目录
Seedance 2.5 能做到一件仍然有些不可思议的事:以一个连续镜头生成最长 30 秒的带声视频,无需拼接,也没有明显接缝。
但长镜头容不得马虎。每一秒都需要有具体内容,否则模型就会开始即兴发挥。下面的 Seedance 2.5 提示词旨在让整个 30 秒按计划进行。示例按用途分类,你可以找到接近自己想法的内容,粘贴后修改细节。
这里的每条提示词都可以在 Seedance 2.5 AI 视频生成器中使用;该生成器位于 AI Studio,让你直接在浏览器里运行 Seedance 2.5。
Seedance 2.5 有什么不同
ByteDance 于 2026 年 7 月 31 日推出 Seedance 2.5。它的一些功能会影响提示词的写法:
- 每次生成最长 30 秒,包含音频。上一版本每次最长为 15 秒。
- 使用时间戳编写提示词。ByteDance 表示,你可以在特定时间段内控制故事、镜头、动作和节奏,例如“0–5s: …”。
- 丰富的参考素材。据 ByteDance 介绍,一次最多可使用 30 张图片、10 段视频和 10 段音频。
- 据 ByteDance 介绍,镜头转场更流畅,不需要的字幕和背景音乐也更少。
- 已知弱点。ByteDance 自己指出,复杂运动的物理表现以及多个主体互动时的场景稳定性仍有改进空间。
写好 Seedance 2.5 提示词的七个习惯
- 按时间段编写。超过约 10 秒的内容,可以把提示词分成几个节拍:“0–8s”“8–18s”“18–30s”。ByteDance 自己的示例提示词就采用这种格式。
- 指明参考素材。如果上传多张图片,要明确说明每张图的作用,例如“参考图 1 中的女子,参考图 2 中的红色背包”。ByteDance 的示例使用 @Image 1 这样的标签。在 AI Studio 中,这类普通描述也很有效。
- 需要连续镜头时明确提出。“single continuous take, no cuts”之类的表述有助于保持镜头连续。
- 用简单的词描述运镜。例如“Slow push-in”“orbit right”“handheld follow”“crane up”。
- 写一行声音要求。音频与画面一起生成,因此要说明你希望听到什么、不希望出现什么:“no music”“no on-screen text”。
- 控制人物数量。考虑到模型处理主体互动时的弱点,一到三个人物较为稳妥。
- 低成本打样,高清完成。在 AI Studio 中使用 480p 生成草稿,再用 1080p 渲染选中的版本。
提示词模板
[Length], [aspect ratio], single continuous take.
[Subject and look] [action] in [setting, time, weather].
Camera: [shot size and movement].
Style: [light, palette, look].
Sound: [ambience, dialogue, music, or "no music"].
0–Xs: [beat one]
X–Ys: [beat two]
Y–30s: [beat three]
产品与电商
1. 沙漠风中的香水
20 seconds, 9:16, single continuous take. An amber glass perfume bottle stands on a sand dune at golden hour as wind lifts fine sand around it.
Camera: low angle, slow 90-degree orbit, ending close on the bottle.
Style: warm gold and copper tones, glowing backlight, luxury commercial look.
Sound: soft wind, a low, breathy synth tone, no voiceover.
0–8s: Sand streams past the bottle in slow motion.
8–15s: The orbit catches the sun through the glass, throwing amber light across the sand.
15–20s: The wind drops, and the sand settles into perfect ripples.
2. 无线耳机与城市通勤
15 seconds, 16:9, single continuous take. A young man in a charcoal coat wears matte white over-ear headphones on a crowded subway platform.
Camera: medium shot, slow push-in toward his face.
Style: cool blue platform lighting with warm highlights on his face, crisp tech-ad look.
Sound: busy platform noise that fades almost to silence as he adjusts the headphones, then a soft piano melody.
0–6s: A train roars in behind him; the crowd rushes past.
6–15s: He touches the ear cup, closes his eyes, and the noise drops away to calm music.
3. 背包徒步测试
25 seconds, 16:9. A hiker carries a forest-green backpack up a rocky mountain trail in the early morning.
Camera: handheld follow from behind, then a side tracking shot, ending in a close-up of the pack's buckles.
Style: crisp alpine light, cool blues and greens, outdoor documentary look.
Sound: boots on rock, wind, a small stream, no music.
0–10s: Follow shot up a steep section of trail.
10–18s: Side tracking shot as she steps across a stream on stones.
18–25s: She stops at a viewpoint, clicks the chest strap, and looks out across the valley.
4. 机械键盘 ASMR
12 seconds, 16:9, single continuous take. Hands type on a mechanical keyboard with cream-colored keycaps on a walnut desk at night.
Camera: macro close-up, slow slide along the keys.
Style: warm desk lamp, soft bokeh from a monitor glow, cozy.
Sound: crisp, deep key clacks, quiet rain on the window, no music.
0–12s: Fingers type steadily as the camera glides from left to right across the keyboard.
5. 海滩防晒霜
15 seconds, 9:16. A woman in a straw hat applies sunscreen from a pale blue tube on a sunny beach.
Camera: close-up on hands and tube, then a pull-back to a medium shot.
Style: bright summer light, turquoise and sand tones, fresh lifestyle look.
Sound: waves, gulls, distant laughter, a light ukulele tune.
0–6s: Close-up as she squeezes lotion onto her palm.
6–15s: The camera pulls back as she smiles, puts on sunglasses, and walks toward the water.
美食与饮品
6. 枫糖浆淋松饼
10 seconds, 1:1, single continuous take. Maple syrup pours onto a tall stack of fluffy pancakes topped with butter and blueberries.
Camera: side angle, slow motion, slight push-in.
Style: warm morning light, golden browns, appetizing food-commercial look.
Sound: a soft drizzle, gentle café ambience.
0–10s: Syrup cascades over the edge and the butter slowly melts down the side.
7. 柴火披萨
20 seconds, 16:9. A pizzaiolo slides a pizza into a glowing wood-fired oven and pulls it out bubbling.
Camera: medium shot at oven height, then a close-up as the pizza comes out.
Style: firelight orange, dark brick, rustic and warm.
Sound: crackling fire, the scrape of the wooden peel, restaurant chatter.
0–7s: He stretches the dough, adds basil, and slides the pizza in.
7–14s: Close on the oven mouth as the crust puffs and blisters.
14–20s: He pulls it out, steam rising, cheese bubbling.
8. 抹茶仪式
15 seconds, 16:9, single continuous take. Hands whisk bright green matcha in a handmade ceramic bowl on a low wooden table.
Camera: top-down view, then a slow tilt to a side view.
Style: soft, diffused daylight, muted earth tones, calm and minimal.
Sound: the bamboo whisk, a kettle, birds outside, no music.
0–8s: Top-down as the whisk turns the matcha into a fine froth.
8–15s: The camera tilts as the hands lift the bowl toward the lens.
时尚与美妆
9. 风中的风衣
15 seconds, 9:16, single continuous take. A model in a camel trench coat walks across a city bridge on a windy, overcast day.
Camera: slow-motion tracking shot from the front, walking backward with her.
Style: muted grey city, the coat as the only warm color, editorial fashion look.
Sound: wind, footsteps, a distant tram bell, minimal electronic beat.
0–15s: She walks toward the camera as the coat billows dramatically behind her.
10. 口红试色
8 seconds, 1:1, single continuous take. A deep red lipstick glides across a white surface in a single swatch.
Camera: macro, top-down, locked off.
Style: clean white studio light, rich saturated red, beauty commercial look.
Sound: a soft, creamy glide, a single chime at the end.
0–8s: The lipstick draws a smooth swatch from left to right, catching a highlight.
11. 牛仔夹克街头造型
12 seconds, 9:16. A skater in an oversized vintage denim jacket leans against a graffiti wall, then pushes off on a board.
Camera: medium shot, then a quick whip pan to follow.
Style: sunny afternoon, punchy colors, 1990s street-style film look.
Sound: street noise, board wheels, an upbeat boom-bap beat.
0–6s: He adjusts his collar and looks at the camera.
6–12s: He drops the board and rolls out of frame, and the camera whips to follow.
旅行与地点
12. 火车窗外的山谷
20 seconds, 16:9, single continuous take. View from a moving train window as it winds through a green mountain valley with waterfalls.
Camera: steady, as if resting on the windowsill; reflections faintly visible in the glass.
Style: soft morning light, lush greens, peaceful travel look.
Sound: train rhythm on the rails, a quiet carriage, no music.
0–20s: Meadows, wooden chalets, and waterfalls slide past as the train enters a sunlit stretch.
13. 陡街上的黄色电车
15 seconds, 16:9. A vintage yellow tram climbs a steep, narrow street of tiled buildings in a hilly old city.
Camera: low angle at the bottom of the hill, slow tilt up as it passes.
Style: bright afternoon sun, blue and white tiles, warm nostalgic tones.
Sound: the tram's bell, metal wheels on rails, street conversation.
0–8s: The tram approaches up the hill toward the camera.
8–15s: It passes close by; the camera tilts to follow it toward the top of the hill.
14. 极光下的小屋
25 seconds, 16:9, single continuous take. A small wooden cabin with a glowing window sits in a snowy field under the northern lights.
Camera: slow push toward the cabin from a distance.
Style: deep night blue, swirling green aurora, warm amber window, magical realism.
Sound: soft wind, snow crunching, a faint crackle from a fireplace.
0–12s: The aurora ripples across the sky as the camera moves forward.
12–20s: Smoke curls up from the chimney.
20–25s: The cabin door opens and warm light spills across the snow.
故事与电影感
15. 地铁小提琴手
30 seconds, 16:9, single continuous take. A young violinist plays alone in a tiled subway corridor as people hurry past.
Camera: slow, continuous dolly toward her from the far end of the corridor.
Style: cool fluorescent light, warm light on her face, cinematic realism.
Sound: footsteps and train rumble, then a slow, beautiful violin melody that grows louder.
0–10s: Commuters rush past without looking; the music is faint.
10–20s: The camera moves closer; one child stops and stares.
20–30s: A small crowd gathers in silence as she plays the final note.
16. 雨中迷路的机器人
20 seconds, 16:9. A small, rusty delivery robot with one flickering headlight rolls down a rainy neon street, looking for an address.
Camera: low tracking shot beside the robot, ending in a close-up.
Style: neon reflections in puddles, pink and blue palette, gentle science fiction.
Sound: rain, distant traffic, soft electronic beeps from the robot.
0–8s: It stops at a doorway and scans the number, beeping uncertainly.
8–15s: It rolls on, its headlight flickering.
15–20s: A door opens and a little girl holds an umbrella over it.
17. 信
15 seconds, 16:9, single continuous take. An older man sits at a kitchen table reading a handwritten letter, his tea going cold.
Camera: slow push-in from medium shot to close-up.
Style: soft grey morning light, quiet, intimate drama.
Sound: a clock ticking, birds outside, paper rustling, no music.
0–8s: He reads, his expression unreadable.
8–15s: His eyes fill with tears, then he smiles and sets the letter down gently.
18. 博物馆盗窃
20 seconds, 21:9. A thief in black slips between red security lasers in a dark museum gallery toward a glowing glass case.
Camera: overhead shot, then a low side angle.
Style: high contrast, red laser light, sleek heist-movie look.
Sound: tense low strings, a faint hum from the lasers, careful breathing.
0–8s: Overhead view as she bends backward under a laser beam.
8–15s: Low angle as she steps over the last beam, inches from the case.
15–20s: She reaches the case and smiles, and then an alarm light starts to spin.
对话与人物
19. 播客片段
15 seconds, 9:16. Two podcast hosts sit at a table with microphones in a cozy studio with acoustic panels and plants.
Camera: medium two-shot, slight push-in.
Style: warm practical lights, natural colors, social clip look.
Sound: clear studio voices, light laughter.
Host 1: "So you're telling me you've never seen the ocean?"
Host 2: "I've seen pictures. That counts, right?"
0–15s: Host 1 asks the question in disbelief, and Host 2 answers with a shrug as both burst out laughing.
20. 科学老师的演示
12 seconds, 16:9. A science teacher with rolled-up sleeves stands behind a lab bench and holds up a beaker of blue liquid.
Camera: medium shot, then a close-up on the beaker.
Style: bright classroom, clean and friendly, educational video look.
Sound: classroom ambience, a fizz, a few kids saying "whoa."
Teacher: "Watch closely. Don't blink."
0–6s: He says the line with a grin.
6–12s: He drops in a tablet and the liquid foams up and turns bright purple.
21. 尴尬的电梯时刻
15 seconds, 16:9, single continuous take. Two coworkers stand in an elevator holding identical coffee cups, avoiding eye contact.
Camera: static wide shot from the corner, slightly high.
Style: flat office lighting, muted colors, deadpan comedy.
Sound: elevator hum, a ding, faint elevator music.
Coworker 1: "...Nice cup."
Coworker 2: "Thanks. You too."
0–15s: Long silence, then the exchange; the doors open and both step out at the same time and bump shoulders.
动作与运动
22. 山地自行车下坡
15 seconds, 16:9. A mountain biker in a bright orange jersey rides down a steep forest trail over roots and rocks.
Camera: chest-mounted point-of-view shot, then a trailside angle for a jump.
Style: dappled forest light, rich greens, high-energy action look.
Sound: tires on dirt, rattling bike, heavy breathing, wind.
0–10s: Fast POV descent weaving between trees.
10–15s: Trailside slow-motion shot as the rider clears a small jump.
23. 清晨拳击
20 seconds, 16:9. A boxer hits a heavy bag in an old gym before sunrise.
Camera: slow circling shot around the boxer and bag.
Style: a single shaft of dusty light from a high window, gritty, warm tones.
Sound: punches thudding into leather, chains rattling, steady breathing.
0–10s: Steady combinations, sweat flying in the light.
10–20s: She stops, catches her breath, and taps her gloves together, focused.
24. 狗与飞盘
10 seconds, 16:9, single continuous take. A border collie sprints across a green park and leaps to catch a red frisbee.
Camera: low tracking shot alongside the dog, in slow motion for the catch.
Style: sunny afternoon, vivid colors, joyful.
Sound: running paws, a happy bark, people cheering.
0–10s: The dog chases the frisbee, launches into the air, and snatches it mid-flight.
音乐与舞蹈
25. 黄昏街舞
20 seconds, 9:16, single continuous take. A dancer in loose white clothes performs a fluid contemporary routine in an empty parking lot at dusk.
Camera: slow orbit around the dancer at waist height.
Style: purple-orange sky, streetlights flickering on, cinematic music-video look.
Sound: a slow, emotional electronic track with a heavy beat.
0–20s: The dancer moves in sync with the beat, spinning, sliding, and freezing on the final drop.
26. 车库乐队
15 seconds, 16:9. A three-piece band (drums, bass, guitar) rehearses in a cluttered garage with fairy lights.
Camera: handheld, moving between members.
Style: warm tungsten light, a little haze, indie music video look.
Sound: an energetic garage-rock riff, cymbals, a shouted "one more time!"
0–8s: The camera weaves from the drummer to the bassist.
8–15s: The guitarist hits the final chord and jumps as the lights flicker.
社交媒体与创作者内容
27. 跟我一起准备出门
15 seconds, 9:16. A young woman gets ready in front of a mirror with round bulbs, doing her makeup and talking to the camera.
Camera: phone propped on a vanity, front-facing, natural.
Style: soft vanity light, natural skin, authentic creator look.
Sound: casual voice, room tone.
Creator: "Okay, five-minute makeup, because I'm late. Again."
0–15s: She speaks the line, swipes on mascara and blush quickly, then grabs her bag and leaves the frame.
28. 房间改造揭晓
20 seconds, 16:9, single continuous take. A plain, empty bedroom transforms into a cozy, plant-filled bohemian room.
Camera: static wide shot from the doorway.
Style: bright daylight, natural colors; the change happens smoothly.
Sound: an upbeat acoustic track, small "pop" sounds as each object appears.
0–6s: Bare white room.
6–16s: A rug, bed, plants, and string lights appear one by one.
16–20s: Golden hour light fills the finished room.
使用参考素材
29. 你的角色在 30 秒连续镜头中
上传一张或多张原创角色的清晰照片(正面、侧面、全身),然后:
30 seconds, 16:9, single continuous take. The young man from reference image 1, wearing the outfit shown in reference image 2, walks through a busy night market.
Keep his face, hair, and outfit consistent throughout.
Camera: handheld follow from behind, then moving to his side.
Style: warm lantern light, steam from food stalls, vibrant colors.
Sound: crowd chatter, sizzling food, a vendor calling out.
0–10s: He weaves through the crowd, looking around.
10–20s: He stops at a dumpling stall and points at the menu.
20–30s: He takes the first bite, burns his mouth, and laughs.
30. 来自照片的产品
上传从多个角度拍摄的清晰产品照片,然后:
15 seconds, 1:1, single continuous take. The ceramic table lamp from reference images 1 to 3 sits on a bedside table in a calm bedroom at night.
Keep the lamp's shape, glaze color, and texture exactly as shown.
Camera: slow orbit around the lamp.
Style: dark room, the lamp as the main light source, warm and premium.
Sound: soft room tone, a gentle click when the lamp turns on.
0–5s: The room is dark.
5–15s: The lamp clicks on and warm light fills the room as the camera orbits.
进阶用户补充:在 ByteDance 自家的应用中,Seedance 2.5 还可以使用白模渲染参考,即用于确定布局、角色姿势和镜头路径的粗略 3D 场景。如果你使用 3D,可以搭建场景、导出,再与提示词一起作为参考。使用前请检查你的平台是否支持此模式。
在 AI Studio 上使用 Seedance 2.5
在 AI Studio 的 Seedance 2.5 页面,你可以设置:
- 时长:4 至 30 秒,可按秒选择
- 分辨率:480p、720p 或 1080p
- 画面比例:16:9、9:16、1:1、4:3、3:4、21:9 或自适应
- 模式:文生视频;使用 1 或 2 张图片的图生视频;使用最多 30 张图片的参考素材生成视频
- 音频:默认开启;如果打算自行添加音乐或旁白,可以关闭
生成按钮会在开始前显示所需积分。视频越长、分辨率越高,消耗的积分越多。
何时改用 Seedance 2.0
Seedance 2.5 并不总是最佳选择。如果你想制作带声音的短图生视频,Seedance 2.0 视频生成器依然很强。根据 GIGAZINE 在 2026 年 8 月报道的 Artificial Analysis 排名,Seedance 2.0 在带音频的图生视频中排名第一。它每次生成约 15 秒,对大多数社交媒体短片来说已经足够。
如果你想看看同一条提示词在完全不同的模型中会产生什么效果,AI Studio 的 AI 视频生成器也支持 Kling 4.0、Gemini Omni 和 Happy Horse 等模型。
来源
- ByteDance Seed,Seedance 2.5 介绍,2026 年 7 月 31 日
- CineD,ByteDance Seedance 2.5 API 上线,2026 年 7 月
- GIGAZINE,MiniMax H3 视频生成 AI,2026 年 8 月 3 日,Artificial Analysis 排名来源
- AI Studio,Seedance 2.5 和 Seedance 页面,核查于 2026 年 11 月 2 日
把创意变成视频
打开视频生成器,开始你的创作。
生成视频


