目录
Kling 4.0 一次最多可以生成 30 秒视频。这是个好消息,但也带来了新问题:适用于 5 秒短片的模糊提示词,到了 30 秒就难以奏效。模型会自行猜测并填补空白,到了第 18 秒,你的咖啡广告可能已经变成了对空厨房的缓慢摇摄。
解决办法是不再像写图片说明一样写提示词,而是把它们写成分镜计划。本指南提供一个适用于任何镜头的七部分公式,以及 25 个可以粘贴并按需修改的例子。
你可以在 Kling5 AI 的 Kling 4.0 AI 视频生成器中尝试本文所有提示词;该平台可在浏览器中运行 Kling 4.0 和 Kling 3.0。首先需要了解:Kling5 AI 是独立的第三方平台,与 Kuaishou 或 Kling 团队没有关联。这些提示词本身适用于所有提供 Kling 4.0 的平台。截至 2026 年 11 月,Kling 5 尚未发布,因此 Kling 4.0 是目前可以使用的最新 Kling 模型。
Kling 4.0 有哪些变化,为什么会影响提示词
根据 Kling 于 2026 年 9 月 30 日发布的指南,Kling 4.0 带来了四项与提示词编写相关的变化:
- 每次生成 3 至 30 秒,无需拼接。
- 一个视频最多支持 10 个关键帧。关键帧是标记视频应到达的时刻的图像,例如开场镜头、揭晓画面或光线变化。
- 每项任务最多支持 15 个参考素材,包括图片、视频、主体和声音参考。
- 提示词最多支持 8,000 个 token,足以容纳详细的时间编码计划。
Kling 的指南将核心概念概括得很简单:关键帧展示重要时刻,提示词解释视频如何在这些时刻之间过渡。想更全面地了解新变化,请参阅我们的 Kling 4.0 功能解析。
关于平台:可用功能因运行模型的平台而异。在 Kling5 AI 的 Kling 4.0 工具中,图生视频接受首帧和可选尾帧;参考生视频最多接受 10 张图片和 5 个视频;输出支持 720p 或 1080p,以及 16:9、9:16、1:1 或 21:9 比例。完整的 10 关键帧工作流在 Kling 自家应用中提供。下面带时间编码的提示词适用于这两种设置。
分镜计划公式
每个优秀的 Kling 4.0 提示词都回答七个问题:
- 主体:画面中是谁或什么?描述要具体,例如“60 多岁、银发松松盘成发髻的女性”,而不是“一个女人”。
- 动作:主体做什么?每个节拍使用一个清晰的动词。
- 场景:何时何地?说明地点、一天中的时间和天气。
- 镜头:景别、运动和镜头感。例如“缓慢向前推镜,中近景,浅景深”。
- 光线和风格:光源、色调和胶片质感。
- 声音:环境音、对白、音乐氛围或静音。
- 时间线:什么时间发生什么事。对于超过约 10 秒的内容,这一点至关重要。
以下是模板:
[Length, aspect ratio]. [Subject] [action] in [setting].
Camera: [shot size], [movement], [lens/depth of field].
Light and style: [light source], [palette], [look].
Sound: [ambience], [dialogue], [music or none].
Timeline:
0–Xs: [beat one]
X–Ys: [beat two]
Y–Zs: [beat three]
减少重新生成次数的五条规则
- 每个节拍只做一项重大变化。Kling 的指南建议每次只改变一个主要元素。不要在同两秒内同时改变地点、服装和时间。
- 保持顺序一致。如果使用关键帧,按关键帧出现的顺序描述事件。
- 使用镜头动作动词,而非形容词。“向前推镜”“向左环绕”“升降机上升”和“手持跟拍”比“动态”和“电影感”更有效。
- 对白单独成行并加引号。每句保持简短。每个节拍中每个角色一句话就足够了。
- 留出空间。30 秒镜头安排 12 个节拍会显得仓促。三个到五个节拍最合适。
25 个 Kling 4.0 提示词示例
可以直接粘贴使用,也可以替换为自己的主体、产品和场景。时长和画面比例仅供参考。
产品和广告
1. 清晨咖啡研磨机(15 秒,16:9)
15 seconds, 16:9. A matte black hand-crank coffee grinder on a pale oak kitchen counter at sunrise.
Camera: macro close-up, slow push-in, shallow depth of field.
Light and style: warm low sun through a window on the left, soft steam haze, natural film grain.
Sound: quiet birdsong outside, the crunch of beans being ground, no music.
Timeline:
0–5s: Beans pour into the grinder in slow motion.
5–11s: A hand turns the crank; fine grounds fall into a glass jar.
11–15s: The camera settles on the grinder as steam drifts through the light.
2. 雨中的运动鞋(10 秒,9:16)
10 seconds, 9:16. A white running sneaker with neon orange laces rotates slowly on a wet black platform as rain falls around it.
Camera: low angle, 180-degree slow orbit, sharp focus on the shoe.
Light and style: cyan and magenta rim lights, raindrops lit like sparks, glossy commercial look.
Sound: heavy rain, a deep bass pulse on each half-turn.
Timeline:
0–6s: The orbit reveals the side profile, water beading on the upper.
6–10s: The camera stops at a three-quarter view; one drop hits the toe in slow motion.
3. 护肤瓶与水(8 秒,1:1)
8 seconds, 1:1. A frosted glass serum bottle stands in shallow, perfectly still water on a white stone surface.
Camera: locked-off, top-down for 3 seconds, then a gentle tilt to eye level.
Light and style: soft diffused daylight, pale sage and cream palette, clean minimal look.
Sound: a single water drop, soft ambient hum.
Timeline:
0–3s: A single drop lands beside the bottle and sends out perfect ripples.
3–8s: The camera tilts down to eye level; light refracts through the bottle onto the water.
4. 从厨房到产品特写的 30 秒广告(30 秒,16:9;适合使用关键帧)
30 seconds, 16:9. A woman in her 30s with curly hair, wearing a mustard sweater, discovers a compact blender in a bright morning kitchen.
Camera: starts wide, then medium shots, ending on a product close-up.
Light and style: bright, airy, lots of white and light wood, warm skin tones.
Sound: kitchen ambience, a short laugh, the blender whirring, a soft upbeat acoustic tune.
Timeline:
0–6s: Wide shot. She enters the kitchen, yawning, and notices a box on the counter.
6–13s: Slow push-in as she opens the box and lifts out the blender, smiling.
13–21s: Medium shot. She drops berries and spinach in and blends; the jar swirls purple-green.
21–26s: Close-up of her first sip; she raises her eyebrows, impressed.
26–30s: Product hero shot on the counter, soft morning light, glass glistening.
*使用关键帧:*以厨房图片作为首帧,产品特写作为尾帧。
5. 智能手表晨跑(12 秒,9:16)
12 seconds, 9:16. A runner in a dark grey jacket jogs along a coastal path at dawn, wearing a slim silver smartwatch.
Camera: handheld follow from behind, then a quick move to a close-up of the wrist.
Light and style: blue hour turning gold, mist over the sea, crisp athletic look.
Sound: footsteps on gravel, waves, steady breathing, a soft chime from the watch.
Timeline:
0–7s: Follow shot as the runner climbs a gentle hill toward the light.
7–12s: The runner glances at the watch; the screen glows with a heart icon as the sun breaks the horizon.
美食与饮品
6. 拉面热气(8 秒,16:9)
8 seconds, 16:9. A bowl of tonkotsu ramen with a soft-boiled egg, chashu, and scallions on a dark wooden counter.
Camera: extreme close-up, slow slide left to right, shallow depth of field.
Light and style: a single warm overhead pendant light, rich browns and golds, moody restaurant look.
Sound: gentle simmering, a distant kitchen clatter.
Timeline:
0–5s: Steam curls up through the light as the camera slides across the toppings.
5–8s: Chopsticks lift a tangle of noodles; broth drips back into the bowl.
7. 夜间塔可摊(12 秒,16:9)
12 seconds, 16:9. A street taco stand at night with a cook flipping tortillas on a sizzling flat-top grill.
Camera: medium shot from the customer's side, slow push-in to the grill.
Light and style: string lights and a bare bulb, orange and teal palette, slight haze.
Sound: sizzling grill, street chatter, a radio playing faint cumbia.
Timeline:
0–6s: The cook flips three tortillas in quick succession, smoke rising.
6–12s: Close-up as he fills a taco with charred meat, onion, and cilantro and slides it onto a paper plate.
8. 倒制冰拿铁(6 秒,9:16)
6 seconds, 9:16. Cold milk pours into a tall glass of ice and espresso on a marble café table.
Camera: side-on macro, locked off, slow motion.
Light and style: bright window light, high contrast, crisp café commercial look.
Sound: ice clinking, a soft pour, café murmur.
Timeline:
0–6s: Milk swirls into the espresso, forming slow marble patterns around the ice.
电影叙事
9. 风暴中的灯塔守护人(20 秒,21:9)
20 seconds, 21:9. An elderly lighthouse keeper in a yellow oilskin coat climbs a spiral staircase during a storm.
Camera: low angle looking up the stairwell, then a slow crane up to the lamp room.
Light and style: cold blue storm light, flashes of lightning, a warm lantern glow, painterly realism.
Sound: howling wind, rain against glass, creaking iron steps, thunder.
Timeline:
0–8s: He climbs steadily, lantern in hand, the stairwell shaking with each gust.
8–15s: He reaches the lamp room and pulls a heavy lever.
15–20s: The great lamp ignites and its beam sweeps across black, churning waves.
10. 宇航员与花(15 秒,16:9)
15 seconds, 16:9. An astronaut in a scuffed white suit walks across a rust-red rocky plain under a hazy pink sky.
Camera: wide tracking shot from the side, then a slow move to a close-up on the ground.
Light and style: dusty low sun, muted reds and oranges, epic science-fiction look.
Sound: muffled breathing inside the helmet, wind hiss, no music.
Timeline:
0–7s: The astronaut walks, then stops abruptly and looks down.
7–12s: They kneel; a single tiny blue flower grows between two rocks.
12–15s: Close-up of a gloved finger hovering near the petals, not quite touching.
11. 黑色电影侦探(12 秒,16:9)
12 seconds, 16:9. A tired detective in a rumpled trench coat sits at a desk in a 1940s office at night.
Camera: slow push-in from the doorway, then a close-up on his face.
Light and style: black-and-white, hard light through window blinds, cigarette-smoke haze, film noir.
Sound: rain outside, a ticking clock, a distant saxophone.
Timeline:
0–7s: He turns over a photograph on the desk and stares at it.
7–12s: The phone rings. He lets it ring twice, then reaches for it.
12. 排水沟里的纸船(10 秒,16:9)
10 seconds, 16:9. A small paper boat made from a newspaper page races down a rain-filled gutter on a city street.
Camera: ground-level tracking shot alongside the boat, wide-angle lens.
Light and style: overcast daylight, wet reflections, saturated red and yellow accents, playful.
Sound: running water, distant traffic, a child laughing off-screen.
Timeline:
0–6s: The boat spins through a small whirlpool and rights itself.
6–10s: It slides toward a storm drain, and a small hand reaches in and scoops it up at the last second.
13. 书店从白天到夜晚(30 秒,16:9;适合使用关键帧)
30 seconds, 16:9. A narrow, cluttered secondhand bookshop with tall shelves and a rolling ladder.
Camera: one continuous shot, a slow dolly from the front door toward the back counter.
Light and style: warm, cozy, dust motes in the light; the light shifts with the time of day.
Sound: a door chime, pages turning, a clock, rain arriving later.
Timeline:
0–8s: Late morning. Sunlight fills the shop; a customer browses near the window.
8–16s: Afternoon. The light turns golden; a cat jumps onto the ladder.
16–24s: Dusk. Lamps flick on one by one; rain starts against the window.
24–30s: Night. The shopkeeper flips the sign to Closed and switches off the last lamp.
*使用关键帧:*依次使用同一家书店早晨、下午、黄昏和夜晚的四张图片。
角色与对白
14. 祖母教做意大利面(12 秒,16:9)
12 seconds, 16:9. A grandmother with flour on her hands teaches her teenage grandson to fold tortellini at a farmhouse table.
Camera: medium two-shot, slight handheld feel, then a close-up of their hands.
Light and style: soft afternoon window light, warm, documentary feel.
Sound: kitchen ambience, a pot simmering.
Dialogue:
Grandmother: "Not like that. Gently. Like you're tucking it into bed."
Timeline:
0–6s: She watches him struggle, then speaks her line.
6–12s: Close-up as she guides his fingers and the pasta folds perfectly. He grins.
15. 屋顶对话(15 秒,16:9)
15 seconds, 16:9. Two friends in their 20s sit on a city rooftop at dusk, legs dangling over the ledge, sharing a bag of chips.
Camera: wide shot from behind, then a slow orbit to a side profile of both.
Light and style: purple and orange sky, city lights blinking on, soft and nostalgic.
Sound: distant traffic, wind, a far-off siren.
Dialogue:
Friend 1: "Do you think we'll still do this when we're old?"
Friend 2: "We'll just need a lot more cushions."
Timeline:
0–6s: Wide shot; they watch the skyline.
6–15s: The orbit reveals their faces as they speak, then laugh.
16. 咖啡师对镜头说话(10 秒,9:16)
10 seconds, 9:16. A cheerful barista with a nose ring and a green apron speaks directly to the camera behind a café counter.
Camera: phone-style selfie framing, slight natural handheld movement.
Light and style: bright café lighting, natural colors, casual social-video look.
Sound: espresso machine hiss, café chatter.
Dialogue:
Barista: "Okay, secret menu time. Ask for it 'upside down.' Trust me."
Timeline:
0–7s: She leans in toward the camera and delivers the line with a wink.
7–10s: She holds up a layered iced drink to the lens.
17. 友善的卡通狐狸(8 秒,1:1)
8 seconds, 1:1. A small, round-faced cartoon fox with big eyes and a striped scarf waves from the doorway of a tree-stump house.
Camera: static medium shot, then a small bounce-in.
Light and style: stylized 3D animation, soft rounded shapes, autumn colors, storybook warmth.
Sound: crunching leaves, a little "hello!" in a bright, playful voice.
Timeline:
0–4s: The fox pops out of the doorway and waves with both paws.
4–8s: It tilts its head, giggles, and a leaf lands on its nose.
动作与运动
18. 日落时分的滑板碗池(8 秒,16:9)
8 seconds, 16:9. A skateboarder in a loose flannel shirt carves through a concrete bowl at sunset.
Camera: low-angle tracking shot, fisheye lens, following the board.
Light and style: golden backlight, long shadows, 1990s skate video look.
Sound: wheels rumbling on concrete, the snap of the tail, distant cheers.
Timeline:
0–5s: He drops into the bowl and carves up the far wall.
5–8s: He pops a frontside air at the lip, silhouetted against the sun.
19. 屋顶跑酷追逐(15 秒,21:9)
15 seconds, 21:9. A runner in a red hoodie sprints across connected rooftops in a dense old city, chased by a figure in black.
Camera: fast handheld follow, then a wide shot for the jump.
Light and style: overcast, desaturated with the red hoodie popping, gritty thriller look.
Sound: heavy breathing, footsteps on tin roofs, a pulsing synth score.
Timeline:
0–6s: Follow shot as she vaults a chimney and slides under a clothesline.
6–11s: Wide shot as she leaps a gap between two buildings.
11–15s: She lands, rolls, and glances back. The pursuer stops at the edge.
20. 在浪管中冲浪(10 秒,16:9)
10 seconds, 16:9. A surfer rides inside the curl of a turquoise wave on a sunny tropical morning.
Camera: starts as a high drone shot, then drops to water level beside the surfer.
Light and style: brilliant sun, clear turquoise water, vivid action-sports look.
Sound: crashing surf, rushing water, then a muffled underwater moment.
Timeline:
0–4s: Drone view of the wave forming and the surfer dropping in.
4–10s: Water-level shot as the surfer crouches inside the barrel, spray catching the light, and shoots out the end.
自然与旅行
21. 黑沙滩(12 秒,21:9)
12 seconds, 21:9. Waves crash onto a black sand beach beneath towering basalt sea stacks on a cold, windy day.
Camera: slow drone push forward at low altitude, then a gentle rise.
Light and style: moody overcast light, white foam against black sand, muted, epic landscape.
Sound: thundering waves, strong wind, seabirds.
Timeline:
0–8s: Low push forward as a wave rushes up the sand toward the lens.
8–12s: The drone rises to reveal the full stretch of coastline.
22. 雨夜小巷(10 秒,9:16)
10 seconds, 9:16. A narrow alley of wooden houses in an old Japanese city on a rainy night, paper lanterns glowing.
Camera: slow walk forward at eye level, steady gimbal.
Light and style: warm lanterns, wet reflective stones, deep blues, romantic travel look.
Sound: rain on rooftops, distant footsteps, a wind chime.
Timeline:
0–6s: The camera glides forward; someone with a clear umbrella turns a corner ahead.
6–10s: The camera stops at a lantern as a raindrop runs down its paper.
23. 沙漠风暴逼近(15 秒,16:9)
15 seconds, 16:9. A massive storm front rolls across a flat desert toward a lone wooden windmill.
Camera: locked-off wide shot, slight time-lapse feel.
Light and style: harsh sun turning to dark, bruised purples, lightning on the horizon, dramatic.
Sound: wind building, distant thunder, the windmill creaking faster.
Timeline:
0–6s: Bright sun; the windmill turns slowly as the dark cloud wall approaches.
6–12s: The sky darkens and dust begins to lift from the ground.
12–15s: A lightning bolt strikes in the distance as the first gust hits the windmill.
社交媒体与 UGC 风格
24. 第一人称开箱(10 秒,9:16)
10 seconds, 9:16. First-person view of hands opening a small kraft paper box on a bed with a knit blanket.
Camera: POV, natural handheld movement, close framing.
Light and style: soft lamp light, cozy, authentic phone-video look.
Sound: paper rustling, a delighted gasp, soft room tone.
Timeline:
0–6s: The hands pull back tissue paper slowly.
6–10s: They lift out a hand-thrown ceramic mug with a speckled glaze and turn it toward the camera.
25. 一镜到底的一天(30 秒,9:16)
30 seconds, 9:16. A young illustrator moves through her small studio apartment over one day, all in a single continuous shot.
Camera: one handheld shot that moves through the rooms with her, never cutting.
Light and style: the light changes naturally from morning to night, warm and lived-in.
Sound: room tone, a kettle, pencil scratching, rain in the evening, a soft lo-fi beat.
Timeline:
0–7s: Morning. She opens the curtains and stretches.
7–15s: Midday. She sketches at her desk, a mug of tea steaming beside her.
15–23s: Afternoon. She holds up a finished drawing toward the window to check it.
23–30s: Night. She turns off the desk lamp and the city glows outside the window.
常见问题排查
| 问题 | 可能原因 | 尝试这样做 |
|---|---|---|
| 视频中段变得模糊 | 相对于时长,指令太少 | 每 6 至 10 秒添加一个节拍 |
| 角色外貌发生变化 | 身份特征细节不足 | 添加参考图片,并在每个节拍中重复关键特征 |
| 过渡看起来像画面故障 | 同时发生的变化太多 | 每个节拍只改变一个主要元素 |
| 镜头运动显得随机 | 镜头描述模糊 | 指定景别和一个具体的镜头运动 |
| 对白听起来不对或互相重叠 | 台词太长或说话者太多 | 每个节拍中每个角色只说一句简短台词 |
| 动作结束得太早 | 动作所需时间不足 | 延长视频或简化动作 |
何时改用 Kling 3.0
Kling 4.0 并不适合所有任务。当你需要 Motion Control(将参考视频中的动作复制到角色身上),或者想要制作 15 秒及以内的短片并通过简单开关控制声音时,Kling 3.0 AI 视频生成器仍然很有用。对于较长且需要规划的场景,则使用 4.0。
在其他模型中尝试相同的提示词
结构清晰的分镜计划并不局限于 Kling。如果想看看相同的提示词在 Seedance 2.5 或 Gemini Omni 等模型中的效果,可以将其粘贴到提供多个视频模型的 AI Studio。一个提示词就能让你了解各个模型的特点。
这些提示词在 Kling 5 中能用吗?
Kling 5 尚未公布,因此还没有针对它的提示词指南。像这样按秒编写的逐镜头提示词并不局限于某一个模型版本;Kling 5.0 发布后,请对照 Kling 发布的说明,按需调整。
资料来源
- Kling AI,Kling 4.0 多关键帧指南,2026 年 9 月 30 日
- Wikipedia,Kling AI,用于查阅版本历史
- Kling5 AI,Kling 4.0 和 Kling 3.0 工具页面及网站常见问题,查阅于 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


