Seedance 2.5 官方用户手册:手把手教你做电影级 AI 视频,公式和案例全能抄

参数开到多大、四个新入口在哪、七类场景各配什么提示词公式,官方把从上手到成片的每一步都写成了可抄的操作手册。

作者:Dreamina 官方使用手册08-03 · 28 分钟⚑ 方法教程
工具教程 · 小互解读
一分钟速览
  • 即梦国际版官方出的 Seedance 2.5 用户手册:参数、界面操作、七类提示词公式带全部案例,手把手从上手到成片。
  • 里面有一个公式专治「AI 脸」,还有一张表把 12 种转场写法一次讲齐,都是能直接抄进工作流的资产。
  • 16 个亮点功能全部配官方演示片,多数还带上一代的同题对比,差距看得见。
⚑ 材料来自 Dreamina(即梦国际版)官方发布的《Seedance 2.5 User Guide》英文文档:提示词一律按英文原文收录、配中文对照译文;操作步骤和功能说明由原文编译成中文;演示视频均为官方成片,本站转码自托管。个别案例的素材与成片在原文档中配对存疑处,图注不做推断。
背景

Seedance 2.5 是什么,先垫个底

字节跳动 Dreamina(即梦国际版)发布了 Seedance 2.5 的官方用户手册:从上传素材的参数上限、界面上四个新入口,到七类场景各自的提示词公式,再到 16 个亮点功能的官方演示,手把手带你从上手一路走到电影级成片。公式和案例提示词全部能直接抄,对用即梦做视频的人,这是一份能当工作台手册用的干货。

先给没摸过这个模型的人垫个底。Seedance 2.5 是字节 Seed 团队 7 月 31 日发布的视频创作模型:单条视频 30 秒直出,一次能组合 50 份参考素材,还能对已有视频做编辑和延长。国内在即梦和豆包专业版可用,海外入口就是这份手册对应的 Dreamina。完整的发布解读和演示片在站内这篇:

相关解读
字节跳动发布 Seedance 2.5:单次 30 秒直出,一条指令能同时指挥 50 个参考素材
发布当天的完整解读:能力、演示片、上线渠道和价格情况。

手册开篇把这一代的升级按用户呼声排成了七件事,先扫一眼有个总图:

01视频延长,连续创作
单次可给 30 秒内的视频再延 4-30 秒,还能「套娃」连续延长,极限一条 60 秒。
02超长视频最长 180 秒
Long Video 模式一次直出 30-180 秒,3 分钟长片不用分段拼接。
03时间戳文字控制
文字里直接写第几秒发生什么,动作和转场能精确到秒,剧本级还原。
04打破语言边界
优先优化中文、英语、西班牙语、印尼语、马来语,全面覆盖泰语、阿拉伯语、葡萄牙语、越南语、日语、韩语,母语直接写提示词。
05去除无关字幕和 BGM
「不要字幕」「不要背景音乐」这类反向指令终于稳定管用,产出干净底片。
06基础生成效果全面优化
削弱「AI 感」:皮肤毛发质感更真、切镜一致性更稳、大量复杂「长尾动作」解锁。
07多模态参考综合优化
一口气升了十项:BGM 分离、迁移创意、局部消除、空间视角修改、音色参考、多人参考、绿幕编辑、白模控制、无缝转场、多宫格分镜。

下面照手册的顺序手把手走一遍:先把参数和界面摸清,再逐类学写法,最后看官方演示。

参数

参数放开到多少:和 2.0 对照着看

上传素材的口子这一代放开了不少,和 2.0 对照着看最直观:

素材类型2.0 上限2.5 上限
图片每次最多 9 张每次最多 30 张(单张 4K 内,jpeg/png/webp/bmp/tiff/gif/heic/heif,单张小于 30MB)
视频最多 3 段,总时长 15 秒内最多 10 段,单段 2-30 秒,总时长 30 秒内(mp4/mov,480p-4K,单个 200MB 内)
音频最多 3 段,总时长 15 秒内最多 10 段,单段 2-30 秒,总时长 30 秒内(wav/mp3,单个 15MB 内)
只传音频必须搭配至少一份图或视频支持纯音频驱动生成

生成端同步放宽:单次生成时长从 4-15 秒放宽到 4-30 秒(97 到 721 帧),输出分辨率 480p 或 720p。

上限归上限,手册同时给了一组小得多的稳定区建议,超出仍能跑,但按官方的说法要「抽卡」(gacha):多生成几次碰运气挑一条好的。

场景稳定区建议
主体图放几个主体1-8 个效果较好;9-12 个可试,稳定性下降
主体音视频放几个主体1-5 个效果较好;6-10 个可试,稳定性下降
主体音视频单段多长5-10 秒效果较好;更长稳定性下降
同一主体要多视角1-5 个主体时单视角、多视角都行;超过 5 个主体优先单视角。要多视角就拆成多张图,多图多视角比单图拼多视角稳
视频编辑用多长原片20 秒内效果较好
视频编辑配几张参考图1-5 张效果较好;6-8 张可试,稳定性下降
上手

界面上四个新入口,各自怎么点

进到 Dreamina 的生成界面,先做基本设置:选 AI Video、选 Dreamina Seedance 2.5 模型,再挑画幅、分辨率和时长。模式菜单里有四个选项:Omni Reference(多模态参考;界面如此拼写,文档文本里多处误拼作 Onmi)、Smart Edit、Long Video、First and last Frames,其中 Smart Edit 和 Long Video 是这一代的新入口。

基本设置:选模型后在模式菜单里切换,Smart Edit 与 Long Video 是新功能。
基本设置:选模型后在模式菜单里切换,Smart Edit 与 Long Video 是新功能。

Long Video:30 到 180 秒一次直出

  1. 写好视频脚本或提示词;
  2. 点开模式下拉菜单(默认可能显示 Omni Reference),选 Long Video;
  3. 点工具栏的时钟图标,拖时间轴或直接输入数字,在 30-180 秒之间任选;
  4. 确认画幅(如 16:9)、分辨率(如 720P+)和时长后点生成,一次出完整长片。
操作截图操作截图
选择 Long Video 模式 → 拖时间轴设定 30-180 秒点图可看大图

Extend Video:给成片续时长,还能「套娃」

  1. 在生成流里找到要延长的视频,点视频下方的 Extend Video 图标(只有 30 秒内的视频有这个入口);
  2. 点时钟图标,拖时间轴或输入数字选延长秒数(单次 4-30 秒);
  3. 在文本框输入提示词,只作用于新加的部分,前面的画面原样保留。比如原片 10 秒、延 5 秒,提示词「男主角落地后看向镜头」只管最后 5 秒;
  4. 点生成,得到原时长加延长时长的完整视频。

套娃玩法:只要当前成片仍在 30 秒内,就能对它再点一次延长,反复叠加。极限操作是 30 秒原片一次延 30 秒,直出 60 秒。

操作截图操作截图操作截图
选中视频 → 设定延长时长 → 输入只管新增段的提示词点图可看大图
操作截图操作截图操作截图
10 秒延成 15 秒,再延成 20 秒,套娃可以一直叠到 60 秒点图可看大图

Smart Edit 三兄弟:文字改、圈着改、改成片

对画面局部不满意、或想往画面里加东西,走这三个入口:

  1. Smart Edit:主界面选 Smart Edit 模式,上传视频,直接用文字描述要改什么;
  2. Edit with marks:上传本地视频后输入框会亮起这个入口,弹出的画面下方有一排标注工具,框选、箭头、锚点,直接在画面上圈出要改的位置(比如红框圈住男孩的裤子),再在下方输入修改要求(「把男孩的裤子改成黑色,全片保持一致」),点 ✅ 加入输入框;
  3. Edit Video:对已生成的视频点 Video Edit 图标做二次编辑,操作和 Edit with marks 相同;
  4. 标注加入后输入框会出现带时间的标签(如 ⏱️00:00 视频帧标注),确认无误点生成。
操作截图操作截图操作截图
Smart Edit 模式入口 → 上传视频解锁 Edit with marks → 打开「画板」标注界面点图可看大图
操作截图操作截图操作截图操作截图
对已生成视频:选中 → 框选/箭头/锚点圈出修改处 → 确认时间戳生成 → 全片完成局部修改点图可看大图

Clay Renderer 插件:从 Maya/Blender 直通渲染

给 3D 从业者的官方插件:在 Maya 或 Blender 里摆好机位路线后,用插件直接导出白模视频并上传到 Dreamina 网页端,让模型参考白模的动作和运镜做高质量渲染。插件安装与使用另有官方文档:【Dreamina】Seedance 2.5 Clay Renderer Plugin User Guide。写法部分见下文白模渲染一节。

写法

基础公式:素材说明、概括、情节、全局补充,四段写完

手册给的基础骨架是四段式:

段落写什么
完整提示词 = 素材说明 + 一句话概括 + 具体情节描述 + 全局补充(收尾)
素材说明按上传顺序给素材编号,写明各自用途(哪份是角色、音色、动作、场景),一份都不能漏。
一句话概括主体 + 地点 + 事件 + 主题/风格 + 特殊运镜。
具体情节描述按时间线或故事线写。每个分镜/时间段里配两类描述,正向:画面内容 + 运镜 + 动作 + 台词 + 音效;反向:不想要的东西(如「不要字幕」「不要 BGM」)。
全局补充(收尾)把全程必须保持的元素再强调一遍:机位、场景特征、整体声音氛围、光线;全局都不想要的(字幕、背景音乐)也在这里再说一次。

官方案例是一条丹顶鹤湿地纪录片,四段式的痕迹在提示词里看得非常清楚:

官方案例 · 丹顶鹤湿地纪录片
Realistic nature documentary style, cinematic realistic lighting. In the morning mist, within a vast wetland reed bed @image1, an elegant red-crowned crane is spreading its wings and dancing. @image2 The red-crowned crane features snow-white feathers, a bright red crown, a slender and upright posture, and light, noble movements. The scene is a shallow wetland in autumn, with sparkling water, surrounded by golden reeds, duckweed, and aquatic plants. The out-of-focus background features distant morning mist and the rising sun. The camera uses a low-angle medium shot with a slight handheld feel, remaining generally steady and keeping the red-crowned crane in the center of the frame at all times.
0s-3s: A red-crowned crane stands quietly in the shallow water with an elegant posture. It slowly begins to unfold its broad black-and-white wings with gentle movements. The airflow from the flapping wings creates ripples on the water surface. The morning breeze blows, and sunlight slants in from the horizon, piercing through the mist to create a Tyndall effect (volumetric lighting).
3s-8s: The red-crowned crane leaps lightly on the water, its feet alternately tapping the surface, splashing crystal-clear water droplets. Its slender neck tilts backward with its beak pointing to the sky. Then, its body lands lightly, folding its wings to return to a standing posture. Its head elegantly looks around, letting out a clear, long call.
Low-angle, slight handheld camera feel, moving up and down slightly to follow the crane's jumping trajectory. Natural depth of field, with the foreground reeds slightly blurred. The main subject, the red-crowned crane, is in sharp focus, while the background morning mist and sunrise are softly blurred. Natural environmental sounds: flowing water, the powerful flapping of wings, and the crane's call. The overall atmosphere is peaceful, ethereal, and highly aesthetic.
中文对照:写实自然纪录片风格,电影级真实光影。清晨薄雾中,在广阔的湿地芦苇荡 @图1 里,一只优雅的丹顶鹤正展翅起舞。@图2 丹顶鹤羽毛雪白、头顶鲜红,体态修长挺拔,动作轻盈高贵。场景是秋日浅水湿地,水面波光粼粼,四周是金色芦苇、浮萍和水生植物,虚焦背景是远处晨雾与初升的太阳。镜头用低角度中景,带轻微手持感但整体稳定,丹顶鹤始终位于画面中心。 0-3 秒:丹顶鹤安静立于浅水中,姿态优雅,缓缓展开宽大的黑白双翼,动作轻柔,振翅气流在水面荡起涟漪。晨风吹拂,阳光从地平线斜射进来,穿透薄雾形成丁达尔光。 3-8 秒:丹顶鹤在水面轻盈跳跃,双足交替点水,溅起晶莹水珠;细长的脖颈后仰、喙尖指天,随后身体轻轻落下收翅回到站姿,头部优雅环顾,发出清亮悠长的鹤鸣。 低角度、轻微手持感,镜头随鹤的跳跃轨迹微微上下移动。自然景深,前景芦苇微虚,主体丹顶鹤清晰锐利,背景晨雾与日出柔和虚化。自然环境音:流水声、有力的振翅声、鹤鸣。整体氛围宁静空灵、极具美感。
参考图参考图
参考图:湿地芦苇荡场景 + 白底丹顶鹤主体图
官方成片:四段式写出来的丹顶鹤纪录片。
写法

真人角色八维公式:专治 AI 脸和双胞胎脸

生成真人角色最常见的两个翻车点:一眼看出是 AI,或者几个角色长成「双胞胎」。手册的解法是把角色拆成八个维度逐一写死,每个维度还有自己的细分公式,动画角色也能套用这一套。

1. 年龄种族 2. 肤色质感 3. 五官细节 4. 眼神灵魂 5. 发型发色 6. 服装质地 7. 体型气质 8. 其他要求 七个实心维度逐一写死,虚线的「其他要求」按需追加
角色装配线:八个维度拼出一个不像 AI 的角色
八维总公式
Character = [Age/Race] + [Skin Color/Skin Texture] + [Facial Details] + [Eye/Soul] + [Hairstyle/Hair Color] + [Clothing/Clothing Texture] + [Body Type/Emotion/Temperament] + [Other Requirements] (if any)
中文对照:角色 = 年龄/种族 + 肤色/皮肤质感 + 五官细节 + 眼神/灵魂 + 发型/发色 + 服装/衣着质感 + 体型/情绪/气质 + 其他要求(如有)

七个维度各自的细分公式和官方示例,示例都是能直接抄的角色描写句(英文原句 + 中文对照):

01年龄 / 种族(定角色基调)
细分公式:[Specific age] + [Specific nationality/race] + [Style adjective] + [Face shape noun]
即:具体年龄 + 国籍/种族 + 风格形容词 + 脸型名词
  • 22-Year-old East Asian woman with a cold and classical bone-like face.
    22 岁东亚女性,清冷古典的骨相脸。
  • 25-Year-old United Kingdom male, with a thin and intelligent British aristocratic face.
    25 岁英国男性,清瘦智性的英伦贵族脸。
  • A 26-year-old African-American woman with an extremely three-dimensional black pearl supermodel face
    26 岁非裔美国女性,立体感极强的黑珍珠超模脸。

02肤色 / 皮肤质感(拒绝假白,强制拟真)
细分公式:[cool and warm tones] + [specific skin tone noun] + [skin texture adjective] + [forced fidelity suffix: retain the true micro pores and skin texture]
即:冷暖调 + 具体肤色名词 + 质感形容词 + 强制拟真后缀「保留真实微小毛孔和皮肤纹理」(可加雀斑/瑕疵)
  • Cold-toned fair skin tone, exuding a delicate and moist jade texture, retaining the true micro-pores and skin texture.
    冷调白皙肤色,透出细腻温润的玉质感,保留真实微小毛孔和皮肤纹理。
  • Warm honey color, with a glossy feeling after being exposed to sunlight. There are faint real freckles on the bridge of the nose, preserving fine pores and skin texture.
    暖蜜色,有晒后光泽感,鼻梁上有隐约真实的雀斑,保留细小毛孔和皮肤纹理。
  • Rough wheat-colored skin, with a sense of wind and frost and fine fine lines, retaining true large pores, stubble marks, and skin texture.
    粗粝小麦色皮肤,带风霜感和细纹,保留真实的粗大毛孔、胡茬痕迹和皮肤纹理。
03五官细节(决定辨识度)
细分公式:[Eye shape description] + [Eyebrow bone feature] + [Nose bridge feature] + [Lip shape feature] + [Jawline/Bone feature] (write at least 3-4 points)
即:眼型 + 眉骨 + 鼻梁 + 唇形 + 下颌线/骨相,至少组合 3-4 项
  • Slender peach blossom eyes (slightly moist eye sockets), small and straight nose bridge, plump lips with a faint smile at the corners of the mouth, and a soft jawline.
    细长桃花眼(眼眶微湿),小巧挺直的鼻梁,饱满唇形嘴角带浅笑,柔和下颌线。
  • Slightly upturned Danfeng eyes, highly folded three-dimensional brow bones, high and restrained nose bridge, slightly protruding cheekbones, and a clear and neat jawline like a knife.
    微微上挑的丹凤眼,高折立体的眉骨,高挺内敛的鼻梁,微凸颧骨,下颌线利落如刀。
  • Round almond eyes, with a sense of innocence under the eyelids, slightly upturned nose with a fleshy feel, full lips, slightly baby fat on the cheeks, and smooth facial lines.
    圆杏眼,眼睑下带无辜感,微翘带肉感的鼻子,丰满嘴唇,脸颊微带婴儿肥,面部线条流畅。
04眼神 / 灵魂(决定情绪)
细分公式:[eye adjective] + [information/metaphor conveyed by eye contact] + [underlying emotion revealed]
即:眼神形容词 + 眼神传递的信息或比喻 + 流露的底层情绪
  • The eyes are extremely affectionate and affectionate, as if there is a pool of spring water in the bottom of the eyes. The gaze is gentle and focused, revealing deep attachment and a hint of reluctance.
    眼神极尽深情,眼底似有一汪春水,目光温柔专注,流露深深依恋和一丝不舍。
  • The eyes are extremely calm and empty, like a bottomless black hole. The gaze is sharp and oppressive, revealing a hint of elusive danger.
    眼神极度平静空洞,像不见底的黑洞,目光锐利有压迫感,透出一丝难以捉摸的危险。
  • The eyes are scattered and dim, with a low and unfocused gaze, revealing a deep sense of exhaustion and numbness towards life.
    眼神涣散黯淡,视线低垂无焦点,流露对生活深深的疲惫与麻木。
05发型 / 发色(拒绝单调黑,写出状态和氛围)
细分公式:[Specific hair color] + [Hair condition/texture] + [Specific hairstyle noun] + [Environmental interaction supplement (such as being blown by the wind/messy)]
即:具体发色 + 头发状态/质感 + 发型名词 + 环境互动补充(被风吹动、凌乱等)
  • The black hair is tied into a casual and elegant classical low bun, fixed only with a plain jade hairpin. A few strands of fine black hair casually fall on the edge of the cheeks, gently brushed by the breeze.
    黑发挽成随意优雅的古典低髻,只用素玉簪固定,几缕碎发垂在脸颊边被微风轻拂。
  • Caramel brown slightly curly long hair presents a natural fluffy and messy feeling like just waking up, with a few strands of hair casually draped over the collarbone.
    焦糖棕微卷长发,呈现像刚睡醒的自然蓬松凌乱感,几缕发丝随意搭在锁骨。
  • Silver-white ear-length short hair with a strong mirror-like highlight texture, sharp edges of the hair, tightly attached to the scalp.
    银白色齐耳短发,带强烈镜面高光质感,发边锋利,紧贴头皮。
06服装 / 衣着质感(从侧面立住人设)
细分公式:[pattern/cut] + [color] + [specific clothing noun] + [fabric material/new and old state] + [wearing method/detail]
即:版型/剪裁 + 颜色 + 具体服装名词 + 面料材质/新旧状态 + 穿着方式/细节
  • A minimalist plain white cross-necked ancient costume, made of soft and elegant silk or light gauze with a sheen. The collar is slightly scattered and there is no extra embroidery.
    极简素白交领古装,柔软素雅的丝绸或轻纱带光泽,领口微散,没有多余刺绣。
  • A loose, worn-out dark brown motorcycle leather jacket with real wear creases and reflective metal zippers on the leather surface, paired with a washed-to-white gray T-shirt.
    宽松做旧的深棕机车皮衣,皮面有真实磨损折痕和反光金属拉链,内搭洗到发白的灰 T 恤。
  • A deep black double- suit with extremely slim tailoring, made of stiff high-end wool blend fabric, paired with a dark red silk tie, is meticulous.
    剪裁极修身的深黑双排扣西装,面料是挺括的高级羊毛混纺,配深红丝质领带,一丝不苟。
07体型 / 情绪 / 气质(定构图和故事感)
细分公式:[Skeleton/Shoulder Features] + [Composition/Shot Requirements] + [Character Action/Sight] + [Overall Atmosphere Words]
即:骨架/肩部特征 + 构图/景别要求 + 人物动作/视线 + 整体氛围词(故事主干里已写过构图和动作时,中间两项可省)
  • Slim figure, slender shoulders. Close-up composition of the half-body above the chest, with direct gaze towards the camera. Overall, it exudes a gentle, graceful, and affectionate classical romantic temperament.
    身形纤细、肩线瘦削;胸口以上半身特写构图,目光直视镜头;整体散发温柔深情的古典浪漫气质。
  • The skeleton is broad and the shoulders are thick and broad. The whole body composition is shot from a low camera position, with the character standing sideways and glancing back at the camera. The whole exudes a hardcore and tragic epic feeling of survival in a desperate situation.
    骨架宽大、肩膀厚实;低机位全身构图,人物侧身回望镜头;整体是绝境求生的硬核悲壮史诗感。
  • The figure is slender and upright. The composition of the face is a close-up shot, with half of the face hidden in the shadow. The whole exudes a mysterious and dangerous, suffocating horror movie protagonist temperament.
    身形瘦长挺拔;面部特写构图,半张脸藏在阴影里;整体是神秘危险、令人窒息的惊悚片主角气质。
八维拼装完整示例
[Age/Race] 22-year-old East Asian woman with a classical and gentle movie face.
[Skin color/skin texture] Cold tone fair skin tone, showing a delicate warm texture, retaining the true micro pores and skin texture.
[Facial Details] slender peach blossom eyes (slightly moist eye sockets), slightly stretched eyebrows, small and straight nose bridge, plump lips and gentle smile at the corners of the mouth, soft jawline.
[Eyes/Soul] The eyes are extremely affectionate, the bottom of the eyes seems to have a pool of spring water, the eyes are gentle and focused, with deep attachment and a hint of reluctance.
[Hairstyle/Hair Color] The black hair is tied into a casual and elegant classical low bun, fixed only with a plain jade hairpin. A few strands of fine black hair hang inadvertently on the edge of the cheeks, gently swaying with the breeze.
[Clothing/Clothing texture] Wearing a minimalist plain white cross-collar ancient costume, the fabric is simple and elegant silk or light gauze with a soft luster, and the collar is slightly scattered.
[Body shape/mood/temperament] Slim figure, slender shoulders, exudes a gentle and affectionate classical romantic temperament. Close-up composition, eye contact with the camera.
中文对照:【年龄/种族】22 岁东亚女性,古典温柔的电影脸。【肤色/质感】冷调白皙肤色,细腻温润,保留真实微小毛孔和皮肤纹理。【五官细节】细长桃花眼(眼眶微湿)、眉头微舒、小巧挺直的鼻梁、饱满唇形嘴角带温柔笑意、柔和下颌线。【眼神/灵魂】眼神极尽深情,眼底似有一汪春水,温柔专注,带深深依恋和一丝不舍。【发型/发色】黑发挽成随意优雅的古典低髻,只用素玉簪固定,几缕碎发垂在脸颊边随微风轻摆。【服装/质感】极简素白交领古装,素雅丝绸或轻纱面料带柔光,领口微散。【体型/气质】身形纤细、肩线瘦削,散发温柔深情的古典浪漫气质;特写构图,与镜头目光相接。
写法

30 秒长视频:一句话定调加三个模块

30 秒的戏事件多、镜头多,手册的写法是「一句话定调 + 三个模块」,把素材、世界观和分镜各管一段:

部分写什么
一句话定调整条视频的基调,如「29 秒一镜到底(古装女子江边送别)」。
模块一 · 素材参考层没传参考素材可略。公式:基于上传的 [素材类型],严格保持 [……] 一致,严格锁定 [……]。全文可以多次 @图/@视频/@音频,越点名越准。
模块二 · 全局设定全局世界观 + 防崩坏,六件事:基础环境与质感(时间+地点+氛围+强调真实物理质感)、视觉风格(电影感/写实/二次元 + 景深 + 光线)、镜头语言(景别/视角 + 细节)、角色造型(套上面的八维公式或直接 @参考图)、表演核心(成片要着重表现什么)、禁止项(从声音、字幕、行为、易崩点列起)。
模块三 · 时间戳剧本分镜把长视频切成若干时间片,每片写:[起秒-止秒] + [核心阶段/主题]、[动作/表情等物理层指令]、[情绪解析或镜头调度](把物理指令背后的深层表达讲给 AI 听),需要切镜就套延长一节的转场写法。

模块二和模块三各有一份官方示范,都来自「江边送别」这条片子:

模块二示范 · 全局设定
[Global Scene Setting]Base Environment: Early morning by the river, with a blurred small boat in the background, a lonely farewell scene. The atmosphere is quiet and restrained.Visual Style: Cinematic, shallow depth of field (simple and blurred background), soft natural light.Camera Language: Consistently maintains a close-up shot of the woman, imitating the subjective POV of "him" standing directly opposite her. Subtle handheld camera movement (breathing effect), no fast cuts.
[Character Styling][Age / Ethnicity] 22-year-old East Asian woman, with a highly classical, gentle cinematic face.[Skin Tone / Texture] Cool-toned fair skin, revealing a delicate and moist texture, retaining realistic fine pores and natural skin texture.[Facial Features] Slender and elegant eyes (slightly moist/teary), eyebrows slightly relaxed, a delicate and straight nose bridge, full lips with a faint, gentle smile at the corners, and a soft jawline.[Eyes / Soul] Extremely affectionate and expressive gaze, eyes shimmering like spring water. Gentle and focused, revealing deep attachment and a hint of reluctance to part.[Hair Style / Color] Jet-black hair pinned into a casual yet elegant classical low bun, secured only by a plain jade hairpin. A few loose strands of black hair casually fall around her cheeks, gently swaying in the breeze.[Clothing / Texture] Wearing a minimalist, pure white traditional cross-collar dress (Hanfu style). The fabric is elegant silk or lightweight chiffon with a soft luster, slightly loosened at the collar.[Physique / Emotion / Aura] Slender figure with delicate narrow shoulders, exuding a gentle, graceful, and deeply affectionate classical romantic aura.
[Core Performance]Restrained and nuanced. Focus on capturing the shifting gaze, the rise and fall of her breathing, slight trembling of the lips, subtle movements of the brow, subtle changes in the nostrils, throat swallowing, and the natural process of a single tear sliding down her cheek.
[Negative Prompts / Restrictions]No exaggerated crying, no fast cuts, no large body movements, no generating extra dialogue or BGM, no runny nose, no premature dropping of tears.
中文对照:【全局场景设定】基础环境:清晨江边,背景有虚化的小船,寂寥的送别场景,氛围安静克制。视觉风格:电影感、浅景深(背景简洁虚化)、柔和自然光。镜头语言:始终保持女子的特写,模拟站在她对面的「他」的主观视角(POV),镜头带轻微手持呼吸感,不快切。 【角色造型】【年龄/种族】22 岁东亚女性,极具古典气质的温柔电影脸。【肤色/质感】冷调白皙,细腻温润,保留真实毛孔与皮肤纹理。【五官】细长清雅的眼睛(微湿含泪)、眉头微舒、精致挺直的鼻梁、饱满嘴唇嘴角带一丝温柔浅笑、柔和下颌线。【眼神】深情至极,眼波如春水,温柔专注,透出深深依恋和一丝不舍。【发型】乌黑长发挽成随意优雅的古典低髻,只用素玉簪固定,几缕碎发垂颊随风轻摆。【服装】极简纯白交领汉服,素雅丝绸或轻纱面料带柔光,领口微散。【体型/气质】身形纤细、肩线瘦削,散发温柔深情的古典浪漫气质。 【表演核心】克制而细腻。着重捕捉眼神流转、呼吸起伏、嘴唇轻颤、眉头细微变化、鼻翼翕动、喉头吞咽,以及一滴泪自然滑落的全过程。 【禁止项】不要夸张大哭、不要快切、不要大幅肢体动作、不要生成多余台词或 BGM、不要流鼻涕、眼泪不要提前掉。
模块三示范 · 时间戳情绪分镜
[Emotion and Action Storyboard (0-29 Seconds Progression)]
Stage 1: 0-3s [Questioning] — Reluctant Waiting
Action/Expression: The camera focuses on her face. She looks directly into the lens (at "him" opposite her). There are no tears in her eyes yet, displaying a quietness that slowly sinks in. She looks at his face, her brows slightly furrowed, her lips slightly parted, and whispers softly: "Are you really leaving?"
Emotion Analysis: Direct gaze (Emotion: She is not bitterly complaining, but waiting for him to personally confirm an answer she already guessed). Soft whisper (Emotion: After the words fall, it feels as if many unspoken words remain stuck in her throat).
Stage 2: 3-10s [Resignation] — Swallowing the Bitterness
Action/Expression: Her gaze slowly shifts away from his face, looking toward the empty space beside him. Her eyelids droop, upper and lower eyelashes overlapping. The corners of her mouth slightly curl into a brief smile, which instantly fades. Her nostrils flare slightly as she takes a highly restrained, quiet breath, her chest rising and falling gently once.
Emotion Analysis: The smile (Emotion: A bitter smile of self-mockery and resignation; an "I knew it" kind of smile). The chest movement (Emotion: Through this action, it seems she is forcefully swallowing her grievances and sorrow back into her stomach).
Stage 3: 11-17s [Remembering] — The Final Engraving
Action/Expression: The camera slowly pushes in (slight zoom). She turns her gaze back to his face, her eyes slowly and carefully scanning his features. Her eye rims are now reddish, but she stubbornly holds back her tears, not letting them fall. Accompanied by a 0.5-second dead-silent pause in the middle, her lips twitch slightly then press together again, her jaw tightens, and her throat swallows slightly.
Emotion Analysis: Her gaze slowly scanning his face (Emotion: At this moment, her gaze is no longer just for bidding farewell, but to engrave and remember this person's appearance in her life for one last time).
Stage 4: 18-23s [Regret] — A Silent Sigh
Action/Expression: She lowers her eyes, and only then do the tears begin to fall. A single tear drops directly without warning, landing on her collar. She does not wipe it away, nor does she sob; she just quietly lets the tear slide down. When she looks up at him again, the emotion in her eyes has changed. Immediately after, her tightly knit brows slowly begin to relax. She shakes her head slightly (a microscopic movement, almost imperceptible).
Emotion Analysis: The change of emotion when looking up (Emotion: Her gaze shifts from trying to make him stay to profound regret, as if saying "We could have been so good together," but she leaves it unspoken). The relaxing of the brows and the slight head shake (The un-furrowing of her brows symbolizes a gradual inner release, while that tiny head shake is an instinctive, final sigh and denial of her helpless fate).
Stage 5: 24-29s [Letting Go] — Relieved Acceptance
Action/Expression: The camera pushes into an extreme close-up. She tries hard to form a gentle smile (very light, very soft, with no self-mockery). Just as the smile reaches the corners of her mouth, a second tear slides from the corner of her eye, past her nose, and stops at her lip. She doesn't cry out loud, she doesn't break down; she just quietly looks at him with tears in her eyes, and says in a barely audible but strongly steadied voice: "You can go." (When saying the word "go", there is an extremely faint tremble in her voice, but she forcefully suppresses the crying tone). After saying this, her smile freezes on her face, tears continue to slide down, and her gaze never wavers, still staring at "him" in front of the lens. The camera finally holds on her tearfully smiling face, her eyes filled with profound regret.
Emotion Analysis: The gentle smile (Emotion: Unlike the previous smile, this one is very light and soft, without self-mockery, only pure reluctance to part and the ultimate tenderness). She doesn't break down; instead, she uses extreme restraint to accomplish the final letting go and fulfillment.
中文对照:【情绪与动作分镜(0-29 秒递进)】 阶段一 0-3 秒【追问】,不舍的等待。动作/表情:镜头聚焦她的脸,她直视镜头(望着对面的「他」),眼里还没有泪,是缓缓下沉的安静。她看着他的脸,眉头轻蹙、嘴唇微启,轻声问:「你真的要走吗?」情绪解析:直视(她没有哭诉,是等他亲口确认一个她已猜到的答案);轻语(话音落下,仿佛还有很多话堵在喉咙里)。 阶段二 3-10 秒【认命】,把苦咽下去。动作/表情:视线缓缓从他脸上移开,望向他身侧的虚空;眼睑低垂、睫毛交叠,嘴角浮起一个转瞬即逝的浅笑,鼻翼轻张,做了一次极克制的安静吸气,胸口轻轻起伏一次。情绪解析:那个笑是自嘲与认命的苦笑(「我早知道会这样」);胸口的起伏像是把委屈和难过硬生生咽回肚子里。 阶段三 11-17 秒【铭记】,最后一次刻进心里。动作/表情:镜头缓缓推近(轻微变焦)。她把视线收回到他脸上,一寸寸缓慢仔细地看过他的眉眼。眼眶已经泛红,却倔强地忍着不让泪掉下来。中间有 0.5 秒死寂的停顿,嘴唇动了动又抿住,下颌绷紧,喉头轻轻吞咽。情绪解析:此刻的注视已不只是道别,是要把这个人的样子在生命里最后刻一遍。 阶段四 18-23 秒【遗憾】,无声的叹息。动作/表情:她垂下眼,泪才开始掉。第一滴泪毫无预兆地直落,落在衣领上。她不擦、不啜泣,只是安静任它滑落。再抬眼看他时,眼里的情绪已经变了。紧蹙的眉头缓缓松开,极轻微地摇了一下头(微小到几乎察觉不到)。情绪解析:抬眼时的变化,从想留住他变成深深的遗憾,像在说「我们本可以很好」,但没说出口;眉头松开象征内心逐渐放下,那一下轻摇是对无奈命运本能的、最后的叹息与否认。 阶段五 24-29 秒【放手】,释然的成全。动作/表情:镜头推成极特写。她努力挤出一个温柔的笑(很轻、很柔、不带自嘲)。笑意刚到嘴角,第二滴泪从眼角滑过鼻侧停在唇边。她不哭出声、不崩溃,只是含泪安静看着他,用几乎听不见却极力稳住的声音说:「你走吧。」(说「走」字时声音有极轻微的颤,但被她硬压住哭腔。)说完笑容定在脸上,泪继续滑落,目光始终没有移开,仍望着镜头前的「他」。镜头最后停在她含泪微笑的脸上,眼里盛满深深的遗憾。情绪解析:这个笑和前面不同,很轻很软,没有自嘲,只有纯粹的不舍与极致的温柔。她没有崩溃,用极致的克制完成了最后的放手与成全。
成片案例 · 29 秒江边送别一镜到底
29-Second One-Shot (Ancient Costume Woman's Riverside Farewell)
[Global Scene Setting]Base Environment: Early morning by the river, with a blurred small boat in the background, a lonely farewell scene. The atmosphere is quiet and restrained.Visual Style: Cinematic, shallow depth of field (simple and blurred background), soft natural light.Camera Language: Consistently maintains a close-up shot of the woman, imitating the subjective POV of "him" standing directly opposite her. Subtle handheld camera movement (breathing effect), one continuous shot with no fast cuts.
[Character Styling][Age / Ethnicity] 22-year-old East Asian woman, with a highly classical, gentle cinematic face.[Skin Tone / Texture] Cool-toned fair skin, revealing a delicate and moist texture, retaining realistic fine pores and natural skin texture.[Facial Features] Slender and elegant eyes (slightly moist), eyebrows slightly relaxed, a delicate and straight nose bridge, full lips with a faint, gentle smile at the corners, and a soft jawline.[Eyes / Soul] Extremely affectionate and expressive gaze, eyes shimmering like spring water. Gentle and focused, revealing deep attachment and a hint of reluctance to part.[Hair Style / Color] Jet-black hair pinned into a casual yet elegant classical low bun, secured only by a plain jade hairpin. A few loose strands of black hair casually fall around her cheeks, gently swaying in the breeze.[Clothing / Texture] Wearing a minimalist, pure white traditional cross-collar dress (Hanfu style). The fabric is elegant silk or lightweight chiffon with a soft luster, slightly loosened at the collar.[Physique / Emotion / Aura] Slender figure with delicate narrow shoulders, exuding a gentle, graceful, and deeply affectionate classical romantic aura.
[Core Performance]Restrained and nuanced. Focus on capturing the shifting gaze, the rise and fall of her breathing, slight trembling of the lips, subtle movements of the brow, subtle changes in the nostrils, throat swallowing, and the natural process of a tear sliding down.
[Negative Prompts]No exaggerated crying, no fast cuts, no large body movements, no extra dialogue, no BGM, no runny nose, no premature dropping of tears.
中文对照:即上面两个模块的拼装成品,「29 秒一镜到底(古装女子江边送别)」+ 全局场景设定 + 角色八维造型 + 表演核心 + 禁止项,一镜到底版本(全文与上两块一致,镜头语言处多一句「一个连续镜头、无快切」)。
官方成片:29 秒一镜到底,古装女子江边送别。
看另外五条 30 秒案例(切镜转场、机甲、武侠、冰川、舞会)
切镜版 · 魔法森林精灵少女
[Global Settings] (Worldview & Anti-Distortion)Base Environment & Texture: An ancient magical forest in the early morning, enveloped in dense morning mist. Emphasize extreme realistic physical textures and optical performance. The Tyndall effect (God rays / volumetric lighting) penetrating the tree canopy, glowing spores floating in the air, and the optical refraction of water droplets on plant leaves must be highly realistic and delicate.Visual Style: High Fantasy epic style. Primary color palette of emerald green and dawn gold. Depth of field changes dynamically with camera cuts. Uses backlighting and side-backlighting (rim lighting) to accentuate the dreamy atmosphere.Camera Language: Montage editing. Includes macro close-ups, crane pull-out wide shots, and 360-degree orbiting tracking shots.
[Character Styling][Age / Ethnicity]: 18-year-old elf girl, with an extremely ethereal, transparent, innocent "first love" cinematic face.[Skin Tone / Texture]: Cool-toned extreme pale skin, revealing a dewy, crystal-clear texture like morning dew. Retains realistic micro-pores, translucent skin texture, and faint, natural light freckles across the bridge of the nose.[Facial Features]: Round, innocent doe eyes with slightly downturned lower eyelids (puppy-dog eyes), soft and gentle brow bone, slightly upturned and slightly fleshy nose tip, plump and delicate lip tubercle, slight baby fat on the cheeks, and iconic long pointed elf ears.[Eyes / Soul]: Extremely clear and innocent gaze, entirely unworldly. Her eyes convey a doe-like naivety, purity, and a strong curiosity and affinity for all natural things.[Hair Style / Color]: Aurora silver-white, waist-length hair. Presents a premium texture, silky smooth with a subtle reflection. The hair falls naturally, with a few tiny vines emitting a faint blue glow intertwined among the strands, gently swaying in the wind.[Clothing / Texture]: A feather-light, cross-collar, off-the-shoulder long dress woven from translucent mint-green chiffon and real emerald-green leaves. The fabric is extremely thin, featuring realistic yarn textures, and appears semi-translucent under the light.[Physique / Emotion / Aura]: Extremely petite skeletal frame, narrow and thin shoulders, with a posture as light as a feather. Exudes an unearthly, sacred, pure, ethereal, and beautifully romantic fantasy fairytale aura.Core Performance: Showcase the innocent feeling of awakening from a deep slumber and her ethereal interaction with the magical environment.[Negative Prompts]: No modern human artifacts. No rapid shaky camera movements (keep it smooth and soft). No distorted facial features or extra fingers. Force mute (no subtitles, no BGM).
[Timestamp Storyboard (with explicit cuts)]
[00:00 - 00:07] Shot 1: [Slumbering Microcosm] (Macro Close-up)Action / Physics Directives: Extreme macro close-up. The frame shows a giant, dew-covered glowing banana leaf. The camera pans slowly, revealing the elf girl curled up and sleeping in the center of the leaf. Her eyes are tightly closed, her silver hair spread across the green leaf, her chest rising and falling smoothly with gentle breathing, and a few glowing blue butterflies resting on the tips of her hair.Camera Subtext: Extreme tranquility and aesthetic beauty. Use a microscopic perspective to draw the audience closer to the character, reflecting her ethereal integration with nature.
[00:08 - 00:15] Shot 2: [Awakening and Touch] (Medium Close-up ➔ Close-up)Action / Physics Directives: Require a natural transition at the cut (no hard/abrupt cuts). Cut to a medium close-up of her face. A beam of morning light happens to hit her cheek. She frowns slightly and slowly opens her eyes (lake-green pupils). She sits up, looking around with innocent bewilderment. Then, the camera follows her gaze, cutting to a close-up of her hand: she extends a slender finger, and a glowing blue butterfly gently lands on her fingertip.Camera Subtext: Showcases the beautiful moment of life awakening. Movements must be extremely gentle and slow; her eyes filled with the pure innocence of first awakening.
[00:16 - 00:23] Shot 3: [Dance and Resonance] (Medium Shot Orbiting Tracking)Action / Physics Directives: Require a natural transition at the cut. The camera pulls back to a medium shot and begins a smooth 360-degree orbiting flight around her. She stands up lightly from the leaf, her toes lightly touching the air, floating as if weightless. As she rises and spins, her light chiffon skirt spreads out like water ripples. Swarms of glowing blue butterflies dance around her in droves, leaving trails of light.Camera Subtext: This is the action climax of the film. Focus on showcasing the environmental interaction and the extreme weightlessness of the flowing garments (cloth simulation must be perfect).
[00:24 - 00:30] Shot 4: [Grand World] (Low-Angle Shot ➔ Rapid High-Altitude Pull-out)Action / Physics Directives: Require a wipe transition (the screen is briefly obscured by a flock of flying butterflies, then revealed). The camera cuts to an extremely low-angle shot. The elf girl flies weightlessly upwards through the forest, heading towards a towering World Tree that reaches into the clouds. Immediately after, the camera rapidly pulls backward and upwards. The field of view instantly becomes infinitely vast: revealing the entire grand magical forest shrouded in morning mist, waterfalls, and glowing plants. The girl becomes a tiny glowing dot in the forest. The frame freezes on this epic, majestic wide shot.Camera Subtext: Conclude with an extreme contrast in perspective (from microscopic to ultimate grand scale), leaving the audience with infinite imagination about this massive magical world. Maximize the epic cinematic feel.
中文对照:【全局设定】(世界观+防崩坏)基础环境与质感:清晨的远古魔法森林,被浓密晨雾笼罩。强调极致真实的物理质感与光学表现:穿透树冠的丁达尔光、空气中漂浮的发光孢子、植物叶片上水珠的光学折射都要高度真实细腻。视觉风格:高魔史诗风,祖母绿与晨金为主色调,景深随切镜动态变化,用逆光和侧逆光(轮廓光)强化梦幻感。镜头语言:蒙太奇剪辑,含微距特写、升降拉远大全景、360 度环绕跟拍。 【角色造型】18 岁精灵少女:空灵透明的初恋脸;冷调极白肌肤带晨露般的通透质感,保留真实微毛孔和鼻梁上淡淡的天然雀斑;圆而无辜的小鹿眼(下眼睑微垂)、柔和眉骨、微翘带肉感的鼻尖、饱满唇珠、脸颊微婴儿肥、标志性的尖长精灵耳;眼神极清澈不谙世事;极光银白及腰长发丝滑带微反光,发间缠绕几缕发着微蓝光的细藤,随风轻摆;薄荷绿雪纺与真实翠绿叶片织成的轻若羽毛的交领露肩长裙,面料极薄带真实纱线质感,逆光下半透明;骨架极娇小、肩窄薄,体态轻盈如羽,散发不食人间烟火的圣洁空灵气质。表演核心:从沉睡中苏醒的无辜感,与魔法环境的空灵互动。【禁止项】不要现代人造物;不要快速晃动镜头(保持平滑柔和);不要五官扭曲和多余手指;强制静音(无字幕无 BGM)。 【时间戳分镜(含明确切镜)】0-7 秒 镜头一【沉睡的微观世界】(微距特写):巨大的挂露发光芭蕉叶上,镜头缓摇,精灵少女蜷睡叶心,银发铺展,胸口随呼吸平稳起伏,发梢停着几只发蓝光的蝴蝶。8-15 秒 镜头二【苏醒与触碰】(中近景→特写,要求切点自然过渡):晨光落在她脸颊,她微蹙眉缓缓睁眼(湖绿色瞳孔),坐起身无辜环顾;镜头随她的视线切到手部特写,纤细手指伸出,一只发光蓝蝶轻落指尖。16-23 秒 镜头三【起舞与共鸣】(中景环绕):镜头拉回中景并围绕她 360 度平滑环飞,她从叶上轻盈起身,足尖点空、悬浮失重,轻纱裙摆如水波散开,成群蓝蝶围绕起舞拖出光轨。24-30 秒 镜头四【宏大世界】(低角度→高空急速拉远,要求蝶群遮罩转场):镜头切至极低角度,精灵少女失重飞升穿过森林,飞向耸入云端的世界树;随后镜头急速后拉上升,视野瞬间无限开阔,晨雾中的宏大魔法森林、瀑布与发光植物尽收眼底,少女化作森林中一个小小光点,画面定格在这个史诗大全景。
官方成片:精灵少女切镜版。
科幻机甲追逐与变形
30-Second One-Shot (Sci-Fi Mecha Chase and Transformation)
[Global Setting]Base Environment & Texture: A cyberpunk-style futuristic cross-sea bridge in 2077. Heavy rain is pouring down, and the road surface is severely waterlogged. Emphasize extreme, realistic physical textures (rain reflections on metal surfaces, water splashed by tires, and the specular refraction of exhaust flames must strictly obey real-world physics).Visual Style: Hardcore sci-fi cinematic feel, high-contrast neon dark tones (cyber-pink and icy blue as primary colors), high-speed shutter (to clearly capture splashing raindrops and sparks).Camera Language: High-speed subjective drone POV, delivering a strong sense of velocity impact and the turbulence of cutting through airflow. One continuous shot.Subject Styling: A silver-and-black concept mecha motorcycle, extremely streamlined with a heavy-industry, heavily armored vibe. Its exhaust pipes are blasting eerie blue plasma flames.Core Performance (Action): Focus on the center-of-gravity shifts during high-speed driving, the physical friction/grip of the tires, and the heavy-industry mechanical aesthetics of precision gears interlocking tightly during mid-air transformation.[Negative Prompts]: No soft-body or mollusk-like twisting/clipping of the mecha structure during movement (must maintain metallic rigidity); no human entities; strictly remove irrelevant subtitles; force mute/no default BGM; avoid copyrighted IP elements like Transformers or Tron.
[Timestamp Storyboard]
[00:00 - 00:08] [Extreme Speed Drive] — Piercing the WindAction / Physics Directives: The camera skims tightly over the waterlogged road, following closely behind and to the side of the mecha motorcycle at extreme speed. The motorcycle races through the heavy rain. The rapidly spinning wide tires kick up massive water curtains several meters high. The blue plasma exhaust flames drag long light trails through the nighttime rain curtain.Camera Subtext: Conveys an extreme sense of speed to the AI, focusing on the physical interaction between the rain, puddle reflections, and engine flames.
[00:09 - 00:16] [Extreme Evasion] — Shift of GravityAction / Physics Directives: The camera instantly pulls up for a high-angle tilt-down shot (spatial perspective switch). A chain-explosion of abandoned vehicle wreckage suddenly appears as an obstacle ahead. The motorcycle leans extremely far over, almost touching the ground. The tires grind against the road, sparking dazzling orange friction sparks, executing a perfect, tight "S" curve to weave through the burning debris.Camera Subtext: Presents the shift in the center of gravity during high-speed maneuvering, showcasing the gravity feedback and spark VFX of extreme tire-to-ground friction.
[00:17 - 00:24] [Mid-Air Reassembly] — Mechanical TransformationAction / Physics Directives: The mecha motorcycle charges off a broken section of the bridge, launching into the air. The camera leaps into the air with it, orbiting the subject. During hang time (slow-mo suspension), the motorcycle's armor rapidly flips, deconstructs, and reassembles. The wheels retract to become thrusters, and the chassis unfolds to extend steel limbs, instantly transforming from a motorcycle form into a heavily armored humanoid combat mecha.Camera Subtext: Showcases pure mechanical heavy-industry aesthetics, emphasizing the precise interlocking and volumetric mass of the parts during assembly, strictly rejecting "noodle-like" soft-body transformation.
[00:25 - 00:30] [Heavy Landing] — Frozen PowerAction / Physics Directives: The camera plummets rapidly to a low-angle extreme close-up (in-your-face shot). The humanoid mecha lands heavily on the highway, dropping to one knee. The massive tonnage impact causes the bridge surface to shatter into a spider-web pattern. The concussive shockwave instantly blows away the surrounding rain curtain (exploding outward in a ring). The mecha slowly raises its head, its tactical eye-lights flashing glowing crimson red, staring directly into the camera as it freezes.Camera Subtext: Creates a visually explosive "Superhero Landing", conveying the mecha's terrifying weight and destructive power to the audience through the shattered road and expelled shockwave.
中文对照:30 秒一镜到底(科幻机甲追逐与变形)。【全局设定】基础环境与质感:2077 年赛博朋克风未来跨海大桥,暴雨倾盆、路面严重积水。强调极致真实物理质感(金属表面的雨水反光、轮胎溅起的水花、尾焰的镜面折射都要严格服从现实物理)。视觉风格:硬核科幻电影感,高对比霓虹暗调(赛博粉与冰蓝为主),高速快门(清晰捕捉飞溅雨滴与火花)。镜头语言:高速主观无人机 POV,强烈的速度冲击感和切风气流的颠簸感,一镜到底。主体造型:银黑配色的概念机甲摩托,极致流线型带重工业重甲气质,排气管喷着诡异的蓝色等离子尾焰。表演核心(动作):高速行驶中的重心变化、轮胎的物理摩擦抓地,以及空中变形时精密齿轮咬合的重工业机械美学。【禁止项】机甲运动中不许软体面条式扭曲穿模(必须保持金属刚性);不要人类实体;严格去除无关字幕;强制静音/无默认 BGM;避开变形金刚、创战纪等版权 IP 元素。 【时间戳分镜】0-8 秒【极速狂飙】,破风前行:镜头贴着积水路面疾掠,紧跟机甲摩托侧后方。摩托在暴雨中飞驰,高速旋转的宽轮胎卷起数米高的水幕,蓝色等离子尾焰在夜雨中拖出长长光轨。9-16 秒【极限规避】,重心偏移:镜头瞬间拉高俯拍(空间视角切换),前方突现连环爆炸的废弃车残骸,摩托极限压弯近乎贴地,轮胎摩擦路面迸出刺目的橙色火花,走出一个漂亮的紧凑 S 弯穿过燃烧残骸。17-24 秒【空中重组】,机械变形:摩托冲出断桥腾空,镜头随之跃起环绕主体;滞空慢动作中,装甲快速翻转、解构、重组,车轮收起变推进器、底盘展开伸出钢铁四肢,瞬间从摩托形态变成重甲人形战斗机甲。25-30 秒【重磅落地】,力量定格:镜头急坠成低角度大特写,人形机甲重重砸在公路上单膝跪地,巨大吨位冲击让桥面碎成蛛网纹,冲击波把周围雨幕环状炸开。机甲缓缓抬头,战术眼灯闪烁猩红光芒,直视镜头定格。
官方成片:机甲追逐与变形。
国风武侠 · 竹林酒馆刺客
[Global Settings] (Worldview & Anti-Distortion)Base Environment & Texture: An open-air wild tavern in a secluded bamboo forest late at night. Emphasize top-tier anime (2D/Cel-shaded) high-quality animation texture. The falling trajectory of bamboo leaves and the ripples in the wine must strictly obey physical laws, possessing the elegant and ethereal flow unique to traditional Chinese aesthetic animation.Visual Style: Top-tier Chinese ancient wuxia anime (Cel-shaded realism / 2.5D rendering), with ink-wash rendered edges. High contrast intertwining of cold and warm lighting (the bleak, cold blue-white moonlight contrasting with the warm yellow candlelight on the tavern table).Camera Language: Employs razor-sharp montage editing and multiple types of seamless physical transitions. Extreme contrast between stillness and explosive motion.Character Styling: A 22-year-old young swordsman, with a handsome, sharply contoured face typical of Chinese anime. Sword-shaped eyebrows and starry eyes, hair tied in a high ponytail with messy, loose strands framing his forehead. Wearing a sleek, white-based, black-edged cross-collar martial arts outfit (Jin-zhuang). A long sword in a black scabbard rests on the table.Core Performance: Focus on the speed of "one-hit kill" in top-tier wuxia, the swordsman's unhurried composure, and the poetic "blank space" (leaving things to the imagination) of Eastern aesthetics.[Negative Prompts]: No greasy 3D feel (strictly maintain pure anime 2D / cel-shaded style). No sluggish, messy combat movements or limb clipping/distortion. No subtitles or extra dialogue. Emphasis: The ink-wash effect must only appear AFTER 25 seconds triggered by the "click" sound, absolutely no premature appearance. The wine bowl must not appear until the full moon has completely faded out.
[Timestamp Storyboard (with explicit cut transitions)]
[00:00 - 00:04] Shot 1: [Drinking in the Still Night] (Match Cut Transition)Action / Physics Directives: The camera is a static high-angle tilt-up, focusing on a massive, bright full moon occupying the dead center of the frame. [Transition Execution]: The full moon in the sky gradually darkens and fades out. A perfectly identical-sized circular wine bowl (viewed from a direct top-down angle of the table) slowly emerges from the moon's position, completely replacing the moon. The wine bowl reflects specks of moonlight and ripples slightly. Requirement: The crossfade between the moon and the wine bowl must be natural; the bowl cannot appear before the moon has faded. After the transition, the camera slowly pulls back and lowers to eye level. The swordsman sits at the tavern table, slowly raising the bowl of wine, preparing to drink.Camera Subtext: An extremely quiet opening. Uses the perfect shape-matching and crossfade between the "bright moon" and the "wine bowl" to create the cold, solitary poetic imagery of wuxia: "The moon reflects in the cup, raising the cup to toast the moon."
[00:05 - 00:10] Shot 2: [Hidden Murderous Intent] (Zoom-in Close-up Transition)Action / Physics Directives: The swordsman's movement pauses slightly, his gaze suddenly turning razor-sharp. [Transition Execution]: The camera pushes forward at extreme speed, infinitely zooming into the swordsman's eye until his pupil occupies the entire screen. The frame seamlessly expands from inside the pupil, transforming into the grand scene reflected within: high above the bamboo forest, an assassin in black clothes, holding a short blade in a reverse grip, dives straight at the camera like an eagle hunting prey.Camera Subtext: Shows the swordsman's extreme alertness. The "Matryoshka-style" eyeball zoom transitions the microscopic eye close-up instantly into the crisis of an impending attack.
[00:11 - 00:18] Shot 3: [Drawing the Sword / Clashing] (Pull-out & Whip Pan Transition)Action / Physics Directives: Connecting to the previous frame, the camera rapidly pulls back out from the assassin reflected in the swordsman's eye, quickly retreating to a full-body shot of the swordsman. The swordsman leaps from his seat and draws his sword. [Transition Execution]: The camera follows the swordsman's sword-swinging arm and whips violently to the right of the frame (Whip Pan), causing severe horizontal motion blur across the entire screen. Within this horizontal blur, the scene instantly shifts to mid-air. The camera stops whipping and freezes: in the exact center of the frame, the swordsman's long sword and the assassin's short blade clash forcefully. A massive orange spark visual effect and a ring-shaped shockwave erupt at the point of weapon intersection.Camera Subtext: Uses rapid pull-outs and horizontal whip pans to connect two different framing shots of combat, giving the audience the visual shock of "speed too fast for the naked eye to catch."
[00:19 - 00:24] Shot 4: [One-Hit Kill] (Wipe/Mask Transition)Action / Physics Directives: The two pass by each other in mid-air. [Transition Execution]: The airflow from the clash sweeps up a storm of bamboo leaves. One giant, emerald-green bamboo leaf flies horizontally across the lens, completely masking the camera and causing a full-screen blackout. Immediately after, from the blackout, the bamboo leaf drifts downward away from the lens, revealing that the frame has already cut to a ground-level shot: in the background, the assassin falls limply to the ground. In the foreground, the swordsman, back facing the camera, slowly sheathes his long sword.Camera Subtext: A classic wuxia "leaving blank space" technique. It avoids directly showing the bloody kill, instead using a bamboo leaf to mask the lens, completing the spatial leap and the pronouncement of life and death with ultimate elegance.
[00:25 - 00:30] Shot 5: [Lingering Echoes] (Ink-Wash Bleed Transition)Action / Physics Directives: The visual cue is the "click" sound of the swordsman fully sheathing his sword. Require an ink-wash transition exactly at the "click" cut (no hard cuts). The entire frame of the bamboo forest and the fallen assassin suddenly acts like a drop of thick ink falling into water—rapidly melting into a black-and-white ink-wash painting, bleeding and diffusing outward. The frame completely freezes and ends in this splashed-ink landscape imagery.Camera Subtext: The sublimation of the artistic conception. The ink-wash transition echoes the supreme aesthetics of traditional Chinese style anime, leaving a long, lingering poetic charm.
中文对照:【全局设定】(世界观+防崩坏)基础环境与质感:深夜幽静竹林中的一座露天野酒馆。强调顶级国漫(2D 赛璐璐)高品质动画质感,竹叶飘落轨迹与酒中涟漪严格服从物理,兼具中式美学动画特有的优雅空灵流动感。视觉风格:顶级中国古风武侠动画(赛璐璐写实/2.5D 渲染),边缘带水墨晕染,冷暖光高对比(清冷月光与酒桌暖黄烛光交织)。镜头语言:凌厉蒙太奇剪辑与多种无缝物理转场,静与爆发的极致对比。角色造型:22 岁年轻剑客,轮廓分明的国漫俊脸,剑眉星目,高马尾配额前碎发,白底黑边的利落交领劲装,黑鞘长剑横在桌上。表演核心:顶级武侠「一击必杀」的速度感、剑客的从容不迫,以及东方美学的诗意留白。【禁止项】不要油腻 3D 感(严格保持纯 2D 赛璐璐);不要拖沓混乱的打斗和肢体穿模;不要字幕和多余台词。强调:水墨效果只能在 25 秒「咔哒」声后出现,绝不许提前;酒碗必须等满月完全淡出后才出现。 【时间戳分镜(含明确转场)】0-4 秒 镜头一【静夜独酌】(相似物叠化转场):镜头静止仰拍,巨大明亮的满月占据画面正中。【转场执行】满月渐暗淡出,一只大小完全相同的圆形酒碗(俯拍视角)从月亮的位置缓缓浮现,完全取代月亮。酒碗映着月光碎点、微微荡漾。要求月亮与酒碗的叠化自然,月亮未淡出前酒碗不能出现。转场后镜头缓缓后拉降至平视:剑客坐在酒馆桌前,缓缓举碗欲饮。镜头潜台词:以「明月映杯、举杯邀月」的形状匹配叠化,立起武侠的清冷孤寂诗意。5-10 秒 镜头二【暗藏杀机】(变焦特写转场):剑客动作微顿,眼神骤然锐利。【转场执行】镜头极速推进,无限放大剑客的眼睛直到瞳孔占满全屏;画面从瞳孔内部无缝展开,化作瞳中倒映的场景,竹林高处,黑衣刺客反握短刃,如猎鹰扑食般直扑镜头。11-18 秒 镜头三【拔剑相击】(拉出+甩镜转场):镜头从瞳中刺客急速拉出退至剑客全身,剑客离座拔剑。【转场执行】镜头随挥剑手臂向右猛甩(甩镜),全屏剧烈横向模糊;模糊中场景瞬移至半空,镜头刹停定格:画面正中,长剑与短刃重重相撞,交击点爆出巨大橙色火花与环状冲击波。19-24 秒 镜头四【一击必杀】(遮罩转场):二人半空错身而过。【转场执行】激斗气流卷起竹叶风暴,一片巨大的翠绿竹叶横飞掠过镜头完全遮挡画面全黑;随后竹叶飘落离开镜头,画面已切至地面视角:背景处刺客软软倒地,前景剑客背对镜头缓缓收剑入鞘。镜头潜台词:经典武侠留白,不直接展示血腥,用一片竹叶完成空间跳跃与生死判定。25-30 秒 镜头五【余韵悠长】(水墨晕染转场):以剑客收剑入鞘的「咔哒」声为视觉提示,要求恰在此处做水墨转场(不许硬切),竹林与倒地刺客的整个画面如浓墨滴水,迅速化开成黑白水墨画向外晕染,画面在泼墨山水意象中彻底定格收尾。
官方成片:武侠竹林一战。
冰川穿越深海幻境
30-Second One-Shot (Glacier Crossing to Deep Sea Illusion)
[Global Setting]Base Environment & Texture: Polar glaciers and the deep ocean. Emphasize "National Geographic" level extreme physical realism. The refraction of glacial meltwater and underwater light dappling (Tyndall effect / volumetric lighting) must strictly obey real-world optical physics.Visual Style: Epic and grandiose narrative landscape, 8K ultra-clear, cool color palette (transitioning from icy blue to deep abyssal blue).Camera Language: A globally smooth, slow, and highly epic "One-Shot" continuous forward drone push. Absolutely no hard cuts or editing transitions.Subject & Core Performance: (No specific character subjects). The core performance lies in the majestic display of the natural environment, the time-lapse changes of sunlight and shadow, and the seamless spatial transition from above water to underwater.[Negative Prompts]: No humans or artificial vehicles (except for the shipwreck). No flickering, stuttering, or visual tearing during the water surface transition. Strictly remove subtitles and default BGM.
[Timestamp Storyboard]
[00:00 - 00:09] [Majestic Polar] — Glacier CruiseAction / Physics Directives: The camera flies forward slowly and smoothly through a massive blue glacial rift valley. Sunlight slants down from the top of the ice valley, refracting blinding specular highlights off the sharp ice surfaces. Occasionally, tiny chunks of ice peel off the cliffs and plummet into the bottomless sea below.Camera Subtext: Creates an extremely pure, grand, and awe-inspiring atmosphere of polar nature.
[00:10 - 00:15] [Passage of Time] — Shifting Light and ShadowsAction / Physics Directives: The camera's forward pushing speed remains constant, but the footage exhibits a "time-lapse photography" effect. Within 5 seconds, the sky rapidly shifts from broad daylight to a brilliant purple sunset, then quickly darkens into a star-studded polar night, with the Aurora Borealis dancing across the sky.Camera Subtext: Uses the rapid cycle of light and shadow to show the passage of time over the glacier, building visual tension for the upcoming transition.
[00:16 - 00:21] [Seamless Crossing] — Diving into Deep BlueAction / Physics Directives: The camera angle suddenly tilts downward and plunges straight into the pitch-black, freezing seawater below (triggering a water-penetration transition). The screen is instantly filled with countless white bubbles rolling upward. Then, the view suddenly clears, entering the deep blue underwater world. Sunlight pierces through the surface, forming sacred "God rays" (Tyndall effect / volumetric lighting).Camera Subtext: This is the most stunning physical transition of the entire film; it must ensure the physical sensation of resistance and the realistic fluid dynamics feedback of the bubbles the moment the camera enters the water.
[00:22 - 00:30] [Underwater Ruins] — The Slumbering Giant ShipAction / Physics Directives: The camera continues to glide forward through the abyssal depths. At the end of the Tyndall light beams, the massive remains of a medieval wooden shipwreck lie quietly on the sandy ocean floor. A few glowing, translucent bioluminescent jellyfish slowly swim past the lens. The camera comes to a halt, focusing on the shipwreck's barnacle-covered giant figurehead, and the frame freezes amidst the deep, dark blue water ripples.Camera Subtext: Concludes with extreme tranquility and mystery. The details of the shipwreck and the bioluminescent jellyfish create a strong chiaroscuro (light and dark contrast).
中文对照:30 秒一镜到底(冰川穿越至深海幻境)。【全局设定】基础环境与质感:极地冰川与深海。强调「国家地理」级的极致物理真实,冰川融水的折射与水下光斑(丁达尔/体积光)严格服从现实光学。视觉风格:史诗宏大叙事风光片,8K 超清,冷色调(从冰蓝过渡到深渊蓝)。镜头语言:全程平滑缓慢、极具史诗感的无人机一镜到底持续前推,绝不硬切、绝无剪辑转场。主体与表演核心:(无特定角色主体)核心是自然环境的雄浑展示、光影的时光流逝变化,以及从水上到水下的无缝空间穿越。【禁止项】不要人类和人造载具(沉船除外);水面穿越瞬间不要闪烁卡顿和画面撕裂;严格去除字幕和默认 BGM。 【时间戳分镜】0-9 秒【极地雄浑】,冰川巡航:镜头平滑缓慢地向前飞过巨大的蓝色冰川裂谷,阳光从冰谷顶端斜射,冰面反射出刺目高光,偶有小冰块从崖壁剥落坠入深不见底的海面。10-15 秒【时光流转】,光影变幻:镜头前推速度不变,画面呈现「延时摄影」效果,5 秒内天空从白昼极速转为绚紫落日,再迅速暗成星斗满天的极夜,极光横贯天际。16-21 秒【无缝穿越】,扎入深蓝:镜头骤然下倾,一头扎进漆黑冰冷的海水(触发穿水转场),全屏瞬间被无数上涌的白色气泡填满;随后视野豁然开朗,进入深蓝水下世界,阳光穿透海面形成圣洁的「上帝光」。22-30 秒【水下遗迹】,沉睡的巨轮:镜头继续在深渊中前滑,丁达尔光束尽头,一艘中世纪木制沉船的巨大残骸静卧海底沙床,几只发光半透明的水母缓缓游过镜头。镜头停驻,聚焦船首长满藤壶的巨大船首像,画面在幽深的深蓝水波中定格。
官方成片:冰川到深海一镜到底。
摄政时期舞会
[Global Scene Setting]Base Environment: A grand manor ballroom in Regency-era England (early 19th century). A massive crystal chandelier hangs from the ceiling, with hundreds of candles emitting a warm yellow glow. Emphasize extreme physical realism and a luxurious ballroom atmosphere.Visual Style: Classic BBC Period Drama style. Primary color palette of warm gold, ivory white, and deep black. High contrast candle-lit illumination, film grain texture.Camera Language: Extremely dramatic cinematic language, including Rack Focus, foreground physical mask seamless transitions, and highly immersive Slow Motion dolly-in shots.Audio & BGM Design: The overall tone is a classical orchestral waltz. The musical emotion must match the camera, progressing from "elegant and restrained social ambient sound" to "extreme fervor after the world is stripped away", synchronized with character movements and SFX.
Character Styling (Female Lead):[Age / Ethnicity] 20-year-old British woman, with classic British bone structure.[Skin Tone / Texture] Cool-toned fair skin, revealing a delicate and moist texture, retaining realistic fine pores and natural skin texture.[Facial Features] Dark brown almond-shaped eyes, high straight nose bridge, a hint of stubbornness at the corners of her mouth, and a soft jawline.[Eyes / Soul] Her gaze shifts from defensive scrutiny to deep immersion. Deep, focused eyes revealing suppressed attachment.[Hair Style / Color] Dark brown long hair pinned into a classical chignon at the back of her head, with a few lively curled tendrils falling around her cheeks, swaying slightly with her movements.[Clothing / Texture] Wearing an Empire waist ivory white silk gown. The fabric reflects a faint candle glow. She wears white silk opera gloves reaching past her elbows.[Physique / Emotion / Aura] Slender and upright posture, graceful demeanor. Exudes an independent, intelligent, and romantic aura, heavily suppressing her emotions beneath extreme restraint.
Character Styling (Male Lead):[Age / Ethnicity] 26-year-old British man, with an angular, cold, aristocratic face.[Skin Tone / Texture] Cool-toned fair skin, revealing a slightly pale, high-intelligence texture, retaining realistic fine pores and natural skin texture.[Facial Features] Deep-set eyes, slightly furrowed brows, a high, restrained nose bridge, and a jawline sharply defined as if carved by a knife.[Eyes / Soul] Deep, cold, focused, and highly aggressive gaze, as if seeing right through her, revealing inevitable fervor and deep affection.[Hair Style / Color] Pure black short hair, meticulously combed with sharp edges.[Clothing / Texture] An extremely well-fitted, deep black, deep V-neck tailcoat made of stiff, premium heavy wool, paired with a crisp pure white shirt and a highly intricately tied pure white cravat.[Physique / Emotion / Aura] Broad and upright skeletal frame, wide shoulders. Exudes an arrogant and restrained, yet incredibly tense masculine oppressive presence.
Core Performance: Showcasing "hidden emotional undercurrents beneath extreme restraint." Focus on the male lead's slight Adam's apple movement and locked gaze, and the female lead's chest heaving due to rapid breathing.[Negative Prompts]: No modern hairstyles or zippered clothing, no modern lighting, no harsh jump cuts, no generated spoken dialogue, no fabrics floating out of thin air, no identical clones in the background.
[Timestamp Storyboard]
[00:00 - 00:06] Shot 1: [Rack Focus & Crane Down]Emotional Subtext: A precise locking of gazes cutting through the noisy crowd, the descent of fate.Audio Design: BGM is a melodious classical waltz with spatial reverb, mixed with the noisy ambient sound of guests softly chatting. When the focus instantly sharpens, the ambient sound suddenly drops, and the main melody of the waltz becomes sharp and clear.Action / Physics Directives: The camera initially starts at a high angle, looking diagonally downward (High Angle Shot), with the frame out of focus. A small section in the right foreground is a massive crystal chandelier hanging from the ceiling, its warm-glowing crystal pendants occupying the upper right corner of the screen. The rest of the frame shows blurred figures of the male and female leads from a top-down perspective. [Transition Execution]: The camera executes an extreme Rack Focus. The foreground crystal chandelier instantly blurs out of focus, while the male and female leads in the lower background instantly become crystal clear. Immediately, the camera executes a Crane Down movement, smoothly and rapidly dropping from high above to an eye-level perspective, displaying the leads from the knees up (Medium Shot). At this moment, the male lead stands before the female lead, elegantly extending his right hand. The female lead raises her eyes, her gaze carrying scrutiny.
[00:07 - 00:13] Shot 2: [Foreground Mask Transition]Emotional Subtext: Fingertips touch, joining the game together.Audio Design: The exact moment the female lead places her hand in his, the music hits a heavy downbeat. Accompanied by the heavy silk friction SFX of a white skirt swooshing past the lens, the BGM tempo suddenly accelerates, and the orchestral section fully opens up.Action / Physics Directives: Continuing the medium shot from the previous frame, the female lead gently places her hand in the male lead's palm. [Transition Execution]: At this moment, a pair of joyfully dancing guests (a gentleman and a lady) holding each other tightly happen to spin rapidly past the lens (the camera is in a medium shot, and this dancing couple dances extremely close to the lens, showing only their waists and lower bodies). In the instant of their spin, the lady's white skirt sweeps horizontally across the lens, completely masking the frame and turning the entire screen pure white for a split second. When the white skirt leaves the camera's field of view, the camera has seamlessly transitioned to a Full Shot of the center of the dance floor, where the male and female leads are gracefully dancing together. Requirement: The other dancing guests in the background must be widely spaced (dancing in pairs), maintaining a spacious feel. The surrounding people's clothing colors (e.g., interwoven dark red, dark green, deep blue), hairstyles, and facial features must all be distinct. Their dancing synchronization should be inconsistent, creating a relaxed, realistic atmosphere and presenting genuine crowd diversity.
[00:14 - 00:20] Shot 3: [Slow Motion Dolly-In & Bokeh Defocus]Emotional Subtext: Emotions lose control, time stands still, and the world is reduced to just each other.Audio Design: As the camera pushes in, the gorgeous waltz and the noisy dance floor are instantly stripped away, creating a strong "tinnitus effect" (underwater muffled sound effect). The BGM strips away all symphonic instruments, leaving only a highly tense solo cello and a "thump, thump" bass drum beat resembling a violent heartbeat.Action / Physics Directives: [Transition Execution]: Connecting to the full shot of the previous frame, the camera begins a smooth, constant-speed forward push (Dolly In) toward the male and female leads, while the footage enters extreme Slow Motion. As the camera continuously draws closer, the male and female leads are magnified from a full shot to an extreme close-up two-shot. During this process, the depth of field rapidly becomes extremely shallow (Shallow Depth of Field). The background ballroom and other guests completely blur out, transforming into massive, flickering, warm yellow Bokeh light spots. In the slow-motion close-up, the male lead's eyes become deeply affectionate, staring fixedly at the female lead, his Adam's apple rolling. The female lead's chest heaves with rapid breathing, appearing somewhat shy. Freeze frame. As the heartbeat drum reaches its climax, the screen cuts to black, and the music abruptly stops.
中文对照:【全局场景设定】基础环境:19 世纪初英国摄政时期,一座宏伟的庄园舞会大厅。巨大的水晶吊灯垂在天顶,数百支蜡烛发出暖黄光。强调极致物理真实与奢华舞会氛围。视觉风格:经典 BBC 时代剧风,暖金、象牙白与深黑为主色调,高对比烛光照明,胶片颗粒质感。镜头语言:极具戏剧性的电影语言,包含变焦点(Rack Focus)、前景实体遮罩无缝转场、沉浸式慢动作推近。声音与 BGM 设计:整体是古典管弦圆舞曲,音乐情绪要随镜头推进,从「优雅克制的社交环境声」到「世界被剥离后的极致炽烈」,与人物动作和音效同步。 角色造型(女主):20 岁英国女性,经典英式骨相;冷调白皙肌肤细腻温润,保留真实毛孔与皮肤纹理;深棕杏眼、高挺鼻梁、嘴角带一丝倔强、柔和下颌线;目光从戒备审视转为深深沉浸,深邃专注,透出被压抑的依恋;深棕长发在脑后挽成古典发髻,几缕灵动卷发垂颊随动作轻摆;帝政腰线的象牙白丝质长裙,面料映着微微烛光,戴过肘白丝歌剧手套;身形纤细挺拔,仪态优雅,散发独立聪慧的浪漫气质,情绪被极致克制地压着。 角色造型(男主):26 岁英国男性,棱角分明的清冷贵族脸;冷调白皙微苍白的高智感肌理,保留真实毛孔;深邃眼窝、眉头微蹙、高挺内敛的鼻梁、下颌线如刀刻;目光深沉冷峻、专注而极具侵略性,仿佛把她看穿,透出不可回避的炽烈与深情;纯黑短发梳理一丝不苟、边缘锋利;极合身的深黑深 V 燕尾服,面料是挺括的高级厚羊毛,配挺括纯白衬衫和系法繁复的纯白领巾;骨架宽阔挺拔、肩线宽,散发傲慢克制却张力十足的雄性压迫感。 表演核心:展现「极致克制下的暗涌」。着重男主喉结的轻微滚动与锁定的目光、女主因呼吸急促而起伏的胸口。【禁止项】不要现代发型和拉链服装、不要现代灯光、不要生硬跳切、不要生成台词、不要凭空飘动的布料、背景不要出现一模一样的克隆人。 【时间戳分镜】0-6 秒 镜头一【变焦点+降落】:镜头从高角度俯拍开场,画面失焦,右前景是暖光水晶吊灯一角,其余是俯视下模糊的男女主身影。【转场执行】镜头执行极速变焦点:前景吊灯瞬间虚化,下方男女主瞬间清晰锐利;紧接着镜头平滑快速下降至平视(Crane Down),中景展示两人,男主立于女主面前,优雅伸出右手,女主抬眼,目光带着审视。声音设计:BGM 是带空间混响的悠扬圆舞曲混着宾客低声交谈,焦点锁定的瞬间环境声骤降、主旋律清晰锐利。7-13 秒 镜头二【前景遮罩转场】:女主把手轻轻放进他掌心的瞬间,音乐落重拍。【转场执行】一对跳得正欢的宾客恰好快速旋转掠过镜头(极贴近镜头,只见腰腹与下身),女宾的白裙横扫过镜头把画面完全遮成纯白一瞬;白裙离开视野时,镜头已无缝切至舞池中央全景,男女主正优雅共舞。要求背景其他宾客间距宽松(成对起舞),衣色(深红、墨绿、深蓝交织)、发型、五官各不相同,舞步不必同步,呈现真实人群的多样性。14-20 秒 镜头三【慢动作推近+焦外虚化】:情绪失控、时间凝滞、世界只剩彼此。【转场执行】接上镜全景,镜头匀速前推(Dolly In),画面进入极致慢动作;随着不断逼近,两人从全景放大到极特写双人镜头,景深急速变浅,背景舞厅与宾客彻底虚化成巨大闪烁的暖黄光斑。慢动作特写中男主深情凝视、喉结滚动,女主呼吸急促、微微羞怯。定格。心跳鼓点到达顶点时画面切黑、音乐戛然而止。声音设计:镜头推进时华丽圆舞曲和嘈杂舞池被瞬间抽走(耳鸣式水下闷响效果),BGM 只剩一把极具张力的大提琴独奏和「咚、咚」如剧烈心跳的低音鼓点。
官方成片:摄政舞会。
写法

超长视频:180 秒一次直出的写法

180 秒的写法和基础公式同构,只是前面多一层全局参数:因为片子太长,画幅和总时长选完参数后,最好在提示词开头再强调一遍。骨架是「全局参数 + 素材说明 + 一句话概括 + 具体情节 + 全局补充」。

官方案例 · 3 分钟橘猫打工 vlog
Full length 3 minutes, @image1 cat working vlog, video aspect ratio 16:9, a day as a waiter with an orange cat as the main character, it goes from waking up brushing teeth and washing face, riding an electric scooter to a street-side small shop, arriving at work early morning wiping tables and clocking in, welcoming guests guiding them to seats and handing menus, scene reference @image6-10, overall restaurant reference @image11, he carefully delivers meals to the tables of customers @image2345, taking a nap and secretly nibbling on dried fish during lunch break, collecting dried fish tips and clearing tables in the evening, finally closing up and curling up to rest by the window, riding the electric scooter, returning to the room reference @image12, washing up, eating, sleeping. No subtitles throughout, no background music.
中文对照:全程 3 分钟,@图1 猫咪打工 vlog,视频比例 16:9。以一只橘猫为主角的服务员的一天:从起床刷牙洗脸、骑电动车到街边小店,清早到店擦桌打卡,迎接客人引座递菜单(场景参考 @图6-10,餐厅整体参考 @图11),他小心翼翼把餐送到顾客桌上 @图2345,午休打盹偷啃小鱼干,傍晚收小鱼干小费、收拾桌子,最后关店蜷在窗边休息,骑电动车回家(房间参考 @图12),洗漱、吃饭、睡觉。全程不要字幕,不要背景音乐。
参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图
12 张参考图:主角橘猫、场景与顾客
官方成片:3 分钟猫咪打工 vlog,叙事完整、动作连贯。
看第二条案例:1 分钟雨天咖啡馆空镜
官方案例 · 雨天咖啡馆空镜
Video duration 1 minute, aspect ratio 16:9, cinematic realistic picture quality, retro healing style. On a rainy afternoon, a fixed camera records the steaming coffee on the table of a retro cafe, and the beautiful rain scene seen through the glass window.
0s - 20s (Quiet opening):The camera position is fixed in front of a wooden coffee table. On the table is a cup of freshly brewed hot coffee, slowly emitting white hot steam upwards. The background is a huge checkered glass window, outside the window it is drizzling, raindrops hit the glass, gather into water drops and slowly slide down. Accompanied by the pattering sound of rain. The camera should not shake, no characters should enter the frame, hard cuts are prohibited.
20s - 40s (Detail push-in):The camera pushes in extremely slowly towards the coffee cup in front. The hot steam continues to swirl. A gust of wind blows, a golden fallen leaf drifts by outside the window and sticks to the outside of the glass window. The focus of the picture is still on the edge of the coffee cup and the hot steam. No complex physical collisions, objects are prohibited from deforming, text is prohibited from appearing.
40s - 60s (Focus shift and fade out):The camera maintains a slight breathing feel, the focus of the picture slowly and smoothly shifts from the coffee cup in the foreground to the rain scene outside the window in the background. The street lights outside the window are blurred by the rain and bloom into warm light spots. The picture gradually darkens and ends along with the sound of rain. The picture should not flicker, no abrupt transitions.
Picture and light/shadow:Throughout the process, maintain a high-level color contrast between the indoor warm yellow light and the outdoor cold rainy night. The camera mainly uses extremely slow push-and-pull and fixed positions, pursuing an extreme sense of stability and atmosphere.
Sound and restrictions:The whole process only needs real natural white noise (the sound of rain hitting the glass, subtle ambient sounds). No background music globally (no bgm), no subtitles globally.
中文对照:视频时长 1 分钟,画幅比例 16:9,电影级写实画质,复古治愈风。雨天午后,固定镜头记录复古咖啡馆桌上冒着热气的咖啡,和透过玻璃窗看到的雨景。 0-20 秒(安静开场):机位固定在木质咖啡桌前。桌上一杯刚冲好的热咖啡,白色热气袅袅上升。背景是巨大的格子玻璃窗,窗外细雨绵绵,雨点打在玻璃上汇成水珠缓缓滑落。伴随淅沥雨声。镜头不要晃动,不要人物入画,禁止硬切。 20-40 秒(细节推近):镜头朝前景的咖啡杯极缓慢推近,热气持续盘旋。一阵风吹过,一片金色落叶飘过窗外贴在玻璃上。画面焦点仍在杯沿与热气上。不要复杂物理碰撞,禁止物体变形,禁止出现文字。 40-60 秒(焦点转移与收尾):镜头保持轻微呼吸感,画面焦点从前景咖啡杯缓慢平滑移到背景窗外雨景。窗外路灯被雨水晕成温暖光斑。画面随雨声渐暗结束。画面不要闪烁,不要突兀转场。 画面与光影:全程保持室内暖黄光与室外冷雨夜的高级色彩对比,镜头以极缓推拉和固定机位为主,追求极致稳定感和氛围感。声音与限制:全程只要真实自然白噪音(雨打玻璃声、细微环境声),全局不要背景音乐(no bgm),全局不要字幕。
官方成片:1 分钟复古咖啡馆空镜。
还有一条 180 秒的校园怪谈长片案例(12 个角色 + 全程时间戳),信息量大,放在下文「亮点演示」的 180 秒一节里完整收录。
写法

延长写法与 12 种转场大全

界面上的延长操作在前面讲过了,这一节解决「延长段的提示词怎么写」。手册分两层:先用常规词把意图说清,再按想要的效果加约束。

层写法
常规词必备词汇:向前延长(forward extension)、向后延长(backward extension)、续写(continuation)等;照常支持追加 @图/@视频/@音频。
效果词 · 顺滑延长延长内容和前片自然连贯、动作延续时,用强制约束钉死:要求自然延长、动作衔接流畅自然、禁止硬切镜、禁止物体凭空出现。
效果词 · 切镜/转场延长延长内容要切场景、换镜头时,套转场公式(见下),也可以列几种转场让模型自己按片子风格挑。
切镜/转场公式
Cut shot setting = [transition type guide] + [basic constraint requirements] + [cut shot logic requirements]

Please choose the most suitable one according to the style of this film from the transition methods such as [natural shot switching/mask transition/ink transition/similar object transition].
中文对照:切镜设定 = 转场类型引导(在切点用什么转场,禁止硬切、禁止凭空出现)+ 基础约束(镜头呼吸感与时长协调,镜头 A 演什么、自然过渡到镜头 B 演什么)+ 切镜逻辑要求(再强调一遍自然切换,可顺带写景别变化,如特写变中景)。也可以直接给几种候选让模型自己挑:请从「自然镜头切换/遮罩转场/水墨转场/相似物转场」等转场方式中,按这条片子的风格选最合适的一种。

接下来是这份手册最值得收藏的一张表:12 种转场,每种都给了「模型自由发挥版」和「具体案例版」两套写法,案例统一用孟母三迁的故事写,前后对照非常好抄。基础五种管常规叙事,创意七种管高光时刻:

〔基础〕自然切换

类似硬切但强调「自然镜头」:画面 A 直接变画面 B,镜头要有呼吸感、时长协调,留足切换时间。适合绝大多数常规叙事,干净利落。

自由发挥版
Need to use "natural lens switching" transition at the cutting point (prohibit rigid cutting, prohibit objects from appearing out of thin air).
Maintain the coordination between the breathing sensation and duration of the lens, leave enough switching time, let picture A transition naturally to picture B, and ensure the coherence of the narrative.
It is required to switch scenes naturally.
中文对照:在切点使用「自然镜头切换」转场(禁止硬切、禁止物体凭空出现)。保持镜头呼吸感与时长的协调,留足切换时间,让画面 A 自然过渡到画面 B,保证叙事连贯。要求场景切换自然。
具体案例版 · 孟母三迁
Need to use "natural lens switching" transition at the cutting point (prohibit rigid cutting, prohibit objects from appearing out of thin air).
Maintain the coordination between the breathing and duration of the camera. Shot A shows Mencius kneeling and bowing beside the cemetery, and then naturally transitions to Shot B, where Mencius' mother stands at the door, frowning at him.
It is required to switch scenes naturally.
中文对照:在切点使用「自然镜头切换」转场(禁止硬切、禁止物体凭空出现)。保持镜头呼吸感与时长协调:镜头 A 是孟子在墓地旁跪拜,随后自然过渡到镜头 B,孟母站在门口皱眉看他。要求场景切换自然。
〔基础〕淡入淡出

画面从全黑(或全白)慢慢浮现,或渐暗到全黑。常用于开头结尾,或中段表示时间流逝、一段剧情彻底结束。

自由发挥版
It is necessary to use "fade in and fade out" transition at the cutting point (forbidding rigid cutting and forbidding objects from appearing out of thin air).
Picture A gradually darkens until it is completely black (or completely white), and then picture B slowly emerges from the darkness (or pure white) and gradually becomes brighter.
It is required to switch scenes naturally.
中文对照:在切点使用「淡入淡出」转场(禁止硬切、禁止物体凭空出现)。画面 A 逐渐变暗直至全黑(或全白),画面 B 再从黑暗(或纯白)中缓缓浮现、逐渐变亮。要求场景切换自然。
具体案例版 · 孟母三迁
It is necessary to use the "fade in and fade out" transition at the cutting point (prohibiting rigid cutting and objects appearing out of thin air).
In shot A, Meng's mother sighed as she closed the door, and the image gradually darkened until it was completely black. Then the image slowly lit up from the darkness, revealing shot B: the noisy market in the early morning, Meng's mother and her family moved to a new home.
It is required to switch scenes naturally.
中文对照:在切点使用「淡入淡出」转场(禁止硬切、禁止物体凭空出现)。镜头 A 里孟母叹气关门,画面渐暗至全黑;随后画面从黑暗中缓缓亮起,露出镜头 B:清晨喧闹的集市,孟母一家搬到了新居。要求场景切换自然。
〔基础〕叠化

画面 A 慢慢变透明的同时画面 B 浮现,两个画面短暂重叠。适合回忆、梦境、时间缓慢流逝这类抒情转场。

自由发挥版
Need to use "stacking" transition at the cutting point (prohibit rigid cutting, prohibit objects from appearing out of thin air).
As picture A slowly becomes transparent, picture B gradually appears, producing a soft overlapping and blending effect of 1-2 seconds between the two pictures.
It is required to switch scenes naturally.
中文对照:在切点使用「叠化」转场(禁止硬切、禁止物体凭空出现)。画面 A 缓缓变透明的同时画面 B 逐渐显现,两个画面之间产生 1-2 秒柔和的重叠交融。要求场景切换自然。
具体案例版 · 孟母三迁
Need to use "stacking" transition at the cutting point (prohibit rigid cutting, prohibit objects from appearing out of thin air).
Shot A shows Mengzi's mischievous appearance as a butcher slaughtering pigs in the market, and the image slowly becomes transparent. At the same time, in Shot B, the image of Mengzi's mother weaving cloth with a worried face under the lamp at night gradually emerges, and the two images overlap and blend for 2 seconds.
It is required to switch scenes naturally.
中文对照:在切点使用「叠化」转场(禁止硬切、禁止物体凭空出现)。镜头 A 是孟子在集市学屠夫杀猪的顽皮模样,画面缓缓变透明;同时镜头 B 里孟母深夜灯下愁容满面织布的画面逐渐浮现,两画面重叠交融 2 秒。要求场景切换自然。
〔基础〕白闪 / 黑闪

画面猛地闪白或闪黑一瞬再切下一个画面。视觉冲击强,常配重音(鼓点、枪声、快门声),多用于卡点视频和紧张气氛。

自由发挥版
It is necessary to use "white flash/black flash" transition at the cutting point (forbidding rigid cutting and forbidding objects from appearing out of thin air).
The screen instantly bursts into a strong white light (or instantly plunges into complete darkness) and flashes by, followed by a rapid transition to the next scene, with a strong visual impact.
It is required to switch scenes naturally.
中文对照:在切点使用「白闪/黑闪」转场(禁止硬切、禁止物体凭空出现)。画面瞬间爆出强烈白光(或瞬间陷入全黑)一闪而过,紧接着快速切入下一场景,视觉冲击强烈。要求场景切换自然。
具体案例版 · 孟母三迁
It is necessary to use "white flash/black flash" transition at the cutting position (forbidding rigid cutting and forbidding objects from appearing out of thin air).
Meng Mu saw Mencius learn from the vendors and shout loudly, angrily smashing the woven shuttle in her hand, and the scene instantly erupted with a strong white flash. Then she quickly cut to the next scene: a cart full of luggage speeding on a bumpy muddy road (showing a decisive second move).
It is required to switch scenes naturally.
中文对照:在切点使用「白闪/黑闪」转场(禁止硬切、禁止物体凭空出现)。孟母看见孟子学商贩大声吆喝,怒而摔碎手中织梭,画面瞬间爆出强烈白闪;随后快切下一场景:满载行李的板车在颠簸泥路上疾行(表现果断的第二次搬家)。要求场景切换自然。
〔基础〕擦除

画面 B 像推拉门一样从某个方向把画面 A「推」出去。偏复古(星球大战式),也适合表现明显的地点转移。

自由发挥版
It is necessary to use the "erase" transition at the cutting point (prohibiting rigid cutting, prohibiting objects from appearing out of thin air).
Picture B smoothly advances from the left (or right/up/down) like a sliding door, completely covering/erasing picture A.
It is required to switch scenes naturally.
中文对照:在切点使用「擦除」转场(禁止硬切、禁止物体凭空出现)。画面 B 像推拉门一样从左(或右/上/下)平滑推进,完全覆盖/擦掉画面 A。要求场景切换自然。
具体案例版 · 孟母三迁
It is necessary to use "erase" transition at the cutting point (prohibiting rigid cutting and objects appearing out of thin air).
Picture A is a noisy market. The mirror is like a sliding door, smoothly advancing from right to left, erasing the market picture and covering it with Picture B: the solemn and quiet gate of the academy.
It is required to switch scenes naturally.
中文对照:在切点使用「擦除」转场(禁止硬切、禁止物体凭空出现)。画面 A 是喧闹集市,镜头像推拉门一样从右向左平滑推进,擦掉集市画面,覆盖上画面 B:庄重安静的学宫大门。要求场景切换自然。
〔创意〕遮罩转场

镜头前进被物体(墙、后背、手掌、柱子)完全挡住变黑,下一镜从黑暗里拉出来已是另一个场景。适合空间穿越。

自由发挥版
It is necessary to use "mask transition" transition at the cutting position (prohibiting rigid cutting and objects appearing out of thin air).
The camera pushes forward and is completely blocked by an object in the picture, such as a big tree or a person's back, causing the full screen to turn black. Then the camera passes through the darkness and directly pulls out to reveal the next brand new scene.
It is required to switch scenes naturally.
中文对照:在切点使用「遮罩转场」(禁止硬切、禁止物体凭空出现)。镜头前推,被画面中的物体(如大树、人物后背)完全遮挡致全屏变黑;随后镜头穿过黑暗,直接拉出下一个全新场景。要求场景切换自然。
具体案例版 · 孟母三迁
It is necessary to use the "mask transition" transition at the cutting position (prohibiting rigid cutting and objects appearing out of thin air).
Camera A follows Meng Mu's moving cart forward, and the cart enters a dark narrow alley. The wall completely blocks the lens, causing the full screen to turn black. Then the camera pulls through the darkness and directly reveals the next scene: the spacious and bright Xuegong Academy.
It is required to switch scenes naturally.
中文对照:在切点使用「遮罩转场」(禁止硬切、禁止物体凭空出现)。镜头 A 跟着孟母搬家的板车前行,板车驶入昏暗窄巷,墙壁完全挡住镜头致全屏变黑;随后镜头穿过黑暗直接拉出下一场景:宽敞明亮的学宫。要求场景切换自然。
〔创意〕相似物转场

上一镜的最后一帧和下一镜的第一帧在形状、轮廓或颜色上高度相似(满月→圆形咖啡拉花)。蒙太奇经典手法。

自由发挥版
It is necessary to use "similar object transition" at the cutting point (prohibiting rigid cutting and objects appearing out of thin air).
Utilize the similarity of shape or color for seamless transition. For example, in the last frame of the previous shot, there is a circular moon in the sky. As the shot advances, the moon seamlessly transforms into a circular coffee latte flower in the cup in the next shot.
It is required to switch scenes naturally.
中文对照:在切点使用「相似物转场」(禁止硬切、禁止物体凭空出现)。利用形状或颜色的相似做无缝切换:上一镜最后一帧是空中的圆月,镜头推进时,圆月无缝变成下一镜杯中的圆形咖啡拉花。要求场景切换自然。
具体案例版 · 孟母三迁
It is necessary to use "similar object transition" at the cutting point (prohibiting rigid cutting and objects appearing out of thin air).
Seamless transition using shape similarity. Shot A is a circular copper coin held by a vendor in the market, with the lens advancing towards the extreme of the coin. The coin seamlessly transforms into the end of the same circular ruler held by Teacher Xuegong in the next shot.
It is required to switch scenes naturally.
中文对照:在切点使用「相似物转场」(禁止硬切、禁止物体凭空出现)。利用形状相似做无缝切换:镜头 A 是集市商贩手中的圆形铜钱,镜头向铜钱极致推进,铜钱无缝变成下一镜学宫先生手中同为圆形的戒尺端头。要求场景切换自然。
〔创意〕动作转场

镜头在场景 A 猛甩(左右或上下)拉出动态模糊,场景 B 以同方向同力度甩入定住。分相似动作(柔)与冲突动作(刚),适合动作片、跑酷、旅行 vlog 的「瞬移感」。

自由发挥版
It is necessary to use "motion transition" at the cutting point (forbidding rigid cutting and objects appearing out of thin air).
Utilize the rapid motion of the lens for connection. In scene A, the lens vigorously swings to the left (or right), creating a strong motion blur in the picture; then in scene B, following the same swinging direction and force, cut in from the blur and freeze the frame.
It is required to switch scenes naturally.
中文对照:在切点使用「动作转场」(禁止硬切、禁止物体凭空出现)。利用镜头的快速运动衔接:场景 A 里镜头向左(或右)猛甩,画面产生强烈动态模糊;场景 B 顺着同一甩动方向和力度,从模糊中切入并定格。要求场景切换自然。
具体案例版 · 孟母三迁
It is necessary to use "motion transition" at the cutting point (forbidding rigid cutting and objects appearing out of thin air).
In shot A, Meng's mother suddenly pulled Mengzi's hand, and the camera followed the action and swung violently to the right, creating a blur. Then scene B followed the same swinging direction and cut in from the blur, freezing the frame where Meng's mother pulled Mengzi into the gate of the academy.
It is required to switch scenes naturally.
中文对照:在切点使用「动作转场」(禁止硬切、禁止物体凭空出现)。镜头 A 里孟母猛地拉起孟子的手,镜头随动作向右剧烈甩动产生模糊;场景 B 顺着同一甩动方向从模糊中切入,定格在孟母拉着孟子迈进学宫大门。要求场景切换自然。
〔创意〕动态接力转场

用角色动作衔接:场景 A 里人物跳起出画,场景 B 里人物保持下落姿势从画面上方落入(衣服或场景已换)。换装视频和跨地域旅拍的常客。

自由发挥版
It is necessary to use "Dynamic Relay Transition" at the cutting position (prohibiting rigid cutting and objects appearing out of thin air).
Use the continuous movements of characters to switch scenes. For example, in scene A, the character jumps up and leaves the frame, and then in scene B, the character maintains a falling posture and falls from the top of the frame into the picture.
It is required to switch scenes naturally.
中文对照:在切点使用「动态接力转场」(禁止硬切、禁止物体凭空出现)。用角色的连续动作切换场景:场景 A 里人物跳起出画,场景 B 里人物保持下落姿势从画面上方落入。要求场景切换自然。
具体案例版 · 孟母三迁
It is necessary to use "Dynamic Relay Transition" transition at the cutting position (prohibiting rigid cutting and objects appearing out of thin air).
In shot A, Mengzi playfully jumps up next to the cemetery and jumps out of the frame; then in scene B, Mengzi maintains a falling posture and falls into the picture from above the frame, but the background has already turned into a noisy market.
It is required to switch scenes naturally.
中文对照:在切点使用「动态接力转场」(禁止硬切、禁止物体凭空出现)。镜头 A 里孟子在墓地旁顽皮跳起、跳出画面;场景 B 里孟子保持下落姿势从画面上方落入,背景已变成喧闹集市。要求场景切换自然。
〔创意〕变焦转场

镜头无限放大场景 A 的某个细节(瞳孔、钥匙孔),放大到极致时该细节展开成场景 B 的全景;反向亦可。科幻片和「画中有画」套娃式叙事最爱。

自由发挥版
It is necessary to use the "zoom-in/zoom-out transition" transition at the cutting point (prohibiting rigid cutting and objects appearing out of thin air).
The camera zooms in on a detail in scene A (such as the pupil of a character) to the extreme. When the pupil occupies the entire screen, the image seamlessly unfolds inside the pupil, becoming a grand panorama of scene B.
It is required to switch scenes naturally.
中文对照:在切点使用「变焦转场」(禁止硬切、禁止物体凭空出现)。镜头把场景 A 的某个细节(如人物瞳孔)放大到极致,瞳孔占满全屏时,画面从瞳孔内部无缝展开,变成场景 B 的宏大全景。要求场景切换自然。
具体案例版 · 孟母三迁
Need to use "zoom-in/zoom-out transition" at the cutting point (prohibit rigid cutting and objects appearing out of thin air).
Camera A is the smiling face of Meng's mother, and the camera magnifies her deep pupils to the extreme. When the pupils occupy the entire screen, the picture seamlessly unfolds inside the pupils, becoming the grand panorama of Scene B: Mengzi is sitting neatly with a group of students in the study hall, kneeling and reading.
It is required to switch scenes naturally.
中文对照:在切点使用「变焦转场」(禁止硬切、禁止物体凭空出现)。镜头 A 是孟母含笑的脸,镜头把她深邃的瞳孔放大到极致;瞳孔占满全屏时,画面从瞳孔内部无缝展开,变成场景 B 的宏大全景:孟子和一众学子在学堂里正襟跪坐诵读。要求场景切换自然。
〔创意〕水墨转场

画面像浓墨滴入水中晕染扩散,前一画面随墨散去、后一画面在墨中浮现。极具东方美学,适合古风、文化景观、历史纪录和悠长回忆。

自由发挥版
It is necessary to use "Ink Transition" at the cutting point (prohibiting rigid cutting and objects appearing out of thin air).
Picture A quickly blends and spreads like thick ink dripping into clear water. As the ink spreads, picture A gradually disappears, and picture B gradually emerges from the scattered ink.
It is required to switch scenes naturally.
中文对照:在切点使用「水墨转场」(禁止硬切、禁止物体凭空出现)。画面 A 像浓墨滴入清水般迅速交融扩散,随墨晕散开逐渐消失;画面 B 从散开的墨迹中逐渐浮现。要求场景切换自然。
具体案例版 · 孟母三迁
It is necessary to use "ink and wash transition" at the cutting point (prohibiting rigid cutting and objects appearing out of thin air).
The noisy scene of the market in picture A quickly fades and diffuses like thick ink dripping into water, gradually disappearing; at the same time, the picture of Lang Lang reading in the school palace B slowly emerges from the scattered ancient ink.
It is required to switch scenes naturally.
中文对照:在切点使用「水墨转场」(禁止硬切、禁止物体凭空出现)。画面 A 集市的喧闹场景像浓墨滴水般迅速淡化扩散、逐渐消失;同时画面 B 学宫里朗朗读书的画面从散开的古墨中缓缓浮现。要求场景切换自然。

三条官方案例演示了「原片 → 顺滑延长 → 转场延长」的完整接力:

第一段 · 原片
The main character is the muscular warrior in @image1, he is walking alone in the dense primeval forest, with steady steps. He looks around with an alert expression, the camera pans forward following him, sunlight sprinkles mottledly through the leaves onto his solid muscles and armor.
中文对照:主角是 @图1 里的肌肉战士,他独自走在茂密的原始森林中,步伐沉稳,神情警觉地环顾四周。镜头跟随他向前平移,阳光透过树叶斑驳地洒在他结实的肌肉和铠甲上。
参考图
参考图:肌肉战士
官方成片:第一段原片。
第二段 · 顺滑延长 5 秒
Reference the content of @video1, extend the video by 5s backwards. Require the extension to be natural, the action connection to be smooth, the action connection to be natural, hard cuts are prohibited, objects appearing out of thin air are prohibited. The muscular warrior in @image1 suddenly notices something unusual, stops his footsteps, and his eyes become sharp. He quickly adopts a defensive posture, staring intently ahead.
中文对照:参考 @视频1 的内容,向后延长视频 5 秒。要求自然延长、动作衔接流畅自然,禁止硬切,禁止物体凭空出现。@图1 里的肌肉战士突然察觉异样,停下脚步,眼神变得锐利,迅速摆出防御姿态,紧盯前方。
参考图
参考图:肌肉战士
延长输入:第一段成片(作 @视频1 喂入)。
官方成片:延长后战士转入戒备。
第三段 · 遮罩转场延长 10 秒
Reference the content of @video1, extend the video by 10s to continue the story. Require to use a "mask transition" at the cut (hard cuts are prohibited, objects appearing out of thin air are prohibited). Shot A shows the muscular warrior in @image1 who is defending (close-up), a gust of wind blows, a normal-sized fallen leaf drifts towards the camera, the leaf completely covers the camera causing the full screen to go dark; then the camera pulls back through the leaf, directly revealing the next scene: the young ranger in @image2 is leaning against a tree, raising his bow and arrow taking aim. Require the scene to be switched naturally. Naturally switch from the warrior's close-up to the ranger's medium shot.
中文对照:参考 @视频1 的内容,延长视频 10 秒续写故事。要求在切点使用「遮罩转场」(禁止硬切、禁止物体凭空出现)。镜头 A 是 @图1 里正在防御的肌肉战士(特写),一阵风吹过,一片正常大小的落叶飘向镜头,完全盖住镜头致全屏变黑;随后镜头穿过落叶后拉,直接拉出下一场景:@图2 里的年轻游侠正倚着树举弓瞄准。要求场景切换自然,从战士的特写自然切到游侠的中景。
参考图参考图
参考图:肌肉战士与年轻游侠
延长输入:第二段成片(作 @视频1 喂入)。
官方成片:落叶遮罩转场引出第二个角色。
写法

智能编辑写法:圈哪里、改什么、几秒生效

三个编辑入口共用一套提示词骨架,核心是把「改哪个、怎么改、几秒生效」写全:

要素写法
编辑公式 = [具体修改对象] + [动作与变化描述] + [生效时序](Edit with marks 再前置一层 [标注工具与方位])
常规动词编辑视频中的 xx、增加/添加、删除/移除、修改/替换/改成。
具体修改对象(锁定目标)写清要改这块区域里的什么:男孩的牛仔裤、画面背景里的枯树、女主手中的咖啡杯、天上的云。
动作与变化描述(下达指令)写清改成什么、加什么、消除什么:替换成深黑色西装裤、变成盛开的樱花树、无痕消除并补全背景、加一道闪电。
生效时序(锁定时间轴)写清这处修改是全程存在还是只在某几秒:全片统一修改保持一致、只在第 5-8 秒变化、从第 3 秒起渐显。配合点确认后输入框出现的系统时间轴使用。
标注工具与方位(仅圈选)标注工具有不同颜色的框、线、箭头、文字、地标点,还有橡皮擦。要写清用什么工具画的、模型该关注工具内侧还是外侧:红框内、红色箭头所指区域、地标点处、红线左侧。
官方案例 · 鲸鱼改白龙
Modify everything entirely from 0 to 8 seconds of the overall video. Change the whale in the video to a white dragon. Remove the BGM.
中文对照:在整体视频的 0-8 秒,全部修改。把视频中的鲸鱼修改为一条白龙。去掉 BGM。
原片:粉色棉花糖云层里的飞天鲸鱼。
官方成片:鲸鱼已改成白龙,BGM 已去除。
看更多编辑案例(白底变绿幕、圈选生成、羊换马)
白底改绿幕
Take the entire video from 0 to 5 seconds, the whole video. Change all the white backgrounds into green screen backgrounds. Remove the sound, keep it muted.
中文对照:把整个视频 0-5 秒、全部画面里的白色背景都变成绿幕背景。去掉声音,保持静音。
原片:白底仓鼠动画。
官方成片:背景已变绿幕,给后期抠像备料。

圈选标注的两条案例(提示词为界面内中文原生输入,随成片文件名保留):

圈选案例一 · 原片:纯白模室内场景。
圈选案例一 · 成片。提示词:红框内的沙发上生成一个白模女生,白模女生正在看着电视;蓝框内生成一只白模猫。
圈选案例二 · 原片:法罗群岛宣传片。
圈选案例二 · 成片。提示词:将红框内的羊替换成一匹棕色的马,动作轨迹与原片保持一致,仅在视频第 7-15 秒期间生效。
写法

白模渲染写法:粗颗粒管动作,细颗粒管质感

白模(blockout)是用灰色几何体或素模型预演动作和机位的参考视频。写法先分清手里是哪一档,手册明说当前粗颗粒的生成效果更好:

粗颗粒白模细颗粒白模
是什么球体、圆柱、方块拼出来的低保真模型视频建模完整、材质结构清楚的 3D 模型视频
它管什么动态信息:运动轨迹、运镜方式、人物站位、光影变化、切镜直接渲染:材质、风格、光影上色
适用场景从零生成新视频,白模只当「动态骨架」场景已完整建模,只差换材质、光线、风格
注意模型带四肢、翅膀时要写全完整动作序列,否则容易僵硬,最好别传带四肢翅膀的上传前去掉轨迹线、坐标轴、相机锥体这类干扰信息
粗颗粒公式
[Reference Statement] + [Corresponding Relationship Mapping] + [Plot Details] + [Scene Processing] + [Overall Conclusion]
中文对照:参考声明(参考 @视频1 的运镜/动作/动线/光影变化/人物站位/音乐音效/画面节奏)+ 对应关系映射(哪个几何体对应哪张图里的谁:把视频 1 里的某色某形模型替换成图 X 的角色,某模型替换成图 Y 的角色,某模型替换成图 Z 的道具/场景)+ 情节细节(按时间或逻辑顺序写动作、表情、台词、光线变化)+ 场景处理(文字描述或引用图 W 的场景)+ 整体收束(整体风格、画质要求、一致性约束、声音处理等补充)。
细颗粒公式
[Rendering Command] + [Segmented Rendering Description] + [Scene Processing] + [Overall Convergence]
中文对照:渲染指令(把 @视频1 的白模动画渲染成成片)+ 分段渲染描述(0-N 秒:渲染要求,背景环境、色调氛围、角色材质、光影效果;N 秒处:转场触发点与方式、转场后新场景的渲染描述)+ 场景处理(文字描述或引用参考图)+ 整体收束(新场景环境与角色渲染保持不变或随场景变化;声音处理等)。
官方案例 · 白模变精灵女王
Reference the dynamic camera movement and physical lighting changes of @Video 1, and accurately replace the white model in the video with the elf queen in @Image 1. The character appearance needs to strictly reference the image, having long silver-white hair, wearing a gorgeous white-gold embroidered robe, and holding a magic staff inlaid with a glowing crystal at the top, maintaining character consistency, with no clipping and no face changing.
The overall scene is an elven hall, retaining the scene in @Video 1.
The camera position remains consistent with the original video, and the elf queen gracefully raises the magic staff following the action of the model in the reference video. During this time, her long silver-white hair and the wide white robe skirt flutter naturally with the airflow, with a silky texture and a sense of drape. Accompanying the action, the lighting effect in the reference video transforms into a sacred light beam rising from the ground, complementing the dazzling golden light emitted by the staff in the queen's hand.
The final video requires cinematic CG realistic picture quality, with real, sacred, and epic lighting and shadows, high-definition visuals, delicate character skin, clear metallic textures on the clothing, rich details, and smooth, natural actions.
中文对照:参考 @视频1 的动态运镜、物理光影变化,将视频中的白色模型精确替换为 @图1 的精灵女王。角色外观需严格参考图片:银白长发,穿华丽的白金刺绣长袍,手持顶端镶发光水晶的法杖,保持角色一致性,不穿模不换脸。整体场景为精灵大殿,保留 @视频1 中的场景。机位与原视频保持一致,精灵女王跟随参考视频中模型的动作优雅举起法杖,期间银白长发和宽大的白袍裙摆随气流自然飘动,有丝绸质感和垂坠感。伴随动作,参考视频中的光效化作从地面升起的神圣光柱,与女王手中法杖迸发的金色光芒交相辉映。最终视频要求电影级 CG 写实画质,光影真实神圣有史诗感,画面高清,角色皮肤细腻,服装金属质感清晰,细节丰富,动作流畅自然。
参考图
参考图:精灵女王设定
参考视频:白模原片。
官方成片:白模渲染成精灵女王。
看更多白模案例(赛博小狗、30 秒糖果厨房 boss 战、香水)
白模小人变赛博小狗
Reference @Image 1 and @Video 1's character actions and camera movements. Completely replace the white humanoid model in the video with the cute puppy from @Image 1. The puppy references @Video 1, walking upright, and the puppy walks naturally following the actions of the video. The background is a cyberpunk city street, and the white tall buildings in @Video 1 are rendered into high-rise buildings. The entire video should have no sound.
中文对照:参考 @图1 和 @视频1 的人物动作和运镜。将视频中的白色人形模型完全替换成 @图1 的可爱小狗。小狗参考 @视频1 直立行走,跟随视频动作自然走位。背景是赛博朋克城市街道,@视频1 中的白色高楼渲染成摩天大楼。整条视频不要声音。
参考图
参考图:小狗
参考视频:白模小人。
官方成片:直立行走的赛博小狗。
30 秒糖果厨房 boss 战
I want to make a 30-second stylized animation boss fight short film. The story takes place in a giant candy kitchen, where a toy-sized little knight, in order to take back the glowing candy, faces a massive, clumsy, and grumpy dessert boss. The little knight cannot defeat the boss head-on and needs to use the jelly, cream, sugar jars, weapons in the scene, and the boss's own attack mistakes to complete an animated fight.
@Video 1 is a white model viewport reference, only used for referencing character size relationships, standing direction, close-up camera distance, and spatial oppression. Do not copy the white model actions, do not just do a simple restoration, and the final footage must not retain the white model texture.
@Video 2 is the little knight's combat action reference, only used for the 20-25 seconds key counterattack segment. You can replicate the body's center of gravity, running rhythm, jumping posture, sword-raising angle, and swinging action inside, but they need to be replaced with the little knight from @Image 1, the weapon and shield from @Image 4, and the scene and art style of the candy kitchen boss fight. Do not bring the white model character, gray scene, or original footage texture of @Video 2 directly into the final film, and do not expand @Video 2 into a full-film action reference.
@Image 1-10 are visual references for the character, boss, scene, small items, weapons, special effects, and key counterattack frames. The final footage needs to strictly reference these images; do not randomly change the settings, and do not add irrelevant important characters or props. For background music, please use @Audio 1.
Storyboard requirements:0-5 seconds: Establish the confrontation. Use a wide or medium shot to show the giant candy kitchen battlefield of @Image 3; the little knight of @Image 1 stands in front of the giant boss of @Image 2, with an obvious size difference. The boss guards the glowing candy weak point on its chest from @Image 5; the little knight tightly grips the weapon from @Image 4, looking up at the boss, creating the oppression of a "weak protagonist facing a giant boss". This segment can reference the close-up spatial oppression and character standing relationship in @Video 1, but do not copy the white model actions.
5-10 seconds: The boss launches the first attack. The boss smashes a giant fist toward the little knight; the little knight quickly rolls, side-jumps, or is knocked away by the impact. The attack shatters the sugar jar or small items from @Image 8 in the scene; sugar powder and fragments scatter, creating the first wave of action impact. This segment focuses on showing the boss's clumsy strength and the little knight's flexible dodging.
10-15 seconds: The little knight's first counterattack fails. The little knight attempts to charge head-on toward the glowing candy weak point on the boss's chest, but because the size difference is too large, the attack has no effect. The boss easily knocks the little knight away, swats them aside, or blows them away with an air wave, showing the sense of failure that "fighting head-on won't work".
15-20 seconds: The little knight discovers a new strategy. The little knight is bounced near the jelly of @Image 6 and discovers that the environment can be leveraged. During the boss's second attack, the little knight deliberately guides the boss to step on the cream of @Image 7, causing the boss to lose balance, while allowing themselves to gain a higher counterattack position. This segment needs to maintain action continuity; do not suddenly reset the character's standing position. Let the position where they were knocked away in the previous segment transition naturally to the position where they discover the jelly and cream mechanisms.
20-25 seconds: Key counterattack. This segment mainly references the actions of @Video 2: the little knight first assumes a ready posture, then tightly grips the weapon of @Image 4 and sprints toward the boss, uses the jelly of @Image 6 to bounce and leap up, raises the weapon in the air to gather power, and then swings the sword to attack the glowing candy weak point on the boss's chest from @Image 5. You can replicate the action rhythm and posture changes of "ready — sprint — jump — raise sword in air — swing" in @Video 2, but the footage must be transformed into a stylized animated fight in the candy kitchen. At this time, because the boss stepped on the cream, it is briefly unbalanced, perfectly exposing the weak point on its chest. The camera can quickly switch between close-up, medium, and wide shots to show the process of the little knight sprinting and jumping, the boss retreating in panic, and the weapon hitting the weak point. At the moment of impact, add the sugar crystal shattering, cream exploding, sugar powder smoke, and glowing shockwave of @Image 9, serving as the biggest explosive point of the whole film.
25-30 seconds: After the boss is hit in the glowing weak point, its body stiffens, and the special effects of sugar crystal shattering and cream exploding spread out. The boss falls backward or kneels on the ground, losing combat capability. The little knight lands steadily, returns to a combat stance similar to the end of @Video 2, and then assumes a victory pose. The camera slowly pushes in from a medium shot to the little knight's victory expression, forming a complete ending.
The boss must always maintain the oppression of a giant size, and the little knight must always maintain a toy size and sense of flexibility. You can add jumping, rolling, shaking, bouncing, slipping, explosions, sugar powder smoke, slow motion, fast camera cuts, and exaggerated animation performances. Do not turn it into a live-action realistic movie; it must remain a stylized animation. Do not let the footage become a pure stacking of special effects; it must be clearly visible who is attacking, who is dodging, and who is counterattacking. Do not directly restore the white model content of @Video 1 into the final footage. @Video 2 is only used for the little knight's action reference in the 20-25 seconds key counterattack segment; do not use it for other segments. Do not reset the standing position in each time period; the entire video must be a continuously occurring boss fight. The final film needs to have a clear beginning, failure, method discovery, key counterattack, and ending.
中文对照:我想做一支 30 秒的风格化动画 boss fight 短片。故事发生在巨大的糖果厨房:玩具大小的小骑士为了夺回发光糖果,对上庞大、笨拙、暴躁的甜品 boss。小骑士没法正面击败 boss,需要借助场景里的果冻、奶油、糖罐、武器和 boss 自己的攻击失误完成一场动画式打斗。 @视频1 是白模机位参考,只用于参考角色体型关系、站位方向、特写镜头距离和空间压迫感。不要照抄白模动作,不要只做简单还原,最终画面不许保留白模质感。@视频2 是小骑士的战斗动作参考,只用于 20-25 秒关键反击段,可以复刻其中的身体重心、奔跑节奏、跳跃姿势、举剑角度和挥砍动作,但要替换成 @图1 的小骑士、@图4 的武器盾牌,以及糖果厨房 boss 战的场景与美术风格。不要把 @视频2 的白模角色、灰色场景或原片质感直接带进成片,也不要把 @视频2 扩展成全片动作参考。@图1-10 是角色、boss、场景、小物件、武器、特效和关键反击帧的视觉参考,最终画面需严格参考这些图,不要乱改设定,不要添加无关的重要角色或道具。背景音乐请使用 @音频1。 分镜要求:0-5 秒 立对峙,全景或中景展示 @图3 的巨大糖果厨房战场;@图1 的小骑士站在 @图2 的巨大 boss 面前,体型差明显;boss 护着胸口 @图5 的发光糖果弱点;小骑士紧握 @图4 的武器仰视 boss,营造「弱小主角面对巨大 boss」的压迫感。此段可参考 @视频1 的特写空间压迫感和角色站位关系,但不要照抄白模动作。5-10 秒 boss 首次攻击,boss 巨拳砸向小骑士;小骑士快速翻滚、侧跳或被冲击掀飞;攻击砸碎场景里 @图8 的糖罐或小物件,糖粉与碎片四溅,做出第一波动作冲击。10-15 秒 小骑士首次反击失败,小骑士试图正面冲向 boss 胸口的发光糖果弱点,但体型差距太大,攻击无效;boss 轻松把他拍开、扫飞或用气浪吹走,表现「正面刚打不过」的挫败感。15-20 秒 发现新策略,小骑士被弹到 @图6 的果冻附近,发现环境可以借力;在 boss 第二次攻击时故意引它踩上 @图7 的奶油失去平衡,自己借势占据更高的反击位。此段要保持动作连续性,不要突然重置角色站位。20-25 秒 关键反击,主要参考 @视频2 的动作:小骑士先摆好预备姿势,紧握 @图4 的武器冲刺,借 @图6 的果冻弹跳跃起,空中举械蓄力,挥剑攻击 boss 胸口 @图5 的发光糖果弱点。可复刻 @视频2「预备—冲刺—跳跃—空中举剑—挥砍」的动作节奏,但画面必须转化成糖果厨房里的风格化动画打斗。此时 boss 因踩到奶油短暂失衡,胸口弱点完全暴露。镜头可在特写、中景、全景间快速切换,展示小骑士冲刺跳跃、boss 慌乱后退、武器命中弱点的全过程。命中瞬间加入 @图9 的糖晶碎裂、奶油爆炸、糖粉烟雾和发光冲击波,作为全片最大爆点。25-30 秒 收尾,boss 被击中发光弱点后身体僵直,糖晶碎裂与奶油爆炸特效扩散,boss 向后倒下或跪地失去战斗力;小骑士稳稳落地,回到与 @视频2 结尾相似的战斗姿态,摆出胜利姿势。镜头从中景缓缓推近小骑士的胜利表情,形成完整结尾。 boss 必须始终保持巨大体型的压迫感,小骑士必须始终保持玩具体型的灵活感。可以加入跳跃、翻滚、震动、弹跳、打滑、爆炸、糖粉烟雾、慢动作、快速切镜和夸张的动画演出。不要做成真人写实电影,必须保持风格化动画。不要让画面变成纯特效堆砌,必须能看清谁在攻击、谁在躲避、谁在反击。不要把 @视频1 的白模内容直接还原进成片。@视频2 只用于 20-25 秒的动作参考,不要用于其他段落。各时间段不要重置站位,整条视频必须是一场连续发生的 boss 战。成片要有清晰的开局、失败、发现方法、关键反击和结尾。
参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图
10 张视觉参考:小骑士、甜品 boss、糖果厨房与关键帧
参考视频一:机位白模。
参考视频二:反击动作示范。
官方成片:风格化动画 boss 战。

细颗粒的香水白模一组(原文档此处复用了精灵女王的提示词文本,直接看白模与成片对照):

参考图参考图参考图
参考图:香水产品与材质
参考视频:香水细颗粒白模。
官方成片:香水白模渲染结果。
白模思路上和 Seedance 通用写法一脉相承:粗颗粒借时序、细颗粒重渲染,模型带四肢时要求完整动作序列。Maya/Blender 用户直接用前面说的官方插件导出白模即可。
演示

亮点演示上:时长、控制力和画质底盘

写法全部讲完,后半程看官方拿这些能力交出的作业。每个功能的讲法一致:能力是什么、哪里用、配哪条案例。先看管时长和画质底盘的五项。

视频延长:从「短平快」到连续叙事

三件事:单次延长上限提到 30 秒内视频再延 4-30 秒;「套娃式」反复叠加,只要当前成片在 30 秒内就能继续延,极限拼到 60 秒;每个延长节点都能用新提示词介入,先让主角跑 10 秒、再指挥他起跳,一步一步导演。适合悬念反转(前 5 秒安静喝咖啡、延长段 UFO 突然飞过)、复杂动作分段接力(拔剑→前刺→转身收剑逐段确认)、以及镜头穿镜换时空的运镜转场。官方案例是一条三段接力的召唤神龙:

第一段 · 12 秒开场(雨崖三人组)
The opening is first an overhead raining empty shot of an environment similar to Image 1, then proceeds with a downward spiral circling camera movement, the two subjects in Image 2 are one with back to the camera and one side-facing the camera, the camera movement speed slows down but continues to maintain camera movement, then the subject of Image 3 appears from below the camera (squatting posture changes to standing) with side face facing the camera (Spielberg shot), the three people's standing position forms an acute triangle, the camera pulls back, all three gradually become facing their backs to the camera and all walk towards the cliff edge, located in the middle, the black-clothed one is located on the left, the white-clothed one is located on the right. Cut 1: The leftmost character looks sideways at the rightmost character (focus on the rightmost, the other two are out of focus), the rightmost one says: "Big brother, will he appear today?"; Cut 2: Frontal low-angle medium shot of the middle character, saying: "He will", then pan right to the leftmost man in black, he says: "He is here"; Cut 3: From the over-the-shoulder perspective of the middle character looking towards the distant water surface, the water surface starts to gradually surge with waves, faintly able to see parts of the body of Image 4, with a large area still in the water;
中文对照:开幕先是类似 @图1 环境的俯拍下雨空镜,然后向下螺旋环绕运镜;@图2 的两个主体一个背对镜头一个侧对镜头,运镜减速但保持移动;接着 @图3 的主体从镜头下方出现(蹲姿变站姿)侧脸对镜头(斯皮尔伯格式入画),三人站位构成锐角三角形;镜头后拉,三人渐成背对镜头一起走向崖边:中间一人居中,黑衣者在左、白衣者在右。切镜一:最左侧角色侧头看最右侧角色(焦点在最右,其余两人虚焦),最右者说:「大哥,他今天会出现吗?」切镜二:中间角色正面低角度中景,说:「他会的」,随后镜头右摇到最左的黑衣男子,他说:「他来了」。切镜三:从中间角色的过肩视角望向远处水面,水面开始涌起波澜,隐约可见 @图4 身体的一部分,大部分仍在水中。
参考图参考图参考图参考图
参考图:环境与三位角色 + 龙的造型
官方成片:12 秒开场。
看后两段延长(龙出水 +5 秒、决战 +13 秒)
第二段 · 延长 5 秒
Reference the content of Video 1 to extend the video by 5 seconds; the dragon comes out of the water and roars angrily at the three people, emitting sound waves that distort space, the three people hurriedly raise their hands to form a light shield for defense, the shape of the dragon references @Image 1.
中文对照:参考 @视频1 的内容,延长视频 5 秒:龙从水里出来对着三人怒号,发出扭曲空间的声波,三人急忙抬手结出光盾防御,龙的造型参考 @图1。
参考图
参考图:龙的造型
延长输入:第一段成片(作 @视频1 喂入)。
官方成片:龙出水怒号。
第三段 · 延长 13 秒
Reference the content of Video 1, extend the video by 13s. The three people's light shields gradually shatter because they cannot withstand the roar, and then they are knocked flying backward by the shock. Cut: Image 1 is still in the middle, and after he is knocked back a certain distance, he stops in a half-kneeling position, pulls out the weapon from his back, and charges forward. The two subjects in Image 2 look at each other and also take out their weapons and charge out. Cut: Panoramic shot, the back perspective of the three people, after sprinting a certain distance towards the dragon, the three people leap high into the air and start fighting with the dragon. During the fight, there must be fire light particles and lightning particle special effects. The shape of the dragon references Image 3. The ending concludes with a panoramic shot of the three people being knocked back again.
中文对照:参考 @视频1 的内容,延长视频 13 秒。三人的光盾因承受不住怒号逐渐碎裂,随后被震飞后退。切镜:@图1 的角色仍在中间,被击退一段距离后半跪刹停,拔出背后武器向前冲锋;@图2 的两个主体对视一眼也掏出武器冲出。切镜:全景、三人背影视角,向龙冲刺一段距离后高高跃起与龙开战,打斗中要有火光粒子和闪电粒子特效,龙的造型参考 @图3。结尾以三人再次被击退的全景收束。
参考图参考图参考图
参考图:三位角色与龙的造型
延长输入:第二段成片(作 @视频1 喂入)。
官方成片:光盾碎裂与空中开战。

超长视频 180 秒:整段叙事一次出

三件事:单次直出最长 180 秒,告别多次延长拼接;3 分钟跨度里角色特征、服装细节、场景建筑首尾一致,拒绝中途「变异」;长时间缓推、环绕、穿越镜头全程平滑。适合超长风光宣传片和完整 MV、情绪短片。前面写法一节埋过的 180 秒校园怪谈就在这里,12 个角色全程时间戳推进:

官方案例 · 180 秒校园怪谈《第十三个名字》
0:00-0:13, the camera slowly pulls back from a close-up of a glass covered with raindrops, revealing an old middle school gate shrouded in dense mist, with dark green moss climbing on the dark gray walls. @Image 1 walks into the frame holding a black umbrella, a loose tie hangs in front of his white shirt, the schoolbag on his back presses down on his shoulder line, he raises his eyes to look at the grayish teaching building, and a trace of unease flashes through his calm eyes. The campus broadcast with static noise slowly sounds: "Welcome, twelve new students." Only the sound of rain, the buzzing echo of the broadcast, and the muffled thunder in the distance are left in the empty campus.
0:13-0:30, the camera smoothly pushes into the teaching building, passes through the corridor and stops at the front door of the classroom, then cuts to a panoramic shot of the classroom. Old fluorescent lights hum overhead, rainwater meanders down the window glass, twelve sets of desks and chairs are neatly arranged, and eleven students have already taken their seats: @Image 2 sits in the front row, turns his/her head to look at the door, and reveals a gentle and somewhat mysterious smile to @Image 1 who just entered; @Image 3 adjusts his/her glasses, lowers his/her head and keeps recording something in a notebook, with a serious and earnest expression; @Image 4 leans against the back of the chair with arms crossed over the chest, tall and straight in figure, silently looking out the window; @Image 5 lies on the desk with a camera hanging around the neck, fingertips fiddling with the camera strap, eyes full of curiosity; @Image 6 sits properly in the seat, the student council armband on the left arm is eye-catching, looking calm and composed; @Image 7 holds a book of urban legends with a worn cover, flipping the pages with shining eyes; @Image 8 tightly grips the schoolbag straps in front of her, twin tails hanging on her shoulders, eyes timidly sweeping the surroundings; @Image 9 leaves his school uniform jacket open, earrings reflecting under the light, leaning in the chair with an impatient face. @Image 1 walks to the empty seat in the last row and sits down, fingertips gently wiping the wet water stains on the desktop.
0:30-0:42, @Image 10 walks onto the podium holding a roll call book, black long hair hanging on the shoulders of her professional suit, with a gentle expression. She adjusts her glasses frame, opens the roll call book, and begins to call names in order. Every time a name is read, it corresponds to a clear response, and the camera sequentially sweeps across the students below the podium following the rhythm of the roll call. After calling the twelfth name, @Image 10's fingertip suddenly stops on the paper page, pauses for two seconds, frowns slightly, and subconsciously reads out a thirteenth name. The classroom instantly falls into dead silence, no one answers, only the sound of rain outside the window becomes exceptionally clear. @Image 10 then closes the roll call book, her tone so natural as if the pause just now never happened: "Alright, everyone do self-study first."
0:42-0:50, the camera pushes to a close-up of @Image 1's face, he frowns, his gaze sweeps across the whole class counting one by one, and after counting, his face shows obvious doubt. The camera slowly pulls away from behind him, sweeping across the entire classroom, in the shadows of the back row, a slender white figure flashes by, so fast it looks like an optical illusion caused by the rain curtain, the suspense of the first act drops here.
0:50-1:12, the corridor between classes is immersed in cold-toned skylight, the rain is still falling. @Image 5 holds up the camera to frame the campus, after pressing the shutter several times in a row, she lowers her head to look through the playback, her frown tightening more and more — in the corner of every photo, stands a blurry figure in white clothes. She pulls @Image 3 who is passing by, hands the camera in front of his eyes, @Image 3 adjusts his glasses, looks carefully for a long time, and shakes his head saying there is nothing. @Image 9 nearby sneers and leans over, glances at it, and mocks her for having blurred vision. @Image 5 tightly grips the camera, looks up again at the end of the corridor corresponding to the photo, there is no one there, only the wind blowing past with threads of rain, and a chill rises on her back.
1:12-1:35, in the school history storage room, dust floats in the slanted light. @Image 7 squats in front of a cabinet piled with old books searching, his fingertips touch a student registry with a charred cover, with a date from forty years ago printed on the cover. He excitedly opens the registry, inside is a list of student victims of a campus fire back then, twelve names come into view one by one, and his smile slowly freezes on his face — the names on the list completely overlap with the names of the twelve classmates currently in the class. @Image 6 just pushes the door open to come in looking for materials, seeing this, he walks over to look through the files, a solemn look also appears on his calm face, fingertips gently tapping the charred cover, falling into silence.
1:35-1:52, night falls, the corridor of the teaching building plunges into pitch black, only the beam of a flashlight sways back and forth on the wall. @Image 11, wearing a blue work uniform, slowly patrols holding the flashlight, the soles of his shoes stepping on the floor making dull sounds. When walking to the door of this class's classroom, he suddenly stops, from inside the door comes blurry voices and the sound of desks and chairs moving. He frowns, turns the handle and pushes the door open, the classroom is empty, only the four large words "Welcome Back" written in white chalk right in the center of the blackboard, with chalk dust still slightly drifting in the air. @Image 4, who stayed in the classroom to train, rushes over upon hearing the sound, standing at the door looking at the words on the blackboard, the muscles in his arms suddenly tightening.
1:52-2:10, cut back to the evening self-study classroom, the overhead light tubes sizzle, the light flickering bright and dark. @Image 8 suddenly lies on the desk and starts crying in a low voice, her shoulders constantly trembling. @Image 2 immediately gets up and walks to her side, gently patting her back to comfort her. @Image 8 speaks while sobbing, her voice trembling: "Every day after school... there is a girl following me, walking very slowly, following me all the way to the bottom of the dormitory building, as soon as I turn my head, she disappears." The classroom instantly quiets down, @Image 9 frowns and mutters a complaint, @Image 3 and @Image 6 exchange a glance, and suppressed unease slowly spreads among the twelve people.
2:10-2:35, @Image 1 sits in his seat, fingertips pinching the newly distributed class group photo. He has repeatedly counted three times, there are always twelve people in the photo, but a strong sense of incongruity lingers. He rubs his swollen temples, inadvertently closes his eyes, but the picture of the group photo clearly floats in his mind — in the empty seat right in the middle of the crowd, stands a girl with long white hair, quietly looking at the camera. @Image 1 suddenly opens his eyes, in the photo there are still only twelve familiar faces. He tries repeatedly a few times, every time he closes his eyes, he can clearly see that figure. He mutters to himself in a low voice, with an imperceptible tremble: "Only by closing my eyes... can I see."
2:35-2:47, the classroom fluorescent lights begin to flicker crazily, alternating between bright and dark, a white figure slowly emerges in the center of the crowd. @Image 12 stands barefoot between the desks, long white hair hanging over the white skirt hem, red eyes calm and wave-less, the corners of her mouth carrying a very faint smile. Everyone stops their actions, only the electric current sound of the light tubes is left in the classroom. @Image 12 speaks softly, a voice as light as threads of rain falling by the ear: "Finally... someone remembers me."
2:47-3:00, the moment the voice falls, the twelve people in the classroom slowly turn their necks, their actions so uniform as if pulled by invisible strings, looking straight towards the direction of the camera together, without any superfluous expressions on their faces. The camera abruptly pushes forward fiercely, and the picture instantly cuts to pure black. Accompanied by a distant and light ring of the class bell, the entire film ends.
中文对照:0:00-0:13,镜头从落满雨珠的玻璃特写缓缓拉开,露出浓雾笼罩的老中学校门,墨绿苔藓爬满深灰围墙。@图1 撑黑伞入画,白衬衫前挂着松垮的领带,书包压着肩线,他抬眼望向灰蒙蒙的教学楼,平静眼底闪过一丝不安。带电流杂音的校园广播缓缓响起:「欢迎,十二位新同学。」空旷校园里只剩雨声、广播的嗡鸣回响和远处闷雷。 0:13-0:30,镜头平滑推进教学楼,穿过走廊停在教室前门,切教室全景。老旧日光灯在头顶嗡鸣,雨水沿窗玻璃蜿蜒而下,十二套桌椅整齐排列,十一名学生已就座:@图2 坐前排回头望向门口,对刚进门的 @图1 露出温和又略带神秘的笑;@图3 扶了扶眼镜低头在笔记本上不停记录,神情严肃认真;@图4 抱臂靠在椅背上,身形高大挺拔,沉默望向窗外;@图5 趴在桌上,脖子挂着相机,指尖拨着相机带,眼里满是好奇;@图6 端坐座位,左臂的学生会袖标很显眼,神色沉稳;@图7 捧着一本封皮磨损的都市传说,翻页时眼睛发亮;@图8 双手紧攥胸前书包带,双马尾垂肩,眼神怯怯地扫视四周;@图9 校服外套敞着,耳钉在灯下反光,靠在椅子里一脸不耐烦。@图1 走到最后一排空位坐下,指尖轻轻抹过桌面的湿水渍。 0:30-0:42,@图10 抱着点名册走上讲台,黑长发垂在职业套装肩头,神情温和。她扶了扶镜框,翻开点名册按序点名,每念一个名字都有清晰的应答,镜头随点名节奏依次扫过台下学生。点完第十二个名字后,@图10 的指尖突然停在纸页上,顿了两秒、眉头微蹙,下意识念出了第十三个名字。教室瞬间死寂,无人应答,只有窗外雨声变得格外清晰。@图10 随即合上点名册,语气自然得仿佛刚才的停顿从未发生:「好了,大家先自习。」 0:42-0:50,镜头推近 @图1 的脸部特写,他皱眉,目光扫过全班逐个数人,数完后满脸疑惑。镜头从他身后缓缓拉远扫过整间教室,后排阴影里一个纤细白色身影一闪而过,快得像雨幕造成的错觉,第一幕的悬念落在这里。 0:50-1:12,课间走廊浸在冷调天光里,雨还在下。@图5 举起相机取景校园,连按几次快门后低头回看,眉头越皱越紧,每张照片角落都站着一个模糊的白衣身影。她拉住路过的 @图3 把相机递到他眼前,@图3 扶了扶眼镜仔细看了半天,摇头说什么都没有。旁边的 @图9 嗤笑着凑过来瞥了一眼,嘲讽她眼花。@图5 攥紧相机,再抬头看向照片对应的走廊尽头,那里空无一人,只有风裹着雨丝吹过,后背窜起一阵寒意。 1:12-1:35,校史储藏室里灰尘在斜光中漂浮。@图7 蹲在旧书堆积的柜前翻找,指尖碰到一本封皮烧焦的学生名册,封面印着四十年前的日期。他兴奋地翻开,里面是当年校园火灾的罹难学生名单,十二个名字逐一映入眼帘,他的笑容慢慢僵在脸上,名单上的名字和现在班里十二个同学的名字完全重合。@图6 恰好推门进来找资料,见状走过去翻看档案,沉稳的脸上也浮现凝重,指尖轻叩焦黑封皮,陷入沉默。 1:35-1:52,入夜,教学楼走廊陷入漆黑,只有手电光束在墙上来回晃动。穿蓝色工装的 @图11 提着手电缓慢巡逻,鞋底踩在地板上发出沉闷声响。走到这个班教室门口时他突然停住,门内传来模糊人声和桌椅挪动声。他皱眉拧动把手推开门,教室空无一人,只有黑板正中用白粉笔写着四个大字「欢迎回来」,粉笔灰还在空气里轻轻飘。留在教室训练的 @图4 闻声赶来,站在门口看着黑板上的字,手臂肌肉骤然绷紧。 1:52-2:10,切回晚自习教室,头顶灯管滋滋作响、忽明忽暗。@图8 突然趴在桌上低声啜泣,肩膀不停发抖。@图2 立刻起身走到她身边轻拍后背安抚。@图8 抽噎着开口,声音发颤:「每天放学……都有个女生跟着我,走得很慢,一路跟到宿舍楼下,一回头她就不见了。」教室瞬间安静,@图9 皱眉低声抱怨,@图3 和 @图6 交换了一个眼神,压抑的不安在十二人之间慢慢蔓延。 2:10-2:35,@图1 坐在座位上,指尖捏着新发的班级合影。他反复数了三遍,照片里始终是十二个人,但强烈的违和感挥之不去。他揉着胀痛的太阳穴无意间闭上眼,合影画面却清晰浮现在脑海,人群正中间的空位上,站着一个白色长发的女生,安静地望着镜头。@图1 猛地睁眼,照片里依旧只有十二张熟悉的脸。他反复试了几次,每次闭眼都能清楚看到那个身影。他喃喃自语,带着难以察觉的颤抖:「只有闭上眼……才能看见。」 2:35-2:47,教室日光灯开始疯狂闪烁,明暗交替间,一个白色身影缓缓浮现在人群中央。@图12 赤足立在课桌之间,白色长发垂在白裙摆上,红色眼瞳平静无波,嘴角带着极淡的笑。所有人停下动作,教室里只剩灯管的电流声。@图12 轻声开口,声音轻得像雨丝落在耳边:「终于……有人记得我了。」 2:47-3:00,话音落下的瞬间,教室里十二个人缓缓转动脖颈,动作整齐得像被无形丝线牵引,一起直直望向镜头方向,脸上没有任何多余表情。镜头猛然向前急推,画面瞬间切至纯黑。伴随一声遥远轻微的上课铃,全片结束。
参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图
12 张角色参考图:十二名学生、老师、保安与白发女生
官方成片:180 秒一次直出的校园怪谈。

时间戳控制:指令下到具体某一秒

很多模型的时间戳只能表达事件先后,画面时长控制不住,节奏赶、切镜早。2.5 原生支持按时间戳控制剧情和镜头区间:在要描述的镜头、台词、动作前直接加时间标记(如 [00.0-05]),按时间顺序写剧本,指令能精确下到秒级。适合多镜头剧情短片(避免剧情「快进」)、卡点视频和音乐 MV、广告分段展示(前 5 秒看外观、中间 10 秒看功能、最后 5 秒看效果)。三条案例连精度口径都不一样,按帧写、按秒写、混着写:

官方案例 · 240 帧武侠琴战(按帧控制)
Chinese anime CG new Chinese-style martial arts style, with no subtitle text throughout, ink particles and flowing light special effects fused together, 240 frames totaling 10 seconds, ancient-style BGM rhythm progresses with the fighting, the climax perfectly synchronized with the ultimate move outbreak. Frame 0 opens with a high-altitude overlook of the ancient courtyard battlefield, dozens of enemies in black tight outfits surrounding the woman in a green shirt kneeling on one knee hugging a guqin in the center from all directions. Frame 32 the camera quickly dives and zooms in, the woman's fingertips move slightly gathering momentum, the guqin strings faintly glow with a light blue light. Frame 48 the woman's fingertips pluck the string for the first time, the first crescent-shaped blue sound blade breaks through the air, the strings vibrating out layers of air ripples. Frame 61 the sound blade knocks away the front row of enemies, wood splinters and clothing fragments splash in mid-air, enemies on both sides continue to close in. Frame 84 the woman's fingers continuously pluck, three sound blades shoot out in a staggered manner suppressing the frontal enemy group, the BGM rhythm gradually rises. Frame 110 the woman spins and leaps up, the guqin hovering in the air following her body, her clothes unfolding and fluttering. Frame 132 the woman's toe kicks the guqin body twice continuously to borrow force, a fan-shaped rain of sound wave arrows spreads out, ink particles spreading along the sound wave trails. Frame 158 the camera cuts to a facial close-up, the woman's eyes are cold, her fingertips hook the thickest string pulling it back with full force until taut and glowing cold. Frame 183 fingertips suddenly release the string, a giant fan-shaped sound wave sweeps out, the picture cuts to high-speed slow motion. Frame 201 the sound wave glides across the ground plowing a deep trench, dust shooting up into the sky, rows of enemies are flipped and smashed down. Frame 217 the woman lands and heavily slaps the guqin onto the blue bricks, pressing the strings with both hands, the remaining enemies all surround to within a few meters in front of her raising swords ready to pounce. Frame 231 the woman's arms fiercely sweep across all strings, the guqin body bursts with blinding white light, a circular energy field instantly explodes, a residual shadow of an ink phoenix shuttles through the battlefield, the picture cuts to a black and white ink filter, the BGM reaches its climax. Frame 240 the enemies fall to the ground uniformly and the dust settles, the woman retracts her fingers with the strings still vibrating, the picture freezes.
中文对照:国漫 CG 新中式武侠风格,全程无字幕文字,水墨粒子与流光特效融合,240 帧共 10 秒,古风 BGM 节奏随打斗推进,高潮与大招爆发完全同步。第 0 帧开场高空俯瞰古院战场,数十名黑衣紧身劲装的敌人从四面八方围住中央单膝跪地抱着古琴的绿衫女子。第 32 帧镜头快速俯冲推近,女子指尖微动蓄势,琴弦隐隐泛起淡蓝光。第 48 帧女子指尖第一次拨弦,第一道新月形蓝色音刃破空而出,琴弦震出层层气浪涟漪。第 61 帧音刃击飞前排敌人,木屑和衣料碎片在半空飞溅,两侧敌人继续合围。第 84 帧女子手指连续拨动,三道音刃交错射出压制正面敌群,BGM 节奏渐起。第 110 帧女子旋身跃起,古琴随身悬空,衣袂展开翻飞。第 132 帧女子足尖连点琴身两次借力,扇形音波箭雨铺开,水墨粒子沿音波轨迹扩散。第 158 帧镜头切面部特写,女子眼神清冷,指尖勾住最粗的琴弦全力回拉,直到绷紧泛出冷光。第 183 帧指尖骤然松弦,巨型扇形音波横扫而出,画面切高速慢动作。第 201 帧音波贴地犁出深沟,尘土冲天,成排敌人被掀翻砸落。第 217 帧女子落地将古琴重重拍在青砖上,双手按弦,残余敌人全部围到身前数米举刀欲扑。第 231 帧女子双臂猛然横扫全部琴弦,琴身爆发刺目白光,环形能量场瞬间炸开,一道水墨凤凰残影穿梭战场,画面切黑白水墨滤镜,BGM 到达高潮。第 240 帧敌人整齐倒地、尘埃落定,女子收指、琴弦仍在震颤,画面定格。
Seedance 2.5 成片:动作连贯,帧级节拍准确。
上一代同题对比:节拍达标,但动作连贯性不足。
看另外两条时间戳案例(手绘科普、720 帧毛毡动画)
太阳能板科普动画(按秒控制)
Take the hand-drawn children's encyclopedia illustration style of @Reference Image 1 - @Reference Image 5 as the unified visual style, maintaining the rounded chibi characters, marker hand-drawn texture, off-white paper texture background, colored stroke text, and infographic layout. The overall use is a vertical popular science animation, picture elements maintain consistency, only completing the explanation through local dynamics, camera pushing in, particle animation, and infographic demonstration, with a brisk animation rhythm.
0-3s | Pose the questionThe camera slowly pushes in from @Reference Image 1, and the title "How do solar panels generate electricity?" gradually enlarges. The sun reveals a smiling face and blinks gently, yellow photon particles continuously fly from the sun to the rooftop solar panels, the panel surface glows with soft blue highlights, the light bulb icon gradually lights up from gray, and the keywords appear on the screen: "sunlight" "photons" "solar panels".
3-6s | Step 1: Photons enter the solar panelThe picture naturally transitions to @Reference Image 2, and the camera pushes in to the cross-section of the solar panel. A large number of glowing photons fly into the solar cell along the arrows like small golden balls, the anti-reflection coating, N-type silicon layer, and P-type silicon layer highlight layer by layer, and the electronic infographic emerges synchronously.
6-9s | Step 2: Electrons are excitedThe camera continues to push in to the microscopic silicon crystal world of @Reference Image 3. Photons strike silicon atoms, blue electrons are "awakened", jump away from their original positions and move at high speed, leaving flashing holes in their original positions. The movement of electrons forms a glowing particle stream, and the entire silicon crystal lattice is accompanied by the diffusion of slight energy ripples.
9-12s | Step 3: Form an electric currentThe picture switches to @Reference Image 4, the built-in electric field in the middle of the PN junction emits a light green energy beam, pushing electrons to the N region and holes to the P region, positive and negative charges flow in different directions, the blue electric current flows quickly in the wire, and the voltage icon gradually lights up. The animation highlights the two key concepts of "electric field pushing" and "forming electric current".
12-15s | Final summaryThe camera pulls away to @Reference Image 5, the solar panel continuously receives sunlight, the blue current flows to the inverter, and is then transmitted to the house, and the light bulb, fan, and mobile phone light up in sequence. Finally freezing on the popular science conclusion of "photons + semiconductors + electric field + circuit = solar power generation", the overall picture is bright, warm, and full of a technological enlightenment atmosphere.
中文对照:以 @参考图1-@参考图5 的手绘儿童百科插画风格为统一视觉风格,保持圆润 Q 版人物、马克笔手绘质感、米白纸纹背景、彩色描边文字和信息图版式。整体是竖版科普动画,画面元素保持一致,只通过局部动态、镜头推进、粒子动画和信息图演示完成讲解,动画节奏轻快。 0-3 秒|抛出问题:镜头从 @参考图1 缓缓推进,标题「太阳能板是怎么发电的?」逐渐放大。太阳露出笑脸轻眨眼,黄色光子粒子不断从太阳飞向屋顶太阳能板,板面泛起柔和蓝色高光,灯泡图标由灰渐亮,画面出现关键词:「阳光」「光子」「太阳能板」。 3-6 秒|第一步 光子进入太阳能板:画面自然过渡到 @参考图2,镜头推进到太阳能板剖面。大量发光光子像小金球一样沿箭头飞入太阳能电池,减反射涂层、N 型硅层、P 型硅层逐层高亮,电子信息图同步浮现。 6-9 秒|第二步 电子被激发:镜头继续推进到 @参考图3 的微观硅晶世界。光子撞击硅原子,蓝色电子被「唤醒」,跳离原位高速运动,原位留下闪烁的空穴。电子的运动形成发光粒子流,整个硅晶格伴随轻微能量涟漪扩散。 9-12 秒|第三步 形成电流:画面切到 @参考图4,PN 结中间的内建电场发出淡绿色能量束,把电子推向 N 区、空穴推向 P 区,正负电荷分向流动,蓝色电流在导线里快速流淌,电压图标渐亮。动画突出「电场推动」和「形成电流」两个关键概念。 12-15 秒|总结收尾:镜头拉远到 @参考图5,太阳能板持续接收阳光,蓝色电流流向逆变器再输送进屋,灯泡、风扇、手机依次亮起。最后定格在「光子 + 半导体 + 电场 + 电路 = 太阳能发电」的科普结论上,整体画面明亮温暖,充满科技启蒙氛围。
参考图参考图参考图参考图参考图
5 张手绘插画参考
Seedance 2.5 成片:节奏基本贴合时间戳,动画呈现生动。
上一代对比:整体快于时间戳,渐变过渡有结构问题。
《奶奶的菜谱》720 帧毛毡动画(帧秒混写)
"Grandma's Recipe" (720 frames)
[Global Constraint Declaration]
Visual Style: The whole film strictly adopts the animated felt style, and any style deviation in any frame is prohibited.
Subtitle Limitation: Except for the specified subtitle fading in at the end (Frames 660-720), generating any other subtitles or text in the whole film is absolutely prohibited.
Reference Materials: Cooking references @Video 1, background music uses @Audio 1, environment and action sound effects use @Sound Effect 2, @Sound Effect 3, and empty shot style references @Image 3.
Frames 0-90 (0s - 3s)@Image 1 old-fashioned kitchen, warm afternoon light shines through the screen window, fine dust floats in the air; a yellowed handwritten recipe @Image 6 is spread out on the mottled wooden table, the camera slowly pushes in on the faded handwriting on the paper surface; nostalgic piano fades in (reference @Audio 1), Frame 60 "tick-tock" sound of a distant wall clock.
Frames 90-240 (3s - 8s)A young girl wearing an apron cuts vegetables according to the recipe (@Image 6), her movements are unfamiliar, @Image 4 the cucumber is cut crookedly (action reference @Video 1), she frowns and recuts; the moment it goes into the pan, oil splatters with a "sizzle" (reference @Sound Effect 3), she is frightened, steps back and shakes her hand, handheld camera shakes slightly, Frame 180 spatula collision sound.
Frames 240-390 (8s - 13s)The picture dissolves into a warm yellow memory flashback — @Image 2 in her childhood, grandma holds her small hands from behind to stir-fry together, grandma smiles and speaks (silent), soft focus halo; Frame 390 dissolves back to reality, the girl stands alone in front of the stove, looking dazed.
Frames 390-540 (13s - 18s)The girl steadies her mind and stir-fries seriously, this time her movements gradually become smooth, the vegetable aroma rises (close-up of hot steam); she plates the dish into a white porcelain plate and sets it on the wooden table, Frame 480 piano melody emotionally advances.
Frames 540-660 (18s - 22s)The girl picks up a bite and tastes it, chews, stops her movement, and her eye sockets slowly turn red; the camera slowly pushes in on her wet eyes, Frame 630 she looks up at grandma's old photo on the wall referencing @Image 5.
Frames 660-720 (22s - 24s)Close-up of the old photo (reference @Image 5), grandma smiles kindly; the girl softly says "It's still this taste" (Note: only vocalization, no subtitles), the picture's warm light softens, and the only subtitle fades in at the bottom "Some tastes, hide a person". Frame 713 picture and music fade out to black, music box trailing sound concludes.
中文对照:《奶奶的菜谱》(720 帧)。【全局约束声明】视觉风格:全片严格采用毛毡定格动画风格,任何一帧都禁止风格漂移。字幕限制:除结尾指定字幕(660-720 帧)淡入外,全片绝对禁止生成任何其他字幕或文字。参考素材:烹饪动作参考 @视频1,背景音乐用 @音频1,环境与动作音效用 @音效2、@音效3,空镜风格参考 @图3。 0-90 帧(0-3 秒):@图1 的老式厨房,温暖午后光透过纱窗照入,细小尘埃浮在空气里;一张泛黄的手写菜谱 @图6 摊在斑驳木桌上,镜头缓缓推近纸面褪色的字迹;怀旧钢琴声淡入(参考 @音频1),第 60 帧远处挂钟「滴答」一声。 90-240 帧(3-8 秒):系围裙的女孩照着菜谱 @图6 切菜,动作生疏,@图4 的黄瓜切得歪歪扭扭,她皱眉重切;下锅瞬间油花「滋啦」溅起(参考 @音效3),她吓得后退甩手,手持镜头轻微晃动,第 180 帧锅铲碰撞声。 240-390 帧(8-13 秒):画面叠化进暖黄色的回忆闪回,@图2 她小时候,奶奶从身后握着她的小手一起翻炒,奶奶笑着说话(无声),柔焦光晕;第 390 帧叠化回现实,女孩独自站在灶前,神情怔忡。 390-540 帧(13-18 秒):女孩定了定神认真翻炒,这次动作渐渐流畅,菜香升腾(热气特写);她把菜装进白瓷盘摆上木桌,第 480 帧钢琴旋律情绪推进。 540-660 帧(18-22 秒):女孩夹起一口尝,咀嚼,动作停住,眼眶慢慢泛红;镜头缓缓推近她湿润的眼睛,第 630 帧她抬头望向墙上奶奶的旧照片(参考 @图5)。 660-720 帧(22-24 秒):旧照片特写(参考 @图5),奶奶笑得慈祥;女孩轻声说「还是这个味道」(注:只出声、无字幕),画面暖光变柔,底部淡入全片唯一字幕「有些味道,藏着一个人」。第 713 帧画面与音乐淡出至黑,八音盒尾音收束。
参考图参考图参考图参考图参考图参考图
毛毡场景与角色参考图
参考视频:毛毡动画质感样片。
Seedance 2.5 成片:基本按「秒」级时间range执行,交互稳定。上一代因素材过多不支持生成。

去除无关字幕和 BGM:反向指令终于管用

老问题:明明写了「不要字幕」「不要 BGM」,成片还是自作主张配了。2.5 大幅优化生成稳定性,乱入字幕和无关配乐的问题得到解决。写法就一句:加上「纯净视频、无字幕、无背景音乐」这类修饰词。适合攒纯净空镜素材(自然风光、城市航拍、产品微距)、对声音设计要求极高的商业广告和 MV「干净底片」、以及展会背景屏这类纯视觉内容。

官方案例 · 电商穿搭视频(全程静音)
Vertical 9:16 e-commerce outfit display video, pure white minimalist seamless studio background, fixed split-column layout remains consistent throughout: the left side of the screen vertically arranges 3 rectangular modules with thin black rounded borders, the top of each module is a black serif number (1, 2, 3) decorated with short horizontal lines on the left and right, corresponding flat-lay images of outfit items are placed inside the modules, and small text "1x" is marked below the items; the right side of the screen is a young female model standing full-body showing the outfit in real scene, full-body eye-level fixed camera position, light soft studio even lighting, no hard shadows, clothing fabric texture is clear and delicate, the model maintains a natural and elegant standing posture, with slight soothing dynamics such as raising a hand to adjust clothes or adjusting the standing posture, the overall style is clean and premium, conforming to the aesthetics of women's clothing e-commerce sales. The reference image serves as a style and layout anchor, keeping the split-column border style, font typography, and light-shadow texture unified throughout, video frame rate 24fps. It is strictly prohibited to have any text subtitles, slogans, watermarks, logos, or extra annotations throughout the process, the entire process has no background music BGM, no human voice-over, no environmental noise, only pure visual pictures in complete silence, the camera is fixed throughout without zooming, panning, or moving, and the layout structure remains unchanged throughout.
中文对照:竖版 9:16 电商穿搭展示视频,纯白极简无缝影棚背景,固定分栏版式全程一致:画面左侧竖排 3 个黑色细圆角边框模块,每个模块顶部是左右带短横线装饰的黑色衬线数字(1、2、3),模块内放对应穿搭单品的平铺图,单品下方标小字「1x」;画面右侧是年轻女模特实景全身展示穿搭,全身平视固定机位,影棚柔光均匀无硬阴影,服装面料质感清晰细腻,模特保持自然优雅的站姿,只有抬手整理衣物、调整站姿这类轻缓动态,整体风格干净高级,符合女装电商审美。参考图作为风格与版式锚点,全程保持分栏边框样式、字体排版和光影质感统一,视频帧率 24fps。全程严禁出现任何文字字幕、标语、水印、logo 或多余标注,全程无背景音乐 BGM、无人声旁白、无环境噪音,只有完全安静的纯视觉画面,镜头全程固定不推拉不摇移,版式结构全程不变。
参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图
版式与单品参考图
官方成片:全程无字幕无声音,版式纹丝不动。

基础生成效果全面优化:给画面「去 AI 感」

三件事:极致物理质感,去掉 AI 视频常见的「油腻感」「塑料反光」,还原真实毛孔与光影,动物毛发蓬松自然,配音的机械电子感也被削弱;「长尾动作」解锁,特殊舞蹈、复杂武打、细微表情这些过去一做就崩的动态如今顺畅自然;时空一致性,单次 30 秒内人物形象和空间物理逻辑保持稳定,切镜过渡自然。写法上:追求写实就写「原生皮肤质感」「发根分明」;描述复杂动作直接写全(「武者雨中完成复杂后空翻,落地后迅速拔出长剑,雨水顺着剑刃滑落」);多镜头保持角色基础特征描述不变。适合美妆特写、宠物广告、写实微电影和街舞武术这类动作向创作。

官方案例 · 三人自由格斗
The bald tattooed boxer of @Image 1, the red-haired female fighter of @Image 2, and the Asian kickboxer wearing a mouthguard of @Image 3 — engaging in a three-way melee in the scene of @Image 4. Opening medium shot: the three people confront each other in a triangular stance, probing each other, their footsteps grinding the ground and kicking up dust. Cut to close-up: the bald boxer takes the lead to punch and attack the red-haired woman, the red-haired woman dodges sideways and counterattacks with a whip kick, and the Asian kickboxer takes advantage of the chaos to launch a swing punch from the flank. Cut to close-up: the moment the bald boxer is hit on the chin, beads of sweat splash, his expression is twisted in pain, and his mouthguard flies out. Cut to medium shot: the red-haired woman turns around to fight the Asian kickboxer while the bald man staggers, and the two engage in close-quarters grappling. Cut to wide high-angle shot: after the three people fall to the ground, they each struggle to get up, the light bulb shakes violently, light and shadow alternately flicker on the three people, and sweat and blood mist form a Tyndall effect in the backlight. The entire footage emphasizes realism, volumetric lighting tracing global illumination, fine fabric textures, film production quality, shallow depth of field, and cinematic color grading.
中文对照:@图1 的光头纹身拳手、@图2 的红发女格斗家、@图3 的戴牙套的亚洲踢拳手,在 @图4 的场景里三方混战。开场中景:三人三角站位对峙互相试探,脚步碾地扬起灰尘。切特写:光头拳手率先出拳攻向红发女,红发女侧身闪避接鞭腿反击,亚洲踢拳手趁乱从侧翼抡出摆拳。切特写:光头拳手下巴被击中的瞬间汗珠飞溅、表情痛苦扭曲,牙套飞出。切中景:红发女趁光头踉跄转身迎战亚洲踢拳手,两人近身缠斗。切高角度全景:三人倒地后各自挣扎起身,灯泡剧烈摇晃,光影在三人身上交替明灭,汗水与血雾在逆光中形成丁达尔效应。整条素材强调写实、体积光全局光照、精细布料纹理、电影级制作质量、浅景深和电影调色。
参考图参考图参考图参考图
三位格斗者与场景参考图
Seedance 2.5 成片:三人身份稳定、打击交互无穿模。
上一代对比:人物一致但打击穿模、分镜站位混乱。
看另外两条基础优化案例(儿童早教、游戏 CG)
儿童早教 3D 动画
Please generate a 3D animated short film for early childhood education within 25 seconds, with the theme "Let's Count Together: 1, 2, 3". Use the lively little boy from @Image 1, the cheerful little girl from @Image 2, and the gentle female teacher from @Image 3. Reference @Image 4 for the scene as a bright suburban street, with blue sky and white clouds, rounded small houses, green lawns, gardens, and soft sunlight. The overall style is bright, safe, cute, rounded, and friendly for young children, suitable for children aged 2–5 to watch.
Shot 1: 0–3 secondsShot Size: Extreme wide shot / Street openingCamera Movement: Slow push inThe bright suburban street appears, the three characters stand in the center of the frame, and the colored numbers 1, 2, 3 pop out like bubbles.Lines: "Let's count together: 1, 2, 3!"
Shot 2: 3–8 secondsShot Size: Medium close-up / Number 1Camera Movement: Slight push inThe little boy runs to the side of the lawn and points to 1 small red ball. The small ball bounces gently, and the number 1 appears.Lines: "1! One small ball."
Shot 3: 8–12 secondsShot Size: Medium close-up / Number 2Camera Movement: Smooth panThe little girl squats by the garden and points to 2 small yellow flowers. The two small flowers sway gently, and the number 2 appears.Lines: "2! Two small flowers."
Shot 4: 12–15 secondsShot Size: Medium shot / Number 3Camera Movement: Slight tilt up and then holdThe three characters look up at the sky together, 3 colored balloons slowly rise, and the number 3 appears.Lines: "3! Three balloons."
Requirements: Horizontal screen 16:9, duration within 15 seconds, high-definition 3D animation, soft natural light, bright colors but not dazzling, simple and clear actions, no dangerous actions, no horror elements, no complex plots, no intense flashing, no text watermarks. The appearance of any IP-related elements, existing animation characters, movie characters, game characters, comic characters, brand logos, trademarks, copyrighted patterns, iconic clothing, iconic props, recognizable brand names, watermarks, or third-party copyrighted elements in the frame is prohibited. All characters, props, and scenes must be original designs.
中文对照:请生成一支 25 秒以内的儿童早教 3D 动画短片,主题「一起数数:1、2、3」。使用 @图1 的活泼小男孩、@图2 的开朗小女孩、@图3 的温柔女老师。场景参考 @图4:明亮的郊区街道,蓝天白云、圆润小屋、绿色草坪、花园和柔和阳光。整体风格明亮、安全、可爱、圆润,对幼儿友好,适合 2-5 岁儿童观看。 镜头一 0-3 秒(大全景/街道开场,缓慢推进):明亮的郊区街道出现,三个角色站在画面中央,彩色数字 1、2、3 像泡泡一样弹出。台词:「我们一起数数:1、2、3!」镜头二 3-8 秒(中近景/数字 1,轻微推进):小男孩跑到草坪边指着 1 个红色小球,小球轻轻弹跳,数字 1 出现。台词:「1!一个小球。」镜头三 8-12 秒(中近景/数字 2,平滑横移):小女孩蹲在花园边指着 2 朵黄色小花,两朵小花轻轻摇摆,数字 2 出现。台词:「2!两朵小花。」镜头四 12-15 秒(中景/数字 3,轻微上摇后固定):三个角色一起抬头看天,3 只彩色气球缓缓升起,数字 3 出现。台词:「3!三只气球。」 要求:横屏 16:9,高清 3D 动画,柔和自然光,色彩明亮不刺眼,动作简单清晰,无危险动作、无恐怖元素、无复杂剧情、无强烈闪烁、无文字水印。画面中禁止出现任何 IP 相关元素、现有动画/电影/游戏/漫画角色、品牌 logo、商标、版权图案、标志性服装道具、可识别品牌名或第三方版权元素,所有角色、道具、场景必须原创设计。
参考图参考图参考图参考图
参考图:两个孩子、老师与街道场景
Seedance 2.5 成片:角色一致,数量与数字精确对应。
上一代对比:多生成了角色,动效不自然。
VRMMORPG 游戏 CG
VRMMORPG virtual game gameplay CG. Deep within a cursed swamp constructed by holographic data streams, a faint blue dark moon with flickering geometric grids hangs high. In the center of the frame, a necromancer (Boss) @Image 1 draped in a shattered dark blue raven feather cloak, is slowly raising a deadwood staff burning with code flames above his head, and suspended above him are a translucent crimson UI health bar and danger warning icons. Sharp golden light instantly bursts from the skeletal face deep within the hood, and the skull decorations on the shoulders and waist rattle after a brief flash of highlighted polygonal wireframes. As he heavily strikes the staff into the mire, the flames surge violently. The camera rapidly pulls back and orbits the wizard close to the ground for a large-scale rotation shot; the moment the staff touches the ground, it triggers a ring of magical shockwaves mixed with glitch art and green digital information streams, causing slight pixel tearing in the surrounding space. Top-tier American Stylized 3D rendering, delicate hand-drawn textures interwoven with holographic neon light and shadows, cyber-gothic fantasy style, presenting a highly immersive virtual reality (VR) game BOSS entrance visual effect.
中文对照:VRMMORPG 虚拟游戏实机 CG。在全息数据流构建的诅咒沼泽深处,闪烁几何网格的幽蓝暗月高悬。画面中央,@图1 披着破碎暗蓝鸦羽斗篷的死灵法师(Boss)正缓缓将燃着代码火焰的枯木法杖举过头顶,头顶悬浮着半透明的猩红 UI 血条和危险警告图标。兜帽深处的骷髅面孔骤然爆出锐利金光,肩部和腰间的骷髅挂饰在高亮多边形线框一闪后咔咔作响。他将法杖重重杵进泥沼的瞬间火焰暴涨,镜头快速后拉并贴地绕法师做大幅旋转运镜;法杖触地瞬间触发混着故障艺术与绿色数字信息流的环形魔法冲击波,周围空间轻微像素撕裂。顶级美式风格化 3D 渲染,细腻手绘纹理与全息霓虹光影交织,赛博哥特奇幻风,呈现极具沉浸感的 VR 游戏 BOSS 登场视觉效果。
参考图
参考图:死灵法师 Boss 设定
Seedance 2.5 成片:风格、氛围与特效表现出色。
上一代对比:整体氛围和配乐节奏都差一档。
演示

亮点演示中:十来种语言直接开口

面对小语种指令或小语种台词,过去常见的结局是理解跑偏、发音不准、断句生硬。2.5 全面支持多语言文字输入:优先优化中文、英语、西班牙语、印尼语、马来语,全面覆盖泰语、阿拉伯语、葡萄牙语、越南语、日语、韩语;台词发音准确度和口型同步也一起升级。用法直接:用你熟悉的语言写提示词;要角色开口,就把目标语言的台词直接写进去。适合出海营销和本地化广告(免拍摄免翻译配音)、全球创作者母语创作、多国角色同框的多语言短片。五条案例横跨五个市场:

官方案例 · 全韩文鸡尾酒广告
Please strictly reference @Image One @Image Two @Image Three @Image Four @Image Five @Image Six to generate a vertical high-end cocktail product commercial video. The overall style is spring fresh, light luxury fashion, cream light green background, soft studio lighting, minimalist premium, commercial ad texture. The protagonists are six tequila cocktails, needing to showcase the close-ups and details of each product one by one: ice cubes, condensation water, salt or chili powder rim, lime, pineapple, lychee, cherry, mint, jalapeño and other decorative elements, highlighting the liquid layers, transparent texture, fresh fruity aroma and high-end mixology style.
The camera movement requires cinematic level, including slow push-ins, macro close-ups, slight orbiting, silky transitions, and can add dynamic details such as ice dropping into the cup, liquid splashing, glass reflection, cold mist feeling, enhancing the sense of coolness and appetite. The overall picture is unified and clean, shadows are soft, the subject is prominent, hyper-realistic, high-definition details, natural and transparent colors, suitable for brand ads, social media short videos, and e-commerce promo video spread.
All text in the final film must be entirely in Korean. This includes the main title, product names, corresponding slogans for each product, subtitles, ending copy, and all on-screen copy. The appearance of Chinese, English, Spanish, or any non-Korean text is prohibited. All product names and slogans need to be transformed into natural, premium, Korean expressions suitable for fashion ad typography based on the reference image content, with a unified overall language style, concise and exquisite, having a high-end brand feel.
中文对照:请严格参考 @图一至 @图六,生成一支竖屏高端鸡尾酒产品广告视频。整体风格春日清新、轻奢时尚,奶油浅绿背景、柔和影棚光、极简高级、商业广告质感。主角是六款龙舌兰鸡尾酒,需逐一展示每款产品的特写与细节:冰块、凝结水珠、盐边或辣椒粉杯口、青柠、菠萝、荔枝、樱桃、薄荷、墨西哥辣椒等装饰元素,突出液体层次、通透质感、新鲜果香和高端调酒风格。 运镜要求电影级,包含缓慢推近、微距特写、轻微环绕、丝滑转场,可加入冰块落杯、液体飞溅、玻璃反光、冷雾感等动态细节,增强清凉感和食欲感。整体画面统一干净,阴影柔和,主体突出,超写实、高清细节、色彩自然通透,适合品牌广告、社交媒体短视频和电商宣传片传播。 成片中所有文字必须全部使用韩语,包括主标题、产品名、每款产品对应的标语、字幕、结尾文案和一切画面文案。禁止出现中文、英文、西班牙语或任何非韩语文字。所有产品名和标语需基于参考图内容转化为自然、高级、适合时尚广告排版的韩语表达,整体语言风格统一、简洁精致,有高端品牌感。
参考图参考图参考图参考图参考图参考图
六款鸡尾酒产品参考图
Seedance 2.5 成片:全片文字严格生成韩语。
上一代对比:以英文为主,只混进了部分韩语。
看另外四条多语言案例(越南语、印尼语、日语、马来语)
越南语 TikTok 带货短片
Reference the marketing style of @Video 1, produce a 15-second viral Vietnam TikTok marketing short video.Target market: VietnamProduct: Free drone @Reference Image 1Video style: Vietnamese local influencer selling goodsStreet interview feelHandheld cameraFast-paced camera cutsStrong conflict marketingAll dialogues must use Vietnamese.All subtitles must use Vietnamese.All product packaging text must use Vietnamese.
The lines are as follows:Fake? Fake? Fake?Boss calm down!Why do people say our free drone is fakeRidiculous, this is absolutely the real appearance, it is very simple to operate, just one button can take off and land, besides you can use a mobile phone to control it, and there are dual 4k HD camerasAbsolutely it's you who hid it, users cannot find the transaction channelJust need to click the video to enter temuThis is completely legalKey sentence: Click the video to download
中文对照:参考 @视频1 的营销风格,制作一条 15 秒越南 TikTok 爆款营销短视频。目标市场:越南。产品:@参考图1 的免费无人机。视频风格:越南本地网红带货、街头采访感、手持镜头、快节奏切镜、强冲突营销。所有对白必须用越南语,所有字幕必须用越南语,所有产品包装文字必须用越南语。 台词如下:假的?假的?假的?/ 老板冷静!/ 为什么有人说我们的免费无人机是假的 / 荒谬,这绝对是真机外观,操作特别简单,一键起降,还能用手机操控,双 4K 高清摄像头 / 绝对是你藏起来了,用户找不到交易渠道 / 只需要点视频进 temu / 这完全合法 / 关键句:点击视频下载。
参考图
参考图:无人机产品
参考视频:营销风格样片。
Seedance 2.5 成片:对白节奏正常、断句流畅。
上一代对比:越南语发音不准。
中文口播转印尼语带货(口型同步)
Convert the explanation video of the Chinese host in @Video 1 into an Indonesian livestream selling video. The host @Image 1 in the livestream room @Image 2 faces the camera, showcases this facial serum product @Image 3 in the frame, and smilingly says this Indonesian line: "This facial serum is super moisturizing! Come and click the little yellow cart below to instantly enjoy a 50% discount!" It is required that in the output video, the character's lip shape must perfectly match the pronunciation of this Indonesian line.
中文对照:将 @视频1 这段中文主播的讲解视频转换为印尼语带货视频。主播 @图1 在直播间 @图2 里面对镜头,展示画面中的精华液产品 @图3,微笑着说出这句印尼语台词:「这款精华液超级保湿!快点击下方小黄车,立享五折优惠!」要求输出视频中人物口型与这句印尼语的发音完美同步。
参考图参考图参考图
参考图:主播、直播间与产品
参考视频:中文口播原片。
Seedance 2.5 成片:语速语气自然,台词发音完整。
上一代对比:语气节奏不自然,台词发音不完整。
全日文拉面广告
The overall style is a real commercial advertisement, cinematic photography, warm yellow lighting, Tokyo night atmosphere, and the visual language is natural and smooth.
Shot 1 (0-3s)The camera slowly pushes in from the Tokyo night street @Image 1 to the entrance of the ramen shop in @Image 2. The shop sign, lanterns, business hours sign, and menu are all in natural and standard Japanese, without any Chinese, English, or garbled text appearing. Other shop signs on the street also uniformly use Japanese.
Shot 2 (3-7s)The camera enters the shop, the shop assistant in @Image 3 stands in the open kitchen, smiles facing the camera, and says in natural and smooth Japanese: "Hello, would you like a portion of homemade tonkotsu ramen?"During speaking, the lip shape is strictly synchronized with the Japanese pronunciation, the voice is natural, and has the characteristics of a real Japanese male voice. The background environment retains real environmental sounds such as boiling noodles in the kitchen, steam, and tableware colliding.
Shot 3 (7-11s)The camera switches to a close-up of the ramen in @Image 4. Hot steam continuously rises, the chashu slightly shakes, and the grease on the soup surface flows naturally. Brand promotional subtitles appear in the frame, and the subtitles all use natural and standard Japanese: "Rich taste, a bowl of happiness."The subtitle font conforms to the design style of Japanese commercial advertisements, without Chinese, English, or garbled text appearing.
Shot 4 (11-15s)@Image 5 customer sits at the table, tastes the ramen, then reveals a satisfied smile, and naturally says to the camera: "It is really delicious, I will definitely come again!"The voice maintains natural Japanese pronunciation, and the lip shape synchronization is accurate.The brand Logo @Image 6 emerges in the frame, and the Japanese Slogan appears in the frame: "Want to eat a bowl every day."
All brand Logos, menus, price tags, posters on the wall, and business information at the entrance all maintain unified Japanese, and mixing in Chinese, English, or garbled text is not allowed; all dialogues in the entire video are in Japanese, and generating Chinese subtitles, Chinese dubbing, or other languages is prohibited. The overall rhythm conforms to the 15-second commercial advertisement style, the camera transitions are natural, the character identities remain consistent, the shop assistant and the customer do not change faces or costumes throughout the whole process, and the picture has a real advertising-level texture.
中文对照:整体风格为真实商业广告,电影级摄影,暖黄色灯光,东京夜晚氛围,视觉语言自然流畅。 镜头一(0-3 秒):镜头从东京夜晚街道 @图1 缓缓推进到 @图2 的拉面店门口。店招、灯笼、营业时间牌和菜单全部使用自然标准的日语,不出现任何中文、英文或乱码文字,街上其他店招也统一用日语。镜头二(3-7 秒):镜头进店,@图3 的店员站在开放厨房里,微笑面对镜头,用自然流利的日语说:「你好,来一份自家制豚骨拉面怎么样?」说话时口型与日语发音严格同步,声音自然,具有真实日本男性的声线特征。背景保留煮面声、蒸汽声、餐具碰撞声等真实环境音。镜头三(7-11 秒):切 @图4 拉面特写,热气持续上升,叉烧微微颤动,汤面油脂自然流动。画面出现品牌宣传字幕,全部使用自然标准的日语:「浓郁口感,一碗幸福。」字幕字体符合日本商业广告设计风格。镜头四(11-15 秒):@图5 的顾客坐在桌前品尝拉面,露出满足的笑,自然地对镜头说:「真的太好吃了,我一定会再来!」保持自然日语发音、口型同步准确。品牌 Logo @图6 浮现,画面出现日语 Slogan:「想每天都吃一碗。」 所有品牌 Logo、菜单、价签、墙上海报、门口营业信息统一日语,不许混入中文英文或乱码;全片对白均为日语,禁止生成中文字幕、中文配音或其他语言。整体节奏符合 15 秒商业广告风格,镜头过渡自然,人物身份一致,店员与顾客全程不换脸不换装,画面有真实广告级质感。
参考图参考图参考图参考图参考图参考图
街景、店面、人物与产品参考图
Seedance 2.5 成片:日语发音自然完整。
上一代对比:画质低、直接复用原图,念词略赶。
马来语游戏预告《低语之岛》
Generate a fantasy healing adventure game worldview teaser short film under 30 seconds, titled Island of Whispers. The short film is intended for release in the Malaysian market. The overall style is high-quality game CG, nature magic, fairy tale fantasy, open-world adventure, and warm healing. The story takes place in a mysterious archipelago floating above a sea of clouds. The world is losing its voice, the wind stops flowing, flowers no longer bloom, and the ancient clock tower stops turning. The protagonist, the "Wind Traveler," needs to find fragments of the wind to awaken the sleeping islands.
The short film's output language should smartly use Malay, with short lines and a game trailer atmosphere; all dialogues must simultaneously be paired with Malay subtitles to facilitate understanding for the Malaysian market audience. The visuals should be suitable for social media, mobile game advertisements, and global game pre-heating promotion, with a fast-paced rhythm. The short film can reference @Audio 1 as the video's theme background melody.
All characters in the short film must appear; even if some characters only appear for 1–2 seconds, they must have clear visual memory points. For scenes, reference @Image 11 Moon Lake, @Image 12 Forest Ruins, and @Image 13 Sea of Clouds Archipelago.
Reference @Image 1 Wind Traveler: The young protagonist, wearing a light beige travel cloak, carrying a small bag, and holding a glowing wind chime or small lantern.Reference @Image 2 Little Wind Spirit: A cute companion, a semi-transparent little cloud or bird-shaped sprite, emitting a white-gold shimmer.Reference @Image 3 Flower Clock Maiden: The NPC guarding the ancient clock tower, with a petal long dress, gear, and flower clock elements, gentle and mysterious.Reference @Image 4 Antler Tree Spirit: A giant deer-shaped forest guardian, with antlers like tree branches, growing glowing flowers and leaves.Reference @Image 5 Cloud Whale: A giant sky guardian floating in the sea of clouds, its body appearing as if made of clouds, mist, and soft light.Reference @Image 6 Lost Postman: A cute little NPC, carrying a giant mailbag, wearing a hat with a feather stuck in it, enthusiastic and flustered.Reference @Image 7 Star Lamp Merchant: A mysterious traveling merchant, wearing a wide hat and robe, carrying a mobile stall hung with star lamps, wind chimes, and small bottles.Reference @Image 8 Silent Stone Statue Boy: A semi-petrified ruin boy, with moss and flowers on his body, and glowing eyes.Reference @Image 9 Moon Lake Witch: An elegant and mysterious mentor, wearing a moon-white robe, with long hair like lake water, holding a crystal lamp or crescent staff.Reference @Image 10 Storm Prince: A sorrowful opposing character, wearing a dark blue cape, a silver hair crown, surrounded by shattered wind blades and dark clouds.
Storyboard Prompts:Shot 1: 0–3 secondsShot Size: Extreme wide shot / Camera Movement: Aerial-style dive and push in for world establishmentAbove the sea of clouds, floating islands slowly appear one after another. Grasslands, waterfalls, ancient clock towers, forests, lakes, and distant ruins float among the clouds. The world is quiet, windmills have stopped turning, and flowers are closed.
Shot 2: 3–6 secondsShot Size: Medium shot / Camera Movement: Low angle slow push inThe Wind Traveler wakes up on the grass, the light beige cloak fluttered by a faint wind. The Little Wind Spirit flies up from beside the backpack, emitting a soft white-gold light, and circles around the protagonist.Little Wind Spirit, English dialogue: "Wake up, traveler."
Shot 3: 6–9 secondsShot Size: Fast montage / Camera Movement: Continuous fast pan + soft transitionThe camera quickly sweeps past the ancient clock tower. The Flower Clock Maiden stands in front of the giant flower clock that has stopped turning, holding a pocket watch in her hand. Then cuts to the forest, the Antler Tree Spirit slowly raises its head, the flowers and leaves on its antlers glowing faintly.
Shot 4: 9–12 secondsShot Size: Extreme wide shot / Camera Movement: Low angle tilt up + slow followA giant Cloud Whale swims past from the depths of the sea of clouds, with small islands, vines, and glowing flowers floating on its back. The protagonist looks up, the picture showing the scale of the world.Voiceover, English: "A silent world is waiting to sing again."
Shot 5: 12–14 secondsShot Size: Medium close-up / Montage Camera Movement: Handheld brisk follow shotThe Lost Postman runs across the grassy slope carrying a giant mailbag, letters and maps flying out; the Star Lamp Merchant lights a star lamp in the floating island market, the stall hung full of wind chimes, small bottles, and glowing seeds.
Shot 6: 14–18 secondsShot Size: Medium shot / Camera Movement: Slow orbitIn the ancient ruins, the Silent Stone Statue Boy half-kneels in front of the stone door, his body covered with moss and small flowers, his eyes lighting up slightly. The cracks on the stone statue boy's body emit a soft light.
Shot 7: 18–22 secondsShot Size: Medium close-up / Camera Movement: Slow push inRipples form on the moonlit lake surface, the Moon Lake Witch appears from the lakeside, her moon-white robe and lake-water-like long hair flowing gently. She holds a crystal lamp, looking towards the protagonist.Moon Lake Witch, Spanish dialogue: "El viento recuerda lo que tú has olvidado."
Shot 8: 22–26 secondsShot Size: Close-up / Camera Movement: Fast push inThe sky suddenly darkens, dark clouds roll up. The Storm Prince stands atop the shattered high tower, his dark blue cape flapping, his silver hair crown flashing, surrounded by wind blades and shattered feathers.Storm Prince, Spanish dialogue: "No todos los mundos merecen despertar."
Shot 9: 26–30 secondsShot Size: Wide shot / Camera Movement: Slow push in from behind, final freeze frame on the titleThe Wind Traveler and the Little Wind Spirit stand on the edge of the cliff, facing the glowing island at the end of the sea of clouds. The wind blows again, the grass, flowers, wind chimes, Cloud Whale, and distant archipelago simultaneously emit a soft light. The game title appears on the screen: Island of WhispersVoiceover, English: "Find the fragments. Awaken the islands."
中文对照:生成一支 30 秒内的奇幻治愈冒险游戏世界观预告短片,名为 Island of Whispers(低语之岛),面向马来西亚市场发行。整体风格是高品质游戏 CG、自然魔法、童话奇幻、开放世界冒险、温暖治愈。故事发生在漂浮于云海之上的神秘群岛:世界正在失去声音,风停止流动,花不再开放,古老钟楼停止转动。主角「风之旅人」要找到风的碎片,唤醒沉睡的岛屿。 短片输出语言智能使用马来语,台词简短、有游戏预告片氛围;所有对白必须同步配马来语字幕,方便马来西亚市场观众理解。画面适合社交媒体、手游广告和全球游戏预热传播,节奏明快。短片中所有角色必须悉数登场,哪怕只出现 1-2 秒也要有清晰的视觉记忆点。场景参考 @图11 月光湖、@图12 森林遗迹、@图13 云海群岛。 角色设定:@图1 风之旅人,少年主角,浅米色旅行斗篷,背小包,手持发光风铃或小灯笼;@图2 小风灵,可爱的同伴,半透明的小云朵或鸟形精灵,泛白金微光;@图3 花钟少女,守护古钟楼的 NPC,花瓣长裙带齿轮与花钟元素,温柔神秘;@图4 鹿角树灵,巨大的鹿形森林守护者,鹿角如树枝,长着发光花叶;@图5 云鲸,漂浮云海的巨大天空守护者,身体如云雾与柔光构成;@图6 迷路邮差,背巨大邮包的可爱小 NPC,帽子上插羽毛,热心又慌张;@图7 星灯商人,神秘旅商,宽檐帽长袍,推着挂满星灯、风铃和小瓶子的移动摊位;@图8 沉默石像男孩,半石化的遗迹男孩,身上长着苔藓与小花,眼睛发光;@图9 月光湖女巫,优雅神秘的导师,月白长袍,长发如湖水,持水晶灯或新月法杖;@图10 风暴王子,悲情的对立角色,深蓝披风、银发冠,周身环绕破碎风刃与乌云。 分镜提示:镜头一 0-3 秒(大全景/航拍式俯冲推进立世界观):云海之上浮岛逐一浮现,草原、瀑布、古钟楼、森林、湖泊与远处遗迹漂浮云间。世界安静,风车停转,花朵闭合。镜头二 3-6 秒(中景/低角度缓推):风之旅人在草地上醒来,浅米色斗篷被微风吹起,小风灵从背包边飞起,泛柔和白金光,绕主角盘旋。小风灵英语台词:「Wake up, traveler.」镜头三 6-9 秒(快速蒙太奇/连续快摇+柔和转场):镜头快速掠过古钟楼,花钟少女站在停转的巨大花钟前手持怀表;切森林,鹿角树灵缓缓抬头,角上花叶微光。镜头四 9-12 秒(大全景/低角度仰拍+缓慢跟随):巨大云鲸从云海深处游过,背上漂着小岛、藤蔓和发光花朵,主角抬头仰望,画面展示世界的尺度。画外音(英语):「A silent world is waiting to sing again.」镜头五 12-14 秒(中近景/蒙太奇手持轻快跟拍):迷路邮差背着巨大邮包跑过草坡,信件和地图飞出;星灯商人在浮岛集市点亮星灯,摊位挂满风铃、小瓶子和发光种子。镜头六 14-18 秒(中景/缓慢环绕):古遗迹里,沉默石像男孩半跪在石门前,身上覆着苔藓与小花,眼睛微微亮起,石像裂缝透出柔光。镜头七 18-22 秒(中近景/缓推):月光湖面泛起涟漪,月光湖女巫从湖边现身,月白长袍与湖水般的长发轻轻流动,她手持水晶灯望向主角。女巫西班牙语台词:「El viento recuerda lo que tú has olvidado.」(风记得你遗忘的一切。)镜头八 22-26 秒(特写/快推):天色骤暗乌云翻卷,风暴王子立于破碎高塔之上,深蓝披风猎猎,银发冠闪光,周身环绕风刃与碎羽。王子西班牙语台词:「No todos los mundos merecen despertar.」(不是每个世界都配得上苏醒。)镜头九 26-30 秒(全景/身后缓推,标题定格):风之旅人与小风灵立于崖边,面向云海尽头的发光岛屿。风重新吹起,草木、风铃、云鲸与远处群岛同时泛起柔光。屏幕浮现游戏标题:Island of Whispers。画外音(英语):「Find the fragments. Awaken the islands.」
参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图
13 张角色与场景设定图
Seedance 2.5 成片:台词节奏好,整体符合预告片调性。上一代因素材过多不支持生成。
演示

亮点演示下:多模态参考十连

最后一组亮点全在「参考」二字上。面对复杂的生成与编辑任务,模型对多模态指令的响应更准,手册把升级拆成十项,按四个方向分组:创意与元素的解构(BGM 分离、迁移创意、无缝转场、多宫格分镜)、精细编辑的控制(局部消除、空间视角)、感官维度的扩展(音色参考、多人参考)、影视工业的纵深(绿幕编辑、白模控制)。逐项看:

① BGM 分离:抽走音乐,留下人声和字幕

原视频人声和音乐糊在一起,想二创只能整段作废,现在一句提示词就能让模型删掉音乐底、只留人物说话声,画面里的原字幕还完好无损。用法:直接写「去除背景音乐」「删除 BGM,只保留人物对白」。适合影视二创重新铺乐、Vlog 采访提纯人声、换语言配音重构。

官方案例 · 去 BGM 加阳光
Remove the background music in @Video 1, retaining the clear human voice. At the same time, reference @Image 1 @Image 2 @Image 3 to add the effect of sunlight shining into the room to the video, ensuring natural brightness and realism. And remove the watermark in the video.
中文对照:去除 @视频1 中的背景音乐,保留清晰的人声。同时参考 @图1 @图2 @图3 给视频加入阳光照进室内的效果,保证亮度自然真实。并去除视频中的水印。
参考图参考图参考图
参考图:阳光效果
原片:带 BGM 的室内视频。
Seedance 2.5 成片:BGM 已除、人声保留,还按图加了阳光。
上一代对比:没有执行去 BGM 指令。

② 迁移创意:从抄动作升级到抄「感觉」

过去传参考视频,AI 只学到表面肢体动作,原片的高级运镜和光影氛围全丢。2.5 实现三层迁移:精准提取镜头语言(推拉摇移、手持晃动的物理轨迹直接继承)、深度复刻情绪与氛围(文艺片清冷感、赛博霓虹的色彩逻辑)、爆款「网感」逻辑迁移(短视频的视觉节奏和排版逻辑)。适合把看中的爆款拍法直接搬到自己素材上。

官方案例 · 划手机创意迁移
Please reference the phone-swiping action in @Video 1, retain the phone style, and center around the visual content in @Image 1, @Image 2, @Image 3, @Image 4, and @Image 5 to create a short video with a coherent visual style and creative expression. The overall style should have a sense of premium quality, creativity, and social media spread; the visuals should be clean, detail-rich, and color-coordinated, avoiding clutter and abrupt changes. Please ensure that the subject elements in the video are stable and consistent, and do not randomly change the core content, spatial structure, and visual style from the reference images.
中文对照:请参考 @视频1 划动手机的动作,保留手机样式,围绕 @图1 至 @图5 的视觉内容创作一条视觉风格连贯、有创意表达的短视频。整体风格要有高级感、创意感和社交媒体传播感;画面干净、细节丰富、色彩协调,避免杂乱和突兀变化。请确保视频中主体元素稳定一致,不随意改变参考图的核心内容、空间结构和视觉风格。
参考图参考图参考图参考图参考图
参考图:五张视觉内容
参考视频:划手机动画。
Seedance 2.5 成片:动作流畅、交互稳定、构图正确。
看另外两条迁移案例(柴犬表情、仓鼠变身)
柴犬神态迁移
Please reference the cute dynamic expressions, demeanor, actions, and creative mode of the Shiba Inu in @Video 1 to create a short video of the black Shiba Inu in @Image 1, requiring an accurate capture of the cute and lively state of the Shiba Inu in the reference video.
中文对照:请参考 @视频1 中柴犬可爱的动态表情、神态、动作及创意模式,为 @图1 的黑色柴犬创作一条短视频,要求精准捕捉参考视频中柴犬可爱活泼的状态。
参考图
参考图:黑色柴犬
参考视频:俏皮柴犬。
官方成片:黑柴犬复刻同款神态。
仓鼠变身 meme
Reference the creative form, camera rhythm, and transition method of @Video 1, a cute hamster spins around in place, gorgeously transforming into the masked little mouse of @Image 1. The video needs to maintain the core appearance features, mask shape, and overall recognizability of the masked little mouse in @Image 1 unchanged. Conduct fun dynamic interpretations around the masked little mouse, making it have the exaggerated reactions and cute comical sense of a meme character.
中文对照:参考 @视频1 的创意形式、镜头节奏、转场方式:一只可爱仓鼠原地转了一圈,华丽变身成 @图1 的蒙面小老鼠。视频需保持 @图1 蒙面小老鼠的核心外观特征、面罩形状和整体辨识度不变,围绕它做有趣的动态演绎,让它具有 meme 角色的夸张反应和可爱喜感。
参考图
参考图:蒙面小老鼠
参考视频:变身创意原片。
官方成片:蒙面小老鼠 meme。

③ 局部消除与编辑:从「全部重画」到「指哪改哪」

画面整体满意、就角落有杂物?两把刷子:智能无痕消除,认清目标元素和环境的关系,边擦边按物理规律和原光影补全背景;定向元素替换,框选区域配文字指令,新元素无缝融进原片的透视和氛围。适合穿帮镜头抢救、低成本换装换道具、局部背景定制。

官方案例 · 消除画面中的两个人
Remove the woman on the left and the man on the right in @Video 1, leaving only the woman in the middle, while keeping the actions unchanged.
中文对照:把 @视频1 里左边的女子和右边的男子消除掉,只留下中间的女子,动作保持不变。
原片:三人同框走秀。
Seedance 2.5 成片:两侧人物已消除(结尾帽子无预警掉落是残留瑕疵)。
上一代对比:消除失败,人物还在。
看四条圈选编辑演示(去物、换色、换武器、换盔甲)
圈选消除:去掉红框框住的包包和筷子,对已生成视频二次编辑,指哪擦哪。
圈选改色:把框住的物体换成蓝色。
圈选换道具:把角色手里的武器换成一把狙击枪。
高级编辑:把框住的人物服装换成全身盔甲、戴上战帽。

④ 空间视角修改:给已有画面换机位

画质满意但机位平庸?模型能深度理解视频里的三维空间关系,像在场景里重新架了一台虚拟摄影机:俯拍、仰拍、平视、特写、全景按指令重构,人物结构与环境比例保持合理;还能把画面外没拍到的空间细节合理外推补全,动态运镜也能平滑生成。写法:直接下镜头指令,「把平视镜头改成高空俯拍缓推」「把第三人称改成第一人称视角穿越森林」。适合强化压迫感的电影机位、航拍大场面、第一人称沉浸体验。

官方案例 · 多机位改固定正面
Replace the multi-camera switching shots in @Video 1 with a fixed frontal shot, and keep the characters and background unchanged.
中文对照:把 @视频1 的多机位切换镜头替换成正面视角的固定镜头,并保持人物和背景不变。
原片:多角度切换的弹唱演出。
Seedance 2.5 成片:机位统一成固定正面,人物背景未变。
上一代对比。
看第二条视角案例(宇航员第一人称恐怖片)
第三人称改第一人称
Change the character perspective of @Video 1 to the first-person perspective of @Video 2. The astronaut pants heavily and constantly looks around on the barren planet. He suddenly turns his head sharply, and an alien from @Image 1 suddenly appears in front of him at an extremely fast speed and opens its big mouth. The astronaut's screams and the alien's shrieks fill the entire screen, maxing out the horror atmosphere.
中文对照:把 @视频1 的角色视角改成 @视频2 的第一人称视角。宇航员在荒芜星球上粗重喘息、不断环顾四周;他突然猛回头,@图1 的异形以极快速度骤然出现在眼前张开巨口,宇航员的尖叫和异形的嘶吼充满整个画面,恐怖氛围拉满。
参考图
参考图:异形
参考视频一:宇航员近距特写。
参考视频二:第一人称视角样例。
Seedance 2.5 成片:头盔的模糊质感强化了第一人称。
另一版本对比。

⑤ 音色参考升级:情绪也能一起复刻

上传了极有个人特色的参考音频,AI 只学会基础音色、丢了情绪起伏?升级后模型精准提取深层声学特征,生成音频与原声高度一致、辨识度拉满;澎湃演讲感、冷峻播报腔、温柔耳语的情绪张力和流动节奏都能迁移。用法:上传清晰无噪的人声参考,配指令「参考 @音频1 说『我找到你了』,保持低沉悬疑的语气」。适合微短剧强情绪演绎和品牌专属声音的规模化生产。官方案例一口气用了四个音色参考:

官方案例 · 斗兽场对决(四音色参考)
The opening of the picture is a relatively dreamy galaxy, then the camera immediately locks onto a star and pushes forward. The camera passes through the atmosphere to see the blurred outline of the continent and continues to push forward. On a cloud-shrouded mountain, there is an arena, the shape of the arena references @Image 1 and @Image 2, and the cheering voices of a boiling crowd can be heard.
Shot 1:Camera: Aerial dive into the circular arena, brushing past the boiling crowd in the stands, and freezing in the center.Picture: The referee @Image 5 stands on a high platform at the edge of the arena, raising his hand and pressing it down.Lines: "Humorous Shark Man: Lan , Sister who needs the most technique: Chizha , This battle will determine life and death! Begin!"Cut: @Image 6 and @Image 10 step on the clouds looking down from the air. @Image 10 says: "Lord Guan Sheng, what do you think", the timbre references @Audio 4. @Image 6 replies: "Alas, let's wait quietly for the result", the timbre references @Audio 3.Cut: Special effects: A red-orange double dragon halo suddenly bursts bright, a pillar of light cleaves down in the center of the arena, the whole audience cheers like a tsunami, the pillar of light dissipates and @Image 3 appears. At the same time, a pillar of light composed of many phantom orange-yellow weapons also falls on the arena, the pillar of light dissipates and @Image 9 appears.
Shot 2:Picture: The flame prosthesis on @Image 3's left arm roars with crimson flames, transforming into a phantom shark, kicking the ground and violently charging forward, the ground shattering and kicking up stone debris. Lines: "Break the void", the timbre references @Audio 1.Action: He dashes at high speed, the flames dragging out a long red afterimage trail, holding double swords @Image 13 in his hands and charging forward.Behind @Image 9, five giant transparent orange-yellow weapons (bow, spear, dagger-axe, halberd, and shu), four of which reference @Image 12, suddenly open like a peacock's tail, crossing to block, fire and blades colliding to burst out dazzling light. Lines: "Shu, spear, dagger-axe, halberd, and bow open together", the timbre references @Audio 2.Special effects: The impact point bursts into a circular shockwave, with sparks and red blade light splashing in all directions.Lines for @Image 3: "A mere Chizha, is also worthy of blocking my void-breaking strike?!" The timbre references @Audio 1.
Shot 3:Picture: @Image 9 sneers coldly, the five weapons behind her detach from her body, hovering in the air to transform into a blade storm, strangling towards @Image 3 from all directions.Action: @Image 3 backflips to dodge, the flame prosthesis slashing, knocking away the incoming blades one by one, every block bursting with sparks and trails; the camera does a 360-degree slow-motion orbit to capture him leaping and spinning.Special effects: The screen is filled with dancing transparent blade afterimages interwoven with crimson flames, under slow motion the sparks are like a galaxy.Lines for @Image 9: "A mere Laojiumen jungler dares to be arrogant! I'm getting angry", the timbre references @Audio 2.
Shot 4:Picture: Both sides pull away and look at each other. @Image 3's whole body ignites with crimson karmic fire, condensing into a giant flame shark; @Image 9's five weapons merge into one, condensing into a glowing giant natal divine bow, raised high overhead, gathering a world-shaking arrow.Spectator reaction: The main audience in the stands @Image 4, @Image 7, and @Image 8 all stand up with different expressions (shown with quick cuts, the shot sizes of the cuts must be different). @Image 11 are background audience characters in the stands.Action: Both roar and charge at the same time, the fire shark roars and pounces, the giant arrow splits into countless small arrows , colliding dead center!Special effects: A screen-devouring white-gold massive explosion erupts in the center of the arena, the shockwave flipping up dust, the light drowning the camera.Lines (@Image 3 and @Image 9, roaring simultaneously): "Deciding victory and defeat , is right now!"
中文对照:画面开场是一片梦幻星河,镜头立即锁定一颗星球推进,穿过大气层看到大陆的模糊轮廓继续推进。云雾缭绕的高山上有一座斗兽场,形制参考 @图1 和 @图2,能听到人群沸腾的欢呼声。 镜头一:镜头航拍俯冲进环形斗兽场,掠过沸腾的看台人群,定格中央。画面:裁判 @图5 站在场边高台上抬手下压。台词:「幽默鲨鱼人:蓝,最需要技术的姐姐:赤刹,此战将决生死!开始!」切镜:@图6 和 @图10 踏云从空中俯瞰。@图10 说:「关胜大人,您怎么看」,音色参考 @音频4。@图6 答:「唉,静待结果吧」,音色参考 @音频3。切镜:特效,红橙双龙光环骤亮,一道光柱劈落场地中央,全场欢呼如潮,光柱散去 @图3 现身;同时一道由众多虚影橙黄兵器组成的光柱也落在场上,光柱散去 @图9 现身。 镜头二:@图3 左臂的火焰义肢咆哮着腾起赤红火焰,化作虚影鲨鱼,蹬地猛冲,地面碎裂溅起石屑。台词:「破空」,音色参考 @音频1。动作:他高速突进,火焰拖出长长的红色残影轨迹,双手持 @图13 的双刀前冲。@图9 身后五件巨大的透明橙黄兵器(弓、矛、戈、戟、殳,其中四件参考 @图12)骤然如孔雀开屏般展开、交叉格挡,火与刃相撞迸出刺目光芒。台词:「殳矛戈戟弓齐开」,音色参考 @音频2。特效:撞击点炸开环形冲击波,火星与红色刀光四溅。@图3 台词:「区区赤刹,也配挡我破空一击?!」音色参考 @音频1。 镜头三:@图9 冷笑,身后五件兵器脱离本体悬空化作刃风暴,从四面八方向 @图3 绞杀。动作:@图3 后空翻闪避,火焰义肢劈砍,将袭来的刀刃逐一击飞,每次格挡都爆出火花与轨迹残影;镜头做 360 度慢动作环绕捕捉他的腾跃旋身。特效:满屏舞动的透明刀刃残影与赤红火焰交织,慢动作下火星如星河。@图9 台词:「区区老九门打野也敢张狂!我要生气了」,音色参考 @音频2。 镜头四:双方分开对视。@图3 全身燃起赤红业火,凝成巨大火焰鲨鱼;@图9 五件兵器合而为一,凝成发光的巨大本命神弓高举过顶,蓄起惊世一箭。观众反应:看台上的主要观众 @图4、@图7、@图8 神情各异地起身(快切呈现,切镜景别必须不同),@图11 是看台背景观众。动作:双方同时怒吼冲锋,火鲨咆哮扑出,巨箭分裂成无数小箭,正中相撞!特效:斗兽场中央爆出吞屏的白金巨爆,冲击波掀翻尘土,光芒淹没镜头。台词(@图3 与 @图9 同时怒吼):「决胜负,就在此刻!」
参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图
13 张角色、场景与兵器参考图
官方成片:四个角色各说各的音色。
看第二条音色案例(老兵短剧《甘愿堕落》)
硬汉短剧 · 全程原声对白
Picture: No subtitles added throughout, relying purely on visual expressiveness.Sound: No background music (No BGM). The whole film relies on real urban ambient sounds (traffic, engines, hard braking), action sound effects (impact, drawing a gun), and the characters' original voice dialogues.
[00:00 - 00:05][Wide shot / Fast paced cut / Day / Exterior](Ambient sound: noisy traffic under the scorching sun, the panting sound of pushing a car)Picture: Bee Doctor is gritting his teeth, pushing a broken-down G-Wagon in the middle of the road.A passing Rookie looks at him, shaking his head disdainfully and sighing.Rookie: Sigh, no future.Just as Bee Doctor straightens up to retort, an engine roar approaches piercingly!(Sound effect: sharp hard braking sound + dull impact sound)An off-road vehicle violently hits Bee Doctor, and Bee Doctor is instantly sprawled spread-eagle on the hood!
[00:05 - 00:15][Medium shot / Inside car perspective / Day / Exterior]The four people in the car are in a mess. Nameless in the driver's seat is pale.Nameless (shouting in panic): It's over, it's over!Bee Doctor gets up from the hood expressionlessly and knocks on the car window.Bee Doctor: ID.The window rolls down, Nameless is stunned, instantly overjoyed, turning around and shouting to Mai Xiaoshu and Uluru in the back seat.Nameless: It's Bee Doctor! Our old captain!Bee Doctor recognizes them and waves his hand helplessly.Bee Doctor: You brats... Hurry up and go, be careful.
[00:15 - 00:22][Close-up / Day / Exterior]Just as the off-road vehicle is about to leave, Weilong in the passenger seat pokes his head out, looks Bee Doctor up and down, and sneers.Weilong (provocatively): Old Bee, how did you end up like this?Mai Xiaoshu and Uluru in the back seat follow with mindless roaring laughter.(Sound effect: abrupt laughter, ambient sound instantly quiets down)The helplessness at the corner of Bee Doctor's mouth instantly freezes, his eyes cooling down in a second, scarily cold.
[00:22 - 00:30][Wide shot to Close-up / Day / Exterior](Sound effect: crisp and clean sound of drawing and cocking a gun)Bee Doctor suddenly draws his pistol and points it straight into the car, bursting with tactical coercion.Bee Doctor (sternly): Everyone out of the car! Hands on your heads!The smiles of the four people in the car disappear instantly, scared into stumbling out of the car and obediently squatting in a row by the side of the road.Nameless (anxiously pointing at Weilong defending): He grabbed the steering wheel!Bee Doctor puts away the gun, condescendingly glances at Weilong, raises his chin, arrogant yet self-deprecating.Bee Doctor: Watch closely, rookie, I'm not doing miserably, it's called willing degradation.
中文对照:画面全程不加字幕,纯靠视觉表现力。声音:无背景音乐,全片靠真实的城市环境音(车流、引擎、急刹)、动作音效(撞击、拔枪)和角色原声对白。 00:00-00:05【全景/快节奏剪辑/白天/外景】(环境音:烈日下嘈杂车流、推车的喘息声)画面:蜂医咬着牙在马路中间推一辆抛锚的大 G。路过的雏儿看着他,不屑地摇头叹气。雏儿:「唉,没前途。」蜂医刚直起身要回嘴,一声引擎轰鸣刺耳逼近!(音效:尖锐急刹 + 沉闷撞击)一辆越野车狠狠撞上蜂医,他瞬间大字型趴在引擎盖上! 00:05-00:15【中景/车内视角/白天/外景】车里四人乱作一团,驾驶座的无名脸色煞白。无名(慌张大喊):「完了完了!」蜂医面无表情地从引擎盖上爬起来,敲了敲车窗。蜂医:「证件。」车窗摇下,无名一愣,瞬间大喜过望,回头冲后座的麦小树和乌鲁鲁喊。无名:「是蜂医!我们老队长!」蜂医认出他们,无奈摆手。蜂医:「你们这帮兔崽子……赶紧走吧,注意安全。」 00:15-00:22【特写/白天/外景】越野车正要离开,副驾驶的委龙探出头,上下打量蜂医,嗤笑。委龙(挑衅):「老蜂,你怎么混成这样了?」后座的麦小树和乌鲁鲁跟着没心没肺地哄笑。(音效:突兀的哄笑,环境音瞬间安静)蜂医嘴角的无奈瞬间凝固,眼神一秒冷下来,冷得吓人。 00:22-00:30【全景转特写/白天/外景】(音效:干脆利落的拔枪上膛声)蜂医突然拔出手枪直指车内,战术压迫感爆棚。蜂医(厉声):「都给我下车!手抱头!」车里四人的笑容瞬间消失,吓得跌跌撞撞下车,乖乖蹲成一排。无名(焦急地指着委龙辩解):「是他抢方向盘!」蜂医收枪,居高临下瞥了委龙一眼,抬起下巴,傲慢又自嘲。蜂医:「看清楚了,雏儿,我这不叫混得惨,这叫甘愿堕落。」
参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图
10 张角色与场景参考图
官方成片:短剧片段,台词情绪完整。

⑥ 多人参考升级:同框不再长「双胞胎脸」

多人同框的老毛病:五官互相「串味」、服装张冠李戴、肢体交叠成「缝合怪」。升级后三件事:角色细节精准绑定(每个人的脸、五官、气质独立解析,不串脸);复杂空间交互(拥抱、对视、激烈对抗的遮挡和互动逻辑正确);专属穿搭独立控制(衣服材质配饰也各归各,不互相污染)。用法:分别上传各角色参考图,指令里写明动作关系,「角色 A 和角色 B 坐在篝火旁交谈,A 递给 B 一杯水」。适合多角色叙事分镜和双模特品牌广告。规模最大的案例一次管住 12 张参考图:

官方案例 · 婚庆宣传片(12 图定位)
Please generate an aesthetic wedding promotional video of about 20–30 seconds based on the materials I provided. The bride in Image 1 and the groom in Image 2 are the main characters; bridesmaids in Image 3 (generate at least 3), groomsmen in Image 4 (generate at least 3); the bride's mother in Image 5, the bride's father in Image 6, the groom's mother in Image 7, the groom's father in Image 8; Images 9 and 10 are the wedding venue backgrounds, and Images 11 & 12 are guests, matched with the music I provided. The shots should be presented in sequence: empty shots of the venue atmosphere → aesthetic entrance of the couple → companionship of bridesmaids and groomsmen → emotional moments of both parents → the climax ceremony of exchanging rings or embracing → a warm and perfect ending. It is required that the characters reference the corresponding images, with clear identities and no face-mixing, accurate quantities, and clear primary and secondary roles. The characters must blend naturally with the venue. The overall feel should be cinematic, romantic, and premium, with the rhythm advancing with the music, suitable for brand promotion of a wedding planning company.
中文对照:请基于我提供的素材生成一支约 20-30 秒的唯美婚礼宣传视频。@图1 的新娘和 @图2 的新郎是主角;@图3 的伴娘(至少生成 3 位)、@图4 的伴郎(至少生成 3 位);@图5 新娘母亲、@图6 新娘父亲、@图7 新郎母亲、@图8 新郎父亲;@图9 和 @图10 是婚礼现场背景,@图11、@图12 是宾客,配合我提供的音乐。镜头按序呈现:场地氛围空镜 → 新人唯美入场 → 伴娘伴郎相伴 → 双方父母的情感瞬间 → 交换戒指或拥抱的仪式高潮 → 温馨圆满收尾。要求人物参考对应图片、身份清晰不混脸、数量准确、主次分明,人物与场地自然融合,整体电影感、浪漫、高级,节奏随音乐推进,适合婚庆策划公司品牌宣传。
参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图参考图
12 张参考图:新人、双方父母、伴娘伴郎、场地与宾客
官方成片:身份各归各位,数量准确。
看另外两条多人案例(四人街舞、奇幻穿越八人组)
复古卡通四人街舞
Vintage American cartoon animation urban street (reference Image 5) joyous short film, the four young characters' images, outfits, and facial features respectively reference @Image 1, @Image 2, @Image 3, and @Image 4. The four people dance a synchronized group dance on the vintage pedestrian street, their limb movements are neat and coherent, the spacing between the characters remains fixed, and the contrasting cool and warm neon light and shadow of the street are unified throughout. Frontal fixed camera position + slow panning camera movement, every detail of the characters' clothing, facial features, and body shape highly matches the reference images throughout, the actions are smooth without clipping or distortion, vintage cartoon animation texture, duration 15 seconds.
中文对照:复古美式卡通动画城市街道(场景参考 @图5)的欢乐短片,四位年轻角色的形象、穿搭和五官分别参考 @图1、@图2、@图3、@图4。四人在复古步行街上跳整齐划一的团体舞,肢体动作整齐连贯,角色间距保持固定,街道冷暖对比的霓虹光影全程统一。正面固定机位 + 缓慢平移运镜,角色服装、五官、体型的每个细节全程与参考图高度一致,动作流畅无穿模无变形,复古卡通动画质感,时长 15 秒。
参考图参考图参考图参考图参考图
四位角色与街景参考
官方成片:四人身份稳定、群舞整齐。
奇幻穿越 15 秒(8 图 + 特效参考)
Generate a fantasy story short video with a total duration of 15 seconds, cinematic feel, focusing on showing the magnificent grandeur of the different world space after crossing over. The crossover intro should be brief, and the confrontation scene of the four-person squad should be restrained and stop at the right point. Strictly follow the timeline below:
0–1.5s: The little girl [@Image 1, the exact same person throughout without deformation] is sucked in and crosses over by the floating, glowing painting in @Image 2 , this crossover special effect must strictly reference Video 1, reproducing its crossover texture and timing of light streaking, color flowing, and accelerating impact, passing through quickly.1.5–3.5s: The girl emerges from the light tunnel and lands on the steps of the different world in @Image 3. The camera shoots from a low-angle looking up, following her as she walks up the steps, showing the towering scale of the steps. The landing point should touch the ground and not float in the air.3.5–8s (Visual climax · Main body): The camera pulls back significantly, tilts up, and slowly orbits, laying out the full view of the floating island universe divine realm in @Image 4 , countless floating islands, surging seas of clouds, golden divine radiance, suspended temples, and waterfalls, as magnificent as Asgard. The girl stands tiny among them, fully showing the shock of scale.8–10.5s: Cut to the little girl's subjective point of view (POV), her gaze slowly sweeps across this magnificent landscape, with a strong sense of immersion.10.5–13.5s: The patrol squad composed of @Image 5, @Image 6, @Image 7, and @Image 8 appears, approaching alertly, forming a semi-encircling group portrait , the captain @Image 7 steps forward to examine, the red-haired girl @Image 8 is curious with a protective intent, and the couple @Image 5 and @Image 6 reveal hostility (the group scene is restrained, no individual close-ups).13.5–15s: The girl is surrounded in the center, with the magnificent divine realm as the background, concluding and freezing on a tense confrontation atmosphere.
Requirements: Shot 1's crossover special effects closely match @Video 1, with smooth and unbroken lighting effects; the different world (@Image 3 steps, @Image 4 full view) is magnificent and spectacular, with a shocking scale, and the camera movements are expansive and have a sense of breathing, serving as the visual main body of the whole film; the girl remains the exact same person throughout without deformation, her landing point touches the ground and does not float; the four squad members strictly correspond to @Image 5, @Image 6, @Image 7, and @Image 8, without mixing faces, their stances are perceptible but their performances are restrained; overall cinematic texture, magnificent colors, no clipping and no distortion; the total duration is strictly controlled to 15 seconds.
中文对照:生成一条总时长 15 秒、电影感的奇幻故事短视频,重点展现穿越后异世界空间的宏伟壮阔。穿越开场要简短,四人小队的对峙收在点上。严格按以下时间线: 0-1.5 秒:小女孩【@图1,全程同一人不变形】被 @图2 悬浮发光的画卷吸入穿越,穿越特效必须严格参考 @视频1,复刻其光线拉丝、色彩流动和加速冲击的质感与时机,快速通过。1.5-3.5 秒:女孩从光之隧道里冲出,落在 @图3 的异世界台阶上。镜头低角度仰拍,跟随她拾级而上,展现台阶的巍峨体量。落点要踩实地面,不能悬空。3.5-8 秒(视觉高潮·主体):镜头大幅后拉、上仰并缓慢环绕,铺开 @图4 浮岛宇宙神域的全貌,无数浮岛、翻涌云海、金色神辉、悬浮神殿与瀑布,如阿斯加德般壮丽。女孩渺小地立于其间,充分呈现尺度的震撼。8-10.5 秒:切小女孩主观视角(POV),目光缓缓扫过这片壮丽景观,沉浸感拉满。10.5-13.5 秒:由 @图5、@图6、@图7、@图8 组成的巡逻小队现身,警惕逼近呈半包围群像,队长 @图7 上前查看,红发少女 @图8 好奇中带保护欲,情侣 @图5 @图6 面露敌意(群像收敛克制,不给单人特写)。13.5-15 秒:女孩被围在中央,以壮丽神域为背景,在剑拔弩张的对峙氛围中收尾定格。 要求:镜头一的穿越特效紧贴 @视频1,光效流畅不断裂;异世界(@图3 台阶、@图4 全景)宏伟壮观、有尺度震撼,运镜大开大合有呼吸感,是全片视觉主体;女孩全程同一人不变形、落点踩实不悬空;四名队员严格对应 @图5-@图8 不混脸,立场可感知但表演克制;整体电影质感、色彩瑰丽、无穿模无变形;总时长严格控制在 15 秒。
参考图参考图参考图参考图参考图参考图参考图参考图
主角、画卷、异世界与巡逻小队参考
参考视频:穿越特效样片。
官方成片:15 秒奇幻穿越。

⑦ 无缝绿幕编辑:合成不再有「贴图感」

影视工业级抠像来了:高精度分割能沿复杂轮廓干净剥离人和物、不留残影;动态场景里人物与背景的相对透视和运动轨迹保持帧级稳定,不闪烁跳动。用法:上传绿幕素材,写清要替换成的新场景,或直接给参考图/视频,「把背景换成中世纪魔法森林,光影神秘」「参考教室图,把绿幕里的人物自然融进去」。适合奇幻科幻低成本大场面、虚拟演播室口播、现代人「穿越」名画和二次元的创意互动。

官方案例 · 导弹三素材合成
Composite @Video 1 and @Video 2 into @Video 3. As the camera pushes in, the missile of @Video 1 streaks across the sky and explodes in the front, the explosion generates the chain reaction of @Video 2, and black smoke spreads.
中文对照:将 @视频1 和 @视频2 合成到 @视频3 里。镜头推进时,@视频1 的导弹从空中划过,在前方发生爆炸,爆炸产生 @视频2 的连锁反应,黑烟蔓延。
合成素材一:导弹划空。
合成素材二:爆炸连锁。
合成素材三:主场景。
Seedance 2.5 成片:三段素材融成一体,基本没有合成感。
上一代对比:绿幕素材融合尚可,但场景延展与原景不匹配。
看第二条绿幕案例(天气 + 丧尸双合成)
天气与丧尸合成
Composite the weather of @Image 1 into @Video 1 with natural angle changes. When the camera tilts down to the city road surface, composite the zombies of @Video 2 onto the city road surface, just passing by the camera. The weather is half cloudy and half sunny, with a little sunlight piercing through the dark clouds, shining a hazy layer on the ground, forming a weird sense of unreality and absurdity.
中文对照:将 @图1 的天气合成进 @视频1,角度变化自然;镜头下摇到城市路面时,把 @视频2 的丧尸合成到路面上,恰好从镜头前经过。天气半阴半晴,一点阳光穿透乌云,在地面照出朦胧的一层,形成诡异的不真实感和荒诞感。
参考图
参考图:天气
参考视频一:绿幕城市街景。
参考视频二:绿幕丧尸动态。
Seedance 2.5 成片:绿幕素材完美融合。
上一代对比:融合失败,画面残留绿调。

⑧ 专业白模控制:给 3D 工作流的工业级入口

专为影视从业者和 3D 动画师准备:镜头语言还原(推拉摇移到复杂环绕升降,模型能理解白模传达的三维运动轨迹)、空间与构图锁定(白模场景里物体间的遮挡逻辑清晰解析)、角色站位与运动轨迹控制(按排好的「走位」演出)。配合前面说的 Maya/Blender 插件使用,适合影视动画前期预演和多角色大调度。写法公式在前面白模一节,这里看三条演示:

官方案例 · 健身对比科普(白模渲染)
Generate a vertical (9:16) fitness action comparison popular science video, adopting a top-and-bottom split-screen comparison layout, strictly retaining the two sets of action sequences of the "correct action" and "wrong action" in it. Render @Video 1 for the upper half of the screen: the character references @Image 2 Beauty B, executing the wrong action in the video, holding the dumbbell from Image 4; lower half of the screen: the character references @Image 1 Beauty A, executing the correct standard action in the video, holding the dumbbell from Image 3. The appearance, figure, and clothing of the two beauties strictly correspond to Image 2 and Image 1 respectively, with no face-mixing, no swapping of actions, and their respective identities running through the entire process. The actions strictly follow the video template, the difference between wrong and standard is clearly distinguishable, limbs are coherent without deformation, and the dumbbells are always held in the hands, consistent in shape, without clipping or drifting. The scene is uniformly the gym environment in Image 5, and the lighting and scene style of the upper and lower screens are consistent. Vertical screen composition, the characters are centered, actions are fully in the frame, mainly frontal medium shots, and can slowly push in to the force-exerting parts. The boundary of the top-and-bottom split screen is clear, with an overall professional fitness popular science texture, clear contrast, and the characters, dumbbells, and scene blend naturally with unified lighting.
中文对照:生成一支竖版(9:16)健身动作对比科普视频,采用上下分屏对比版式,严格保留白模视频里「正确动作」和「错误动作」两组动作序列。渲染 @视频1:画面上半部分角色参考 @图2 的美女 B,执行视频中的错误动作,手持 @图4 的哑铃;下半部分角色参考 @图1 的美女 A,执行视频中的正确标准动作,手持 @图3 的哑铃。两位人物的外貌、身材、服装分别严格对应 @图2 和 @图1,不混脸、不互换动作,各自身份贯穿全程。动作严格跟随视频模板,错误与标准的差异清晰可辨,肢体连贯不变形,哑铃始终握在手中、形态一致、不穿模不漂移。场景统一为 @图5 的健身房环境,上下屏光线与场景风格一致。竖屏构图、人物居中、动作完整入画,以正面中景为主,可向发力部位缓慢推近。上下分屏边界清晰,整体专业健身科普质感,对比清晰,人物、哑铃与场景自然融合、光照统一。
参考图参考图参考图参考图参考图
两位人物、两副哑铃与健身房参考
参考视频:上下分屏白模。
Seedance 2.5 成片:白模按提示词精准渲染,特效与场景参考图还原到位。
上一代对比:白模渲染效果不理想,与原视频偏差大。
看另外两条白模演示(海浪灯塔、四段接力渲染)
海浪灯塔 · 细颗粒渲染
Render the white model animation of [Video 1] into the final film.
0–7s: Realistic ocean night scene rendering. Dark gray-blue stormy sea surface, giant waves continuously surging, overall adopting cold tones and low-saturation cinematic atmosphere; waves have real volume sense, fluid details, and foam layers, the floating red and white striped lighthouse maintains metal and worn paint material, naturally undulating and tilting with the waves; the sky is full of heavy dark clouds, accompanied by fine rainfall, faint lightning fill light occasionally appears in the distance; overall adopting HDR lighting, volumetric fog, real environment reflection, wet highlights, and dynamic shadows, strengthening the lighting changes of the waves and the lighthouse, ensuring the continuity and stability of materials and lighting during the camera movement.
[Scene Processing]Take the stormy sea area as the sole scene, the sea surface occupies the main frame, the distant horizon naturally connects with the dark clouds; the waves maintain a continuous flowing and undulating rhythm, the lighthouse always floats near the center of the sea surface, generating real pitching and rolling motions driven by the surge, without deformation or size change; raindrops, wave splashes, and environmental fog maintain a unified wind direction and physical logic, the camera can slightly float up and down following the waves, enhancing immersion.
[Overall Conclusion]Keep the stormy sea area scene unchanged, the overall rendering style, lighthouse material, wave fluid effects, and lighting performance remain consistent; character rendering: remains unchanged (no characters).Audio processing: Retain continuous wave crashing sounds, howling wind sounds, rain sounds, and occasional low thunder, the environmental sound naturally changes with the camera distance, do not add background music or voice-over.
中文对照:将 @视频1 的白模动画渲染成成片。0-7 秒:写实海洋夜景渲染。暗灰蓝的风暴海面,巨浪持续翻涌,整体冷色调低饱和电影氛围;海浪要有真实的体积感、流体细节和泡沫层次;漂浮的红白条纹灯塔保持金属与磨损油漆材质,随浪自然起伏倾斜;天空乌云密布伴细雨,远处偶有微弱闪电补光;整体采用 HDR 光照、体积雾、真实环境反射、潮湿高光和动态阴影,强化海浪与灯塔的光影变化,保证运镜过程中材质与光照的连续稳定。 【场景处理】以风暴海域为唯一场景,海面占画面主体,远处海平线与乌云自然衔接;海浪保持连续的流动起伏节奏,灯塔始终漂浮在海面中央附近,被涌浪推动产生真实的俯仰与摇摆,不变形、不改变大小;雨滴、浪花与环境雾保持统一的风向和物理逻辑,镜头可随海浪轻微起伏增强沉浸感。 【整体收束】保持风暴海域场景不变,整体渲染风格、灯塔材质、海浪流体效果与光照表现一致;角色渲染:保持不变(无角色)。声音处理:保留连续的海浪拍击声、风声、雨声和偶尔的低沉雷声,环境音随镜头距离自然变化,不加背景音乐和旁白。
参考图参考图参考图参考图
参考图:灯塔与海面材质
参考视频:灯塔白模。
官方成片:风暴海域渲染结果。
四段接力渲染(潜水员/宇航员/神将/月面)
The actions throughout reference the actions of @Video 1, fixed camera throughout, no background music, with sound effects.The first frame of the rendered video references @Image 1, producing a 0-5 seconds video. A heavy diver with a retro copper helmet drags a yellow-green glowing spherical sample cabin, swimming forward slowly, and the helmet headlamp shoots out a clear Tyndall beam; the giant ice-blue glowing jellyfish in the upper right of the picture does slow breathing-style contractions and expansions, its tentacles floating gently with the water flow; schools of glowing small fish cruise at a constant speed around, and small jellyfish slowly float and sink;
Hard cut, fixed camera throughout, the first frame of the rendered video references @Image 2, 6-10 seconds video. A lonely astronaut wearing a weathered futuristic spacesuit trudges laboriously from left to right on the cracked and barren wasteland. The character's body leans deeply forward, struggling against massive gravity and hurricane-level headwinds. The long, damaged mechanical cables and fabric behind him flutter violently backward. In the foreground, sand, dust, and floating crushed stones fly rapidly past the camera. In the background, a giant, glowing flaming black hole accretion disk rotates slowly, and flashes with intense orange light. The cape and the black cables on his back are pulled backward by the fierce wind. Surrounding crushed stones suspend, slowly rotate, and sweep past the front of the camera, while the distant black hole accretion disk rotates at high speed.
Hard cut, fixed camera throughout, the first frame of the rendered video references @Image 3, 11-15 seconds video. Every step of the protagonist is very heavy, stepping on the cracked earth and kicking up dust; the hem of the robe, hair, and shoulder fabric are blown backward by the fierce wind. The chain in the hand is pulled tight, and the giant cyan-green divine weapon behind is dragged forward on the ground, rubbing against the ground to splash crushed stones and green energy sparks. The golden sunset in the distance provides a strong backlight, the ground runes slowly flicker, and orange-red light shines through the cracks. The giant cyan-green celestial body in the sky slowly rotates, golden rune halos flow, and floating crushed stones and broken statues drift slowly around the celestial body.
Hard cut, fixed camera throughout, the first frame of the rendered video references @Image 4, 16-20 seconds video. Every step of the protagonist astronaut carries a sense of low gravity, kicking up fine lunar dust when the soles of the feet land; the lunar dust slowly floats and then falls. The space capsule behind him shakes slightly when dragged, its bottom rubbing against the lunar surface, bringing up a gray dust track; the rope remains taut at all times and slightly vibrates. The crashed spaceship wreckage and satellite in the distance remain stationary, and the foreground rocks do not move. The giant Jupiter in the sky slowly rotates on its axis, the slender planetary rings flow slightly, and the starry sky flickers faintly.
中文对照:全程动作参考 @视频1,全程固定镜头,无背景音乐,带音效。 渲染视频首帧参考 @图1,产出 0-5 秒:戴复古铜盔的重装潜水员拖着黄绿发光的球形样本舱缓慢向前游,头盔头灯射出清晰的丁达尔光束;画面右上巨大的冰蓝发光水母做缓慢呼吸式收缩扩张,触手随水流轻摆;发光小鱼群匀速巡游,小水母缓慢浮沉。 硬切,固定镜头,首帧参考 @图2,6-10 秒:孤独的宇航员穿着风化的未来宇航服,在龟裂荒原上从左向右艰难跋涉;身体深深前倾,对抗巨大重力与飓风级逆风,身后破损的机械线缆和织物猛烈向后翻飞;前景沙尘和悬浮碎石快速掠过镜头,背景巨大的燃烧黑洞吸积盘缓慢旋转、闪烁强烈橙光。 硬切,固定镜头,首帧参考 @图3,11-15 秒:主角每一步都极沉重,踩裂大地扬起尘土;袍摆、头发和肩部织物被狂风向后吹起,手中锁链绷紧,背后巨大的青绿色神兵在地上拖行,摩擦地面溅起碎石与绿色能量火花;远处金色落日强逆光,地面符文缓缓闪烁,橙红光从裂缝中透出;天空巨大的青绿色天体缓慢自转,金色符文光环流动,漂浮的碎石与残破雕像绕天体缓慢漂移。 硬切,固定镜头,首帧参考 @图4,16-20 秒:主角宇航员每一步都带低重力感,落脚扬起细腻月尘缓慢飘落;身后拖拽的太空舱轻微晃动,底部摩擦月面拖出灰色尘迹,绳索始终绷紧微颤;远处坠毁的飞船残骸和卫星静止,前景岩石不动;天空巨大的木星缓慢自转,纤细行星环轻微流动,星空微闪。
参考视频:接力白模。
官方成片:四段场景接力渲染。

⑨ 无缝视频转场:两段片之间自动补桥

拼素材硬切伤连贯?传两段视频,模型自动分析画面内容、空间纵深和运动趋势,在中间生成过渡画面补齐缺口:支持变焦转场、旋转换装、俯冲转场等高冲击玩法;原素材保持不动,用动作触发、视线牵引、遮挡切换这些高级视听语言完成衔接。适合创意混剪卡点挑战和剧情时空的高级过渡。四条案例四种玩法:

官方案例 · 水洼霓虹变阳光湖面
Seamlessly connect [@Video 1] and [@Video 2], but do not modify [@Video 1] and [@Video 2]. The perspective of [@Video 1] continues to push in quickly towards the circular neon reflection in the puddle. After the circular reflection fills the screen, seamlessly and naturally transition to [@Video 2]. During the continuous movement of the camera, the blue-purple neon reflection slowly turns into a golden sun reflection, and the puddle texture of the rainy city night gradually changes into the water ripples of a forest lake, with the entire scene changing accordingly. At the same time, keep the position, size, and movement direction of the central circular reflection consistent, and do not change the two uploaded videos.
中文对照:无缝衔接【@视频1】和【@视频2】,但不要修改两段原视频。【@视频1】的视角持续快速推向水洼中的圆形霓虹倒影,圆形倒影充满画面后,无缝自然过渡到【@视频2】。镜头持续运动中,蓝紫色霓虹倒影缓缓变成金色太阳倒影,雨夜城市的水洼质感渐变为森林湖泊的水波涟漪,整个场景随之改变。同时保持中央圆形倒影的位置、大小和运动方向一致,不改变上传的两段视频。
前段素材:雨夜水洼霓虹。
后段素材:森林湖泊。
官方成片:圆形倒影为轴心的无缝过渡。
看另外三条转场案例(红伞变花瓣、纸飞机穿越、水母变星云)
红伞擦镜变花瓣
Seamlessly connect [@Video 1] and [@Video 2], but do not modify [@Video 1] and [@Video 2]. After the red umbrella in [@Video 1] wipes across the camera from left to right and covers the entire screen, seamlessly and naturally transition to [@Video 2], with the red umbrella surface continuing as red petals or red fabric moving away from left to right. As the red area swipes across the screen, the black and white city street is gradually wiped into a colorful spring garden, and the scene changes accordingly. Keep the wiping direction, speed, and the position of the red obscuring object consistent, and do not change the two uploaded videos.
中文对照:无缝衔接【@视频1】和【@视频2】,但不要修改两段原视频。【@视频1】里红伞从左到右擦过镜头盖满全屏后,无缝自然过渡到【@视频2】,红色伞面延续成红色花瓣或红色织物从左向右移出。红色区域扫过画面的同时,黑白城市街道被逐渐擦成缤纷的春日花园,场景随之改变。保持擦拭方向、速度和红色遮挡物的位置一致,不改变上传的两段视频。
前段素材:黑白街道红伞。
后段素材:春日花园。
官方成片:红伞擦镜换世界。
纸飞机穿越时空
Seamlessly connect [@Video 1] and [@Video 2], but do not modify [@Video 1] and [@Video 2]. The white paper airplane thrown by the little girl in [@Video 1] flies towards the camera, gradually becoming larger and covering the entire screen; then seamlessly and naturally transition to [@Video 2], the same white paper airplane flies away from the front of the camera, bringing out the summer country road and the running boy. The paper airplane maintains consistent shape, position, size, and movement direction at the connection point, naturally transitioning from "flying towards the camera" to "flying away from the camera". The scene naturally changes from grayish-yellow, low-saturation war ruins to a bright and warm summer countryside, and the mood transitions from tense and restrained to healing and nostalgic, while not changing the main content of the two uploaded videos. The camera movement remains continuous and smooth, without black screens, frame skipping, hard cuts, flickering, text, logos, or subtitles.
中文对照:无缝衔接【@视频1】和【@视频2】,但不要修改两段原视频。【@视频1】里小女孩掷出的白纸飞机飞向镜头、逐渐变大盖满全屏;随后无缝自然过渡到【@视频2】,同一只白纸飞机从镜头前飞远,带出夏日乡间道路和奔跑的男孩。纸飞机在衔接点保持形状、位置、大小和运动方向一致,自然完成从「飞向镜头」到「飞离镜头」的过渡。场景从灰黄低饱和的战争废墟自然变为明亮温暖的夏日乡野,情绪从紧张压抑过渡到治愈怀旧,同时不改变两段视频的主体内容。镜头运动保持连续平滑,无黑屏、跳帧、硬切、闪烁、文字、logo 或字幕。
前段素材:战争废墟中的纸飞机。
后段素材:夏日乡道。
官方成片:一只纸飞机缝合两个时空。
深海水母变星云
Seamlessly connect [@Video 1] and [@Video 2], but do not modify [@Video 1] and [@Video 2]. In [@Video 1], the glowing jellyfish swarm diffuses, and after blue-purple fluorescent particles fill the screen, seamlessly and naturally transition to [@Video 2], the fluorescent particles gradually turn into nebula dust and starlight, and the deep-sea environment transitions into vast space. The diver's floating posture naturally turns into the astronaut's weightless floating posture, bubbles turn into stardust, and the flashlight beam turns into a nebula light band. The transition uses particle diffusion, halo extension, slow rotation, and spatial stretching; overall dreamy, silky, and shocking, do not change the two uploaded videos.
中文对照:无缝衔接【@视频1】和【@视频2】,但不要修改两段原视频。【@视频1】里发光水母群扩散、蓝紫色荧光粒子充满画面后,无缝自然过渡到【@视频2】,荧光粒子渐变成星云尘埃与星光,深海环境过渡成浩瀚太空。潜水员的漂浮姿态自然转为宇航员的失重漂浮,气泡变星尘,手电光束变星云光带。转场使用粒子扩散、光晕延展、缓慢旋转和空间拉伸;整体梦幻、丝滑、震撼,不改变上传的两段视频。
前段素材:深海水母。
后段素材:太空宇航员。
官方成片:从深海一路漂到星云。

⑩ 多宫格分镜:火柴人分镜图直出成片

专为需要精确还原分镜设计的生产场景准备:分镜参考低门槛,简单线稿或火柴人分镜图就行;提示词按四步写,先说明各宫格代表什么(是角色还是场景)、再引参考图详细立主体设定、然后按故事概要逐格写构图景别运镜动作、最后补全局风格和禁止项。适合影视动画的工业化预演(纸面分镜 1:1 转动态预览)和强逻辑连贯叙事短片。

官方案例 · 九宫格流浪猫短片
[Image 1] is the complete film storyboard script, comprising 9 continuous shots. Each storyboard represents a complete shot, rather than a single character or scene, forming a complete, warm, and healing cinematic narrative.
The middle-aged woman references the three-view drawing of [Image 2], a 40-50-year-old middle-aged female, 3D animated film CG style, fair skin, dark brown short curly hair, big eyes, gentle and kind face, always maintaining a natural smile. Wearing a light yellow knitted cardigan, white inner wear, blue ankle-length skirt, and off-white flat shoes, the overall color scheme is soft and warm, a home-staying mother image, and the character's appearance, clothing, and proportions remain consistent throughout.
The kitten references [Image 3], a rounded chibi orange and white tabby kitten, 3D animated film CG style, big eyes, pink nose, fluffy and soft fur. Wearing a yellow knitted hat, off-white striped top, and ginger-yellow overalls, carrying a green drawstring cloth bag, the overall color tone is warm and healing, a well-behaved and cute image, and the character's fur color, clothing, and proportions remain consistent throughout.
Shot 1 (0-3s)In an urban residential area at dusk, the camera adopts a medium shot from a low angle slowly pushing in. A stray kitten [Image 3] gently jumps onto a street-side garbage can, uses its front paws to lift the lid, and pokes its head into the garbage can to look for food. Residential buildings, streetlights, and trees are visible in the background, warm sunset shines on the street, overall creating a slightly lonely but real life atmosphere.
Shot 2 (3-5s)The camera switches to a close-up, slightly pushing in towards the inside of the garbage can. The kitten grips the edge of the can with both paws, its body leans into the garbage can, earnestly rummaging for food, its ears twitching gently, eyes focused, highlighting the state of hunger and helplessness.
Shot 3 (5-9s)The scene turns to the door of the residential building, the camera adopts a medium fixed shot. A middle-aged woman [Image 2] stands in front of the door holding food, discovering the kitten in the distance. The kitten looks back at the woman, its body slightly alert but not running away, the two establish eye contact for the first time, and the background door plate has a "Food" signboard, enhancing the breath of life.
Shot 4 (9-12s)The camera switches to a full-body medium shot, the kitten stands in the open space of the yard, slowly approaching the door, with hopeful eyes.
Shot 5 (12-16s)The camera lowers to the cat's perspective. The middle-aged woman squats down and gently places the food bowl with cat food in front of the kitten. The kitten immediately lowers its head to eat, the action is natural and coherent, its tail gently swaying, the middle-aged woman watches the kitten with a smile, and the overall picture is full of warmth.
Shot 6 (16-20s)A little later in time, the camera adopts a medium tracking shot. The middle-aged woman slowly walks out of the yard carrying a basket of fresh vegetables, the kitten quietly follows by the woman's side, always keeping a half-step distance, showing that the two have established an initial companionship relationship.
Shot 7 (20-23s)The camera switches to the indoor doorway. The woman slowly opens the door preparing to enter the house, warm yellow indoor light spills out from inside the door, forming a cool and warm light contrast with the dusk outdoors, creating the warm atmosphere of home.
Shot 8 (23-26s)The camera adopts a medium shot from the side and rear. The middle-aged woman turns back, smiles, and gently waves her hand, signaling the kitten to go home together. The kitten stands at the door looking up at the woman, ears perked up, tail gently swaying, showing hesitation before gradually accepting the invitation.
Shot 9 (26-30s)The camera slowly pulls back. The middle-aged woman stands at the door waiting patiently, the kitten finally steps forward and follows the old person into the house. The warm yellow light elongates their silhouettes, the door slowly closes, the picture stays on the warm and quiet courtyard, and the story ends with "the stray cat has a new home."
Adopt Pixar / DreamWorks animated film CG style, characters maintain rounded chibi proportions, PBR physical materials, high-quality global illumination, soft volumetric lighting, cinematic depth of field, warm dusk lighting matched with indoor warm yellow light, the overall color is soft, warm, and healing. Camera movement is natural and smooth, maintaining the cinematic narrative rhythm, character actions conform to real animal behaviors, and time, space, character positions, and lighting remain continuous between shots.
Character face changing, deformation, and proportion errors are prohibited; kitten breed, fur color, and body size changes are prohibited; middle-aged woman clothing, age, and appearance random changes are prohibited; adding irrelevant characters, animals, or props is prohibited.
中文对照:【@图1】是完整的影片分镜脚本,包含 9 个连续镜头。每格分镜代表一个完整镜头(不是单个角色或场景),构成一部完整、温暖、治愈的电影化叙事。 中年女性参考【@图2】的三视图:40-50 岁中年女性,3D 动画电影 CG 风格,皮肤白皙,深棕色短卷发,大眼睛,温和善良的脸,始终保持自然微笑。穿浅黄针织开衫、白色内搭、蓝色及踝长裙、米白平底鞋,整体配色柔和温暖,居家母亲形象,外貌、服装、比例全程一致。 小猫参考【@图3】:圆润 Q 版橘白虎斑小猫,3D 动画电影 CG 风格,大眼睛粉鼻头,毛发蓬松柔软。戴黄色针织帽,穿米白条纹上衣和姜黄背带裤,背绿色抽绳布袋,整体色调温暖治愈,乖巧可爱形象,毛色、服装、比例全程一致。 镜头 1(0-3 秒):黄昏的城市居民区,镜头低角度中景缓慢推进。流浪小猫【@图3】轻轻跳上路边垃圾桶,用前爪掀开盖子,把头探进垃圾桶找食物。背景可见居民楼、路灯和树木,暖色夕阳洒在街道上,整体营造略孤独但真实的生活氛围。镜头 2(3-5 秒):切特写,向垃圾桶内部轻微推进。小猫双爪扒着桶沿,身体探进垃圾桶认真翻找食物,耳朵轻轻抖动,眼神专注,突出饥饿与无助的状态。镜头 3(5-9 秒):场景转到居民楼门口,中景固定镜头。中年女性【@图2】端着食物站在门前,发现远处的小猫。小猫回头看她,身体微微警觉但没有逃走,两者第一次建立目光接触,背景门牌处有「Food」招牌,增强生活气息。镜头 4(9-12 秒):切全身中景,小猫站在院子空地上,缓缓靠近门口,眼神带着期待。镜头 5(12-16 秒):镜头降到猫的视角。中年女性蹲下身,把盛着猫粮的食盆轻轻放在小猫面前。小猫立刻低头进食,动作自然连贯,尾巴轻轻摇摆;女性微笑注视,画面充满温情。镜头 6(16-20 秒):时间稍后,中景跟拍。女性提着一篮新鲜蔬菜缓缓走出院子,小猫安静跟在她身边,始终保持半步距离,表现两者建立起初步陪伴关系。镜头 7(20-23 秒):切室内门口。女性缓缓开门准备进屋,暖黄的室内灯光从门内漫出,与黄昏室外形成冷暖对比,营造家的温暖氛围。镜头 8(23-26 秒):侧后方中景。女性回头微笑、轻轻招手,示意小猫一起回家。小猫站在门口抬头望她,耳朵立起、尾巴轻摆,从迟疑到逐渐接受邀请。镜头 9(26-30 秒):镜头缓缓后拉。女性站在门口耐心等待,小猫终于迈步跟进屋里。暖黄灯光拉长两者的剪影,门缓缓合上,画面停在温馨安静的院落,故事以「流浪猫有了新家」收尾。 采用皮克斯/梦工厂动画电影 CG 风格,角色保持圆润 Q 版比例,PBR 物理材质、高质量全局光照、柔和体积光、电影景深,黄昏暖光配室内暖黄灯光,整体色彩柔和温暖治愈。运镜自然流畅,保持电影叙事节奏,角色动作符合真实动物行为,镜头之间时间、空间、人物位置和光线保持连续。禁止角色换脸、变形和比例错误;禁止小猫品种、毛色、体型变化;禁止中年女性服装、年龄、外貌随机变化;禁止添加无关角色、动物或道具。
参考图参考图参考图
分镜脚本 + 两位主角的设定图
官方成片:九格分镜按序成片。
收尾

结尾:配上提示词写法,两条腿走路

到这里,这份手册就走完了:参数上限、四个新入口、七类写法公式、16 个亮点演示,公式和案例都能直接搬进自己的工作流。日常用的时候按场景查,操作和界面问题回「上手」一节,各类任务的专用公式回「写法」七节,想看某个能力的边界就翻对应的演示。

写法层面还想再进阶一层,单条提示词怎么把素材职责、结束状态、可观察表现写扎实,Seedance 官方另有一份提示词指南,站内有完整解读,两份配合使用刚好补上「通用心法」这块拼图:

配合使用
字节官方 Seedance 2.5 提示词指南:从一句话公式到 30 秒长片,模板全可直接抄
通用提示词写法的官方模板全集:素材职责、阶段化长视频、编辑与延长的标准写法。
两点如实交代:这份手册面向 Dreamina 国际版(dreamina.capcut.com),与国内即梦的功能是否完全对齐,原文档未作说明;文档文本里的「Onmi Reference」为原文误拼,界面实际拼写是 Omni Reference,正文已按界面拼写。
来源
【Dreamina】Seedance 2.5 User GuideDreamina(字节跳动)·dreamina.capcut.com
本站说明
提示词均按官方英文原文收录并配本站中文对照译文(个别圈选案例的中文提示词随官方成片文件名保留);操作步骤与功能说明由英文原文编译;122 条官方演示视频经压缩转码后自托管,图表与截图来自原文档。
↑ 正文结束
✓ 本篇全文免费开放

会员可复制、下载全文 Markdown,还能让你的 AI 订阅和提问。

阅读原文 ↗
📮 新解读第一时间进邮箱

读完觉得有用?留个邮箱,每篇新解读发布就送达,随时退订。