限时特惠爆款 AI 视频提示词大全查看价格方案
AI 视频提示词库

爆款 AI 视频提示词大全

为当下正在被翻拍的热梗整理的“参考视频生视频”提示词,可直接复制。每条提示词都写清楚哪段视频是动作参考(@Video 1)、哪张照片是主角(@Image 1):脸来自你的照片,动作来自原梗。

这些提示词怎么用

@Video 1 = 动作

原梗视频片段。模型只学它的动作、机位和节奏,绝不沿用原表演者的脸。

@Image 1 = 主角

你(或宠物)的照片,保留脸、发型和服饰;双人梗再加 @Image 2。

按节拍写动作

动作逐段写清,让每个动作都落在原曲的节拍上。

按热梗分类的入门提示词

适用于 Seedance、Kling 等支持参考视频的模型,提示词为英文;时长与画幅和原片一致。

Toc Toc 敲门舞 AI

Toc Toc 敲门舞 AI

1 张照片 · 11s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the dancer in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. In a bare white room with a fixed full-body camera: one-foot hops with a knee up, a fist knocking on an invisible door, alternating kicks, then a wide-arm finish facing the camera. vertical 9:16, 11 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
Unforgettable 24 牌甩车 AI

Unforgettable 24 牌甩车 AI

1 张照片 · 11s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the person lip-syncing close to the lens in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Sunset parking lot, selfie framing: lip-sync, hold up a card with a 2, then a 4, flick them at the lens; when the cards clear, a black sports car is parked behind; finish with both hands on the head. landscape 16:9, 11 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
黑猫鼓掌 AI 视频

黑猫鼓掌 AI 视频

1 张照片 · 6s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the cat in @Video 1 with the subject from @Image 1 and keep @Image 1's breed, fur colour, markings and face; do not blend in the original performer's face. Top-down phone view of a beige tiled floor with the owner’s knees at the bottom of the frame: the pet stands on its hind legs, grins and slowly claps its front paws, eyes on the camera. vertical 9:16, 6 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
Hotel Lobby 橙墙双人说唱 AI

Hotel Lobby 橙墙双人说唱 AI

2 张照片 · 10s

@Video 1 is the motion reference: use only its movement, positions, camera work, timing and audio. @Image 1 defines Character 1 and @Image 2 defines Character 2 (face, hair or fur, skin tone, clothes). Character 1 always stands left of the hanging mic, Character 2 always right. Keep the two characters separate for the whole clip: never swap, mix or merge their faces, hair or clothes. Orange wall, one hanging mic, fixed mid shot: Character 1 opens the verse, Character 2 answers, both trade hand gestures, then a wide shot of both moving to the beat. landscape 16:9, 10 seconds, synced to the original audio. Only these two characters, no third person, no text, no watermark.
Crazy Frog 隐形摩托追逐 AI

Crazy Frog 隐形摩托追逐 AI

1 张照片 · 15s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the rider in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Dusk overpass, low camera: the rider crouches in mid-air on an invisible motorbike with fists forward and leans into the curve, a pursuer on a white bike with red lights closes in, then a fast sweep along the curving road. Nothing under the rider. landscape 16:9, 15 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
Alors on Danse 白棚键盘小品 AI

Alors on Danse 白棚键盘小品 AI

2 张照片 · 29s

@Video 1 is the motion reference: use only its movement, positions, camera work, timing and audio. @Image 1 defines Character 1 and @Image 2 defines Character 2 (face, hair or fur, skin tone, clothes). Character 1 plays the laptop and mini keyboard on the left; Character 2 is the hype partner on the right. Keep the two characters separate for the whole clip: never swap, mix or merge their faces, hair or clothes. White seamless studio: Character 1 taps the keys deadpan, Character 2 points, claps, laughs, “coaches” the keys and leans right into the lens, then both walk toward the camera dancing. landscape 16:9, 29 seconds, synced to the original audio. Only these two characters, no third person, no text, no watermark.
NEON 双人开场舞 AI

NEON 双人开场舞 AI

2 张照片 · 15s

@Video 1 is the motion reference: use only its movement, positions, camera work, timing and audio. @Image 1 defines Character 1 and @Image 2 defines Character 2 (face, hair or fur, skin tone, clothes). Character 1 opens the intro alone in front; Character 2 starts hidden behind and steps out. Keep the two characters separate for the whole clip: never swap, mix or merge their faces, hair or clothes. Lilac seamless dance studio, fixed full-body camera: Character 2 steps out and hugs Character 1 from behind with arms raised, sway, hair whips, finger guns at the lens, a spin into a high kick, then a high-low pose. landscape 16:9, 15 seconds, synced to the original audio. Only these two characters, no third person, no text, no watermark.
Victoria’s Secret 维密天使走秀 AI

Victoria’s Secret 维密天使走秀 AI

1 张照片 · 10s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the runway model in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Pink runway under glowing arches, giant white feathered wings on the back: walk out from the far end, wave, strut toward the camera, spin once, pose with a hand on the hip, blow a kiss as sparkler fountains fire behind. vertical 9:16, 10 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
特斯拉飞天跑车 AI 视频生成器

特斯拉飞天跑车 AI 视频生成器

1 张照片 · 12s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the driver in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Sunset rush-hour jam: the driver of a white Roadster smiles and presses a glowing blue button, thrusters fire with white smoke, the car tips upright and rolls in mid-air, then flies off over the traffic toward the sunset. vertical 9:16, 12 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
Sawadika 泰式合十舞 AI

Sawadika 泰式合十舞 AI

1 张照片 · 15s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the dancer in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. In the room from @Image 1: palms together above the head, a quick turn, sharp diagonal arm accents, kicks and hip sways, finishing with a Thai wai at the chest. vertical 9:16, 15 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
Coldplay Kiss Cam 演唱会大屏 AI

Coldplay Kiss Cam 演唱会大屏 AI

2 张照片 · 12s

@Video 1 is the motion reference: use only its movement, positions, camera work, timing and audio. @Image 1 defines Character 1 and @Image 2 defines Character 2 (face, hair or fur, skin tone, clothes). Character 1 is in front; Character 2 hugs from behind. Keep the two characters separate for the whole clip: never swap, mix or merge their faces, hair or clothes. Night stadium, round jumbotron kiss cam: the two sway in a hug, look up and see themselves on the screen, Character 1 covers their face and turns away, Character 2 ducks behind the railing while the crowd cheers. vertical 9:16, 12 seconds, synced to the original audio. Only these two characters, no third person, no text, no watermark.
Bby Wow 咬面包男 AI

Bby Wow 咬面包男 AI

1 张照片 · 9s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the man in the white T-shirt in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Party lights: dance up to the camera, bite a pink concha, chew with puffed cheeks and stare into the lens dead serious. vertical 9:16, 9 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
Ozil 厄齐尔绝杀进球 AI

Ozil 厄齐尔绝杀进球 AI

1 张照片 · 15s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the striker in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Stadium at night: run at the keeper, chip the ball over the dive, control it on the chest, dummy two sliding defenders, then roll it into the empty net. vertical 9:16, 15 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
Burning Bridges 晚宴名场面

Burning Bridges 晚宴名场面

1 张照片 · 15s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the man at the head of the table in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Chandelier banquet with friends around a long table, handheld camera: raise a silver goblet, sing along, wave and cover the eyes with a hand on the beat. vertical 9:16, 15 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
Smalltown Boy 双猫尖叫舞

Smalltown Boy 双猫尖叫舞

2 张照片 · 21s

@Video 1 is the motion reference: use only its movement, positions, camera work, timing and audio. @Image 1 defines Character 1 and @Image 2 defines Character 2 (face, hair or fur, skin tone, clothes). Character 1 starts in front screaming; Character 2 dances behind. Keep the two characters separate for the whole clip: never swap, mix or merge their faces, hair or clothes. Character 1 holds their head and screams while Character 2 dances, they swap places on the beat, dance side by side, then both rush toward the camera. vertical 9:16, 21 seconds, synced to the original audio. Only these two characters, no third person, no text, no watermark.
Baby Knight 荒原小骑士

Baby Knight 荒原小骑士

1 张照片 · 10s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the tiny knight in @Video 1 with the subject from @Image 1 and keep @Image 1's face, fur or hair and proportions; do not blend in the original performer's face. Snowstorm on a dark plain: a tiny figure in silver armor holding a sword stands its ground while a giant spectral guardian rises behind it. vertical 9:16, 10 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
Bulin 47 街头抖肩舞

Bulin 47 街头抖肩舞

1 张照片 · 15s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the man dancing in the center in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Street party under red lights, crowd around: dembow shoulder bounces and footwork to the boombox beat, facing the camera. vertical 9:16, 15 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
Hit the Road Dude 皮卡老哥

Hit the Road Dude 皮卡老哥

1 张照片 · 15s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the driver of the white pickup truck in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Country road: lean out of the pickup window, say the private-property line to the camera with a stern face and point down the road. landscape 16:9, 15 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
万圣节骨架吸血鬼舞

万圣节骨架吸血鬼舞

1 张照片 · 10s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the dancer in the vampire cape in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Full moon night: in a black cape with red lining, a goofy skeleton-style dance with loose arms and knee wobbles. vertical 9:16, 10 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
Topo Gigio 足球捂耳庆祝

Topo Gigio 足球捂耳庆祝

1 张照片 · 8s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the player celebrating in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Stadium floodlights: after the goal, put both hands behind the ears and lean toward the crowd. near-square 6:5, 8 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.
末日丧尸情侣虐恋

末日丧尸情侣虐恋

2 张照片 · 26s

@Video 1 is the motion reference: use only its movement, positions, camera work, timing and audio. @Image 1 defines Character 1 and @Image 2 defines Character 2 (face, hair or fur, skin tone, clothes). Character 1 is the survivor; Character 2 is the zombie partner. Keep the two characters separate for the whole clip: never swap, mix or merge their faces, hair or clothes. Ruined city: a tense standoff, a sudden embrace, then a soft flashback of the couple before the outbreak. vertical 9:16, 26 seconds, synced to the original audio. Only these two characters, no third person, no text, no watermark.
Closed on Sunday 唱诗班 AI

Closed on Sunday 唱诗班 AI

1 张照片 · 14s

@Video 1 is the motion reference: use only its movement, camera work, timing and audio. @Image 1 is the only appearance reference. Replace the person standing in the center in @Video 1 with the subject from @Image 1 and keep @Image 1's face, hair, skin tone and clothes; do not blend in the original performer's face. Sunlit hillside: the subject stands perfectly still while hundreds of robed choir members sway, clap and sing around them. landscape 16:9, 14 seconds, every move on the beat of the original audio. One subject only, no extra people, no text, no watermark.

用提示词还是用模板?

提示词需要你准备参考视频、找到支持参考视频的模型,并多试几次才能对准节奏。下面每个热梗在 TrendSwap 都有现成模板,参考视频、提示词和原声都已配好:上传照片即可生成带声音的视频。

提示词常见问题

这些提示词适用于哪些 AI 视频模型?

适用于支持“动作参考视频 + 1–2 张照片”的参考生视频模型,例如 Seedance 2.x,或带参考视频功能的 Kling。纯文生视频模型会忽略 @Video、@Image 标记。

@Video 1 从哪里来?

使用你有权使用的原梗片段。没有参考视频,模型会自己编动作,节奏也会跑偏。TrendSwap 模板已内置参考视频。

为什么提示词是英文的?

视频模型对英文提示词的执行最稳定。页面其他内容已翻译,提示词本身保留英文。

双人梗怎么避免两张脸混在一起?

每张照片只对应一个角色,并把角色固定在一侧;保留“两人绝不互换、绝不融合”的防互串语句。每张照片只放一个人,不要用合照。