精选 Seedance 2.5 视频提示词库——复制任意提示词,改编成你的场景,直接在 VidAndPic AI 视频生成器中生成。
A pristine sneaker rotating 360 degrees on a matte studio turntable, soft gradient gray background, three-point studio lighting with crisp edge highlights, subtle floor reflection, dust-free seamless loop, high-end e-commerce product video, square format.
模型: seedance-2-5Aerial drone shot pushing forward over a fog-filled mountain valley at sunrise, peaks piercing through a sea of clouds, golden light grazing the ridgelines. Smooth forward flight with a slow upward tilt, vast scale, cinematic color grading, ultra realistic landscape footage.
模型: seedance-2-5Animate this portrait photo: subtle natural blinking, gentle head micro-movements, hair swaying slightly, soft breathing motion in the shoulders, consistent facial identity and lighting preserved, background stays static, realistic and lifelike result, no distortion of facial features.
模型: seedance-2-5An office worker in business attire walks away from a desk in exaggerated slow motion while papers fly into the air behind them, dramatic wind machine effect, deadpan facial expression, colleagues reacting in the blurred background. Low angle camera, epic action-movie lighting applied to a mundane office, comedic contrast, vertical video.
模型: seedance-2-5Extreme slow motion of hot espresso pouring from a moka pot into a clear glass cup, crema swirling in golden-brown ribbons, steam rising in soft backlight. Macro lens, shallow focus shifting from stream to cup, warm kitchen morning light, hyperrealistic liquid physics, appetizing commercial food cinematography, vertical format.
模型: seedance-2-5Anime style: a high school student in uniform stands on a school rooftop, wind lifting their hair and collar, cumulus clouds drifting across a bright summer sky. Slow pan from sky down to the character, hand-drawn cel shading, soft bloom, lens flare from the sun, gentle emotional atmosphere, classic Japanese animation aesthetic.
模型: seedance-2-5A fluffy orange cat waking up on a windowsill as warm morning sunlight streams through sheer curtains. The cat stretches, yawns widely, then blinks slowly at the camera. Gentle handheld camera with a slight drift, soft golden light, dust particles floating in the air, cozy domestic atmosphere, realistic fur detail, vertical video for social media.
模型: seedance-2-5A lone figure in a long weathered coat stands in a narrow cyberpunk alley at night, neon signs reflecting in rain puddles. The figure slowly turns to face the camera as the camera pushes in from a low angle. Volumetric haze, drifting steam, flickering holographic signage, cinematic teal and magenta color grading, subtle rain particles, moody sci-fi atmosphere.
模型: seedance-2-5Luxury perfume bottle standing on black polished stone, dramatic rim lighting, golden mist slowly swirling around the glass. Macro camera glides from left to right, capturing refracted light inside the liquid, droplets sliding down the glass surface. High-end commercial product cinematography, deep shadows, crisp highlights, ultra clean composition, no text.
模型: seedance-2-5A cinematic close-up of a young adult fashion model standing on a quiet rooftop during blue hour, natural skin texture, dark hair moving gently in the wind, soft city lights glowing behind her. She looks away from camera, then slowly turns toward the lens. The camera performs a subtle slow dolly-in, shallow depth of field, realistic fabric and hair motion, cool ambient light mixed with warm practical lights, premium fashion campaign, restrained natural movement, realistic cinematic footage.
模型: seedance-2-5Seedance 2.5 最吃「分镜清单式」的 Prompt:一个明确的主体、一段连续的动作,再加清晰的镜头与灯光指令。上方提示词库里的 10 条都是真实生成过的实例——缩略图就是 Seedance 2.5 真实输出的抽帧。任意复制使用,或按本指南的模板写出你自己的 Prompt。
一条强有力的 Seedance 2.5 Prompt 按顺序回答七个问题:画面里有什么(主体)、在做什么(动作)、发生在哪里(环境)、镜头如何运动(运镜)、场景如何布光(灯光)、画面随时间如何变化(动态)、最终画面应该是什么质感(风格)。
把这七项写成一个连贯的段落,而不是逗号分隔的清单——Seedance 2.5 对自然语言的理解最好。主体和动作放在第一句,因为靠前的词权重最高;皮肤质感、织物摆动、空气氛围这类细节放在结尾,让它们修饰画面而不是抢主指令的戏。
主体 + 动作 + 环境 + 运镜 + 灯光 + 动态 + 风格示例:一位时装模特的电影感特写(主体)在蓝调时刻的天台缓缓转向镜头(动作),镜头缓慢推近(运镜),身后城市灯光柔和发光(灯光),发丝随微风轻动(动态),高级时装广告质感、写实电影画面(风格)。
Seedance 2.5 认识标准摄影术语,请直接点名运镜:slow dolly-in(缓慢推近)、orbit around the subject(环绕主体)、handheld follow(手持跟随)、crane reveal(升降揭示)、static tripod shot(固定机位)。一段 4–8 秒的片段最多组合两种运镜——「推近 + 主体转身」远好于让模型硬凑四种叠在一起的运镜。
景别同样重要:面部和产品用特写、互动用中景、风景揭示用大远景或航拍。需要特定的镜头性格就写出来:浅景深、35mm 质感、微距镜头。模型奖励精确而非形容词——「slow dolly-in, shallow depth of field」永远好过「beautiful cinematic camera work」。
灯光是让生成画面显高级的最低成本手段。点名时间和光源:blue hour(蓝调时刻)、golden hour(黄金时刻)、阴天窗光、夜晚霓虹、暗室单灯柔光箱。再写光如何作用在主体上——发丝轮廓光、背景实用灯发光、湿沥青路面的反光。
产品类场景用影棚语言:无缝背景、柔和渐变阴影、受控高光;生活类场景写有动机的光源:窗光、路灯、壁炉光。避免「明亮的晴天夜晚」这类自相矛盾的描述——冲突的光线通常会生成又平又浑浊的画面。
用图生视频动画一张照片时,参考图已经决定了主体和环境——Prompt 只需要描述运动。保持极简和物理化:自然眨眼、头部微动、发丝轻摆、保持面部一致性,然后明确写出什么不该动:背景保持静止。
图生视频最常见的失败是「过度导演」:让一张人像同时走路、说话、转身,模型会为了同时满足三个指令而扭曲面部。每段片段给一到两条运动指令,就能得到最真实的效果——参考提示词库里的 Bring Portrait Photos Alive 实例。
四个错误造成了绝大多数翻车画面。第一,多场景 Prompt:「她走进来,然后切到海滩,然后出现 Logo」描述的是剪辑而非镜头——一条 Prompt 等于一个连续镜头,场景切换留给后期。第二,矛盾的运动:同一片段里既慢动作又快动作,两者会互相抵消。
第三,画面内文字:要求招牌、Logo 或字幕会生成乱码字形,文字请在剪辑软件里加。第四,空洞的风格词:「very beautiful, 8K, masterpiece」没有信息量——直接点名具体质感,如 35mm 胶片颗粒、柔和粉彩色调、纪录片写实主义。
先用 480p 廉价迭代构图与运动,确定最佳一版后再用 720p 出成片。实验阶段把片段控制在 4–8 秒——更短的生成更快更省,Seedance 2.5 在这个窗口内对单一动作的把控最好。
复制上方任意提示词卡片,或打开卡片点 Generate,即可带着预设的模型、画面比例和时长直接进入 VidAndPic AI 视频生成器。每次只改一个变量——先换环境、再换灯光、最后换运镜——你才能确定到底是哪个改动让画面变好。
准备好创作了吗?打开Seedance 2.5 视频生成器,浏览AI 视频生成器,或查看套餐与价格。
Choose whether to allow analytics cookies to help us understand how the site is used.