Man vs. Pudding

3D Pixar-style animation, 8K resolution, vertical portrait orientation, 3:4 aspect ratio. A cozy, luxurious salon ASMR hair treatment session sequence. A cute young animated woman with long, silky black hair wearing a plush white bathrobe is reclining comfortably at a sleek white shampoo bowl. A cheerful stylist with a neat messy bun and apron gently washes, shampoos, and massages her hair. The clip features close-up details of water pouring from a golden faucet, rich white foam lathering, a wooden scalp massage brush working through soft foam, drops of hair oil being massaged into the scalp, smooth wooden combing, and drying with a white towel. Ending with the woman sitting up, smiling brightly at the camera, showcasing super shiny, voluminous, soft hair. Cozy aesthetic with warm glowing candles, marble countertops, white orchids, and golden accents. Cinematic lighting, soft depth of field, ultra-smooth camera movement, realistic fluid physics, relaxing and soothing atmosphere. --ar 3:4

[ImageReference] shows the FINAL frame of this video, not the first. The video must end exactly matching [ImageReference]. The video begins with the same location, same man, same camera angle and same lighting as [ImageReference], but with the parking space in front of the man completely EMPTY: only grey asphalt with faded white parking lines. No car is present at the start. Shot opens locked off in the over-the-shoulder position, the man seen from behind. He presses the key fob button with a small thumb movement. A deep rumble builds above. The camera tilts up fast, sweeping off the asphalt and up into the overcast sky, leaving the man's shoulder at the bottom edge of frame. High above, breaking down through the low grey clouds, a heavy dark metallic robot is falling head-first toward the ground, limbs tucked tight against its body, growing rapidly larger as it drops toward the lens. The camera tilts back down, following the robot's descent, the ground rushing back into frame. The robot slams into the empty parking space in a crouched landing, both fists and one knee hitting the asphalt. The impact cracks the asphalt in a tight radius, a burst of grey dust and small debris blows outward low across the ground, and the frame shakes hard from the force. Immediately the robot begins folding down into itself with mechanical precision. Its limbs retract and telescope inward, its back and shoulder armor plates slide over each other and lock flat, its torso compresses downward and widens. Panels rotate into position and seal along clean seams. The silhouette lowers and lengthens until it settles into the exact shape of the dark charcoal grey sports coupe seen in [ImageReference]. As the fold completes the camera settles back into the exact same over-the-shoulder framing, height and angle it started from. The car drops the last few centimeters onto its suspension and rocks once, then stops completely. The dust settles back onto the asphalt. The final second is completely still and matches [ImageReference] exactly: camera locked, car parked motionless, engine off, the man standing exactly as he was. Overcast flat diffused daylight throughout, no sun, no lens flare, no glowing energy, no magic particles, no teleportation. Photorealistic heavy machinery, practical VFX realism, grounded physics, desaturated color grade. Ambient sound only, no music.

0:00–0:05 — LONELY HIGHWAY Wide cinematic establishing shot of a remote two-lane highway in the late 1990s during golden hour. A vintage late-1990s sedan drives alone through an isolated rural landscape. Cut to the interior. The woman drives calmly, one hand on the steering wheel. Warm sunlight moves naturally across her face through the windows. Her long dark-brown wavy hair is loose, falling naturally over her shoulders. The dashboard, cassette player, steering wheel, and interior are authentically late-1990s. Camera: smooth tracking shot → interior close-up → profile shot. 0:05–0:09 — THE CAR STOPS The engine suddenly sputters. She looks toward the dashboard. The car loses power. She calmly steers onto the roadside and stops. Silence. She gets out and walks toward the front of the car. The environment feels completely empty. She opens the hood and looks inside. 0:09–0:13 — SOMEONE IS BEHIND HER A dark late-1990s vehicle slowly pulls onto the highway behind her. She hears it. She turns her head. Three men step out of the vehicle. They begin walking toward her. She watches them carefully. One man says: MAN: "Car trouble?" She doesn't answer. Her expression becomes cautious. The men continue approaching. 0:13–0:17 — THE HAIR-TIE MOMENT IMPORTANT: Make this action extremely clear and uninterrupted. The woman realizes the men are a threat. She slowly closes the hood. She reaches into the front pocket of her jeans and deliberately pulls out a small black rubber hair tie. She holds it visibly between her fingers. She gathers her long loose hair with both hands. She quickly ties her hair into a high practical ponytail using the black rubber band. Her face changes from cautious to completely focused. She looks directly at the approaching men. Close-up: ponytail tightening. Close-up: determined eyes. Final shot: she drops her hands and stands ready. This should feel like a classic cinematic "now I'm ready" moment. Do NOT skip the hair-tie action. Do NOT magically change her hairstyle. Show the complete action from loose hair → rubber band → ponytail. 0:17–0:25 — THE FIGHT The first man suddenly rushes toward her. She immediately sidesteps his attack and counters with a sharp strike. He stumbles backward. The second man attacks. She blocks him, grabs his arm, uses his momentum against him, and throws him onto the hood of the car. The third man charges from behind. She turns at the last second, ducks his attack, and delivers a powerful counter that knocks him to the ground. Fast but readable choreography. Every movement has realistic weight and momentum. No superhero powers. No exaggerated impossible movements. No gore. The woman remains calm and highly skilled rather than wildly aggressive. Camera: dynamic handheld tracking, low-angle action shots, quick close-ups, wide shots that clearly show the choreography. 0:25–0:30 — WALK AWAY The three men are left defeated on the roadside. The woman calmly adjusts her leather jacket. She walks back to her car. Gets inside. Turns the key. The engine starts. She looks into the rearview mirror. A tiny confident smirk. She accelerates away down the empty highway. Final wide shot from behind as the vintage car disappears into the golden sunset. CUT TO BLACK. CINEMATOGRAPHY Authentic late-1990s Hollywood action-thriller cinematography. 35mm film look, subtle natural film grain, realistic lens characteristics, shallow depth of field for close-ups, cinematic wide shots for the highway, natural golden-hour lighting, realistic shadows, restrained color grading. Camera movement should feel intentional and professional, never chaotic. Use smooth tracking shots during the driving scenes and controlled handheld movement during the fight. REALISM & CONTINUITY Photorealistic human movement. Realistic vehicle physics. Realistic fighting choreography. Consistent lighting and geography. The car must remain the same vehicle throughout. The woman must remain the exact same character throughout. No face changes. No outfit changes. No hairstyle changes before the designated hair-tie scene. No hairstyle changes after the ponytail is created. No extra fingers or distorted hands. No duplicated characters. No modern objects. No random scene changes. No teleporting. No sudden cuts that skip the hair-tying action. The overall result should feel like a lost 1998 Hollywood action-thriller scene, not an AI-generated video

Main Subject: Young Korean woman, 23, natural everyday appearance, shoulder-length dark brown hair worn loose with soft windswept bangs, realistic skin texture, minimal makeup, calm and thoughtful personality. Wearing a light beige trench coat over a white knit top, straight-fit blue jeans, white sneakers, a small brown leather shoulder bag, and a silver wristwatch. Maintain consistent identity, clothing, hairstyle, and appearance throughout. Location: A small coastal train station in Busan during a bright autumn morning. Quiet outdoor platform overlooking the sea, weathered benches, yellow safety lines, overhead power cables, station signs in Korean, distant green hills, and gentle ocean waves beyond the tracks. Only a few commuters are waiting. Visual Style: Ultra-realistic documentary realism. Authentic slice-of-life, genuine candid behaviour, natural body language, soft autumn sunlight, rich environmental details. Camera Style: Early-2000s Sony MiniDV camcorder. Heavy handheld shake, imperfect framing, autofocus hunting between passing trains and her face, exposure pumping under changing sunlight, faded colours, soft contrast, slight motion blur, MiniDV compression artifacts, no stabilization. 00:00–00:03 She stands near the platform edge, holding a warm takeaway coffee while watching the distant railway tracks. The camera briefly focuses on the station sign before finding her face. 00:03–00:06 A cool sea breeze blows through her hair as she checks the train schedule and quietly smiles. The operator accidentally zooms in too far before correcting. 00:06–00:09 She sits on a wooden bench, taking a sip of coffee while looking toward the ocean. Autofocus shifts between her profile and the sparkling water behind her. 00:09–00:12 The train slowly arrives with a soft metallic rumble. She stands, adjusts her shoulder bag, and steps toward the open doors. 00:12–00:15 Before boarding, she notices the camcorder, gives a small genuine wave and smile, then enters the train. The doors begin to close as the camera drifts downward, and the recording abruptly cuts to black. Audio: Natural ambience only–ocean waves, seagulls, light wind, distant station announcements, footsteps, train brakes, doors opening, quiet commuter conversations. No music. No narration. Goal: A forgotten MiniDV home video from autumn 2004, capturing a peaceful morning at a small coastal train station in Busan–simple, candid, nostalgic, and completely believable.

Bayou and forest environment. No music. No film grain. Quiet cinematic scene at the edge of a still black bayou channel deep in the Louisiana woods on a hot clear summer morning. Shot on Arri Alexa 35 with anamorphic lenses. One single continuous shaky handheld shot, no cuts, held close and low. Brilliant hard sunlight dappled through moss-draped canopy, deep shadows, the water flat and black. Sound design: dense cicadas and insect noise, birdsong, water lapping faintly, laboured breathing. 0–6s: Shaky handheld close shot low at the waterline on the woman kneeling in the shallows, soaked hair plastered flat across her forehead, water running down her face, breathing hard. She looks off at something ahead of her and speaks in a soft, warm, unhurried rural Louisiana accent: "I walked the whole bank lookin' for them." A beat, her breath catching. "I stayed out there till it got dark." Her head drops slightly, her voice going quieter and steadier. "I don't know where they went."

一、项目定位电影级写实质感，纯古风中国仙侠美学，采用克制的关系型喜剧、优雅古典人物调度、Arri Alexa 电影机观感、清晰稳定的面部细节、细腻胶片颗粒与自然体积光。本段核心机制为：同一句话，被两个人完全不同地理解。笑点只来自人物关系、理解偏差与克制反应。两名角色始终聪明、克制、有尊严。不得使用法术失败、背景事故、隐藏怪物、人物出丑或羞辱式包袱。二、参考控制 1. 人物参考 @图片 1 仅作为角色 ID A 的身份锚点 @图片 2 仅作为角色 ID B 的身份锚点 2. 环境参考本轮新上传的所有背景与地点参考图共同视为同一套环境 DNA 正式构图前，先无声执行一轮环境规划综合推演彼此兼容的：实际地形建筑语言空间尺度材质年代天气状态植被生态水体关系云层与山雾主光方向反射关系整体彩色系统空气纵深真实可行走路线再将这些信息重新组合成一个本轮独有、完整统一、物理合理的新空间 3. 环境运行原则环境必须始终鲜活，但保持叙事中立。风、水、山雾、植被、云层、远处光线、反光和空间环境声都要持续自然变化，但它们：不触发剧情不改变人物目标不替人物制造或解决误会不参与包袱整体按照 Seedance 2.0 官方统一多模态音视频联合架构进行组织，充分利用其对文字、图像、音频和视频参考的支持。三、角色设定角色 ID A｜剑仙师姐同一个 @图片 1 中的剑仙师姐。 25–30 岁东亚女性，椭圆脸，白皙肤色，深色杏眼，黑色长发半挽，以白玉簪固定，身材高挑纤细。始终穿同一套：白色刺绣真丝汉服半透明宽袖银色腰封玉佩白色布靴持唯一一柄银色长剑。角色 ID B｜小师妹同一个 @图片 2 中的小师妹。 20–25 岁东亚女性，圆润灵动的脸，黑发编辫，身形娇小。始终穿同一套：青绿色亚麻汉服深色腰带木簪黑色布鞋持唯一一柄深色钢剑。四、分镜脚本 0–5s｜全景或远景在根据本轮参考图重新规划出的地点中，两人相隔三步停下。同一个白衣剑仙神情郑重地说道： “昨日之事，到此为止。” 同一个青衣小师妹身体微微一僵，只回答： “好。” 随后她直接转身开始离开。要求：两人从鞋到头完整可见空间关系清楚环境保持持续自然运动误会从台词语义差开始建立 5–10s｜中景或牛仔景保持同样两名女子、同样服装、同样长剑和完全一致的空间地理关系。白衣女子看着她越走越远，明显有些不解，终于问道： “你去哪？” 青衣女子停下，却没有回头： “不是逐我出师门？” 完整停顿半拍。白衣女子回答： “我是说，别吵了。” 要求：误解在这一段清楚揭晓节奏克制表演依靠停顿、视线与语气背景依旧自然运动，但绝不参与误会形成或化解 10–15s｜特写或极近特写同一个小师妹缓慢转回身，尴尬只通过眼神和极轻微的面部变化表现出来，说道： “你早说。” 同一个剑仙师姐平静回答： “我以为你听得懂。” 小师妹从她身旁走过，小声嘀咕： “你每次都说得像遗言。” 极近特写：师姐始终端庄克制的表情终于出现一个极其轻微的笑意小师妹余光发现以后，也努力压住自己的笑最后两人自然并肩，朝同一个方向离开要求：收尾温和关系反转落在人与人之间的默契恢复不要夸张喜剧表演最后的笑意必须非常轻五、技术规格总时长：15 秒画幅：16:9 横屏镜头数：三个连续清晰镜头原生同步普通话对白不生成字幕始终只出现两名可见角色人物面部、服装、长剑和地理关系全程稳定身体动作、真丝衣料、头发以及所有环境物理效果真实自然环境持续鲜活，但保持叙事中立六、Negative blurry, bad quality, low quality, low resolution, noisy, jpeg artifacts, watermark, text, error; deformed, mutated, bad anatomy, poorly drawn hands, bad composition, out of frame, disfigured; inconsistent character, changing clothes, face morphing, background shift, glitching cuts, disappearing props

Photorealistic live-action video, shot on cinema camera, natural skin texture, realistic lighting, 16:9 horizontal. Setting: A cozy indoor room with warm ambient lighting, posters on wall, hanging lamp, modern furniture. Subject: A young blonde American woman with shoulder-length golden blonde hair, fair skin, wearing a gray hooded jacket. She faces the camera from waist up. The video has TWO distinct phases — do NOT repeat the same motion throughout. === FIRST HALF (seconds 0–5): Horizontal spring stretch === - Both hands at face height. Thumb and index finger on each hand form L-shaped viewfinder corners facing inward. - Purely horizontal motion: hands close together in front of her face, then pull outward left and right to shoulder width, then pull back to center. Repeat this lateral spring cycle 2–3 times. - Smooth, elastic sideways pull. Forearms stay level. Elbows move outward as hands spread. === SECOND HALF (seconds 5–10): Varied finger-frame gestures === - Transition naturally into different finger-frame poses. She rotates and flips her wrists so the thumb-index frame orientation changes. - Always keep thumb and index finger extended upright forming the corner of a viewfinder on BOTH hands. - End with hands at a relaxed wide finger-frame pose. NOT animated, NOT cartoon, NOT anime, NOT fully CGI — real filmed footage. Stable camera, no zoom, no cuts.

Seedance 提示词｜离山与放手一、项目定位电影级写实质感，纯古风中国仙侠美学。整体气质以克制的情感关系戏为核心，带一点极轻的冷面温度。风格要求：优雅古典的人物调度 Arri Alexa 电影机质感清晰稳定的面部微细节自然体积光真实可触的真丝与风化材质克制、安静、有分量的情绪推进核心反转：一句最初听起来像“把人推开”的话，随着人物真实行动逐渐显露出另一层含义。它不是疏远，而是允许对方真正离开自己的影子，长成她自己。硬性限制：不允许法术失败不依赖误会梗不允许环境事故所有反转只能来自对白、人物动作，以及两个人没有直接说出口的情感二、叙事核心这段戏的真正主题不是“送别”，而是放手。表面上：小师妹要下山师姐说“不拦” 听起来像冷淡、像放开、像不在意实际上：师姐不是不要她师姐是在承认她已经该去走自己的路她给出的不是束缚，而是退后一步后的保护整段戏要完成的情绪变化：第一拍：一句冷硬的话制造轻微刺痛第二拍：一个细小但真实的照料动作推翻表层理解第三拍：一句真正点明关系的话，让“放手”变成“成全” 三、参考控制 1. 人物身份锚点 @图片 1 严格作为角色 ID A 剑仙师姐身份锚点 @图片 2 严格作为角色 ID B 小师妹身份锚点 2. 环境参考本轮新上传的所有背景与地点参考图共同视为同一套环境 DNA。正式构图前，先无声执行一轮环境规划，综合推演彼此兼容的：真实地形建筑语言空间尺度材质年代植被生态水系走向天气云层山雾运动主光方向反光表面整体彩色空气纵深人物真实可行走路线再重新组合成一个本轮独有、完整统一、此前从未出现过的新空间。要求：不机械复制单张参考图保持同一世界身份三个镜头共享同一套地理逻辑环境鲜活但叙事绝对中立四、角色设定角色 ID A｜剑仙师姐同一个 @图片 1 剑仙师姐。 25–30 岁东亚女性，椭圆脸，白皙自然肤色，深色杏眼，黑色长发半挽，以白玉簪固定，身材高挑纤细。固定造型：白色刺绣真丝汉服半透明分层宽袖银色腰封玉佩白色布靴唯一一柄银色长剑人物气质：克制安静说话短情绪藏在动作里真正重要的部分，不靠表白，而靠处理细节角色 ID B｜小师妹同一个 @图片 2 小师妹。 20–25 岁东亚女性，圆润灵动的脸，黑发编辫，身形娇小。固定造型：青绿色亚麻汉服深色腰带木簪黑色布鞋唯一一柄深色钢剑一个符合纯古风世界的小型行囊人物气质：敏感但不脆弱理解师姐，但也会被她的话刺到内心柔软，表面尽量克制这一段要从“失落”走向“真正懂了” 五、环境原则环境必须始终鲜活，但在叙事上保持绝对中立。可持续存在的环境生命包括：风水体流动山雾游走云层缓慢变化植被响应自然风倒影随摄影机角度改变远处细微活动空间环境声持续存在但必须满足：绝不触发剧情绝不解决问题绝不打断对白绝不替角色表达情绪绝不制造包袱环境只负责：提供真实存在感提供时间流动感提供空间层次提供电影呼吸感六、核心道具连续性本段核心道具：小师妹的小型行囊小师妹的深色钢剑师姐交给她的唯一一枚小巧玉哨要求：行囊从开场就已准备好剑始终为同一柄玉哨必须清楚可见，成为第二拍与第三拍的情感锚点玉哨从交到掌心到结尾被握紧，连续性必须稳定所有手部动作必须清楚易读且真实克制七、分镜脚本 0–5s｜全景或远景在根据本轮参考图真实地理结构自然选择出的离山空间中，小师妹已经整理好行装准备出发。同一个小师妹看向师姐，轻声问： “师姐，我这次下山，你真不拦我？” 同一个白衣剑仙没有犹豫，回答： “不拦。走远些。” 小师妹眼神明显黯了一点。镜头要求两人从鞋到头完整可见空间要安静、开阔，有“离开”的意味第一拍必须成立 “走远些”听起来要像推开、像冷、像不留小师妹受伤感只通过极轻微眼神变化体现环境正常运行，但绝不参与这一拍的情绪落点 5–10s｜中景或牛仔景保持同样两名女子、同样服装和完全一致的地理空间。同一个小师妹转身准备离开。就在她还没有迈出第二步的时候，同一个剑仙师姐主动伸手：替她整理松开的肩带重新收紧剑上的绳结然后把唯一一枚小巧玉哨放入她掌心同时平静说道： “迷路就吹，受伤别硬撑。” 小师妹低头看了一眼同一个玉哨，又重新抬眼看向师姐。镜头要求第二拍必须清楚成立师姐的动作要比对白更重要整理肩带、收紧绳结、放玉哨，都是安静却非常具体的照料这些动作不能急，要像她早就想好了摄影机要让手部互动、玉哨进入掌心的动作清晰可见背景持续鲜活，但叙事中立 10–15s｜特写或极近特写同一个小师妹轻声问： “你不是让我走远些吗？” 同一个剑仙师姐这一次真正与她对视，平静回答： “走远，是让你看看自己的路。” 完整保留一个具有情绪重量的半拍停顿。随后她补上一句： “不是让你找不到回来的路。” 极近特写继续推进：小师妹眼底出现极轻的湿润却终于笑了出来慢慢握紧掌心中同一枚玉哨她又问： “那我回来时呢？” 同一个剑仙师姐嘴角出现极轻微的笑意： “带点自己的故事回来。” 小师妹认真点头，随后转身向前走，这一次不再回头。师姐平静站在原处看着她走远，表情不是失落，而是一种极克制的欣慰。镜头要求第三拍必须真正落地 “看看自己的路”是主题句 “不是让你找不到回来的路”是关系重定义句 “带点自己的故事回来”是收尾句，必须温柔但克制结尾不是诀别，而是承认成长师姐的欣慰不能做得太大，只要让观众感到她真正在放手八、表演要求师姐的表演逻辑表面冷内里很稳说的话比情绪更短行动永远比语言更柔软真正重要的话，要在最后才给重点表演点：说“不拦。走远些。”时的平静替她整理肩带与剑绳时的自然熟练把玉哨放进她掌心时的克制分量说“看看自己的路”时的真正对视结尾那一点几乎压住的笑意与欣慰小师妹的表演逻辑先被刺到再被照料再慢慢意识到师姐真正的意思最后在安心里生出轻轻的笑重点表演点：听到“走远些”后的眼神一暗发现师姐在替她整理东西时的迟疑与重新观察看到玉哨时的情绪回转听到“不是让你找不到回来的路”时的真正松动最后笑出来不是撒娇，而是懂了以后终于轻下来九、手部与道具互动要求这一段手部戏非常关键，必须真实、克制、清楚。必须清晰呈现的动作：小师妹背着行囊准备离开师姐替她整理松开的肩带师姐收紧剑上的绳结玉哨被放入掌心小师妹低头看玉哨小师妹在结尾握紧玉哨要求：手部动作自然、不夸张肩带、剑绳、玉哨之间的关系清楚触碰必须有分寸感师姐的照料不是亲昵表演，而是熟悉到近乎本能小师妹的掌心接住玉哨时，要让观众明确感受到情绪转折十、摄影与镜头要求 16:9 横屏三个连续清晰镜头 Arri Alexa 电影观感观察式、克制、稳定全景建立离山空间与人物关系中景放大动作与道具互动特写完成情绪落点强调真实视差强调面部微表情不使用煽情式夸张推拉不依赖环境戏剧化变化让情绪从人物之间自然流出十一、声音要求原生同步普通话对白精准口型真实空间环境声配乐克制，不能喧宾夺主风声、水声、远处环境声持续自然存在对白之间允许有安静停顿 “不是让你找不到回来的路”后必须留半拍小师妹笑出来的那一下要轻、短、真结尾让环境继续存在，但不要替情绪说话十二、技术规格总时长：15 秒画幅：16:9 横屏三个连续清晰镜头针对 Seedance 2.0 Fast 强化参考图人物身份与核心道具连续性原生同步普通话对白精准口型面部微表情清楚易读手部互动克制真实真丝衣料、头发、玉石、水体、山雾以及所有符合参考图世界逻辑的环境运动自然可信前景、中景、远景保持真实视差和空间环境声不生成字幕始终只出现两名可见人物整段持续锁定人物、环境和声音连续性十三、Negative blurry, bad quality, low quality, low resolution, noisy, jpeg artifacts, watermark, text, error; deformed, mutated, bad anatomy, poorly drawn hands, bad composition, out of frame, disfigured; inconsistent character, changing clothes, face morphing, background shift, glitching cuts, disappearing props