Seedance 2.5 在 Capcut 中，提示词 ->

Seedance 2.5 in Capcut, prompt -> FORMAT 8 seconds, 16:9, 24fps. One generation, ONE CONTINUOUS HANDHELD MOVE, zero cuts — the arc, the whip and the hold are one single take, DRIVEN BY @video1: its camera path and its clock are the law of the clip. LANGUAGE: ENGLISH. One woman's voice, OFF SCREEN for the whole clip, American English, a native speaker; production ambience beneath it. VOICE — EVE (off screen, every spoken word in the clip) The speaker is EVE, heard off screen for the whole clip — only her voice is in it. Her voice: a young adult Black American woman, Gen Z — gentle above all. She sounds like she is smiling while she speaks: light, airy, warm, the tone of a late-night conversation between close friends, a bedtime story told to someone she loves. Softer and higher-pitched than a broadcast narrator, with a subtle rasp riding the edges of her phrases; low in volume but recorded close to the microphone so it sits close and large in the mix; a light breathiness under every line. The passion is in the care she gives each word while the volume stays low; the delivery stays light and rounded, the vowels warm, every sentence landing softly. An even, slow pace, around three words per second inside a line; consonants soft; a small audible inhale before the sentence and a falling, settled landing at the end of it; micro-pauses between clauses. HER ACCENT IS AMERICAN: General American vowels, an American r, American intonation. THE WORDS AND THEIR CLOCK EVE V.O.: {Yet no matter how much we try, we can't create anything close to the precision of the universe.} (0.30s–5.80s) — the full line, off screen; on screen the dancer dances with her lips closed, and the robot's face is a smooth blank shell. ASSETS @image1 as character reference — THE ROBOT, full body: a humanoid robot skinned in pale cream synthetic fabric panels over articulated joints, visible seams, its head a smooth blank shell with two small dark sensor eyes catching the light. Its body, material, colour and head exactly as this image, in every frame it appears. @image2 as location, light and character reference — THE ROOM: a wide modern living room in warm afternoon sun, floor-to-ceiling windows with trees and haze beyond, cream sofas, pale wood floor, a dark rug; THE DANCER in her white leotard, layered white tulle skirt and pink pointe shoes with ribbons; THE ROBOT standing by the tall window at the right where the sunlight pours in. The place, the backlight and both their looks come from this plate. THE POSE IN THE PLATE BELONGS TO THE PLATE ALONE: in this clip the dancer is UP AND DANCING from the first frame — on pointe, turning. @video1 as blocking reference — THE CAMERA AND THE CHOREOGRAPHY, frame-aligned with this clip: a grey-box previz on a checkerboard floor. THE BLUE FIGURE (a cylinder with a ball head) IS THE DANCER; THE RED FIGURE (the same build with two arms) IS THE ROBOT. @video1 supplies the camera 1:1 — the low start at her feet, the rise held below her head, the WHIP RIGHT at 2.0s–2.9s that TILTS UP as it lands on the robot so the floor leaves the frame and stays out, the slow drift and the hold — and the robot's clock: the arms rising to the T at 3.3s–4.6s, its turn at 4.5s–5.8s, the lean at 6.1s–6.7s, the body below frame by 7.0s, the hold to the end. The checkerboard, the wall and the grey-box surfaces belong to the blocking; every visible surface and light comes from @image1, @image2 and the text. SUMMARY One handheld move in a sunlit room: an arc around a dancer's pirouette that rises early into the backlight, breaks away, and lands on the robot — which dances its own pirouette for one clean second, tips past its balance point, and falls out of the bottom of the frame. The camera holds on the window where it stood. LENS AND GRADE Photorealistic live-action cinema: filmic grain, natural backlit sunlight through the glass, lens flare living in the rise and the pan, real-world physics in tulle, hair, fabric panels and floor contact.

Seedance 2.5 在 Capcut 中，提示词 -> 格式 8 秒，16:9，24fps。一次性生成，一条连续的、持续运动的手持镜头，零剪辑 —— 弧线、甩动与定帧都是同一镜完成，由 @video1 驱动：它的摄影机轨迹与时间轴就是本片的法则。语言：英语。一名女性的声音，全片画面之外（OFF SCREEN），美式英语，母语者；其下垫有制作环境声。 声音 —— EVE（画面之外，本片每一个被说出的词都由她完成） 说话者是 EVE，全片在画面之外被听到 —— 本片里只有她的声音。她的声音：一位年轻的成年非裔美国女性，Z 世代 —— 首要特质是温柔。她说的时候听起来像在微笑：轻盈、通透、温暖，是挚友之间深夜交谈的语气，是讲给她所爱之人听的睡前故事。比播音员更柔、更高一些，词句边缘游走着一丝轻微的颗粒感；音量低，但录制时贴近话筒，因此在混音中贴耳且占比大；每一句底下都带着轻柔的气息。她对每个字所倾注的关切就是热情所在，音量始终压低；吐字保持轻柔圆润，元音温暖，每一句都柔和落地。均匀、缓慢的语速，一行之内约每秒三个词；辅音柔软；句首有一次可辨的轻吸气，句尾是下降、落定的收束；分句之间有微停顿。她的口音为美式：通用美式元音，美式卷舌 r，美式语调。 台词与时间轴 EVE 旁白：{Yet no matter how much we try, we can't create anything close to the precision of the universe.}（0.30s–5.80s） —— 完整一句，画面之外；画面中，舞者闭着嘴唇跳舞，机器人的脸是一个光滑的空白壳面。 资产 @image1 作为角色参考 —— 机器人，全身：一个类人机器人，外覆浅奶油色合成织物面板，覆盖在可活动关节之上，接缝可见，头部为光滑的空白壳面，两只深色小型传感眼睛捕捉光线。它的身体、材质、颜色与头部必须与此图像一致，在它出现的每一帧中保持一致。 @image2 作为场景、光线与角色参考 —— 房间：一间宽敞的现代客厅，沐浴在温暖的午后阳光中，落地窗，窗外是树木与薄雾，奶油色沙发，淡木色地板，一块深色地毯；舞者身着白色紧身衣、层叠白色薄纱裙与系缎带的粉色足尖鞋；机器人站在右侧高窗旁，阳光倾洒而入。地点、逆光以及两者的造型都来自此版。版中的姿势只属于该版：在本片中，舞者从第一帧起就起身在跳 —— 立于足尖，正在旋转。 @video1 作为走位参考 —— 摄影机与编舞，与本片帧对齐：在棋盘格地板上的灰盒预演。蓝色人形（带球状头部的圆柱体）是舞者；红色人形（同样造型但有两条手臂）是机器人。@video1 1:1 提供摄影机 —— 从她脚部开始的低位镜头、升至其头部下方的跟随、在 2.0s–2.9s 向右甩动并在落到机器人身上时向上倾斜，使地板离开画面并保持在画面之外、缓慢的漂移与定帧 —— 以及机器人的时间轴：3.3s–4.6s 双臂抬升至 T 位，4.5s–5.8s 转身，6.1s–6.7s 倾斜，7.0s 时身体位于画面下方，一直保持至结束。棋盘格、墙面与灰盒表面属于走位；每一个可见的表面与光线都来自 @image1、@image2 与文字。 概要 在阳光洒落的房间里的一条手持运动：环绕舞者 pirouette 的一道弧线早早升起进入逆光，旋即脱开，落在机器人身上 —— 机器人用一秒钟干净地跳出它自己的 pirouette，越过平衡点后，从画面底部跌出。摄影机在它曾站立的窗户处定帧。 镜头与调色 照片级真实的真人电影：胶片颗粒，透过玻璃的自然逆光日光，在升起与横移中游动的镜头光晕，薄纱、头发、织物面板与地面接触中的真实物理感。