan), and (Beauty 1 and Beauty 2) are all defined and bound to their respective reference pictures to maintain facial structure, clothing, and hairstyle consistency as requested. The specific clothing constraints for Gu Nian (gas station uniform) and Zhan Yuan (casual clothes) are explicitly retained in their subject definitions.\n\ndetailed_description:\n[Shot 1] (15.0s) Over-the-shoulder medium close-up with a slow push-in. The scene is set at the entrance of a mountaintop restaurant during the day, outdoors, with strong sunlight, road reflections, and slightly raised dust. The environment features a recognizable foreground and background with tables, chairs, lighting, and parked vehicles, including a small scooter, a red Harley, and a blue Harley, with a crowd at the restaurant entrance watching attentively. No on-screen text, captions, or subtitles of any kind. Do not reduce image quality. (Zhan Yuan) walks forward two steps, looking (Xiao Ran) up and down. (2.0s) -> (S1) looks smug, speaking in a side profile or over-the-shoulder medium close-up, with a focus on the changes in the corners of his mouth, eyes, and breathing, saying [Chinese]小子,现在后悔还来得及。跪下来给爷爷磕三个响头,再让你身边这小妞陪本少爷睡一晚, (Delivery: rough, arrogant male voice, smug state, strong emotion, synchronized with actions, avoid precise lip-sync dependence) (6.5s) -> (S1) continues looking smug in the same medium close-up, emphasizing the corners of his mouth, eyes, and breathing changes, saying [Chinese]今天这事儿我就不跟你计较了,怎么样? (Delivery: rough, arrogant male voice, smug state, strong emotion, synchronized with actions, avoid precise lip-sync dependence) (6.5s).\n\noverall_soundscape:\nUnspecified beyond the scripted dialogue. No background music.\n\nnon_diegetic_music:\nN/A", "original_prompt": "禁止BGM,禁止任何文字。 角色:萧然萧然 角色锁定,保持脸部结构、服装、发型一致,沈墨沈墨 角色锁定,保持脸部结构、服装、发型一致。顾念顾念(加油站工作服) 角色锁定,保持脸部结构、服装、发型一致。战渊战渊(休闲便装) 角色锁定,保持脸部结构、服装、发型一致。美女1和美女2美女1,2 角色锁定,保持脸部结构、服装、发型一致。 场景: 山顶饭店四视图 。 道具:小踏板、红色哈雷、蓝色哈雷小红蓝 。 4-2 这是上一段剧情的视频,严格按照这个视频的角色站位和场景位置继续生成接下来的剧情,不要降低画质,仅参考视频的剧情走向和站位动作,不需要参考画质。 剧情:山顶饭店门口中,战渊往前走了两步,上下打量着萧然,随后推进到战渊得意,以侧脸或过肩近景说出“今天这事儿我就不跟你计较了,怎么样?”,重点表现嘴角、眼神和呼吸变化。 高光:战渊往前走了两步,上下打量着萧然。 背景:日/外 山顶饭店门口(中景特写)+ 日光强烈 + 路面反光 + 尘土微扬 环境:饭店门口或包间内人群视线集中,桌椅、灯光和停放车辆形成可辨识的前后景。 🎬 镜头 1:持续15秒 场景:【过肩近景缓推】镜头起幅,战渊往前走了两步,上下打量着萧然。(2s)→ 战渊得意,以侧脸或过肩近景说出“小子,现在后悔还来得及。跪下来给爷爷磕三个响头,再让你身边这小妞陪本少爷睡一晚,”,重点表现嘴角、眼神和呼吸变化(6.5s)→ 战渊得意,以侧脸或过肩近景说出“今天这事儿我就不跟你计较了,怎么样?”,重点表现嘴角、眼神和呼吸变化(6.5s)。 【对话模块】 台词:“战渊:小子,现在后悔还来得及。跪下来给爷爷磕三个响头,再让你身边这小妞陪本少爷睡一晚,” 配音指令:战渊|粗粝嚣张男声|状态(得意)|语气(情绪强烈,与画面动作同步,避免精确口型依赖) 对应镜头:镜头1,时间点 002-008 台词:“战渊:今天这事儿我就不跟你计较了,怎么样?” 配音指令:战渊|粗粝嚣张男声|状态(得意)|语气(情绪强烈,与画面动作同步,避免精确口型依赖) 对应镜头:镜头1,时间点 008-014 【物理接戏】 本组从【战渊嚣张大笑,以侧脸或过肩近景说出“哈哈哈!听见没有?连你女人都比你有自知之明!”,重点表现嘴角、眼...】承接,最终落在【战渊得意,以侧脸或过肩近景说出“今天这事儿我就不跟你计较了,怎么样?”,重点表现嘴角、眼神和呼吸变化】。下一组可从【顾念气急,以侧脸或过肩近景说出“你们嘴巴放干净点!”,重点表现嘴角、眼神和呼吸变化】接起,保持动作、位置和情绪连续。", "reference_images": ["references/90997ff520849ba90e88ef1bc2393eeb327bf8bd6f75dcb692b3f32465a3bfec.png", "references/28f4468e8819acd3034f7e74711bbcdb480bb159d05863f4b682cb8c8963541d.png", "references/9194da52c40d49cd8e865eb5522fdd0ecdc812958729bd42c6c530bddd25c5ac.png", "references/bfeae89db23426d9a81682a99c6b755e1e91c126c5cfea924005a987a35eef59.png", "references/7bf771b8685b853c6f399267c12cc01c06c32f3907b3bd88b8e4493a62ac2cf2.png"], "reference_sha256": ["90997ff520849ba90e88ef1bc2393eeb327bf8bd6f75dcb692b3f32465a3bfec", "28f4468e8819acd3034f7e74711bbcdb480bb159d05863f4b682cb8c8963541d", "9194da52c40d49cd8e865eb5522fdd0ecdc812958729bd42c6c530bddd25c5ac", "bfeae89db23426d9a81682a99c6b755e1e91c126c5cfea924005a987a35eef59", "7bf771b8685b853c6f399267c12cc01c06c32f3907b3bd88b8e4493a62ac2cf2"], "workflow": "ref2va", "prompt_sha256": "fde49f3ee7a996d2f59f976558dc22202fbe567e0c0b5b608e2a0d9fad32ea4b", "original_prompt_sha256": "589bde8a014efd9d606103e653edbe9f947d637d639a6624a782aead8fe78fd7", "video_sha256": "3f9646b1e52173073500dfe084949043798f05b8ec9bd83a2dfd4f738e2b6e1d", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 328, "duration_seconds": 13.666666666666666, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_080.json", "verification": "verification/sample_080.json"} {"file_name": "videos/sample_081.mp4", "sample_id": "#81", "prompt": "subject_definitions:\n: The character from .\n\nsummary:\n[reference generation] The video is set at the location from . The shot begins with the first frame from . slightly lowers the walkie-talkie, then picks it up again, and coldly orders into it.\n\nretention_analysis:\nThe location from is fully preserved as the setting. The character identity and appearance from are fully preserved as . The first frame from is fully preserved as the starting frame of the shot.\n\ndetailed_description:\n[Shot 1] Insert close-up, static shot. The shot begins with the first frame from . (S1) slightly lowers the walkie-talkie, then picks it up again, and coldly orders into the walkie-talkie: [English]Team B, cover the back exit. No one gets out.\n\noverall_soundscape:\nUnspecified\n\nnon_diegetic_music:\nN/A", "original_prompt": "Location:图片1\n1.插入特写 | 静止镜头 |以图片3为首帧, 图片2稍微放下对讲机,随后又拿起,对对讲机冷声下令:'Team B, cover the back exit. No one gets out.", "reference_images": ["references/41064f9b16360a8ce6194ba0554095789469c54076411e0a519930c806e7f66f.jpg", "references/460c6fc24b02bd15229406210749a02fe825ca170486e4c292e135cd98080818.jpg", "references/2eef7b52d7b599e82ac18446138fa3cb2b70f76810812dde20b530d9452649ab.jpg"], "reference_sha256": ["41064f9b16360a8ce6194ba0554095789469c54076411e0a519930c806e7f66f", "460c6fc24b02bd15229406210749a02fe825ca170486e4c292e135cd98080818", "2eef7b52d7b599e82ac18446138fa3cb2b70f76810812dde20b530d9452649ab"], "workflow": "ref2va", "prompt_sha256": "3dcba7a856bfad7e4e0351ea3010bacf4105e531361460e394470c4969d27290", "original_prompt_sha256": "28e9ae7d38163872c3ad441b1a3f6445d946ac1bd8f5247db9f8b69cbf0b363a", "video_sha256": "0273bdf194a69e53568880c3960a149a21438bf5f9d1c8db67c678563979f8d3", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 107, "duration_seconds": 4.458333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_081.json", "verification": "verification/sample_081.json"} {"file_name": "videos/sample_082.mp4", "sample_id": "#82", "prompt": "integrated_multimodal_description:\n[Shot 1] (Duration: 6s) Fixed realistic style, real portrait photography, vertical composition (Portrait Mode), desk lamp lighting and shadow. (Character A) and (Character B) maintain completely consistent appearance, clothing, hairstyle, and makeup throughout. Indoor scene. and are seated separately. No subtitles. Over-the-shoulder shot, over 's shoulder filming , framing the upper body of . looks at the expectation in mother's eyes. Cut to over-the-shoulder shot, over 's shoulder filming , framing the upper body of . 's eyes show expectation, hands nervously twisting together. Cut back to , silent for a moment, eyelashes slowly lowering, lips slightly pursed. High-definition cinematic texture, rich details, natural and soft lighting and shadow; stable facial features without deformation, clear facial features, normal human body structure, natural and smooth movements without stiffness, no stuttering, flickering, or clipping in the picture.\n\n[Shot 2] (Duration: 5s) Fixed realistic style, real portrait photography, vertical composition (Portrait Mode). and maintain completely consistent appearance, clothing, hairstyle, and makeup throughout. Indoor scene. and are seated separately. No subtitles. medium close-up, (S1) nods, smiles helplessly, and says [Chinese]好,听妈的。 with a tone of compromise, compliance, and a soft voice. High-definition cinematic texture, rich details, natural and soft lighting and shadow; stable facial features without deformation, clear facial features, normal human body structure, natural and smooth movements without stiffness, no stuttering, flickering, or clipping in the picture.\n\n[Shot 3] (Duration: 4s) Fixed realistic style, real portrait photography, vertical composition (Portrait Mode). and 's hands enter the frame, maintaining completely consistent appearance, clothing, hairstyle, and makeup throughout. Indoor scene. and are seated separately. No subtitles. Medium close-up, with a close-up of hand movements quickly tilting up to a medium close-up of 's face. 's hand covers the back of 's hand and pats twice, smiling with relief. High-definition cinematic texture, rich details, natural and soft lighting and shadow; stable facial features without deformation, clear facial features, normal human body structure, natural and smooth movements without stiffness, no stuttering, flickering, or clipping in the picture.\n\noverall_soundscape:\nReal environmental sound effects and character action sound effects are retained. No background music.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【风格定位】固定写实风格,真实人像摄影,竖屏构图(Portrait Mode),【台灯光影】\n【角色锚定】图1图2全程保持角色外观、服装、发型、妆容完全一致\n【场景环境】图3\n【人物站位】图1坐在绿色方框位置,图2坐在黄色方框位置。\n【其他要求】无字幕、无BGM音乐,保留真实环境音效与人物动作音效\n【时长配置】总时长6秒,\n【分镜时序(时间轴拆解)】\n镜头19:过肩镜头,过图1的肩膀拍摄图2,镜头框柱图2的上半身,图2看着母亲眼中的期盼,切过肩镜头,过图2的肩拍摄图1,镜头框柱图1的上半身,图1眼神期盼,双手紧张的绞在一起,切回图2沉默片刻,睫毛缓缓低垂,嘴唇微抿,\n【画质与强约束】高清电影质感,细节丰富,光影自然柔和;人物面部稳定不变形、五官清晰、人体结构正常、动作自然流畅不僵硬、画面无卡顿无闪烁无穿模\n镜号21(第24场)\n【风格定位】固定写实风格,真实人像摄影,竖屏构图(Portrait Mode),\n【角色锚定图1图2全程保持角色外观、服装、发型、妆容完全一致\n【场景环境】图3\n【人物站位】图1坐在绿色方框位置,图2坐在黄色方框位置。\n【其他要求】无字幕、无BGM音乐,保留真实环境音效与人物动作音效\n【时长配置】总时长5秒\n【分镜时序】\n镜头21:近景,图2点头,无奈一笑\n【对白/口型】图2音频1+妥协、顺从、声音轻柔:\"好,听妈的。\"\n【画质与强约束】高清电影质感,细节丰富,光影自然柔和;人物面部稳定不变形、五官清晰、人体结构正常、动作自然流畅不僵硬、画面无卡顿无闪烁无穿模\n镜号22(第24场)\n【风格定位】固定写实风格,真实人像摄影,竖屏构图(Portrait Mode),\n【角色锚定】图1,图2手部入画,全程保持角色外观、服装、发型、妆容完全一致\n【场景环境】图3\n【人物站位】图1坐在绿色方框位置,图2坐在黄色方框位置。\n【其他要求】无字幕、无BGM音乐,保留真实环境音效与人物动作音效\n【时长配置】总时长4秒\n【分镜时序(时间轴拆解)】\n 镜头22:近景+特写手部动作快速上摇至图1面部近景+图1的手覆上图2手背轻拍两下,欣慰一笑\n【画质与强约束】高清电影质感,细节丰富,光影自然柔和;人物面部稳定不变形、五官清晰、人体结构正常、动作自然流畅不僵硬、画面无卡顿无闪烁无穿模", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "e21685e3d041dafb760c4f11e067c48847f0099533133ea0c95d8b7a0a3208ef", "original_prompt_sha256": "da68ea3e65663fb34a5d331de8b14588236e0b8868e71ce90e5d040dcb73b84c", "video_sha256": "5d645efa34552e323ef98c6f29cb2e8b25f760ec8bb8adcf76f08d29431e7c5a", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_082.json", "verification": "verification/sample_082.json"} {"file_name": "videos/sample_083.mp4", "sample_id": "#83", "prompt": "subject_definitions:\n (Lin Feng) is the male character whose appearance, facial features, bone structure, eye spacing, and jaw geometry are 100% identical to . (Su Wan) is the female character whose appearance and facial features are 100% identical to . The system panel's visual design is based on .\n\nsummary:\n[reference generation] The video is a live-action style generation set in a meeting room (based on ). It features (Lin Feng) and (Su Wan) standing by a table. The camera focuses on close-ups and medium shots, capturing micro-expressions and body details with high contrast, rich cinematic lighting, and hyper-realistic 8K textures. A system panel (based on ) appears between them, displaying specific text. The video includes natural blinking, chest heaving from breathing, and eye movement, with no beautification or age alteration. Only ambient sounds and inner monologues are present.\n\nretention_analysis:\nfully_preserved: (Lin Feng) from and (Su Wan) from are the main characters and must retain 100% facial and structural consistency. The system panel from is retained as a prop. The meeting room scene from is retained as the background. partially_preserved: is requested to be used as the standing position reference for Lin Feng, Su Wan, and other people simultaneously, which is a conflicting assignment for a single image.\n\ndetailed_description:\n[Shot 1] Inside a meeting room, (Su Wan) stands by the table with a smile. In a medium shot, (Lin Feng) stands by the table looking at Su Wan. (S1) says in an inner monologue [Chinese]系统,对她使用“欲望洞察。. Immediately, a system panel appears between the standing Lin Feng and the standing Su Wan. The camera cuts to Lin Feng's subjective point of view. The system panel displays the following visible text: 【系统面板】\n【姓名:苏婉】\n【评级:96(S级诡异)】\n【当前最强烈的欲望:希望看到一名身材健硕、肌肉线条流畅的男学生,身上带着薄汗,在午后的阳光下,认真地为她擦拭办公室那面巨大的落地窗。】. In a close-up shot, Lin Feng smiles slightly. (S1) says in an inner monologue [Chinese]”擦窗户?这位S级的导员老师…口味还真特别。“. The video maintains a live-action style with close-ups and medium shots, minimizing redundant backgrounds to focus on micro-expressions and body details. The lighting is cinematic, high contrast, with a mix of warm and cold tones, and soft, layered light transitions. The texture is hyper-realistic with extreme 8K details. The character exhibits natural blinking, chest heaving from breathing, and flowing eye movements. The image remains sharp in focus with natural bokeh in the background. HDR high dynamic range is applied. No beautification or age alteration is allowed. The action is realistic and fluid, with natural emotions and a steady rhythm. The ending is natural with complete action logic. Negative constraints: no subtitles, no textures, no voiceover, no background music, no scene jumping, no character deformation, no lip-sync errors, no scene continuity errors, and no shot may repeat any framing from the original storyboard.\n\noverall_soundscape:\nOnly ambient environmental sounds and the characters' inner monologues are present. No other sound effects are included.\n\nnon_diegetic_music:\nN/A", "original_prompt": "保持人物与参考图完全一致,面部特征100%相同,保持相同的骨骼结构、眼间距和下颚几何形状,人物有自然的眨眼、呼吸时的胸腔起伏、以及眼神的流转,焦内锐利,焦外自然虚化,HDR高动态范围,禁止美化,禁止改变年龄。负面提示词:无字幕、无贴图、无旁白、无背景音乐。无任何变化,负面提示词:无字幕、无贴图、无旁白、无背景音乐、无画面跳变、无人物变形、无口型错位、无场景穿帮。 风格:真人实拍风格,镜头以近景、特写为主,弱化冗余背景,重点捕捉人物微表情与肢体细节,层次极具张力,与原有分镜无任何重复镜头。高对比度,光影层次丰富,随场景切换自然过渡。超写实质感,8K极致细节,电影级光影,暖冷交织,高对比度,柔和光影过渡,光影柔和且有层次。 视频节奏:动作写实流畅。情绪自然,节奏平稳过渡。 视频仅出现环境音和对白 其他声音无 画面收尾自然,动作逻辑完整,无多余违和元素。\n场景:[图片5] \n林枫: [图片1] \n苏婉 :[图片2] \n系统面板: [图片3] \n林枫站位参考图:[图片4] \n苏婉站位参考图: [图片4] \n其他人的站位参考:[图片4] \n\n会议室内,苏婉面带微笑地站在桌边,中景林枫站在桌边看着苏婉,内心独白心里想:系统,对她使用“欲望洞察。紧接着站着的林枫与站着的苏婉的中间出现一个系统面板,镜头切林枫主观视角,系统面板上显示如下文字:\n【系统面板】\n【姓名:苏婉】\n【评级:96(S级诡异)】\n【当前最强烈的欲望:希望看到一名身材健硕、肌肉线条流畅的男学生,身上带着薄汗,在午后的阳光下,认真地为她擦拭办公室那面巨大的落地窗。】\n特写林枫微微一笑,林枫心里想:”擦窗户?这位S级的导员老师…口味还真特别。“", "reference_images": ["references/8ea4ad6d2d5f128074a85f38c76b74fb317711e3fda1598f27f7fc7f74d0b914.jpg", "references/b258456acd490d0188c73e2e710b126132a90b9b10ef89bb7b758ff3966d3233.jpg", "references/7d9da739ffc15bd04d3ae3e35cffbd5329ffdcac6a7ff91f2e704706441cd689.jpg", "references/b35c227c7cf669e96ee61f3cf51bbb40dd378399db438912a779109437bc7eed.jpg", "references/10377281a77ba41803282efe315f38f86c6406e16c9c71cc3dcf554211b7f1c9.jpg"], "reference_sha256": ["8ea4ad6d2d5f128074a85f38c76b74fb317711e3fda1598f27f7fc7f74d0b914", "b258456acd490d0188c73e2e710b126132a90b9b10ef89bb7b758ff3966d3233", "7d9da739ffc15bd04d3ae3e35cffbd5329ffdcac6a7ff91f2e704706441cd689", "b35c227c7cf669e96ee61f3cf51bbb40dd378399db438912a779109437bc7eed", "10377281a77ba41803282efe315f38f86c6406e16c9c71cc3dcf554211b7f1c9"], "workflow": "ref2va", "prompt_sha256": "5b2a94f089e03877f0c150c507a7a8fe1b222c1ff864158d2f9c6be36565065f", "original_prompt_sha256": "56e0d7ef6b7e954d4a38604323e3d47bef60c5f5bf4d61e1bf5d838783bd8927", "video_sha256": "c1ec5bae074af46bd8cb94a1df2220ab6be4d9df0d5d2527311ecc8fa4797866", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 243, "duration_seconds": 10.125, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_083.json", "verification": "verification/sample_083.json"} {"file_name": "videos/sample_084.mp4", "sample_id": "#84", "prompt": "subject_definitions:\n: Male lead Lu Zhen, appearance and clothing from .\n: Character Gu Yanzhi, appearance and clothing from .\n: Inner sect senior brother 1, appearance and clothing from .\n: Inner sect senior brothers several 2, appearance and clothing from .\n: Iron Arm Martial Arts disciple 2, appearance and clothing from .\n: Iron Wall Martial Arts Xiong Yue, appearance and clothing from .\n: Character Yan Shanshan 1, appearance and clothing from .\n: Character Zhang Lei 1, appearance and clothing from .\n\nsummary:\n[reference generation] The video features a martial arts courtyard scene with cinematic lighting and retro color grading. In the first shot, several characters practice boxing in a sunlit courtyard. In the second shot, the male lead walks steadily into the frame from the left.\n\nretention_analysis:\nThe subjects through are requested to be retained with their appearances and clothing from through respectively. is requested to be retained as the scene background. is provided but not explicitly used in the shot descriptions.\n\ndetailed_description:\nCinematic quality, strong light-dark contrast, warm golden highlights, cool cyan shadows, low saturation retro tones, Chinese spy drama color grading, cinematic lighting, 4K ultra-clear, realistic style. The entire video has no watermark, no subtitles, no logo, and no music. The scene is locked to the environment from .\n\n[Shot 1] wide shot, high-angle looking down, slow lateral trucking. The martial arts courtyard is bathed in sunlight. , , , , , and are practicing boxing. This shot occupies 0-2s.\n\n[Shot 2] At 00:02.000, medium shot, eye-level, slow push-in, camera tracking. walks steadily through the courtyard gate from the left into the frame. His posture is straight as a pine, and the hard sunlight outlines his confident and calm facial contours. This shot occupies 2-4s.\n\noverall_soundscape:\nUnspecified; music is prohibited.\n\nnon_diegetic_music:\nN/A", "original_prompt": "电影级质感,强烈明暗对比,暖金色高光,冷青色暗部,低饱和度复古色调,国谍战剧调色,电影级打光,4K超清,写实风格。\n画面:全程无水印,无字幕,无logo ,无音乐。\n锁定 @图片9 为场景。\n镜号1,0-2s | 全景,俯视 | 缓慢横移 | 武馆庭院阳光明媚, @顾言之 @铁臂武馆弟子2 @内门师兄弟1 @内门师兄弟若干2 @张雷1 @严珊珊1 正在练拳。\n镜号2,2-4s,中景,平视| 缓慢推进| 镜头跟拍| @陆真-男主 穿步伐沉稳地穿过院门从左侧走入画面。他的身姿挺拔如松,硬质阳光勾勒出他自信从容的面部轮廓。\n\n参考图片说明:\n图1: 角色: 陆真-男主\n图2: 角色: 顾言之\n图3: 角色: 内门师兄弟1\n图4: 角色: 内门师兄弟若干2\n图5: 角色: 铁臂武馆弟子2\n图6: 角色: 铁壁武馆 熊月\n图7: 角色: 严珊珊1\n图8: 角色: 张雷1", "reference_images": ["references/f3ee784b5cf26015d712e999cef2abdb6988bff5cdf7fbd480b5a0cb62fd789d.jpg", "references/02d588b222734845927b31c09f23d7b29ada97e76fc03c78d583811cbbda855c.jpg", "references/3afb0ab9e8b0c373ae3278dbbbcf81eb8d1babfdc26c9b4e517fcc39bec9af2d.jpg", "references/dde94746d2989dfbf246b982b4a995818e9f8e5778171970a544e5c278a49bc8.jpg", "references/bcade4f78b43e070e9c883aadb9a5cf0c322a3762323ca7833fe6a522450940d.jpg", "references/d3b474ab7e13f8f3f96fcd434ea80b700c87a49a656546875f4fd5fb25b67c22.jpg", "references/2ad660b6d160a22c0ecfa3090ae104aa83236df4aa0f126be3bfb97746dff47d.jpg", "references/0ffae513ad1fe1c45270adce4e671f611bda3d47abc97bcd7f3653a7d1c7c47e.jpg", "references/3ceee1d085ffeeac72c6d5b3ca06b4352486afb1537bf63509640e5c950b3874.jpg"], "reference_sha256": ["f3ee784b5cf26015d712e999cef2abdb6988bff5cdf7fbd480b5a0cb62fd789d", "02d588b222734845927b31c09f23d7b29ada97e76fc03c78d583811cbbda855c", "3afb0ab9e8b0c373ae3278dbbbcf81eb8d1babfdc26c9b4e517fcc39bec9af2d", "dde94746d2989dfbf246b982b4a995818e9f8e5778171970a544e5c278a49bc8", "bcade4f78b43e070e9c883aadb9a5cf0c322a3762323ca7833fe6a522450940d", "d3b474ab7e13f8f3f96fcd434ea80b700c87a49a656546875f4fd5fb25b67c22", "2ad660b6d160a22c0ecfa3090ae104aa83236df4aa0f126be3bfb97746dff47d", "0ffae513ad1fe1c45270adce4e671f611bda3d47abc97bcd7f3653a7d1c7c47e", "3ceee1d085ffeeac72c6d5b3ca06b4352486afb1537bf63509640e5c950b3874"], "workflow": "ref2va", "prompt_sha256": "1242a18ba389eef0d17a35715da49ff4891c5e83b089b4c644a7c49ea645c7f4", "original_prompt_sha256": "e869f5599badb41f5c46d4af6c190a45d0a3b69cd874442a1b52f7cb847a91ae", "video_sha256": "59d3b25ed7917d07f2da426977ced4abf1c3f2c177a1306374dc35df66cf58e8", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 124, "duration_seconds": 5.166666666666667, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_084.json", "verification": "verification/sample_084.json"} {"file_name": "videos/sample_085.mp4", "sample_id": "#85", "prompt": "subject_definitions:\n is the staff from . The other subjects are the group on the right and the white-armored performer described in the action.\n\nsummary:\n[reference generation] A glowing golden staff attacks the people on the right. A white-armored performer blocks it with a glowing weapon, producing shockwaves and gold particles. The staff then presses against the performer, who struggles to resist it.\n\nretention_analysis:\nfully_preserved: supplies the appearance of the staff, which remains the same staff throughout all three shots. The other actors are defined by the textual white-armor and screen-position requirements.\n\ndetailed_description:\nGlobal style: max70mm film texture, grainy realism, epic scale, low saturation. No subtitles. [Shot 1] From 0 to 2 seconds, the staff from glows all over with smoky, particle-like golden light and occupies one eighth of the image. It extends diagonally into view from the upper-left sky toward the group on the right. [Shot 2] At 00:02.000, continuing until 5 seconds, the white-armored performer in the center of the frame radiates intense platinum-white light over the entire body and raises a weapon emitting particle-like platinum-white light to block the staff. The staff collides with the performer’s weapon. A shockwave spreads in all directions, and golden particles spray outward from the collision point. [Shot 3] At 00:05.000, continuing until 7 seconds, the staff presses against the white-clad performer on the right. Light surges over the performer’s whole body as the performer struggles to resist the staff using the weapon. Sparks fly where the weapon and staff collide.\n\noverall_soundscape:\nSpecial sound effects, weapon collisions and sparks flying. No music.\n\nnon_diegetic_music:\nN/A", "original_prompt": "[Reference Aliases]\n@图片1 = 金箍棒\n\n【全局基础设定】\n使用@图片1 作为棒子参考。\n@视频1 为原始拍摄素材(包含基础演员表演与构图)\n百分百原样复刻 @视频1 原生构图、机位视角、所有演员完整表演动作、肢体节奏、面部神情,原视频所有运镜轨迹、速度、幅度全程丝毫不改,精准还原演员神态与原生表情动作。\n风格要求:max70mm胶片质感,颗粒写实,史诗规模低饱和度\n音频要求:无音乐、无字幕、加入特效声音,兵器对撞声,火星四溅声。\n【时间片分镜脚本】\n0-2秒,镜头1,完全保留@视频1 (原始素材)的原有构图与演员表演动作,一根棒子通体冒着粒子烟雾般的金光(比例为占画面八分之一)从画面左上角的天空中斜着伸长入画,袭向画右的众人。运镜与@视频1 原有运镜保持完全一致。\n2-5秒,镜头2,完全保留@视频1 (原始素材)的原有构图与演员表演动作,画面中心穿白色盔甲的演员全身白金光芒大盛,举起武器(散发粒子白金色光芒)格挡棒子,棒子与白衣演员的武器相撞,冲击波相四面八方散去,金色颗粒从碰撞的位置向着四周飞溅。运镜与@视频1 原有运镜保持完全一致。\n5-7秒,镜头3,完全保留@视频1 (原始素材)的原有构图与演员表演动作,棒子压制着画右的白衣演员,白衣演员全身光芒大盛,努力用兵器抗衡棒子,兵器和棒子对撞的位置火星四溅。运镜与@视频1 原有运镜保持完全一致。\n【画质与防崩坏约束】\n1:1 精准还原 @视频1 原版人物所有表演动作、肢体挪动姿态、面部神态变化,动作无偏移、无走形、无二次创作,原运镜轨迹完全沿用,同步匹配演员。", "reference_images": ["references/d7c7e25344a83f0af315dcdf1e820759763839576fb180451ea5f3c39e882bf6.jpg"], "reference_sha256": ["d7c7e25344a83f0af315dcdf1e820759763839576fb180451ea5f3c39e882bf6"], "workflow": "ref2va", "prompt_sha256": "546ff97b9bbceb4914d129513909fb375292a2af222f6c14b565019c17e785af", "original_prompt_sha256": "0f5fa49246e7a065b5af1e4fa615dab00a6894d714e21262c2dd8fad172fc0ae", "video_sha256": "acbfe04eb638d5d4aa630605fb2d7d80dfeed98597666d635071b01999652591", "task_adapted": true, "width": 1344, "height": 768, "fps": 24, "num_frames": 175, "duration_seconds": 7.291666666666667, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_085.json", "verification": "verification/sample_085.json"} {"file_name": "videos/sample_086.mp4", "sample_id": "#86", "prompt": "integrated_multimodal_description:\n[Shot 1] A woman in her thirties is inside a tomato greenhouse, picking peaches. The tomatoes are Provence tomatoes, and there is a basket of tomatoes in front of her. The woman (S1) says [Chinese]你是否还在寻找一款好吃的西红柿?有网友想知道普罗旺斯口感究竟是如何!.\n\noverall_soundscape:\nAmbient sounds of a greenhouse, rustling of leaves, and the physical sounds of picking peaches and handling tomatoes.\n\nnon_diegetic_music:\nN/A", "original_prompt": "一个三十多岁的女士,在西红柿大棚里面在摘桃子,西红柿是普罗旺斯西红柿,面前一筐子西红柿,说话音频为:你是否还在寻找一款好吃的西红柿?有网友想知道普罗旺斯口感究竟是如何!", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "b203e5d1005573da593359cc3f59dacef497d508105d31548681b3e2d259de34", "original_prompt_sha256": "5309a56ad3fee5b24bcae525d8cf1350989d0d6be06d88977fc9c0eb8083a3c7", "video_sha256": "7a889efc3fb4937a54fc644b4b3e816d82ab4241fdd1da0df51219e7cf1cce84", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 243, "duration_seconds": 10.125, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_086.json", "verification": "verification/sample_086.json"} {"file_name": "videos/sample_087.mp4", "sample_id": "#87", "prompt": "subject_definitions:\n: The model. Face and identity from (virtual face, high cranial vault with large wavy curly hair, exquisite party makeup, flawless skin without tattoos). Clothing strictly from (no changes or color differences allowed in color, shape, or details). Standard figure.\n\nsummary:\n[reference generation] A 12-second e-commerce marketing video for women's clothing to promote purchase conversion. The model, wearing the exact clothing from Picture 2 and featuring the face from Picture 1, walks and poses in the background from Picture 3, showcasing the outfit's details and textures with dynamic camera movements.\n\nretention_analysis:\nThe requested subjects and pictures are fully preserved. The model's face and identity from are retained. The exact clothing from is strictly retained without any changes. The background from is retained.\n\ndetailed_description:\n[Shot 1] Full-body long shot. confidently walks into an exquisite interior to a light and upbeat rhythm. Warm environmental light and shadow sweep across, realistically capturing the matte texture of velvet and the mirror-like highlights of patent leather, showcasing the overall styling's sense of power. The camera follows steadily. Background from . [Shot 2] At 0:04.000, medium shot. elegantly stops beside a sofa and turns slightly sideways, making natural hair-touching and looking-back movements, dynamically displaying the heart-shaped cut and back details. Under the light, the rhinestone edging flashes with flowing light. The camera translates laterally and orbits. [Shot 3] At 0:08.000, close-up of details transitioning to full-body. The camera pushes in to focus on the top, naturally shifts her center of gravity, showcasing the high elasticity and fit of the knitted fabric and its craftsmanship texture, with clear fabric texture. Finally, the camera pulls back to reveal the complete styling, with movements being clean and natural. Negative constraints: Must conform to basic physical laws. Strictly prohibit clipping (character penetrating objects like tables and chairs) or unreasonable scenes like the model standing on tables and chairs. Hands are strictly prohibited from being in pockets throughout the entire video.\n\noverall_soundscape:\nUnspecified.\n\nnon_diegetic_music:\nPure instrumental music, no vocals, upbeat and rhythmic modern electronic jazz, with rhythm points matching the model's footsteps.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。服装严格使用图二,颜色/版型/细节不允许有任何改变或色差;背景直接使用图三。模特采用图一虚拟人脸,标准身材,高颅顶大波浪卷发配精致派对妆,皮肤无瑕无纹身。符合基本物理规律,禁止出现穿模(人物穿透桌椅等物体)或模特站在桌椅上等不合理画面。双手全程禁止插兜。 分镜设计: 1.【0-4s 全身远景】模特伴随轻快节奏自信步入精致室内,温润的环境光影拂过,真实捕捉丝绒的哑光质感与漆皮的镜面高光,展现整体穿搭力量感。镜头稳定跟随。 2.【4-8s 中景】模特在沙发旁优雅停步并微侧身,做自然拨发与回眸动作,动态展示心形剪裁与背部细节,光线下水钻包边闪烁流光。镜头横移环绕。 3.【8-12s 局部近景/全身】镜头推进聚焦上衣,模特自然移动重心,展现针织面料的高弹贴合度与工艺质感,面料纹理清晰。最后拉远展现完整造型,动作利落自然。 配乐:纯音乐,无人声,明快且富有节奏感的现代电子爵士,节奏点适配模特步伐。\\n\\n素材说明:图1=模特参考(asset-20260224202629-db8dr);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/b6a086feeefbfc1368067945dee0421d9e7aa21ed0db1e841c16e809c66f2874.jpg", "references/384092e3ac61091a200c9fe65ae14d1c77f64a9693477a1e9a66e5e82d648ee0.jpg", "references/e79907f7456a3dcb21f4ab1e84a3aa9e6dba09286585301c9f0f77afaf9e2430.jpg"], "reference_sha256": ["b6a086feeefbfc1368067945dee0421d9e7aa21ed0db1e841c16e809c66f2874", "384092e3ac61091a200c9fe65ae14d1c77f64a9693477a1e9a66e5e82d648ee0", "e79907f7456a3dcb21f4ab1e84a3aa9e6dba09286585301c9f0f77afaf9e2430"], "workflow": "ref2va", "prompt_sha256": "0434fa3f2c5cf7d5944629d36392911cf22c8bd7b9f300ef8f26459d008d0a73", "original_prompt_sha256": "f50444258a1b411c2fa800908f1a9b21762868e417b856c62ecb33e265831820", "video_sha256": "a9d7595f7f21e635c1718baa7d1cba27917bbf3e612be2b92c68d918e912692d", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_087.json", "verification": "verification/sample_087.json"} {"file_name": "videos/sample_088.mp4", "sample_id": "#88", "prompt": "subject_definitions:\n is the character Shen Wanwan, with appearance and clothing from . is the character Jailer, with appearance and clothing from . The scene setting is the daytime Chang'an Street, as depicted in .\n\nsummary:\n[reference generation] A cinematic 16:9 horizontal video in the style of a viral female revenge drama, featuring a winter prison cart scene on Chang'an Street with two characters and five specific shots.\n\nretention_analysis:\nfully_preserved: is retained as the daytime Chang'an Street scene setting throughout the video. fully_preserved: is retained as the character identity and appearance of Shen Wanwan, appearing in shots 1 through 5. fully_preserved: is retained as the character identity and appearance of the Jailer, appearing in shots 1 and 4.\n\ndetailed_description:\n[Shot 1] Medium shot, fixed camera. Two prison carts roll forward on the snow-covered bluestone road of the daytime Chang'an Street, wheels crushing the snow. The vermilion palace walls extend into the distance, cold wind blows snow dust, and scattered snowflakes fall. (Shen Wanwan) is in the foreground left, leaning on the prison cart's wooden bars, wearing a cangue. (Jailer) is in the midground right, holding a whip, standing by the prison cart. The background shows the moving prison cart and silhouettes of onlookers. The lighting is winter side-backlight, with snow reflecting cold white diffuse light, forming strong light and dark contrast. Sharp hard rim light outlines the prison cart and characters. Low, oppressive string music plays, with the sound effect of wheels crushing snow.\n\n[Shot 2] Extreme close-up, fixed camera. (Shen Wanwan) grips the prison cart's wooden bars with all ten fingers, wood splinters embedding into her flesh. The realistic skin texture shows pores and blood vessel details. The string music tightens, with the sound effect of nails scratching wood.\n\n[Shot 3] Close-up, fixed camera. (Shen Wanwan) has disheveled hair, squeezing her tear-stained face between the bars. (S1) says [Chinese]姐姐…… 姐姐我错了! in a female voice, voice breaking. There is no background music and no sound effects.\n\n[Shot 4] Medium shot, quick cut. (Jailer) strikes the prison cart's bars with his whip handle. (S2) says [Chinese]给我老实点! in a male voice, rough and brutal. The background music is the whip crack, with the sound effect of the whip cracking.\n\n[Shot 5] Medium shot, tracking shot. The prison cart speeds up. (Shen Wanwan) extends her entire forearm through the bars, fingers grasping wildly in the air. The string music becomes urgent, with the sound effect of the wooden bars trembling.\n\nThroughout the entire video, the spatial relationship is strictly maintained with (Shen Wanwan) on the left and (Jailer) on the right to prevent axis jumping. The visual style is cinematic quality, 8K ultra HD, with realistic skin texture (pores and blood vessel details maximized), cold color grading, and Dolby Vision HDR. Blood drops form deep red reflection points on the snow. There are no subtitles, no watermarks, and no logos. Characters have no deformation and no facial collapse. The same character is not duplicated in the same frame, and there are no multiple people with the same face.\n\noverall_soundscape:\nThe soundscape includes the sound of wheels crushing snow, nails scratching wood, a whip cracking, and wooden bars trembling. The background music consists of low, oppressive string music that tightens and becomes urgent.\n\nnon_diegetic_music:\nLow, oppressive string music that tightens and becomes urgent, with a specific whip crack sound effect.", "original_prompt": "画面质感和风格:抖音爆款女频复仇爽剧 16:9 横屏,电影级质感,8K 超高清,真实皮肤纹理 (毛孔、血丝细节拉满),冷色调调色,杜比视界 HDR。光影设定:冬日侧逆光,积雪反射冷白色漫射光,形成强烈明暗对比;囚车和人物边缘形成锐利硬轮廓光;血珠在雪地上形成深红色反光点。场景:@白天长安街,积雪覆盖的青石板路,朱红宫墙延伸至远方,寒风卷起地上的雪沫,空中飘着零星雪花。角色站位及装备态:画面前景左侧 @沈婉婉 扑在囚车木栏上,枷锁加身;画面中景右侧 @差役 持鞭站在囚车旁;背景是行驶的囚车和围观百姓剪影。防跳轴指令:全程保持沈婉婉在左、差役在右的空间关系。分镜内容:镜头 1,中景,固定镜头,两辆囚车辘辘前行,车轮碾过积雪,(BGM:低沉压抑的弦乐),< 音效:车轮碾雪声 >,台词:无镜头 2,大特写,固定镜头,@沈婉婉 十指死死抠住囚车木栏,木屑嵌进肉里,(BGM:弦乐收紧),< 音效:指甲刮木头声 >,台词:无镜头 3,近景,固定镜头,@沈婉婉 披头散发,将满是泪痕的脸挤进栅栏之间,(BGM:无),< 音效:无 >,@沈婉婉 撕心裂肺喊:{姐姐…… 姐姐我错了!}(女声,破音)镜头 4,中景,快速切镜,@差役 一鞭杆敲在囚车栅栏上,(BGM:鞭响),< 音效:鞭子抽打声 >,@差役 呵斥:{给我老实点!}(男声,粗暴)镜头 5,中景,跟拍镜头,囚车速度加快,@沈婉婉 将整条小臂从栅栏间伸出去,手指在空中乱抓,(BGM:弦乐急促),< 音效:木栏震颤声 >,台词:无保持无字幕,避免画面生成字幕。不要生成水印,不要生成 logo,人物形象无变形、无面部崩坏。视频全程不要在同一画面中复制相同人物,不要多人同脸。 图片1是白天长安街,图片2是沈婉婉,图片3是差役", "reference_images": ["references/a886455e082843e26262b9c5ef4b81a11f4f73426bf3e0dca91a36ded4dd5f51.jpg", "references/5c8352c9a86106e2df6b7e01ee4be45f9ace784958838671e2216f02c1da3090.jpg", "references/28f7c0a59c16bc7d34a5f0e079601074b9fad6240f8a457d21c883a6469bb6d1.jpg"], "reference_sha256": ["a886455e082843e26262b9c5ef4b81a11f4f73426bf3e0dca91a36ded4dd5f51", "5c8352c9a86106e2df6b7e01ee4be45f9ace784958838671e2216f02c1da3090", "28f7c0a59c16bc7d34a5f0e079601074b9fad6240f8a457d21c883a6469bb6d1"], "workflow": "ref2va", "prompt_sha256": "ee7bb3803e8052753675ac3321ac30d55780b50ca0baed54aed4a279382b4132", "original_prompt_sha256": "a23a79c7e3b4d17bcfbaba652d43364e943dc9e3203ed24632138a3549da471e", "video_sha256": "41b46738bbbda3739915e8e1713fe7647b775d285829b2c8a66eb9e2ef98c31d", "task_adapted": false, "width": 1344, "height": 768, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_088.json", "verification": "verification/sample_088.json"} {"file_name": "videos/sample_089.mp4", "sample_id": "#89", "prompt": "integrated_multimodal_description:\n[Shot 1] Photorealistic live-action style, lifelike and detailed, realistic lighting and natural colors, accurate perspective, rich texture, rigorous composition, simple style with key details emphasized. The first frame directly shows the complete scene with both characters in their initial positions — no white background, transition frame or progressive loading. Throughout the video, no duplicated characters appear in the same frame and no two people share the same face. Medium shot with center composition in the Xie Mansion flower pavilion corridor. Scattered skylight, front lighting, warm light, soft without shadows. Xie Jingheng (S1) stands on the left-center, turned sideways to the right; Ling Xueyin (S2) stands on the right-center, turned sideways to the left. The two walk side by side through the corridor. Ling Xueyin (S2), a young woman with a melon-seed face, delicate small features, soft facial muscles, gentle undertone and elegant lady temperament, suddenly trips slightly on her skirt while walking, her body leaning forward slightly as she loses balance. Xie Jingheng (S1), a young man with sword eyebrows and starry eyes, distinct contours, tight muscles, cold undertone and cold noble young master temperament, reacts swiftly and immediately reaches out to steadily support her arm. Ling Xueyin (S2) shows a startled and panicked expression facing her crush: eyes slightly widened, eyebrows slightly furrowed, lips slightly parted, a gasp not yet voiced. Xie Jingheng (S1) shows a calm and concerned expression facing his fiancée: eyes focused on the woman, brows drawn slightly together, thin lips pressed tight, jawline tense.\n\n[Shot 2] Medium close shot with rule-of-thirds composition. Scattered skylight, front lighting, warm light, soft without shadows. Ling Xueyin (S2) steadies herself with his support, cheeks flushing red, shyly lowering her head. Xie Jingheng (S1) lowers his head to look at her and opens his mouth to ask. Ling Xueyin (S2) shows a shy expression with heart racing facing her crush: eyes lowered and evasive, eyes flowing, cheeks slightly red, corners of mouth cannot help but slightly curve up, lightly biting lower lip. Xie Jingheng (S1) shows a relaxed and concerned expression facing his fiancée: eyes gently gazing, eyebrows relaxed, corners of mouth slightly relaxed. Xie Jingheng (S1) says in a cold domineering president voice, relaxed, concerned and comforting, with moderate volume, slow pace, steady breath, soft tail sounds and gentle concern [Chinese]没事吧?\n\n[Shot 3] Close-up with rule-of-thirds composition. Scattered skylight, front lighting, warm light, soft without shadows. Ling Xueyin (S2) slightly raises her eyes to look at him, corners of mouth carrying a shy smile as she softly responds. Ling Xueyin (S2) shows a shy and joyful expression facing her crush: raising eyes affectionately, eyebrows curved, apple cheeks slightly puffed, pulling a smile, corners of mouth spreading a sweet arc. Ling Xueyin (S2) says in a sweet coquettish little fox voice, relaxed, shy and responding, with slightly small volume, light pace, breath with smile, upturned tail sounds and sweet shyness [Chinese]没事。\n\n[Shot 4] Medium close shot with center composition. Scattered skylight, front lighting, warm light, soft without shadows. Xie Jingheng (S1) slowly releases the hand supporting her, eyes gently looking at her as he softly reminds her. Xie Jingheng (S1) shows a relaxed and soft expression facing his fiancée: eyes containing a smile, gaze focused, facial muscles completely relaxed, lips slightly moving. Xie Jingheng (S1) says in a cold domineering president voice, relaxed, concerned and reminding, with moderate volume, slow pace, steady breath, clean tail sounds and gentle reminder [Chinese]下次小心点。 No subtitles, dialogue cards, screen text or bottom captions.\n\noverall_soundscape:\nUnspecified beyond the scripted dialogue.\n\nnon_diegetic_music:\nN/A", "original_prompt": "【全局强制约束】\n视频全程不要在同一画面中复制相同人物,不要多人同脸。\n首帧必须直接是完整场景 + 全部角色初始站位(禁止白底/过渡帧/逐帧加载)\n\n场景:[图片1](谢府花亭 · 连廊)\n角色:[图片2](凌雪吟 - 基础形态).[图片3](谢景珩 - 基础形态)\n风格:真人写实风格,逼真精细,光影真实,色彩自然,透视准确,质感丰富,构图严谨,细节丰富,画风简洁但关键细节突出\n光线:散射天光,顺光,暖光,柔和无阴影\n位置:[图片3](谢景珩 - 基础形态)在左侧偏中,侧身向右;[图片2](凌雪吟 - 基础形态)在右侧偏中,侧身向左,两人并肩走在游廊中\n描述:中景+居中构图|散射天光,顺光,暖光,柔和无阴影,[图片2](凌雪吟 - 基础形态)走动时脚下突然被裙摆绊了一下,身体微微前倾失去平衡,[图片3](谢景珩 - 基础形态)反应迅速,立刻伸出手稳稳扶住她的胳膊。\n【表情指令】[图片2](凌雪吟 - 基础形态)|女/青年/瓜子脸/五官精致小巧/面部肌肉柔和/温婉底色/大家闺秀气质|惊吓/慌乱/面对心上人|眼睛微微睁大,眉毛微蹙,唇部微张,惊呼未出声\n【表情指令】[图片3](谢景珩 - 基础形态)|男/青年/剑眉星目/轮廓分明/肌肉紧实/冷峻底色/清冷贵公子气质|冷静/关切/面对未婚妻|眼神聚焦在女方身上,眉头微收,薄唇紧抿,下颌线紧绷\n描述:近景+三分构图|散射天光,顺光,暖光,柔和无阴影,[图片2](凌雪吟 - 基础形态)借着搀扶站稳身子,脸颊泛起红晕,害羞地低下了头,[图片3](谢景珩 - 基础形态)低头看着她,开口询问。\n【表情指令】[图片2](凌雪吟 - 基础形态)|女/青年/瓜子脸/五官精致小巧/面部肌肉柔和/温婉底色/大家闺秀气质|害羞/心跳加速/面对心上人|眼神低垂躲闪,眼波流转,脸颊微红,嘴角忍不住轻轻上扬,轻咬下唇\n【表情指令】[图片3](谢景珩 - 基础形态)|男/青年/剑眉星目/轮廓分明/肌肉紧实/冷峻底色/清冷贵公子气质|放松/关切/面对未婚妻|眼神温柔注视,眉毛舒展,嘴角微松\n【配音指令:[图片3](谢景珩 - 基础形态)|VoiceID:冷酷霸总音|状态:放松/关切/安抚|语气特点:音量适中/语速平缓/气息稳/尾音轻柔/温和关心。对白:没事吧?】\n描述:特写+三分构图|散射天光,顺光,暖光,柔和无阴影,[图片2](凌雪吟 - 基础形态)微微抬眼看向对方,嘴角带着羞涩的笑意轻声回应。\n【表情指令】[图片2](凌雪吟 - 基础形态)|女/青年/瓜子脸/五官精致小巧/面部肌肉柔和/温婉底色/大家闺秀气质|害羞/喜悦/面对心上人|抬眼含情脉脉,眉眼弯弯,苹果肌微鼓,扯笑,嘴角漾开甜蜜的弧度\n【配音指令:[图片2](凌雪吟 - 基础形态)|VoiceID:绿茶甜心撒娇小狐狸音|状态:放松/羞涩/回应|语气特点:音量偏小/语速轻快/气息带笑/尾音上挑/甜蜜羞涩。对白:没事。】\n描述:近景+居中构图|散射天光,顺光,暖光,柔和无阴影,[图片3](谢景珩 - 基础形态)缓缓松开扶着她的手,眼神温柔地看着她,轻声叮嘱。\n【表情指令】[图片3](谢景珩 - 基础形态)|男/青年/剑眉星目/轮廓分明/肌肉紧实/冷峻底色/清冷贵公子气质|放松/柔和/面对未婚妻|眼神含笑,目光专注,面部肌肉完全放松,唇部微动\n【配音指令:[图片3](谢景珩 - 基础形态)|VoiceID:冷酷霸总音|状态:放松/关切/叮嘱|语气特点:音量适中/语速平缓/气息稳/尾音收干净/温柔叮咛。对白:下次小心点。】\n\n附加限制要求:\n1. 画面中不要出现任何字幕、台词字卡、屏幕文字说明或底部文案。\n2. 不要添加背景音乐、配乐或音乐性音轨。", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "4453d0838455c86213c20a4a913c0f3fba61711fdaf15a2b8531e6b0ee1be1f9", "original_prompt_sha256": "ba8f9471e0e57b81c7eb1f60c8ca46981ad7e8c0a307488d139788e688bf80fc", "video_sha256": "c3892cfb2f12097e773c5eb70c81df566b1b3a6ab200ccee50d9d9cd751e4de3", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_089.json", "verification": "verification/sample_089.json"} {"file_name": "videos/sample_090.mp4", "sample_id": "#90", "prompt": "integrated_multimodal_description:\nGlobal Style: 8K ultra-realistic live-action cinematic texture, 60fps, ancient Chinese royal high-couture political drama. Clear facial pores, skin texture, hair layers, gold-woven fabric texture, jewelry cut gloss, metal hairpin texture, fox fur flow, python robe dark patterns. Cinematic large aperture shallow depth of field, unified scene lighting logic integrating soft light, top light, side backlight, rim light, and 10%-35% Tyndall effect. Ultra-sharp texture, high-contrast ancient style color grading (torment scenes low-saturation cold cyan-gray, counterattack scenes cold white high-sharp, highlight scenes warm gold, palace scenes gorgeous gold-red, northern border cold silver-white, tomb-sweeping scenes cold plain white). Natural dark vignette transition, enhancing broken, satisfying, high-end, and fatalistic feelings. Hyper-realistic human texture, clear pores, age and status-appropriate skin texture, no AI feel, enhanced real skin texture and high-end makeup. Consistent costumes and props throughout, strict royal etiquette, no cheap or incongruous elements. Maintain stable 180-degree axis for eyelines. Merciful action aesthetics: warm white soft light + sharp white contrast, high-healing kind color tone, large light ratio redemption contrast. Narrative emotion: firm, compassionate, capable, merciful. Soft light + side light mainly, 18% Tyndall effect, warm light for redemption.\n\n[Shot 1] Medium shot. Camera: shoulder-mounted handheld narrative breathing feel + cinematic calm breathing shot + slow pull back, extremely subtle soothing float. Scene: secluded alley next to a horse-drawn carriage, cold cyan-gray rainy night, continuous drizzle, clear cold rain light falling, heavy shadows on both sides of the alley, muddy puddles on the ground, 12% Tyndall effect penetrating the rain curtain. Characters: Shen Ruowei, Cousin-in-law, Shadow Guard. Prop: horse-drawn carriage. Composition: characters in the center. Positioning: Shen Ruowei stands at the carriage door, 2 meters from the camera; Cousin-in-law and Shadow Guard walk away into the distance. Action: Shen Ruowei watches the Cousin-in-law disappear into the night under the escort of the Shadow Guard. Emotion: calm eyes, with a trace of relief. Layered scene: Foreground: Shen Ruowei's back inside the carriage; Midground: Shadow Guard following the Cousin-in-law walking in the alley; Background: pitch-black alley, distant lanterns. Constraint: precise undistorted facial features, clear pores and skin texture, distinct hair strands, realistic fabric texture, no clipping/distortion, stable face, normal body structure, costumes identical to previous shots, no obstructing subtitles, no subtitles, no background music.\n\n[Shot 2] Action close-up. Camera: shoulder-mounted handheld narrative breathing feel + cinematic calm breathing shot + slow pull back to a full view of the carriage, small slow float. Scene: secluded alley next to a horse-drawn carriage, cold cyan-gray rainy night, continuous drizzle, clear cold rain light falling, heavy shadows on both sides of the alley, muddy puddles on the ground, 12% Tyndall effect penetrating the rain curtain. Character: Shen Ruowei. Prop: carriage curtain. Composition: hand and curtain in the center. Positioning: Shen Ruowei's hand is 0.5 meters from the camera. Action: slowly reaches out a hand, gently closes the carriage curtain. Emotion: slow movement, calm expression. Layered scene: Foreground: Shen Ruowei's back inside the carriage; Midground: Shen Ruowei's hand, carriage curtain; Background: pitch-black night. Constraint: no clipping, no flicker, clear details, no obstructing subtitles, no subtitles.\n\noverall_soundscape:\nNo redundant background BGM; minimal atmospheric SFX for key plot beats. Retain environmental soundscapes such as footsteps, fabric friction, jade pendant clinks, palace gate opening/closing, hoofbeats, armor clashing, wind and snow, plus this shot's howling cold wind, sound of fine rain dripping, distant dog barking, and slight sound of the carriage curtain closing.\n\nnon_diegetic_music:\nN/A", "original_prompt": "8K 超写实真人电影质感,60 帧,古风王侯高定权谋爽剧,脸部毛孔、皮肤纹理、发丝层次、织金面料肌理、珠宝切割光泽、金属钗环质感、狐裘毛流感、蟒袍暗纹清晰可见;无冗余背景 BGM,关键剧情配极简氛围感音效,无遮挡字幕,仅保留台词与环境声场(脚步声、布料摩擦、玉佩碰撞、宫门开合、马蹄声、甲胄碰撞、风雪声);电影级大光圈浅景深,场景化光影逻辑统一,融合柔光、顶光、侧逆光、轮廓光与 10%-35% 丁达尔光效;对白切换近景,人物对峙与情绪爆发用近景 / 特写,关键调度采用全景 / 中景,重要动作、道具、设计稿、面料、礼服使用特写,镜头自带情绪呼吸感,景别切换遵循爽剧节奏,保持 180° 轴线稳定;超清锐化质感,高对比古风色调,虐戏低饱和冷青灰、逆袭戏冷白高锐、高光戏暖金鎏光、宫廷侯府戏华贵金红、北疆战场凛冽银白、祭墓戏清冷素白,画面暗角自然过渡,强化破碎感、爽感、高级感与宿命感;超写实真人质感,毛孔清晰,具有符合相应年龄段与身份的肌理感,去掉人物 AI 感,增强皮肤真实质感与妆面高级感,高清写实人物质感,8K 分辨率;人物服化道全程统一不穿模,王侯礼制严谨,无廉价感与违和元素。\n慈悲出手美学,暖白柔光 + 银票锐白对冲,高治愈善良色调,大光比救赎反差;叙事情绪为坚定、怜悯、利落、慈悲;调度以赌场威逼与女主出手摆平对比、表嫂走投无路与绝境逢生对比;柔光 + 侧光为主,丁达尔光效 18%,暖光塑救赎、锐白点银两;运镜银票特写、女主扶起孕妇慢镜头、暗卫领命全景、表嫂感恩怼脸;暖白铺满强化善意救赎,锐白银票、厚实披风为标志性色彩,光影温暖凸显无辜不被辜负。\n【时长】2.5 秒\n【场次】第 42 场\n【镜号】31\n【景别】中景\n【摄法】肩扛手持叙事呼吸感 + 电影级平静呼吸镜头 + 缓慢拉远,极微小幅度舒缓浮动\n【场景】偏僻巷弄图1 - 马车旁图2,冷青灰雨夜,细雨绵绵,清冷雨光洒下,巷弄两侧阴影浓重,地面泥泞积水,12% 丁达尔光效穿透雨幕\n【出镜人物】沈若微图3、表嫂图4、暗卫图5\n【道具】马车\n【画面描述】\n1、构图:人物位于画面中央\n2、人物站位:沈若微图3站在马车门口,距离镜头 2 米;表嫂图4和暗卫图5向远处走去\n3、人物动作:核心动作:沈若微看着表嫂在暗卫的护送下消失在夜色中\n4、情绪描述:眼神平静,带着一丝释然\n5、分层场景:前景层:在马车内沈若微的背影;中景层:,暗卫跟着表嫂走在巷弄里;后景层:漆黑的巷弄,远处的灯笼\n【对白】无\n【音效】寒风呼啸声、细雨滴落声、远处的狗吠声\n【特效】无\n【约束】人物五官精准不崩坏,人物五官与参考保持一致,脸部毛孔皮肤纹理清晰,发丝根根分明,衣物纹理写实,无穿模无扭曲,人物面部稳定不变形,人体结构正常,服装样貌与前文完全一致,无字幕,无背景音乐\n【时长】2.5 秒\n【场次】第 42 场\n【镜号】32\n【景别】动作特写\n【摄法】肩扛手持叙事呼吸感 + 电影级平静呼吸镜头 + 缓慢拉远至马车全景,小幅缓慢浮动\n【场景】偏僻巷弄图1 - 马车旁图2,冷青灰雨夜,细雨绵绵,清冷雨光洒下,巷弄两侧阴影浓重,地面泥泞积水,12% 丁达尔光效穿透雨幕\n【出镜人物】沈若微图3\n【道具】马车车帘\n【画面描述】\n1、构图:手和车帘位于画面中央\n2、人物站位:沈若微的手距离镜头 0.5 米\n3、人物动作:核心动作:缓缓伸出手,将马车的车帘轻轻合上\n4、情绪描述:动作缓慢,神情平静\n5、分层场景:前景层:在马车里沈若微的背影;中景层:沈若微的手,车帘;后景层:漆黑的夜色\n【对白】无\n【音效】车帘合上的轻微声响、细雨滴落声\n【特效】无\n【约束】无穿模,无闪烁,细节清晰,无字幕", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "ccbc9487afa9b7c7c9edd30cbe0f6cbaf74f75b93a0e6f62342408763cdb230c", "original_prompt_sha256": "f2b69d792e8b8432331f67387208425b49ab93981c29ba8aa5bdcb89d724e605", "video_sha256": "03d02f25dca599ebf24034392857b5fb30d9b94c4203e10f586b5f0abb39ad0c", "task_adapted": true, "width": 768, "height": 1344, "fps": 24, "num_frames": 124, "duration_seconds": 5.166666666666667, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_090.json", "verification": "verification/sample_090.json"} {"file_name": "videos/sample_091.mp4", "sample_id": "#91", "prompt": "integrated_multimodal_description:\n[Shot 1] A renovation quotation sheet is shown. The text \"据实结算\" is written on it. An unspecified-gender narrator (S1) says in an off-screen voiceover [Chinese]装修报价单上写“据实结算”——这四个字等于“后期随便加钱”。水电加8000,防水加3000,美缝单独再收2000,最后预算翻一倍.\n\noverall_soundscape:\nUnspecified.\n\nnon_diegetic_music:\nN/A", "original_prompt": "装修报价单上写“据实结算”——这四个字等于“后期随便加钱”。水电加8000,防水加3000,美缝单独再收2000,最后预算翻一倍", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "848bd12df53f97ac0d327388c0488d1825301138de2bce75de20ff8dcf2abfb5", "original_prompt_sha256": "d389e4ea3cc2b1a18fa0ad7fa90f084bf23b1caad1cf3deb66673afca536aed6", "video_sha256": "2f751ef06aabaf196b03c5de45bfc99f2851381920692be19280818a4b8f1cc8", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_091.json", "verification": "verification/sample_091.json"} {"file_name": "videos/sample_092.mp4", "sample_id": "#92", "prompt": "subject_definitions:\n is a gasoline lawn mower, appearance and details from through .\n\nsummary:\n[reference generation] A professional product display video of a gasoline lawn mower, demonstrating cold start with one-pull success, engine close-ups, smooth idling, body welding details, thick metal shell, high-quality parts, and complete accessories in an outdoor setting with natural daylight, styled as a practical industrial tool for the Russian market.\n\nretention_analysis:\nweak_reference: The source prompt describes a gasoline lawn mower product display but does not specify how the 7 supplied images should be used, bound, or retained across shots. All 7 pictures are included as unspecified references.\n\ndetailed_description:\n[Shot 1] Realistic outdoor shooting, professional product display of , a gasoline lawn mower. Cold start demonstration, one pull and start successfully. Close-up of the engine metal structure, smooth idling operation, no black smoke. Show body welding details, thick metal shell, high-quality parts, complete accessories. Natural daylight, clear picture, high definition, 4K, real shooting texture, no exaggerated filter, industrial tool style, practical and reliable, Russian market product style.\n\noverall_soundscape:\nStable engine sound during idling operation, outdoor ambient environment.\n\nnon_diegetic_music:\nN/A", "original_prompt": "Realistic outdoor shooting, professional product display, gasoline lawn mower, cold start demonstration, one pull and start successfully. Close-up of the engine metal structure, smooth idling.\noperation, no black smoke, stable sound. Show body welding details, thick metal shell, high-quality parts, complete accessories. Natural daylight, clear picture, high definition, 4K, real shooting.\ntexture, no exaggerated filter, industrial tool style, practical and reliable, Russian market product style.", "reference_images": ["references/cebc964dd3be0499992f73c3fe92b1df38e9c9c8307727d205bd5e2300fae2c2.png", "references/5fd5596de09587326c37951aeab6ee948acbe91479a06fa88f1d3144220dee8f.png", "references/d3453e4768a53e6c916dd9a8e091a05d9c516ff629466414120ebaf641f85b8a.png", "references/87ed765ddf8ed1855ffe56b6ef6c46f0baf7e0c930896884e663f1b3fc50da12.png", "references/f25d1affda9e8c9f57b06d3ec340299d0759244684fea02c33845d1025362707.png", "references/4eda45537d2970dd4e3e7e24bc5ac9ab75c7f663d283e16fe3d5aaccb5290cf8.png", "references/916c9112e1fb525e204c23bef10b31fbb40ddf4302dd562d8424a000957ba8af.png"], "reference_sha256": ["cebc964dd3be0499992f73c3fe92b1df38e9c9c8307727d205bd5e2300fae2c2", "5fd5596de09587326c37951aeab6ee948acbe91479a06fa88f1d3144220dee8f", "d3453e4768a53e6c916dd9a8e091a05d9c516ff629466414120ebaf641f85b8a", "87ed765ddf8ed1855ffe56b6ef6c46f0baf7e0c930896884e663f1b3fc50da12", "f25d1affda9e8c9f57b06d3ec340299d0759244684fea02c33845d1025362707", "4eda45537d2970dd4e3e7e24bc5ac9ab75c7f663d283e16fe3d5aaccb5290cf8", "916c9112e1fb525e204c23bef10b31fbb40ddf4302dd562d8424a000957ba8af"], "workflow": "ref2va", "prompt_sha256": "de3ed1b7327c82ba02265ff317c107135e753282c520681f1f98a2b88fb28639", "original_prompt_sha256": "06d9a15516c75a1c2ab787ddbee5c56ac2b36355f2b6f7c4fa7d9016d4a52dc1", "video_sha256": "81215d95a4cc2b1b6158fb47e58aca2e3982b05ed07cec0b63f1d6e001a41063", "task_adapted": false, "width": 768, "height": 768, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_092.json", "verification": "verification/sample_092.json"} {"file_name": "videos/sample_093.mp4", "sample_id": "#93", "prompt": "integrated_multimodal_description:\n[Shot 1] From 00:00.000 to 00:03.000, hard cut. medium close-up, eye-level, handheld slight shake. The corridor is crowded with people, illuminated by cold white light. Xu Jia holds her textbooks tightly against her chest, her shoulders hunched, taking extremely light steps, trying to pass through the edge of the crowd. She lowers her head and quickens her pace, wrapping her arms tighter around the textbooks. She walks close to the wall, the textbooks pressed to her chest, her hair swaying slightly from the flow of the crowd. In the foreground, Chen Qian and a group of classmates form a semi-encirclement, their gazes focused on Xu Jia. The background is blurred into patches while the characters remain in sharp focus. Characters blink naturally at a rate of 15 times per minute, without staring. No subtitles are displayed. The scene transitions from the damp memory of the restroom to the public humiliation in the corridor, with the emotion continuing to sink.\n[Shot 2] From 00:03.000 to 00:06.000, no transition. medium close-up, slightly low-angle, slow push-in. Chen Qian stands in front of the crowd, extending his arm with exaggerated movements, a contemptuous expression on his lips, with classmates behind him forming a blocking background. Chen Qian takes the lead in pointing at Xu Jia, and the crowd's gazes follow. His pointing finger and Xu Jia's lowered head form a strong directional contrast. Chen Qian raises his hand to point in Xu Jia's direction, and the surrounding crowd joins in in taunting and booing, making the air in the corridor feel even more oppressive. The background is blurred into patches while the characters remain in sharp focus. Characters blink naturally at a rate of 15 times per minute, without staring. No subtitles are displayed. Characters are speaking and taunting.\n\noverall_soundscape:\nCorridor footsteps, snickering, textbook rustling, scolding voices, booing and taunting voices, corridor echoes. Characters are speaking and taunting.\n\nnon_diegetic_music:\nN/A (Music is explicitly prohibited).", "original_prompt": "【场景分析】\n场景:(现代学校走廊/日内喧闹)\n承接:厕所的阴湿记忆未断,画面顺势转入走廊的公开羞辱\n过渡:硬切,增强围观压力与空间压迫感\n人物表演底层逻辑总纲:【陈乾处于带头挑衅与群体煽动状态;许葭处于低头忍受与快速穿行状态】\n【分镜承接】\n前一分镜末尾画面:厕所潮湿角落中,许葭抱紧自己\n本分镜开场画面:喧闹走廊里,许葭抱着课本从人群边缘穿过\n承接逻辑:同一记忆链条中的空间转换,情绪持续下坠\n【镜头1】0.0-3.0秒\n转场: 硬切\n镜头: 近景/平视/手持微晃\n画面描述:走廊人流拥挤,光线冷白,许葭紧抱课本,肩膀缩起,脚步极轻,试图从人群边缘穿行。\n核心动作: 她低头加快步伐,双臂把课本抱得更紧。\n对手戏设计: 前景中陈乾与一群同学形成半包围,视线集中在许葭身上。\n人物底层逻辑:试图以最小存在感穿过公开羞辱。\n主体动作:许葭贴着墙侧前行,课本压在胸前,发丝被人流带起轻微晃动。\n音效:走廊脚步声、窃笑声、课本摩擦声\n镜头2】3.0-6.0秒\n转场: 无\n镜头:近景/略仰视/缓慢推近\n画面描述: 陈乾站在人群前方,手臂伸出,动作夸张,嘴角带着轻慢神情,身后同学形成围堵式背景。\n核心动作: 陈乾带头指向许葭,众人目光跟随。\n对手戏设计: 陈乾的手指与许葭低垂的头形成强烈方向对冲。\n人物底层逻辑:通过公开指责建立群体优势。\n主体动作: 陈乾抬手指向许葭的方向,周围人群随之起哄,走廊的空气被压得更紧。\n音效: 指责声、起哄声、回廊回声\n资产对照表(视频中涉及的角色、场景与物品如下(保持一致性)):\n视频场景对照表:现代学校走廊=【@现代学校走廊/日内喧闹】\n视频角色对照表:许葭=【图1】;陈乾=【图2】\n禁止字幕和音乐,无字幕,但要有人物台词和音效。\n背景虚化成斑块,人物清晰。\n人物眨眼频率每分钟15次。自然不刻意。不瞪眼。", "reference_images": [], "reference_sha256": [], "workflow": "t2va", "prompt_sha256": "1a53d67d855bc554406ded620b005727ada1485d86596d6e84b037478b77058c", "original_prompt_sha256": "cd5e5e24e1572c9ddb75a308b655d374da9a0ff20dad98da865b27d1381e5301", "video_sha256": "616eb0e8fe6cca112a417b0644da12c4ad506e54541261e212c1fc3f874d35b3", "task_adapted": true, "width": 768, "height": 1024, "fps": 24, "num_frames": 158, "duration_seconds": 6.583333333333333, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_093.json", "verification": "verification/sample_093.json"} {"file_name": "videos/sample_094.mp4", "sample_id": "#94", "prompt": "subject_definitions:\n (Yang's mother): appearance and clothing from .\n (Yang Xiaomei): appearance and clothing from .\n (Hu Yating): appearance and clothing from , with a slightly protruding belly.\n (Zhao Shuyue): appearance and clothing from .\n\nsummary:\n[reference generation] A strong-emotion, high-conflict short drama scene set in a courtyard at dawn (early morning twilight). The scene uses indoor natural light with warm tones. Visual and structural constraints: 8K ultra-HD resolution, ultra-high detail, cinematic picture quality, rich detail, clean and stable image; 24 fps cinematic feel, smooth natural motion without stutter or motion smear; no model breakdown, no goofs, no flash-black or dissolve transitions, cuts only at specified shot boundaries; no distortion, no clipping, no extra structures or parts, stable unified forms, no deformation, no floating or broken elements, and no watermarks. (Yang's mother) and (Yang Xiaomei) initially stand side by side at the main room door (position reference from ). (Zhao Shuyue) is initially inside the kitchen. (Hu Yating) is initially inside the right-side room. The sequence shows scolding about money, walking out with morning sickness, and responding politely from the kitchen. No subtitles, no logos, no garbled text, no background music. Strict 100% scene element restoration with spatial logic. No duplicate characters or same-face multiples in any single frame.\n\nretention_analysis:\n (Yang's mother) from is fully_preserved as a speaking character with specific blocking and dialogue. (Yang Xiaomei) from is fully_preserved as a silent character receiving a glare. (Hu Yating) from is fully_preserved with the added detail of a slightly protruding belly, performing physical actions and speaking. (Zhao Shuyue) from is fully_preserved as a speaking character in the kitchen. is partially_preserved as a positional reference for the initial standing arrangement of and at the main room door. has an unspecified role as it is supplied but not explicitly referenced in the source prompt.\n\ndetailed_description:\n[Shot 1] Medium close shot, fixed camera. (S1) glares at and says [Chinese]钱钱钱!家里哪有那么多钱?找你二哥要去!. The courtyard is lit by early morning natural light with warm tones. and are positioned at the main room door.\n\n[Shot 2] Medium close shot, fixed camera. (S2) with a slightly protruding belly slowly walks out from the right-side room and deliberately dry heaves several times in the courtyard. No dialogue.\n\n[Shot 3] Close-up, fixed camera. (S2) delicately wipes her mouth and says [Chinese]这早起反应就是大。弟妹,早饭做好了吗?我饿得心慌。 in a delicate manner.\n\n[Shot 4] Medium close shot, fixed camera. (S3) peeks her head out from the kitchen with a subtle 'respectful' smile on her face. No dialogue.\n\n[Shot 5] Medium close shot, fixed camera. (S3) with a smile on her lips says [Chinese]快了,嫂子再等会儿。 in a calm tone. The shot ends as a reaction medium close shot of (S3).\n\noverall_soundscape:\nEarly morning courtyard ambient sounds including light environmental noise. Character dialogue and physical sounds such as dry heaving and footsteps are present. No background music.\n\nnon_diegetic_music:\nN/A", "original_prompt": "无文字,无 logo,无字幕,无乱码,8K 超高清,超高分辨率,极致细节,电影级画质,细节丰富,画面干净稳定风格:流畅自然,无卡顿,24 帧电影感,动作顺滑无卡顿:无扭曲,无穿模,无多余结构:形态规整,结构完整,无变形无穿模,无断裂无漂浮,无多余部件,造型稳定统一无水印。《80年代,我不当软柿子》11-1\n强情绪强冲突短剧风格。不要自动生成字幕和bgm,只保留环境音和角色对话,目前的场景为院子,目前的天色为清晨(天蒙蒙亮)。全程无字幕,无背景音乐、无卡顿、拖影、模型崩坏、穿帮,无闪黑、叠化等转场;严格执行场景元素 100% 还原规则,人物与场景位置关系符合空间逻辑。光影:室内自然光,暖色调。赵舒悦【图片2】初始位于灶房内,杨母【图片3】跟杨晓玫【图片4】初始位于堂屋门口并肩站立(位置参考【图片6】)。胡雅婷【图片5】初始位于右侧屋内。\n近景,固定机位,杨母(瞪了杨晓玫一眼):“钱钱钱!家里哪有那么多钱?找你二哥要去!”(发声人必须为杨母)\n近景,固定机位,胡雅婷挺着微凸的肚子从右侧屋内慢悠悠走出来,故意在院子里干呕了几声。(无台词)\n特写,固定机位,胡雅婷(娇弱地擦了擦嘴):“这早起反应就是大。弟妹,早饭做好了吗?我饿得心慌。”(发声人必须为胡雅婷)\n近景,固定机位,赵舒悦从灶房探出头,脸上带着一丝“恭顺”的笑。(无台词)\n近景,固定机位,赵舒悦嘴角挂着微笑(语气平静):“快了,嫂子再等会儿。”(发声人为赵舒悦)(结束镜头为赵舒悦的反应近景镜头),视频全程不要在同一画面中复制相同人物,不要多人同脸", "reference_images": ["references/e2ec16395656c475fed926335b43fefbc74aca51d8e2af21d299917ea58493ca.jpg", "references/d27b0e937d03a1895006a6fa030c6c595d332ce98b1899233f1ed89c264d1bc0.jpg", "references/a130fd3d45fefb1c7dc9169f3baeada19b6d18ae8f120d20e91ab3479795b759.jpg", "references/d2540235d47fecbf108c439038a84aa129b0fa2651541e6d8bf611229795c760.jpg", "references/9b25504506321b00587c36befc5eb807e5b1dbb47c5061e7175dbb2c71e019c8.jpg", "references/768b2367d711bf9448bd75d5015b221d00a9c35089393d74a574beafe54e0d34.jpg"], "reference_sha256": ["e2ec16395656c475fed926335b43fefbc74aca51d8e2af21d299917ea58493ca", "d27b0e937d03a1895006a6fa030c6c595d332ce98b1899233f1ed89c264d1bc0", "a130fd3d45fefb1c7dc9169f3baeada19b6d18ae8f120d20e91ab3479795b759", "d2540235d47fecbf108c439038a84aa129b0fa2651541e6d8bf611229795c760", "9b25504506321b00587c36befc5eb807e5b1dbb47c5061e7175dbb2c71e019c8", "768b2367d711bf9448bd75d5015b221d00a9c35089393d74a574beafe54e0d34"], "workflow": "ref2va", "prompt_sha256": "9e846ffeda41763b6331524e976955c2e927adea47380f8c46a1b9e9f43c8ba9", "original_prompt_sha256": "87984d6b6c15e6dfeef3a92fe282b3e0bf4c2b6df42866645862451183b71ae5", "video_sha256": "a67ebb81787dc46e5691118cbf07e7999597ee56a4459b73aa028e7efbf8a0ed", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_094.json", "verification": "verification/sample_094.json"} {"file_name": "videos/sample_095.mp4", "sample_id": "#95", "prompt": "subject_definitions:\n is Hua Tongtong, whose appearance and clothing are from . is Feng Xiao, whose appearance and clothing are from . is Lu Shibai, whose appearance and clothing are from . is Assistant Li, whose appearance and clothing are from . The setting is the auction house from .\n\nsummary:\n[reference generation] A multi-shot sequence at an auction house featuring Hua Tongtong, Feng Xiao, Lu Shibai, and Assistant Li. The scene establishes spatial positioning, followed by a tense dialogue exchange where Feng Xiao confronts Hua Tongtong, Lu Shibai provocatively interrupts, and Feng Xiao insults Lu Shibai, ending with Hua Tongtong's dazed reactions.\n\nretention_analysis:\nfully_preserved: The identities and appearances of Hua Tongtong (), Feng Xiao (), Lu Shibai (), Assistant Li (), and the auction house setting () are fully retained as character and location references throughout the shot sequence.\n\ndetailed_description:\n[Shot 1] 0-2s: Wide shot, fixed camera. Warm top lighting, luxurious and solemn. The image is clean and high-end, with clear foreground, midground, and background depth progression, extreme emotional tension, and physical detail realism. No watermarks, no subtitles, no logos. stands on the left side of the black seats facing . sits in the deep background. stands next to in the deep scene. Other guests in high-end clothing stand around. Establishes space and character positions.\n\n[Shot 2] 2-5s: Medium shot, eye-level, fixed camera. Warm top lighting, luxurious and solemn. adjusts her hairstyle with one hand while slowly sitting down, her throat bobbing as she swallows saliva, her eyes dodging as she stutters, her chest heaving from nervousness. (S1) says [Chinese]为……为我而来,我正忙着呢。 in a stuttering, surprised, and guilty voice. Shot-reverse-shot dialogue, axis unchanged.\n\n[Shot 3] 5-7s: Close-up, low angle, fixed camera. Warm side lighting, luxurious and solemn. frowns slightly (20% intensity), mouth corners pressed down (10% intensity), peripheral vision scanning like a blade towards the left out of frame. (S2) says [Chinese]你忙什么?忙着抛下老公和野男人私会? in an extremely angry, questioning, cold, and hard voice. Oppressive sound effect. Note: source requests high angle to highlight male lead's oppression.\n\n[Shot 4] 7-9s: Medium shot, eye-level, lateral camera translation. Warm top lighting, luxurious and solemn. is on the left, stands in front of , 's back blocking . The camera translates right to leaning lazily against the seat back, legs crossed, mouth forming a shallow smile of interest. Eyes full of provocation and playfulness, one eyebrow raised, fingers rhythmically tapping the seat armrest. (S3) says [Chinese]好酸的味道啊!桐桐,你闻到了吗? in a teasing and provocative voice. Slight clothing rustle sound. Spatial continuity.\n\n[Shot 5] 9-10s: Close-up, eye-level, slow push in. Warm top lighting, luxurious and solemn. sits upright in the center, eyes rolling left and right, brows slightly furrowed, mouth slightly open, expression dazed and at a loss. (S1) says [Chinese]啊? in a dazed and confused voice. Reaction shot.\n\n[Shot 6] 10-13s: Close-up, over-the-shoulder perspective, low angle. Warm side lighting, luxurious and solemn. Over-the-shoulder shot including 's shoulder. 's right shoulder is out of focus on the left side of the frame. on the right stands staring dead at down in the lower left, mouth forming an extremely cold and disgusted arc. (S2) says [Chinese]我看,是有人死皮赖脸,臭虫一样追着别人老婆, in an extremely disgusted and mocking voice. Over-the-shoulder confrontation establishes power dynamics.\n\n[Shot 7] 13-15s: Close-up, eye-level, slow push in. Warm top lighting, luxurious and solemn. remains in her original seat, eyes slightly widened (10% intensity), gaze shifting restlessly, subconsciously leaning back, eyes full of confusion and guilt. (S2) says in an off-screen voiceover [Chinese] @花桐桐 ,你闻到臭味了吗? in a flat, slightly cold voice. (S1) says [Chinese]啊? in an even more dazed voice. 's second line of dialogue lands on the female lead's reaction.\n\noverall_soundscape:\nAmbient noisy crowd sounds in the first shot, an oppressive sound effect during the third shot, and slight clothing rustle sounds during the fourth shot. No other specific sound effects are mentioned.\n\nnon_diegetic_music:\nN/A", "original_prompt": "人物站位: @花桐桐 端坐在 @拍卖会现场 左侧的黑色座椅上, @陆时白 坐在 @花桐桐 旁边画面深处, @凤骁 与 @李特助 站在右侧压迫感十足地逼近。周围站着其它服装高级的宾客\n要求:画面干净又高级;画面里前中远景递进关系分明。极具情绪张力与物理细节真实感。\n画面:无水印,无字幕,无logo 。\n镜号:1 | 场次:日内 | @拍卖会现场 | 0-2s | 全景| 固定镜头 | 暖色调顶光,奢华庄重, 画面左边 @花桐桐 站着面对 @凤骁 , @陆时白 坐在后侧, @李特助 站在 @凤骁 旁边场景深处, | 台词 & 音效:Dialogue:无 SFX:环境嘈杂声 | 备注:建立空间与人物方位\n镜号:2 | 场次:日内 | @拍卖会现场 | 2-5s | 中景,平视 | 固定镜头 | 暖色调顶光,奢华庄重, @花桐桐 一只手调整发型,一边慢慢坐下,喉结滚动吞咽了一口口水,眼神闪躲着结巴,胸口因为紧张而起伏。 | 台词 & 音效:Dialogue:[用结巴诧异的声音说] @花桐桐 (惊诧/心虚/结巴):\"为……为我而来,我正忙着呢。\" SFX:无 | 备注:对话正反打,轴线不变\n镜号:4 | 场次:日内 | @拍卖会现场 | 5-7s | 近景,仰拍 | 固定镜头 | 暖色调侧光,奢华庄重, @凤骁 眉头微皱程度20%,嘴角下压程度10%,余光如刀刃般扫向画外左侧。 | 台词 & 音效:Dialogue:[用极度愤怒的声音质问] @凤骁 (愤怒/质问/冷硬):\"你忙什么?忙着抛下老公和野男人私会?\" SFX:压抑音效 | 备注:俯视凸显男主压迫感\n镜号:5 | 场次:日内 | @拍卖会现场 | 7-9s | 中景,平视 | 横向平移 | 暖色调顶光,奢华庄重, 花桐桐在左,凤骁站在花桐桐前面,凤骁背影遮住了花桐桐,镜头向右平移至 @陆时白 慵懒地靠在座椅靠背上,双腿交叠,嘴角勾起一抹饶有兴致的浅笑。 眼神充满挑衅与玩味,挑起一边眉毛,手指有节奏地敲击着座椅扶手。 | 台词 & 音效:Dialogue:[用调侃挑衅的声音说] @陆时白 (调侃/挑衅):\"好酸的味道啊!桐桐,你闻到了吗?\" SFX:轻微的衣物摩擦声 | 备注:空间连贯\n镜号:6 | 场次:日内 | @拍卖会现场 | 9-10s | 特写,平视 | 缓慢推进 | 暖色调顶光,奢华庄重, @花桐桐 端坐在中间,左右来回转了一下眼球,眉头微蹙,嘴巴微微张开,表情呆萌且不知所措。 | 台词 & 音效:Dialogue:[用发懵的声音说] @花桐桐 (发懵/疑惑):\"啊?\" SFX:无 | 备注:反应镜头\n镜号:8 | 场次:日内 | @拍卖会现场 | 10-13s | 近景,过肩视角 | 仰拍镜头 | 暖色调侧光,奢华庄重, 带着 @陆时白 肩膀的过肩镜头。@陆时白 的右肩膀在画面左侧虚焦,右侧的 @凤骁 站着死死盯着画左下的 @陆时白 ,嘴角扯出极其冰冷的嫌恶弧度, | 台词 & 音效:Dialogue:[用极其厌恶的声音说] @凤骁 (厌恶/嘲讽):\"我看,是有人死皮赖脸,臭虫一样追着别人老婆,\" SFX:无 | 备注:越肩对峙建立权力感\n镜号:9 | 场次:日内 | @拍卖会现场 | 13-15s | 特写,平视 | 缓慢推进 | 暖色调顶光,奢华庄重, @花桐桐 依旧坐在原位,眼睛微微瞪大程度10%,视线飘忽不定,下意识向后靠了靠,眼底写满了懵逼与心虚。 | 台词 & 音效:Dialogue:[用平淡微冷的声音说] @凤骁 (VO):\" @花桐桐 ,你闻到臭味了吗?\" / [用更加懵逼的声音说] @花桐桐 (懵逼):\"啊?\" SFX:无 | 备注: @凤骁 后半句台词落到女主反应上\n\n参考图片说明:\n图1: 角色: 花桐桐\n图2: 角色: 凤骁\n图3: 角色: 陆时白\n图4: 角色: 李特助\n图5: 场景: 拍卖会现场", "reference_images": ["references/13697ebd85fbc2ce230f803bfc077e4942d34bac0ca27a3ec9dc1fa5375ee34d.jpg", "references/cbc34d5d1a11bfc1a611b5ce1384072df6efd3003cdb1c14630c3b68a6e8a402.jpg", "references/1ecc687fce2a6f2e93b1b4be2c47f2a898449b3b9f78819283ab1a1875e49bfa.jpg", "references/38f8868015e4b08bade18fc419cce7bef4d71fc56fb820d751ce7f98ffbd1f9c.jpg", "references/0f85c1d50ac8e43f0a30efdfaea2506227c01884308bf0b286c3758003c9ff0b.jpg"], "reference_sha256": ["13697ebd85fbc2ce230f803bfc077e4942d34bac0ca27a3ec9dc1fa5375ee34d", "cbc34d5d1a11bfc1a611b5ce1384072df6efd3003cdb1c14630c3b68a6e8a402", "1ecc687fce2a6f2e93b1b4be2c47f2a898449b3b9f78819283ab1a1875e49bfa", "38f8868015e4b08bade18fc419cce7bef4d71fc56fb820d751ce7f98ffbd1f9c", "0f85c1d50ac8e43f0a30efdfaea2506227c01884308bf0b286c3758003c9ff0b"], "workflow": "ref2va", "prompt_sha256": "ce9f191a51e0c50e8845d95037828eece1ad10e6dc787a5543aac4bf249cfcba", "original_prompt_sha256": "21dcd33955f24055bca9495654a8bcf89d83f70fd46751ab781b22c38660cf1f", "video_sha256": "0130fe0e96847ce020a11b9a41be5a901aec0ee18566b679084b89cd9548d04e", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_095.json", "verification": "verification/sample_095.json"} {"file_name": "videos/sample_096.mp4", "sample_id": "#96", "prompt": "subject_definitions:\nSubject 1: The female protagonist, appearance and clothing from .\nSubject 2: The male alpha Tristan Veyne, appearance and clothing from .\n\nsummary:\n[reference generation] A blurred memory sequence of a young werewolf couple transitions into reality, where the female protagonist reflects on her past while touching her neck.\n\nretention_analysis:\n is fully preserved as the female protagonist, appearing as a silhouette in the memory and in the reality scene. is fully preserved as the male alpha Tristan Veyne, appearing as a silhouette in the memory. is partially preserved as the implied background setting of the ceremonial hall.\n\ndetailed_description:\n[Shot 1] Medium shot silhouette, eye-level angle, slow lateral camera translation. The screen displays a blurred, overlapping memory, revealing the hazy silhouettes of a young werewolf couple. The male alpha, , is bowing down in the classic posture to mark the female protagonist, , as his pack mate. The lighting and color tone feature soft and hazy light, a light gold filter, and blurred highlights, creating an atmosphere of yearning, a sense of destiny, and a beautiful past, rendered in a hyper-detailed cinematic blur with a dreamy effect and romantic foresight. returns from the phantom image to reality, slowly raising her right hand, her fingertips gently brushing over her empty right neck. The lighting shifts to daytime with hard side window light and a cold cyan color tone. A young female voice (S1) says in an off-screen voiceover with a mocking state and a cold, hard tone, [English]The day @Tristan Veyne would mark me as his Luna. No subtitles are displayed.\n\noverall_soundscape:\nNo sound effects.\n\nnon_diegetic_music:\nN/A", "original_prompt": "分镜序号: 18,\n镜头: 【中景剪影,平视,缓慢平移镜头】 画面虚化叠忆,浮现一对少年狼人男女的朦胧剪影,@Tristan Veyne 男主俯身欲对女主 @Protagonist 完成族群伴侣标记的经典姿势光影色调:柔光朦胧、浅金滤镜、虚化高光氛围:憧憬、宿命感、美好过往\nAI提示词:blur memory silhouette, young werewolf couple, male alpha bowing to mark his luna, hazy golden soft light, dreamy blurred effect, romantic foresight, hyper detailed, cinematic blur @Protagonist 从虚影中回到现实缓缓抬起右手,指尖轻轻抚过自己空荡荡的右侧脖颈。(光线:白天,侧窗硬光,冷青色调) \n\n\n音效: 无\n配音指令: 【旁白VO|年轻女声|状态(嘲讽)|语气(冷硬):The day @Tristan Veyne would mark me as his Luna.】\n备注: no srt, no music\n\n参考图片说明:\n图1: 角色: Protagonist\n图2: 角色: Tristan Veyne\n图3: 场景: 仪式大厅", "reference_images": ["references/af080c5fe4e218d9bd55e7087bd58a9078934836dda80321290aed2d4f0610b6.jpg", "references/3f580bbbe3f7c7777e9ca3785e73086f5bfdaf032734ceb2e3566ca670fd8710.jpg", "references/ac6daa3a06168314694c47c546e7ecf071efdf00eaa684704f4f3a66cdf2c02d.jpg"], "reference_sha256": ["af080c5fe4e218d9bd55e7087bd58a9078934836dda80321290aed2d4f0610b6", "3f580bbbe3f7c7777e9ca3785e73086f5bfdaf032734ceb2e3566ca670fd8710", "ac6daa3a06168314694c47c546e7ecf071efdf00eaa684704f4f3a66cdf2c02d"], "workflow": "ref2va", "prompt_sha256": "8f29dd319bda254c5b8a9c4d7c3db5441d0935e19be7ed029dee101c45939988", "original_prompt_sha256": "f44ea202a5cfd72a5994db43682e0f29640b0a3ea8d3a54683778ee469295a60", "video_sha256": "1db51cb885db3211a42c0cea43704269b027d8f34a67732402edce071894321c", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 345, "duration_seconds": 14.375, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_096.json", "verification": "verification/sample_096.json"} {"file_name": "videos/sample_097.mp4", "sample_id": "#97", "prompt": "subject_definitions:\n: The female lead. Her appearance and clothing are derived from . She has white hair, pale skin, and wears a white shirt with lace cuffs. A strict global constraint applies to her face: her mouth must remain completely closed and still throughout the entire video, with absolutely no lip movement, opening, closing, teeth showing, or tongue movement. Any facial expression changes must not affect her lips or jaw.\nA separate off-screen female voice, independent of the visible female lead, delivers the specified voiceover lines and is designated (S1).\n\nsummary:\n[reference generation] A high-quality, semi-realistic anime style video featuring a female lead with white hair lying in bed. The video consists of three fixed shots showing her waking up, testing her hand movement, and reacting to an off-screen sound, all while her mouth remains strictly closed. A female off-screen voiceover delivers the specified lines independently of the visible female lead. The visual style features clean line art, soft cel shading, and high-definition details. Strictly no subtitles or on-screen text are allowed.\n\nretention_analysis:\n is fully preserved as the reference for the scene, lighting, atmosphere, and overall style details. is fully preserved as the reference for the female lead's appearance, identity, and clothing.\n\ndetailed_description:\nGlobal constraint: No subtitles or on-screen text are allowed throughout the entire video. [Shot 1] The camera is positioned directly above, looking vertically down at lying on a large bed pillow. The camera is fixed with no movement. The scene, lighting, and atmosphere are derived from . Her white hair is scattered across the pillow. Her eyes slowly open slightly, her gaze hazy and dazed. A deep purple velvet quilt covers her up to the abdomen. Her exposed arms wear a white shirt with lace cuffs. Cool moonlight outlines her facial contours. Her lips remain completely closed and still, with no opening, closing, or movement; only her eyes open and her gaze changes. A separate off-screen female voice (S1) says in an off-screen voiceover in a low voice [Chinese]……这里是哪里 while the on-screen female lead's lips remain completely closed and motionless.\n\n[Shot 2] The camera is positioned horizontally at the side of the bed, focusing on 's hand exposed outside the deep purple velvet quilt. The hand has lace cuffs. The background is the deep purple velvet quilt surface. The camera is fixed with no movement. Her hand tentatively slightly clenches into a fist, then her wrist slowly rotates once, emphasizing that her wrist is completely bare with no binding marks. The light highlights her pale skin texture. Her mouth remains completely closed and still, maintaining the global constraint. A separate off-screen female voice (S1) says in an off-screen voiceover softly [Chinese]身体……能动了 while the on-screen female lead's lips remain completely closed and motionless.\n\n[Shot 3] The camera cuts back to a direct vertical top-down view of 's face, with composition consistent with Shot 1. The camera is fixed with no movement. First, she maintains the same static state as Shot 1. Then, an extremely soft, lazy male cough is heard from off-screen in the upper diagonal direction. Upon hearing this, 's facial expression suddenly tightens, her eyes flashing with alertness and slight tension. Her lips remain completely closed and completely still. She then maintains a guarded expression and extremely slowly turns her head to the right side of the frame, her long hair gently rubbing against the pillow. The warm light from a wall lamp and the cool shadows of the moonlight interweave and flow across her turning face. A separate off-screen female voice (S1) says in an off-screen voiceover in a tense whisper [Chinese]……有人 while the on-screen female lead's lips remain completely closed and motionless.\n\noverall_soundscape:\nIn Shot 3, an extremely soft, lazy male cough comes from an off-screen upper diagonal direction. As the female lead slowly turns her head, her long hair softly rubs against the pillow. No other ambience or physical sounds are specified.\n\nnon_diegetic_music:\nN/A", "original_prompt": "**全程约束(优先级最高)** \n\n- 女主:嘴部始终保持完全闭合、静止,**绝无任何口型、嘴唇开合、牙齿露出或舌头动作**。面部任何表情变化(警惕、紧张、转动视线等)均不得牵动嘴唇。 \n\n- 声音:所有台词均为**女性画外音(女声OS)**,不来自画面内女主。禁止出现任何画面字幕。 \n\n- 风格:高质量二次元插画,干净线稿,半写实动漫风,柔和赛璐璐阴影,高清细节。 \n\n- 参考:@图片1 提供场景、光影、氛围、风格细节参考;@图片2 确定女主形象。\n\n**镜头1:客观面部特写(俯拍)** \n\n- 机位:摄像机从正上方垂直俯拍,镜头固定,无运镜。 \n\n- 画面:女主躺在⟨场景1⟩大床枕头上,白发散乱铺枕,双眼缓缓微睁,眼神迷离恍惚。深紫色天鹅绒被子盖至腹部,露出的手臂穿白衬衫、蕾丝袖口。清冷月光勾勒面部轮廓。 \n\n- 女主动作:**嘴唇全程紧闭,无任何开合**。只可睁眼、眼神变化。 \n\n- 画外音(女声OS,无字幕):低声「……这里是哪里」\n\n**镜头2:侧面平拍局部特写** \n\n- 机位:镜头从床侧水平平视,聚焦女主露在被子外的手(带蕾丝袖口),背景为深紫色天鹅绒被面。镜头固定,无运镜。 \n\n- 画面:手试探性微握成拳,随后手腕缓慢转动一下,强调手腕空无一物(无束缚痕迹)。光线突出苍白皮肤质感。 \n\n- 女主动作:**嘴部依然完全闭合静止**(镜头虽不拍脸,但必须确保全片统一约束)。 \n\n- 画外音(女声OS,无字幕):轻声「身体……能动了」\n\n**镜头3:客观面部特写(俯拍)** \n\n- 机位:切回正上方垂直俯拍女主面部,构图与镜头1一致。镜头固定,无运镜。 \n\n- 画面: \n\n - 先保持与镜头1相同的静态。 \n\n - 听到斜上方画外传来的**一声极其轻微、慵懒的男声轻咳**(此声源为画外,非女主)。 \n\n - 女主听到后,面部表情骤然紧绷,眼神闪过警惕与微微紧张。**嘴唇保持完全闭合,纹丝不动**。 \n\n - 随后她保持防备神情,**极为缓慢地将头转向画面右侧**,长发在枕上轻轻摩擦。壁灯暖光与月光冷影在转动的脸庞上交错流转。 \n\n- 画外音(女声OS,无字幕):紧绷低语「……有人」\n\n**全片额外强调** \n\n- 女主的任何表情(迷离、紧张、警惕)及转头动作,均**不得引发嘴唇或下颌的任何运动**。 \n\n- 若生成过程中出现口型或嘴唇开合,视为严重错误,必须避免。 \n\n- 无字幕,无任何画面文字。", "reference_images": ["references/0e68efb35bd6aa68a3f5112faa4c39432d352a0b0f7a7d6a8187270cde8635c9.jpg", "references/0e68efb35bd6aa68a3f5112faa4c39432d352a0b0f7a7d6a8187270cde8635c9.jpg"], "reference_sha256": ["0e68efb35bd6aa68a3f5112faa4c39432d352a0b0f7a7d6a8187270cde8635c9", "0e68efb35bd6aa68a3f5112faa4c39432d352a0b0f7a7d6a8187270cde8635c9"], "workflow": "ref2va", "prompt_sha256": "a17769a29e81f44eaeb60ff626cb7a9ccdc2578004a263433c89fd6215cac5e5", "original_prompt_sha256": "90343b589c47a1c6615b07f39444b2fb0cb71abc1247ed36e66c59ac7e4198f5", "video_sha256": "50a603242e356c5a2a2d3bae0a15c421f66026e6a71fa6e0ffe2e64982b125c6", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 277, "duration_seconds": 11.541666666666666, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_097.json", "verification": "verification/sample_097.json"} {"file_name": "videos/sample_098.mp4", "sample_id": "#98", "prompt": "subject_definitions:\n: The Fox Ancestor (head only, body completely hidden in smoke), with eyes burning like flames, appearance based on .\n: A small fox that transforms into a yellow-clothed woman, appearance based on .\n: The environment and overall visual reference for the Fox Ancestor, based on .\n\nsummary:\n[reference generation] IMAX 70mm film aesthetic, annual box-office-champion epic positioning. 4K HD, cinematic-grade texture, smooth natural character motion, no distortion or collapse, realistic physical logic, follows natural physical friction. A cinematic, epic scene in low saturation with a large aperture, following the visual style of . A small fox looks up at a pillar as dark smoke gathers into a vortex at the top, revealing the head of the Fox Ancestor with burning eyes. The Fox Ancestor blows a pale red halo at the fox, transforming it into a yellow-clothed woman, while the Fox Ancestor watches her from the vortex. Follow practical on-location light sources.\n\nretention_analysis:\n (Fox Ancestor's head) is fully preserved as the central figure emerging from the smoke. (small fox) is fully preserved as the initial subject that transforms into a yellow-clothed woman. (environment reference) is fully preserved to guide the overall visual style and setting. All character identities, transformations, and specific visual references are strictly retained.\n\ndetailed_description:\n[Shot 1] At 00:00.000, a side full shot begins. In the starting frame, is very small in volume, with its head facing the left side of the frame, looking up at a pillar on the left. The sky features rolling dark clouds. Gray, thick particle smoke and dust gather from the bottom of the pillar towards the top, converging at the top of the pillar to form a circular vortex (the vortex faces the right side of the frame). (only the head, body completely hidden in the smoke) with eyes burning like flames, appears from the thick fog (face facing the right side of the frame towards ). opens its mouth and gently blows a pale red halo. The halo flies towards , and transforms into a yellow-clothed woman.\n[Shot 2] From 00:07.000 until 00:10.000, (only the head, body completely hidden in the smoke) appears and disappears within the circular vortex formed at the top of the pillar. Its eyes stare at the yellow-clothed woman. The shot size remains unchanged. No subtitles.\n\noverall_soundscape:\nNatural environment sounds and on-site sound effects only. No music. Background sounds are moderate and do not interfere with the scene, with atmospheric sound effects adding a sense of space.\n\nnon_diegetic_music:\nN/A", "original_prompt": "[Reference Aliases]\n@图片1 = 景汐AI_图片_20260528_125023_gyoj.jpg\n@视频1 = 狐狸变人.mp4\n@图片2 = 狐狸.png\n@图片3 = 参考图1\n\n【全局基础设定】@视频1 为原始拍摄素材(包含环境与构图)\n风格要求:保留@视频1 光照及色彩参数。\n以@图片3 作为狐妖老祖与环境的参考。\n音频要求:无音乐、无字幕、仅保留自然环境音和现场音效。符合年度票房冠军定位imax70mm胶片、遵循实景光源,遵循自然物理摩擦力,电影史诗级规模,低保和度,加入自然环境音和现场音效,保持对白清晰,背景音适度,不干扰人物对话,氛围音效加入场景感,所有画面必须是合理自然适配的使用大光圈音频要求:无音乐、无字幕、仅保留自然环境音和现场音效。\n【时间片分镜脚本】\n0-7秒,侧面全景镜头,参考@视频1 的构图和环境不要改变,起始画面中将黄衣女子更改为@图片2 ,@图片2 体积很小,约为画面中原黄衣女子体积的30%,@图片2 的头部朝向画面左侧,抬头看着画面左侧的柱子,内容:天空中乌云滚动,灰色浓稠的粒子烟尘从柱子底部向着柱子顶端汇聚,烟尘在柱子顶端汇聚,汇聚成圆形漩涡(漩涡朝向画右),狐妖老祖(仅头部,身体完全隐藏在烟尘中)@图片1 眼睛呈火焰燃烧状态,从浓雾中显现(脸部朝向画面右侧的@图片2 )狐妖老祖@图片1 张嘴轻轻吐出一道淡红色的光晕,光晕飞向@图片2 ,@图片2 变为黄衣女子。\n7-10秒,完全使用@视频1 的原有构图与演员表演动作,内容:狐妖老祖(仅头部,身体完全隐藏在烟尘中)在柱子顶端形成的圆形漩涡中若隐若现,眼神盯着看着黄衣女子。保持景别不变,黄衣女子的表演与@视频1 保持不变。运镜与@视频1 原有运镜保持完全一致。\n【画质与防崩坏约束】4K高清,电影级质感,人物动作流畅自然,画面无畸变、无崩坏,符合现实物理逻辑", "reference_images": ["references/fa221841cdb0e8b8a1f83a2e9d3745130a8f0f755052e58273435c8fe86111e1.jpg", "references/ad1188afcb106a84e83b49311ef3ea94c436a0b448e9066ab88ac699d1696251.jpg", "references/66b48c03ea98b247d251dee932829b732551f1564d8d3c7b9aed2282d10bb7f6.jpg"], "reference_sha256": ["fa221841cdb0e8b8a1f83a2e9d3745130a8f0f755052e58273435c8fe86111e1", "ad1188afcb106a84e83b49311ef3ea94c436a0b448e9066ab88ac699d1696251", "66b48c03ea98b247d251dee932829b732551f1564d8d3c7b9aed2282d10bb7f6"], "workflow": "ref2va", "prompt_sha256": "ae12df619ee4c7b544fe999835a873785bb62f88d4363f3a99f35e33bc8750ce", "original_prompt_sha256": "e9d1b94dd7237c01c1bc87cbc173895f9c188337d4e121275098c97291e3befb", "video_sha256": "7a911e2d08c750ccdde9410aee2bc15a7ffeea9eaae555768b1fcafb2e94a729", "task_adapted": true, "width": 1344, "height": 768, "fps": 24, "num_frames": 243, "duration_seconds": 10.125, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_098.json", "verification": "verification/sample_098.json"} {"file_name": "videos/sample_099.mp4", "sample_id": "#99", "prompt": "subject_definitions:\n is a male character aged approximately 28-35, with appearance and clothing from . is a premium durian with appearance from . The scene takes place in the location from .\n\nsummary:\n[reference generation] A realistic cinematic scene set at noon under a scorching sun, shot with an ARRI ALEXA MINI camera and 35mm lens in natural high-contrast light with a 2.35:1 aspect ratio. The sequence follows leaning against a stone pillar in the location, with a slow push-in, a focus pull to a premium durian in his hand, and an extreme close-up of his face showing subtle contempt.\n\nretention_analysis:\nfully_preserved: (from ) is the central character throughout all three shots, maintaining identity and screen presence. (from ) is the durian prop featured in the second shot. The location is the scene setting for the entire sequence. All three references are explicitly required and retained.\n\ndetailed_description:\n[Shot 1] Medium close-up, shot from a 45-degree side angle at eye level, with the camera slowly pushing in towards . leans his body weight against a stone pillar, his posture relaxed but his gaze focused. The noon heat haze causes slight air distortion, and a hot wind continuously blows the hem of his shirt and his hair. A torn piece of paper on the ground (the waiver of final payment thrown by Zhao Qiang) is swept up by the wind and quickly brushes past his feet. 's gaze does not follow the paper; instead, he maintains a steady stare ahead, his Adam's apple moving slightly, his breathing steady. (S1) says [Chinese]赵强要办国际狂欢节,要招待外宾, in a low, thick, slightly hoarse and rough voice with clear articulation, conveying a faint sense of understanding and control.\n\n[Shot 2] Continuing from the previous shot, a close-up where the focus quickly pulls from 's face to the premium durian () held in his hand in the foreground. occupies the main frame; its green-yellow fruit shell displays delicate highlight texture under the strong light, its fresh stem is emerald green, and its surface faintly shows fine water vapor or natural texture, hinting at ultimate freshness. The background is blurred into soft light spots at the entrance of an office building. (S1) says [Chinese]他必须要是刚从树上摘下来,果壳带清香、果肉如冰激凌的极品新鲜榴莲。 in the same low, thick, slightly hoarse and rough voice with clear articulation.\n\n[Shot 3] Continuing from the previous shot, an extreme close-up of 's face, shot from the front at eye level. The muscles at the corners of 's mouth turn down extremely subtly, then immediately return to calm. A clear mocking smile flashes in his eyes, but all other facial muscles remain restrained and still. His eyes stare directly toward the camera direction (simulating looking at Zhao Qiang or his representative), his pupils stable, with no large blinks, conveying certainty and contempt through subtle eye changes. Light and shadow form natural transitions between light and dark on his face, highlighting the three-dimensionality of his facial contours and the suppressed tension of his emotions. No subtitles appear.\n\noverall_soundscape:\nThe sound of a hot wind continuously blowing, rustling the shirt hem and hair of , and sweeping a piece of paper across the ground. Ambient noon heat haze atmosphere.\n\nnon_diegetic_music:\nN/A", "original_prompt": "场景:@图片3场景,正午烈日。写实电影风格,ARRI ALEXA MINI摄影机,35mm镜头,自然光高对比度,画面清晰锐利,2.35:1画幅。\n\n片段1:中近景,侧方45度角平视拍摄,镜头缓慢向@图片1人物推进。@图片1人物身体重心倚靠在石柱上,姿态松弛但眼神聚焦。正午热浪导致空气轻微扭曲,热风持续吹动他的衬衫下摆和发丝。地上一张碎纸片(赵强扔的放弃尾款声明)被风卷起,快速擦过他脚边,@图片1人物视线未随纸片移动,而是保持凝视前方,喉结微动,呼吸平稳,语气带着淡淡的了然与掌控感,@图片1人物(音色设定:男性,年龄区间约为28-35岁,声音偏低沉厚重,咬字清晰,带着常年户外劳作的微哑粗粝感,说标准普通话。)说:“赵强要办国际狂欢节,要招待外宾,”\n\n片段2:接上一个镜头,特写镜头,焦点从@图片1人物脸部快速拉焦至前景手中的@图片2的极品榴莲。榴莲占据画面主体,青黄色果壳在强光下呈现细腻的高光质感,新鲜果柄翠绿,表面隐约可见细微水汽或天然纹理,暗示极致新鲜。背景虚化为写字楼门口的模糊光斑。@图片1人物(音色设定:男性,年龄区间约为28-35岁,声音偏低沉厚重,咬字清晰,带着常年户外劳作的微哑粗粝感,说标准普通话。)说:“他必须要是刚从树上摘下来,果壳带清香、果肉如冰激凌的极品新鲜榴莲。”\n\n片段3:接上一个镜头,脸部大特写,正面平视。@图片1人物嘴角肌肉极其细微地向下一撇,随即恢复平静,眼底闪过一丝清晰的嘲弄笑意,但面部其他肌肉保持克制静止。眼神直视镜头方向(模拟看向赵强或其代表),瞳孔稳定,没有大幅眨眼,通过微小的眼神变化传递出智珠在握的笃定与轻蔑。光影在面部形成自然的明暗过渡,突出面部轮廓的立体感与情绪的压抑张力。,不要出现字幕,禁止出现字幕,禁止出现背景音乐", "reference_images": ["references/3b3598044e39fa571e76bc46958564fd48c8b5cf0edcb3edada64e0b32a6c30c.jpg", "references/5c88790ddadf17453a56e94928c5df1018318b608ba16357926f8be579903454.jpg", "references/6ea1906b6c602f9202585b5f8d7c3850197bf8401891ae947703834037bf0dad.jpg"], "reference_sha256": ["3b3598044e39fa571e76bc46958564fd48c8b5cf0edcb3edada64e0b32a6c30c", "5c88790ddadf17453a56e94928c5df1018318b608ba16357926f8be579903454", "6ea1906b6c602f9202585b5f8d7c3850197bf8401891ae947703834037bf0dad"], "workflow": "ref2va", "prompt_sha256": "bd8ca307c595d3246f5760941d86b03b94af690815f7f1bdd4237a7be870c937", "original_prompt_sha256": "49ca406f58cebf2f41e9bdde0e756ecc6d8a567473ded80420b43979b50604bc", "video_sha256": "cd2d60000e4eca6e9b1bc86bc0ead298e1f7a14a5b2ab2454f9ea574ee7e493a", "task_adapted": false, "width": 768, "height": 1344, "fps": 24, "num_frames": 209, "duration_seconds": 8.708333333333334, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_099.json", "verification": "verification/sample_099.json"} {"file_name": "videos/sample_100.mp4", "sample_id": "#100", "prompt": "subject_definitions:\n is a virtual female model with a 165-168cm plus-size hourglass figure, extremely full and round hips, a curved and smooth waist, and lively natural eyes, with appearance from . is the clothing, strictly matching the color, style, and details from with no alterations permitted. is the background scene, directly matching .\n\nsummary:\n[reference generation] A 12-second women's clothing e-commerce marketing video in 9:16 aspect ratio featuring wearing in , showcasing the garment through a three-shot sequence with specific camera movements, accompanied by bright electronic jazz music.\n\nretention_analysis:\nfully_preserved: (model identity and body type from ), (clothing color, style, and details from ), and (background from ) are all explicitly required to be retained throughout the video. The source mandates strict adherence to all three references with no alterations to clothing or background.\n\ndetailed_description:\n[Shot 1] A full-body long shot (0-4s). , wearing with hairstyle of French lazy slightly curly hair and high-end clean makeup, confidently walks into from one side and smoothly completes a 360-degree turn, using large-scale body movements to showcase the high-elastic wrapping feel of the knitted fabric. The camera smoothly pushes in. The lighting in serves the true color presentation of , with no color deviation or darkening of the clothing due to environmental light. [Shot 2] At 0:04.000, a medium shot (4-8s). continues walking while lightly touching her hair and looking back over her shoulder, her hips moving naturally to showcase the tension of the hip-hugging silhouette and the side stacked pleats. The camera moves laterally at a constant speed to capture the matte texture of the fabric. [Shot 3] At 0:08.000, a close-up (8-12s) focusing on the upper body and hem. naturally slows and walks deeper into the scene, showcasing the layered rhythm of the diagonal shoulder cut and wrapped hem in dynamic motion. The camera pushes in diagonally downward. All movements comply with physical laws, with no clipping, standing on tables, or hands in pockets; clothing coverage is complete with a smooth outline and no bulges; the image is decent and healthy.\n\noverall_soundscape:\nNo dialogue or speech. Ambient sounds consistent with the background scene from .\n\nnon_diegetic_music:\nPure instrumental music with no vocals, in a bright and light electronic jazz style.", "original_prompt": "[{\"role\": \"user\", \"content\": \"女装电商商品详情页营销视频,促进购买转化。图一虚拟模特,165-168cm大码沙漏身材,臀部极度丰盈圆润,腰部有弧度且平顺,眼神生动自然。服装严格使用图二,颜色、版型及细节禁止任何改变;背景直接使用图三。符合物理规律,禁止穿模、站立桌面或插兜;服装覆盖完整、轮廓平整无凸点;画面端庄健康。服装颜色以图二为准,背景光照服务于服装真实颜色呈现,禁止因环境光致服装偏色或暗化。发型为法式慵懒微卷,妆容高级干净,配合场景自然光影。 【12秒分镜设计】: 0-4s:全身远景,模特从一侧自信步入并顺势完成一次360度流畅转身,利用大幅度肢体动作展现针织面料的高弹包覆感,运镜平稳推近。 4-8s:中景,模特持续行进中轻抚发丝并侧身回眸,通过胯部自然律动展示包臀廓形与侧边堆叠褶皱的张力,运镜匀速横移捕捉面料哑光质感。 8-12s:近景聚焦上半身及裙摆,模特自然收势并向场景深处行走,展现斜肩剪裁与围裹下摆在动态下的层次律动,运镜斜向下沉推近。 配乐:纯音乐,无人声,明快轻盈的电子爵士风。\\n\\n素材说明:图1=模特参考(asset-20260224202454-g6xs6);图2=服装参考;图3=背景参考。\\n\\n生成参数:{\\\"ratio\\\":\\\"9:16\\\",\\\"duration\\\":12,\\\"resolution\\\":\\\"720p\\\",\\\"generate_audio\\\":true,\\\"watermark\\\":false}\"}]", "reference_images": ["references/b1f1f77fa1ebc09ba350eab9e2e5dfcdaf181b16b11038f325a7bab1c5552909.jpg", "references/fa500d02154081b00d2d30431081b87dd0c6da64de490a2eec8bf1db9611be59.jpg", "references/1f648934260b187e8488113f000d9d39c6b1350e71fb321c18d0c6bef7c405d6.jpg"], "reference_sha256": ["b1f1f77fa1ebc09ba350eab9e2e5dfcdaf181b16b11038f325a7bab1c5552909", "fa500d02154081b00d2d30431081b87dd0c6da64de490a2eec8bf1db9611be59", "1f648934260b187e8488113f000d9d39c6b1350e71fb321c18d0c6bef7c405d6"], "workflow": "ref2va", "prompt_sha256": "3e1c695aead133b9ad86d0d25f4cc481153e0992573fbaae0fa71f200be32d61", "original_prompt_sha256": "36c0e3ae212e36ac9835ac415abf428dde56be88db66db314cf80b185a1d68ea", "video_sha256": "d1a8396ddc1d542273c1b8cc6fc593ffa455802759abc449164ea75d2f1b087e", "task_adapted": false, "width": 768, "height": 1024, "fps": 24, "num_frames": 294, "duration_seconds": 12.25, "seed": 42, "num_inference_steps": 6, "generation_request": "generation/requests/sample_100.json", "verification": "verification/sample_100.json"}