Croissant Baking Animation

Use the attached THE MONKEY VS THE SELFIE character sheet as the exact visual reference throughout. The couple are Pixar 3D man and woman, late 20s, matching tropical tourist outfits, sunglasses, sun hat on the woman. The monkey is a small brown Pixar 3D monkey, wide grin, bright curious eyes, full charisma. Full Pixar 3D render. No realistic humans or animals. THE SETTING: Bright lush wildlife park. Vivid green trees, warm golden sunlight, blue sky through the canopy. Maximum brightness. No dark frames. THE PHONE: Slim black smartphone. Screen always facing the camera — never the back of the phone. When the couple scrolls through photos the camera is either directly above them looking down at the screen or at their eye level looking down at the screen. Phone screen always clearly visible. MUSIC: Light cheerful tourist music at start. Playful comedy track the moment the monkey takes the phone. Silence when the phone drops. Surprised sting when the couple sees the photos. 0:00–0:02 — Wide shot. Bright lush park. Couple standing together, man holding phone out for a selfie, both smiling wide. Perfect setup. SOUND: Cheerful music. Birds. Leaves. 0:02–0:03 — A small hand drops from the tree above and snatches the phone mid-selfie. Couple freeze. Both look up slowly. SOUND: Quick snatch. Music stops. Silence. 0:03–0:06 — Monkey up in the tree, phone in hand. Turns it to face himself. Holds it out, chin raised, wide grin, one eyebrow up. Takes a photo. Adjusts angle. Takes another. Both arms wide, huge grin directly at the camera, tongue slightly out. Photo after photo. Completely in his element. SOUND: Playful comedy track. Camera shutter with each photo. Monkey chattering. 0:06–0:08 — Wide shot. Couple below, arms out, completely helpless. Monkey above still posing — side profile, one hand on chin, very serious expression. Another click. SOUND: Comedy track. Shutters. Couple's quiet disbelief. 0:08–0:10 — Monkey looks down. Satisfied. Drops the phone. It spins through the air. Man lunges and catches it. SOUND: Phone dropping. Couple's gasp. Catch. 0:10–0:13 — Camera directly above the couple or at eye level looking down — phone screen fully visible facing camera. They open the camera roll. Every single photo is the monkey. Different poses, full charisma in every frame. Couple's faces shifting from relief to confusion to pure disbelief with each swipe. Woman's hand over her mouth. Man's thumb frozen mid-swipe. SOUND: Surprised sting. Swipe sounds. Stunned silence. 0:13–0:15 — Wide shot. Couple standing still, staring at the phone. Camera pulls back slowly to show the monkey on the branch above watching them. Completely unbothered. Looks away. SOUND: Silence. One playful note. Fade out. Camera: wide park shot, close snatch from above, close monkey posing in tree, wide tree shot with couple below, phone drop and catch, directly above or eye-level looking down at screen for scrolling, wide final pullback. Bright vivid Pixar colors throughout.

Create a highly realistic, relatable comedy video in a modern cozy apartment, 9:16 vertical format. Use the reference girl as the main character and maintain perfect character consistency. 0–3 sec: The girl suddenly realizes her phone is missing. Her expression changes from relaxed to confused and then panicked. She quickly checks her pockets and looks around the room. 3–6 sec: She starts frantically searching the apartment—checking the sofa cushions, coffee table, kitchen counter, and under a pillow. She looks increasingly stressed. 6–9 sec: She gets on the floor and searches underneath the couch and bed. She is using a bright flashlight to look around, but the flashlight is actually coming from the missing phone that she is unknowingly holding in her other hand. 9–12 sec: She continues searching faster, moving around the room while holding the phone up like a flashlight. She suddenly stops. Slowly looks down at the phone in her hand. 12–15 sec: Close-up of her face as she realizes what happened. She gives an awkward, embarrassed smile and looks directly into the camera. Add a tiny comedic pause. Voice-over: “I spent five minutes looking for my phone… while using my phone’s flashlight.” Final text: “Modern problems require… no brain at all. 😂” Style: Photorealistic, natural acting, authentic expressions, realistic apartment lighting, handheld camera feel, quick comedic cuts, subtle camera shake, cinematic depth of field, realistic skin texture, natural movement, high detail, 4K. No exaggerated cartoon behavior, no background music—only natural room sounds and subtle comedic timing.

14s photorealistic cinematic sci-fi mystery short, 2.39:1 aspect ratio, 4K 24fps, subtle 35mm film grain, Hollywood thriller color grading: cool desaturated tones, deep blacks. 0:00-0:03: Slow tracking dolly shot follows young photographer stepping into moonlit abandoned photo shop, dust motes drifting in volumetric silver moonlight through cracked window. Slow push-in as his hand lifts the weathered black camera off a dust-caked oak table, shallow depth of field. 0:03-0:06: Medium shot, he raises the camera toward empty dark hallway. Sharp flash burst cuts the dark. Cut to tight close-up: his fingers curl around the developed print, eyes widening from quiet curiosity to cold shock as he spots the distant figure on the photo that is not there in the hall. 0:06-0:10: He glances back at the still-empty hallway, then snaps back to the print — the figure has moved three meters closer. Trembling hand lifts the camera again, second flash bleaches the room. 0:10-0:14: He stares at the new print: himself, from this exact angle, with a tall shadow figure standing directly at his back. Slow pan follows as he turns his head over his shoulder. The room remains empty. Final extreme close-up: his wide, terrified eyes, breath fogging faintly in the cold air. Smooth cinematic camera movement, physically accurate lighting, natural human expressions, no text/overlays.

Continue seamlessly from the previous scene. Maintain the exact same characters, clothing, facial features, lighting, mountain environment, and CGI animated feature film quality 3D animation style, NOT photorealistic. SET LAYOUT: Girl remains fixed center-frame. Squirrel's photobombs escalate in speed and silliness, still entering from the same fixed branch area, ending seated calmly on her shoulder for the final shot. She snaps again — this time the squirrel photobombs with an even more exaggerated pose, tongue out, tail puffed dramatically. Girl (exasperated shout): "SERIOUSLY?!" Rapid montage of three quick snap attempts in fast succession, each interrupted by an increasingly ridiculous squirrel face — cross-eyed, mid-jump, cheeks stuffed with an acorn. The girl finally throws her hands up, defeated, sitting down on a rock with a sigh. Girl (muttering): "Fine. You win." The squirrel calmly hops up and settles onto her shoulder, striking a proud little pose. She glances at it, surprised, then can't help but smile genuinely. She holds up the phone one last time — girl and squirrel both grinning naturally at the camera — and snaps the photo. It's clearly the best shot yet. Girl (delighted, laughing): "...Okay, THIS one's perfect." Camera quick zoom into the phone screen showing the adorable photo. Cut to black.

USE THE ATTACHED MARCO CHARACTER SHEET as the exact visual reference for Marco throughout. Match his face, dark tousled hair, forest green apron, white long-sleeve shirt with rolled sleeves, gold stud earring, and warm medium-brown skin in every shot. Create a 15-second 16:9 anime animated short. Moody cozy atmosphere throughout — warm amber interior light, rain on windows, dark NYC night outside, soft steam. No dialogue, no text on screen. Emotion told entirely through movement, light, and atmosphere. Rules: •Marco looks identical to the character sheet in every shot •No cuts to anything outside the coffee shop •Rain audible but soft throughout — never overpowering •Every shot warm and intimate, never cold or bright Shot sequence: 0:00–0:02 — Outside. Wide shot of a rainy NYC street at night. A small coffee shop window glows warm amber against the dark wet street. Rain streaks down the glass. City lights blur through the water. Camera slowly drifts forward toward the window. 0:02–0:04 — The Sign. Interior. Marco alone behind the counter, back slightly to camera. He reaches up and flips the door sign from OPEN to CLOSED. Turns back toward the espresso machine. Pause. One more. 0:04–0:07 — The Craft. Close-up on Marco’s hands — he tamps the portafilter with quiet precision, locks it into the machine, presses the button. Espresso streams into a small white cup. Rich dark crema forming on top. Steam rises slow. Camera stays tight on his hands and the cup. 0:07–0:10 — The Pour. Marco lifts a small steel milk pitcher. Slow controlled pour into the cup. Latte art blooms — a clean leaf pattern. He sets the pitcher down gently and looks at it. Just looks. 0:10–0:13 — The Sip. Marco leans back against the counter. Picks up the cup with both hands. Takes one slow sip. Eyes close. Rain taps the window behind him. Amber light wraps around him. Total stillness. 0:13–0:15 — Fade. Camera slowly pulls back. Marco stays still, cup in hand, small against the warm glow of the empty shop. Rain outside gets slightly louder. Scene fades gently to black. Style: Anime, Ghibli-influenced warmth, cel shading, cinematic amber and deep navy color palette, soft grain, rain audio throughout, no music needed.

【生成模式与总要求】必须使用 Seedance 2.0 Fast 生成，严格生成完整 15 秒视频，16:9 横屏，三个连续清晰镜头，原生同步普通话对白。大师级电影写实质感，Arri Alexa 电影机观感，带克制而诗意的仙侠冷面喜剧气质。画面锐利清晰，体积晨光自然可见，表演克制，人物运动真实但具有电影感，真丝衣料、头发、纸张、水流、山雾与自然光具有可信物理反馈。采用“参考图驱动的活世界系统”：所有参考图共同决定地点与视觉世界，包括地形逻辑、建筑语言、空间尺度、材质年代、植被生态、水流走向、天气状态、云雾运动、主光方向、反光表面、综合色彩和空气纵深。正式生成前先无声完成一轮环境规划，从这些共同环境 DNA 中重新组织出一个全新、统一、物理合理的空间。背景必须鲜活、独特并持续自然运行，但只负责环境存在感，不主动触发包袱，不改变人物目标，不抢夺剧情中心。【角色与核心道具】角色 ID A｜剑仙师姐始终是 @图片1 中同一个 25–30 岁东亚女性，严格保持同一张脸、椭圆脸型、白皙自然肤色、深色杏眼、黑色长发半挽、白玉簪、高挑纤细身材比例、白色刺绣真丝汉服、半透明分层宽袖、银色腰封、玉佩和白色布靴。角色 ID B｜小师妹始终是 @图片2 中同一个 20–25 岁东亚女性，严格保持同一张脸、圆润灵动脸型、黑发编辫、娇小身材比例、青绿色亚麻汉服、深色腰带、木簪和黑色布鞋。核心道具全程只有同一卷密封竹简。竹简在三个镜头中的位置、开合状态和人物持有关系必须连续稳定。【0–5s｜低角度全景，缓慢推轨】雨后清晨的山门回廊石台。下方是云海山谷，雕花石栏、湿润石板、狭窄水渠、古松、浅色开花枝条、远处瓦顶殿宇、缓慢游走的山雾与持续变化的湿地倒影共同建立一个真实存在、持续呼吸的环境。同一个白衣剑仙师姐与同一个青衣小师妹位于石案两侧。白衣剑仙师姐把同一卷密封竹简轻轻放到两人之间的石案上，平静说道： “三年前，你落下一物，今日归还。” 环境保持克制运动：水渠持续流动，薄雾缓慢穿过回廊，树枝和衣摆响应同一晨风，雨后石面反射随摄影机推轨自然变化。【5–10s｜牛仔景中景】同一个青衣小师妹缓慢伸手去取同一卷竹简，另一只手仍自然停在剑旁。同一个白衣剑仙师姐保持完全平静的神情，补上一句： “我已替你批过。” 近处叶片自然掠过画面，形成真实前景视差；水渠持续流动，山雾在回廊之间缓慢游走。背景始终只维持世界存在感，不主动制造新的剧情事件。小师妹取到竹简后开始展开，为下一镜头保持连续动作。【10–15s｜特写转极近特写】同一卷竹简完整展开。里面并不是秘籍，而是同一个青衣小师妹三年前写下的一首旧情诗；纸面上密密写满工整、优雅的红色批注。同一个青衣小师妹瞬间僵住，低声问： “你全看了？” 同一个白衣剑仙师姐平静回答： “还改了三处韵脚，第三句最好。” 极近特写捕捉小师妹耳尖明显泛红。剑仙师姐依旧保持端庄、平静、毫无取笑意味的神情。此时檐角一滴雨水自然落入同一条水渠，发出一声轻微而清楚的水响，作为最后的冷面节拍。严格在 15.0 秒结束。【执行硬约束】必须使用 Seedance 2.0 Fast 必须生成完整 15 秒，不缩成 10 秒，不提前结束 16:9 横屏严格三个镜头原生同步普通话对白普通话口型准确人物身份、面部、发型、身体比例、服装全程稳定同一卷竹简连续稳定，不复制、不消失、不更换前中后景具有真实视差环境持续活着，但绝不能抢戏真丝、纸张、水流、山雾、头发、花枝、雨滴和光线物理真实始终只出现两名可见角色不生成画面字幕不出现额外人物【Negative】 blurry, bad quality, low quality, low resolution, noisy, jpeg artifacts, watermark, text, subtitles, error; deformed, mutated, bad anatomy, poorly drawn hands, bad composition, out of frame, disfigured; inconsistent character, changing clothes, face morphing, hairstyle change, background shift, architecture mutation, lighting jump, glitching cuts, disappearing props, duplicated bamboo scroll, changing scroll design, extra characters, background crowd, environment stealing focus, random environmental gag, exaggerated comedy acting, slapstick reaction, overacting, modern elements, frozen background, static fog, static water, no parallax, abrupt camera jump, dialogue desync, lip-sync error

【生成模式】 Seedance 2.0 【视频规格】严格总时长 10 秒 16:9 横屏 24fps 严格两个干净镜头 10.0 秒必须准确结束，不生成 10–15 秒内容【整体风格】电影级写实质感，采用优雅动作喜剧节奏、高级商业广告灯光、清晰的港式动作片空间关系、克制冷面反应，以及全新的“技能互换”机制。没有任何一方被打脸或失败，而是让两个人发现：自身熟练的骑行经验，只能部分迁移到另一种完全不同的“坐骑”上。喜剧核心来自：熟练习惯 → 错误迁移 → 对方提醒 → 同时修正 → 意外建立默契。【场景设定】清晨空旷的高架滨河广场。淡金色阳光、修长建筑阴影、拉丝混凝土地面、玻璃护栏、远处城市天际线、轻微晨雾和自然吹动衣料的风。一条约二十米长的训练路线清楚可见。空间必须开阔、简单、可读，确保两名角色、自行车、飞剑以及后续左右运动轨迹始终清楚。【角色锁定】角色 ID A｜剑仙姐同一个来自 @图片 1 的 25–30 岁东亚剑仙姐。严格保持参考图中的：同一张脸锐利深棕色眼睛黑色长直发高挑纤细身材比例白色布靴白色刺绣真丝汉服半透明层叠宽袖银色腰饰玉簪同一柄银色飞剑角色 ID B｜单车姐同一个来自 @图片 2 的 25–30 岁东亚单车姐。严格保持参考图中的：同一张脸棕色马尾同一身材比例黄色外套蓝色牛仔裤白色运动鞋饰品同一辆自行车【核心道具】同一柄银色飞剑同一辆自行车不得复制、替换、消失或临时生成第二辆自行车、第二柄飞剑。【镜头 1｜0–5s｜低角度全景缓慢推轨】摄影机低角度全景缓慢推轨，先完整建立清晨高架滨河广场和约二十米长的训练路线。同一个剑仙姐与同一个单车姐面对面站立。同一辆自行车和同一柄悬浮飞剑位于两人中间。单车姐伸手指向飞剑。剑仙姐同时指向自行车。单车姐问： “换？” 剑仙姐回答： “换。” 两个人保持无比严肃、像进行正式交接仪式一样的表情，随后交换位置：单车姐走向飞剑剑仙姐走向自行车动作直接、克制、清楚，不加入额外笑料。【镜头 2｜5–10s｜牛仔景中景侧面跟拍】摄影机切换至牛仔景中景侧面跟拍。两名角色分别使用对方原本熟悉的坐骑，并沿相反方向运动，最终在画面中交错经过。单车姐｜骑飞剑同一个单车姐踩上同一柄银色飞剑。因为长期骑自行车形成的肌肉记忆，她本能地：屈膝压低身体双脚做出踩踏动作试图踩动根本不存在的脚踏板她不是踩在剑身上行走，而是双脚已经站稳在飞剑上，却下意识做出骑自行车式踩踏动作。飞剑受到身体动作影响，立刻向前加速。剑仙姐｜骑自行车与此同时，同一个穿白色刺绣真丝汉服的剑仙姐骑上同一辆自行车。因为长期御剑形成的动作习惯，她本能地松开一只手，并用御剑时的双指指诀试图操控自行车。自行车并不会响应指诀，因此前轮出现一次轻微偏移。她仍坐稳车身，不摔倒、不失控，只是必须重新理解车把的实际作用。【交错与对白】两人沿相反方向继续前进，并在同一条训练路线中交错经过。剑仙姐大喊： “脚别蹬剑！” 单车姐立刻回喊： “手要扶把！” 【同步修正】两人在同一瞬间理解对方提醒，并同步修正动作。单车姐：停止踩踏动作双脚稳定站在飞剑上调整重心恢复平稳御剑姿态剑仙姐：立即重新双手稳定扶住车把修正前轮方向保持正常骑行平衡两个人都没有失败，也没有任何一方占绝对上风。她们只是发现：对方的坐骑不能完全按照自己的旧习惯来控制。【10秒结束构图】修正完成后，两人继续稳定前进。她们一边保持各自新的稳定驾驶状态，一边侧头互相看向对方，以一种克制、意外认可的眼神完成确认。画面必须在 10.0 秒准确结束于一个干净的平行运动构图：两名角色同时处于稳定运动状态同一辆自行车正常前进同一柄飞剑平稳滑行两人已经能够自信控制交换后的坐骑两人的运动方向与摄影机轴线清楚不发生停顿、摔倒、撞击或额外反转 10.0 秒直接结束。【后续扩展槽位】 10–15 秒仅作为后续扩展槽位。本次视频绝对不要生成这一部分。不要提前续写。不要自动补结尾。不要生成第三镜头。不要把 10 秒剧情压缩后再额外生成 5 秒。【动作与空间硬约束】两名角色的左右运动轨迹连续两人在第二镜头沿相反方向运动并清楚交错单车姐始终使用同一柄飞剑剑仙姐始终使用同一辆自行车飞剑必须是可站立、可稳定滑行的实体银色飞剑自行车必须保持真实车轮、车把和重心逻辑单车姐的错误动作仅是“骑车肌肉记忆迁移到飞剑” 剑仙姐的错误动作仅是“御剑手势习惯迁移到自行车” 两人均在对方提醒后立即修正不允许摔倒不允许撞车不允许法术失灵不允许飞剑失控不允许自行车损坏不允许任何一方成为单独笑柄【声音设计】原生同步普通话对白。保留：自行车链条声轮胎滚动声银色飞剑低沉剑鸣晨间自然风声衣料掠风声轻微城市远景环境声对白口型准确。声音与动作同步。【物理与连续性】角色身份与服装全程稳定。自行车平衡、车轮旋转、飞剑加速、人物重心变化、头发、真丝衣料、晨雾和风力物理效果真实。人物面部、身体比例、发型、服装、饰品、飞剑、自行车均不得发生漂移或变化。不生成字幕。不增加额外人物。【负面词】 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

风格：2秒真实生活感手机实拍 → 位置匹配硬切 → 8秒静态唯美海报（2.5D视差 + 播放器UI），两段质感形成强烈反差角色：主角@ 图片1 —————— 【镜头一 00:00-00:02｜真实居家·转唱片】中景，固定机位，无运镜。室内走廊，背景是米白色墙面和一扇深棕色木门（门轴合页清晰可见），偏暗的室内暖调灯光，画面有真实手机拍摄的噪点和轻微曝光波动。主角穿一件白底黑色碎花印花上衣，戴细框眼镜，素颜居家状态，站在画面右半边正对镜头，表情平静内敛，眼神直视镜头，身体几乎不动。 00:00-00:00.6 她抬起左手向身体左侧伸出，左手食指穿过一张黑色唱片正中间的圆孔，把唱片举到**画面左侧偏中间的位置**（不在画面正中，而是明显偏左），唱片盘面完全正对镜头，像一个圆盘悬在画面左边。 00:00.6-00:02 她伸出右手，指尖搭到唱片盘面边缘，手腕发力猛地一拨——唱片立刻绕着左手食指高速自转。右手拨完立即收回，唱片继续独自旋转，盘面被转成一圈圈环形运动模糊，边缘反光被甩成一道旋转光弧。唱片始终停在画面左侧那个位置高速旋转，大小和位置都不变。 00:02 硬切（HARD CUT，位置匹配）。就在唱片转到最快的那一帧，画面直接切断——下一帧已经是完全不同的场景，而画面左侧那个旋转唱片的位置上，刚好接上了播放器 UI 里同样在旋转的黑胶唱片控件，位置、大小、旋转方向完全对齐，视觉上像是同一张唱片一直在转，只是周围的世界整个换掉了。 —————— 【镜头二 00:02-00:10｜唱片封面海报·2.5D视差】近景人物特写。这一段不是实拍视频，而是一张精修静态海报被赋予了 2.5D 视差动态——没有物理运镜，只有极缓慢的伪推镜头感：人物层与背景层之间产生轻微的相对位移，前景落叶层移动最快、人物层次之、背景层最慢，营造立体纵深。整段视觉节奏极慢、极舒缓。画面右半边是同一位主角，但已完全变成海报里的形象：妆容精致（通透白皙的肤色、奶油色眼影、上挑眼线、饱满水润的唇），身穿镶钻细吊带礼服，锁骨处一串珍珠链，长发盘成松散高髻、几缕碎发垂落。她没有任何肢体动作，只是面向镜头微微含笑，眼神温柔而略带忧郁，神情静谧唯美，像一张定格的杂志封面。背景：秋日金黄的银杏大道，阳光穿过树叶洒下光斑，远处有虚化的行人剪影和长椅，浅景深把背景化成一片流动的暖金光晕。【粒子特效】整个画面覆盖一层金黄色落叶飘落特效，叶片从画面上方缓缓旋转飘下——靠近镜头的几片落叶带明显景深虚化，大而模糊地从画面前方划过；远处的叶片小而清晰，强化立体纵深。【UI 图层（固定在画面上，不随视差移动）】 - 左上角：金色衬线大字"FASHION"，下方小字"SELF-PORTRAIT"，再下一行年份数字"1901" - **画面左侧偏中间（承接上一镜头唱片的位置）**：一个立体黑胶唱片形状的播放器控件，唱片持续缓慢顺时针旋转，圆心内嵌一张她自己的圆形头像 - 左侧下方：极简音乐播放器进度条，带播放/暂停、上一首、下一首按钮，进度条随时间缓慢向右推进 - 左侧另有淡黄色和浅灰色的杂志风色块条作为版式装饰 - 画面上不出现任何歌词或抒情文案文字【整体色调】镜头一是真实偏暗的室内暖调，日常写实；镜头二是浓郁的秋日暖金色调，阳光感强、肤色通透，商业平面海报质感。【音效】镜头一是真实室内环境音+唱片旋转的轻微塑料摩擦和风声；硬切瞬间声音一并切断；镜头二直接切入完整的抒情音乐（带黑胶底噪的温暖质感），背景是极轻的风声和落叶摩擦