成片拆解
来源 · X

粉色夜城说唱 MV · MiniMax H3
12 秒真人风短片,怎么做出来的?

粉色霓虹夜城说唱 MV:每句歌词换一个雨夜场景,大字歌词印在她身后的墙、围栏和楼上。

套路 · 字效·片头套路 · 角色表演12秒 · 说唱 MV16:9 横屏11 个镜头歌词大字排版角色卡 + 音频驱动

钩子在哪

时间为抽帧估计

结构:每句歌词一个场景 → 歌词大字落在她身后的真实表面 → 卡在拍点上硬切

看完成片再对照:这条片子靠什么在几秒内抓住人。点时间可跳到成片对应位置。

开场钩子

第 0 秒她站在砖墙小巷里,身后一道竖向粉色霓虹;0.25 秒起「PINK AFTER DARK」三行大字从积水倒影里升起,贴到左边墙上。

场景怎么换

约 1.75s 切到地铁站近景,她整理马甲领口;2.5s 砖墙上贴出「CITY ON MY SKIN」海报;3.25s 铁丝网横幅「YOU CAN'T PULL ME IN」;3.75s 雨夜斑马线「I WALK LIKE THUNDER」,脚下一圈粉色涟漪。

中段

约 5s 天台粉烟里浮出「BLACK SKY BURNING」;6.25s 地下车库,「HEARTBEAT IN THE STREET」外套银色心电圆环;7.25s 玻璃大楼窗格拼出「I OWN THE NIGHT」,她摸头顶墨镜,随后窗格一排排熄灭。

结尾怎么收

约 8.5s 地下通道壁画上斜着落下「MOVE TO MY RHYTHM」;9.75s 切到隧道口;10–12.3s 近景,身后巨型「CONCRETE」和小字「WATCH ME LIGHT THE」,她看着镜头唱完最后一句。

跟做照抄这一点

把字写成场景的一部分:规定字体(超窄粗黑体、粉色填充 + 细银边)、字在她身后的哪种表面上(墙、横幅、窗格、壁画),并要求字离镜头比人远、不挡脸、拼写正确。

三步搞懂

别急着上视频模型。这个片子的秘密是:先把角色和场景锁死,再一次生成 11 个镜头。

1

第一步:准备角色卡、歌词场景图和音轨

作者在主帖放了三张图:一张角色卡(全身 + 脸部特写 + 半身,粉色马甲造型)、一张 3×3 歌词场景图(9 个场景各配一句歌词大字),和一张提示词长图。角色卡和 9 张场景图都已收录。提示词里的 @Audio1 是说唱音轨,作者没有公开。

2

第二步:写一段三次生成都用的公共设定

开头的 SHARED BLOCK 每次生成都贴:@Image1 只管脸、发色、妆和服装,@Audio1 只管节奏和口型;每个场景只有一个酸性玫瑰粉霓虹光源、积水从下往上补光;色调约 70% 黑 + 27% 粉 + 3% 银;写明服装细节、字体规则和机位范围;音乐 160 BPM、每个切点落在小节第一拍、不要模型自己生成音乐。

3

第三步:分三次生成,每次 3 个镜头

RUN A、B、C 各写 3 个镜头(5 秒、5 秒、6 秒),每个镜头都写:机位和镜头焦段、灯光变化、运镜、动作和要唱的歌词、字怎么出现、怎么转到下一个镜头(硬切、甩镜、光扫、烟雾转场、匹配剪辑、推镜爆闪)。结尾统一贴技术限制:只出现列出的字、不要字幕、水印、Logo、多余的人、不要换装换脸。

步骤一 · 做出这 10 张参考图

推荐用 GPT Image / 同等图模,比例 4:3。每张点开即可复制完整英文提示词。

@Image1

@Image1 · 角色卡 Nastya

主帖第一张图原件 HQ0fbK0WUAAsL00(1536x1024)· 全身 + 脸部特写 + 半身,AI 角色卡

第 1 步
@Image1 · 角色卡 Nastya
▸查看完整出图提示词(点击复制区)
复制
作者未公开出图提示词。视频提示词中的用法:@Image1 = character card. Controls face, grey blue eyes, ash brown hair, makeup and wardrobe only; the backdrop of @Image1 stays out of the output.
1

歌词场景图 · PINK AFTER DARK

砖墙小巷 + 竖向粉色霓虹;主帖第二张 3×3 拼图裁出

第 2 步
歌词场景图 · PINK AFTER DARK
▸查看完整出图提示词(点击复制区)
复制
作者没有公开这张图的出图提示词。
2

歌词场景图 · CITY ON MY SKIN

黑砖墙 + 三条粉色海报;主帖第二张 3×3 拼图裁出

第 3 步
歌词场景图 · CITY ON MY SKIN
▸查看完整出图提示词(点击复制区)
复制
作者没有公开这张图的出图提示词。
3

歌词场景图 · I WALK LIKE THUNDER

雨夜斑马线 + 粉色涟漪;主帖第二张 3×3 拼图裁出

第 4 步
歌词场景图 · I WALK LIKE THUNDER
▸查看完整出图提示词(点击复制区)
复制
作者没有公开这张图的出图提示词。
4

歌词场景图 · YOU CAN'T PULL ME IN

铁丝网 + 粉色横幅;主帖第二张 3×3 拼图裁出

第 5 步
歌词场景图 · YOU CAN'T PULL ME IN
▸查看完整出图提示词(点击复制区)
复制
作者没有公开这张图的出图提示词。
5

歌词场景图 · BLACK SKY BURNING

天台 + 粉色烟雾;主帖第二张 3×3 拼图裁出

第 6 步
歌词场景图 · BLACK SKY BURNING
▸查看完整出图提示词(点击复制区)
复制
作者没有公开这张图的出图提示词。
6

歌词场景图 · HEARTBEAT IN THE STREET

地下车库 + 银色圆环心电线;主帖第二张 3×3 拼图裁出

第 7 步
歌词场景图 · HEARTBEAT IN THE STREET
▸查看完整出图提示词(点击复制区)
复制
作者没有公开这张图的出图提示词。
7

歌词场景图 · I OWN THE NIGHT

玻璃大楼窗格亮字;主帖第二张 3×3 拼图裁出

第 8 步
歌词场景图 · I OWN THE NIGHT
▸查看完整出图提示词(点击复制区)
复制
作者没有公开这张图的出图提示词。
8

歌词场景图 · MOVE TO MY RHYTHM

地下通道几何壁画;主帖第二张 3×3 拼图裁出

第 9 步
歌词场景图 · MOVE TO MY RHYTHM
▸查看完整出图提示词(点击复制区)
复制
作者没有公开这张图的出图提示词。
9

歌词场景图 · WATCH ME LIGHT THE CONCRETE

隧道口粉色天际 + 巨型字;主帖第二张 3×3 拼图裁出

第 10 步
歌词场景图 · WATCH ME LIGHT THE CONCRETE
▸查看完整出图提示词(点击复制区)
复制
作者没有公开这张图的出图提示词。

步骤二 · 11 镜头故事板(人话版)

方便你对照成片理解每一镜在干什么。

  1. Shot 1 — 0–1.6s 砖墙小巷,竖向粉色霓虹;「PINK AFTER DARK」从积水倒影里升起贴到墙上。
  2. Shot 2 — 1.75–2.4s 地铁站近景,背后列车掠过,她整理马甲领口。
  3. Shot 3 — 2.5–3.1s 黑砖墙,「CITY / ON MY / SKIN」三条粉色海报贴在她身后。
  4. Shot 4 — 3.25–3.6s 铁丝网前,粉色横幅「YOU CAN'T PULL ME IN」。
  5. Shot 5 — 3.75–4.9s 雨夜斑马线,她朝镜头走来,「I WALK LIKE THUNDER」斜体大字,脚下粉色涟漪扩散。
  6. Shot 6 — 5–5.9s 天台低角度,粉色烟雾里浮出「BLACK SKY BURNING」,风吹起头发。
  7. Shot 7 — 6–7.1s 地下车库粉色灯管,「HEARTBEAT IN THE STREET」外套银色圆环和心电线。
  8. Shot 8 — 7.25–8.3s 玻璃大楼窗格拼出「I OWN THE NIGHT」,她摸头顶墨镜,窗格一排排熄灭。
  9. Shot 9 — 8.5–9.6s 地下通道几何壁画,「MOVE / TO MY / RHYTHM」斜着落下,她甩手腕转身。
  10. Shot 10 — 9.75s 隧道口,远处粉色天际,隧道灯依次亮起。
  11. Shot 11 — 10–12.3s 近景,身后巨型「CONCRETE」和小字「WATCH ME LIGHT THE」,她看镜头唱完最后一句。
关键约束:只出现列出的字,拼写正确,不要字幕、水印、Logo;字不挡脸;只有她一个人,不换装不换脸;不要模型自己生成音乐。与成片不符:提示词写 9:16、分三次生成(5+5+6 秒),成片是一条 12 秒 16:9 横屏;成片里的字是整句歌词(如「YOU CAN'T PULL ME IN」),提示词只写了部分关键词(如「PULL ME IN」);「CITY ON MY SKIN」在成片里是地铁近景后另切一个砖墙镜头,不是贴在地铁瓷砖上;成片把铁丝网镜头放在斑马线之前。缺口:@Audio1 音轨没有公开;9 张歌词场景图怎么用(是否作为参考图上传)作者没说;口型和歌词没有逐句核对。

步骤三 · 完整视频提示词

直接全选复制到 Seedance。@Image 1–10 必须对应上面 REF 编号。

GEN 01

Pink After Dark · Rap Music Video

MiniMax H3 · 提示词写 9:16(成片 16:9)· 英文完整提示词(主帖第三张图,逐字转录)

复制
=== SHARED BLOCK – paste at the top of all three runs ===

REFS:
@Image1 = character card. Controls face, grey blue eyes, ash brown hair, makeup and wardrobe only.
@Audio1 = the track for this segment. Controls tempo, flow, phrasing and every lip position only.
ACTIVE REFERENCES: @Image1 100%; @Audio1 100%. Framing is fresh in every shot, the backdrop of @Image1 stays out of the output.

GLOBAL STYLE NOTES:
lighting_philosophy: one acid rose neon source per location, the brightest thing in frame, wet ground returning it as fill from below, everything outside its throw falling to black
color_grade: 35mm film still, push processed, fine grain, halation off every pink source. ~70% black and carbon in asphalt, brick and concrete + ~27% acid rose in the neon, the lettering and the wet reflections, powder pink on her wardrobe + ~3% chrome silver on rings and letter outlines
setting: rain soaked black city at night, a new location on every cut, real time
ambience: rain on asphalt at three distances, tyre hiss, vent hum, dripping concrete, swapping on the same frame as every cut
characters: <Nastya> corresponds to @Image1. Powder pink cropped puffer vest open over a powder pink long sleeve crop top, oversized powder pink parachute trousers, pink and grey plaid shirt knotted at her hips, grey faux fur boots with pink laces, black sunglasses pushed up on her head, silver rings, long ash brown layered hair loose in the wind, visible pores. Keep identical in every shot. She is the only person in the video.
typography: ultra condensed heavyweight sans in capitals, acid rose fill with a thin chrome silver outline. The words live on a real surface behind her, painted, printed, projected or lit, always further from the lens than she is. They land in the first half of the shot and hold crisp and correctly spelled to the cut. Nothing crosses her face.
camera_character: handheld throughout, operator breath, hand tremor, small live reframes in every shot
location_map: she holds the centre of the vertical frame, x 25% to 75%, the lettering and the neon behind her in the upper and outer thirds

MUSIC_LAYER:
instrumental: supplied by @Audio1, 160 BPM half time, minor key, sub bass, tight kick, dry clap. Generate no music of your own.
lyrics: rapped to @Audio1, one line per shot as listed
vocal_style: rapping, percussive consonants, cold and unhurried, hitting the beat rather than sliding across it
audio_visual_sync: every cut lands on the first beat of a bar, 36 frames apart, kick on each cut, the words locking on the clap
lip_sync_mode: rapping, English, every consonant matched to @Audio1


=== RUN A ===

PINK AFTER DARK, SEGMENT A: IGNITION > CLAIM > SURGE
SHOT STRUCTURE: 3 shots, 5 seconds, 9:16, 24 fps, exactly as listed, no added shots

SEQUENCE LIST:
SHOT 1 (0 to 1.5s) HOOK
camera: waist up, eye level, three quarter 45. lens: 47 standard, no focal drift
camera_motion: slow push in
action_visual: brick alley, one vertical acid rose slit burning at the far end, rain falling through three depth planes into breaking puddles. She rolls one shoulder forward, turns into the lens and raps: 'Pink after dark'
text: 'PINK AFTER DARK' stacked in three lines, rising out of the wet pavement reflection and locking upright on the brick wall at frame left
exit: her chin drops on the last syllable
(HARD CUT TO)
SHOT 2 (1.5 to 3s) SETUP
camera: medium close up, eye level, frontal 0. lens: 29 short telephoto
LIGHT SHIFT: fluorescent tube overhead, one flicker, pink bleeding off wet subway tile
camera_motion: small arc around her left shoulder
action_visual: black subway tile, a train passing as a blurred band behind glass. She straightens her vest collar and raps: 'City on my skin'
text: 'CITY', 'ON MY' and 'SKIN' printed on three pink poster strips that slide in from three directions and flatten onto the tile behind her, corners lifting in the airflow
exit: the bottom poster corner peels and whips past the lens
(WHIP PAN TO)
SHOT 3 (3 to 5s) RISING
camera: knees up medium wide, hip height, profile 90 turning frontal. lens: 65 moderate wide
LIGHT SHIFT: distant pink signals, long reflections running the wet crosswalk
camera_motion: retreating dolly, she closes the gap
action_visual: rainwater sheeting across crosswalk stripes, mist off a street vent. She steps into the lens, shoulder rolling through it, and raps: 'I walk like thunder'
text: 'THUNDER' in heavy oblique capitals slams down onto the wet crosswalk behind her, one acid rose ripple leaving the impact
exit: her boot lands and the ripple widens past the frame


=== RUN B ===

PINK AFTER DARK, SEGMENT B: RESISTANCE > ELEVATION > PULSE
SHOT STRUCTURE: 3 shots, 5 seconds, 9:16, 24 fps, exactly as listed, no added shots

SEQUENCE LIST:
SHOT 1 (0 to 1.5s)
camera: waist up, eye level, three quarter 45 from her right. lens: 47 standard
LIGHT SHIFT: pink headlight streaks sweeping behind chain link
camera_motion: lateral track right, holding her frame left
action_visual: rain beading down black wire, the fence shivering as a vehicle passes. She turns one shoulder away, looks straight back into the lens and raps: 'You can't pull me in'
text: 'PULL ME IN' printed across a pink banner zip tied to the chain link behind her, stretching taut with the fence then snapping flat on the clap
exit: a headlight streak slices the frame horizontally
(LIGHT WIPE TO)
SHOT 2 (1.5 to 3s) SIGNATURE
camera: medium close up, low angle, frontal 0. lens: 29 short telephoto
LIGHT SHIFT: acid rose city glow from below, matte black cloud above
camera_motion: slow rise, tilting with her chin
action_visual: rooftop under fast carbon cloud, pink smoke rolling along the deck, wind lifting her hair and collar. She raises her chin into it and raps: 'Black sky burning'
text: 'BLACK SKY' above 'BURNING', revealing bottom to top through the rolling pink smoke behind her
exit: smoke rushes the lens and fills the frame
(SMOKE WIPE TO)
SHOT 3 (3 to 5s)
camera: waist up, eye level, three quarter 45. lens: 47 standard
LIGHT SHIFT: pink ceiling tubes pulsing down the garage in sequence
camera_motion: handheld hold, breathing only
action_visual: black concrete pillars, vent fog crossing the floor, puddles trembling on the bass. She pulses both shoulders twice on the kick and raps: 'Heartbeat in the street'
text: 'HEARTBEAT' inside a thin chrome silver ring behind her, the ring expanding once on the kick and freezing around the word
exit: the tubes darken toward her one by one


=== RUN C ===

PINK AFTER DARK, SEGMENT C: OWNERSHIP > RELEASE > SIGNATURE
SHOT STRUCTURE: 3 shots, 6 seconds, 9:16, 24 fps, exactly as listed, no added shots

SEQUENCE LIST:
SHOT 1 (0 to 1.5s) CLIMAX
camera: tight medium close up, eye level, frontal 0. lens: 29 short telephoto
LIGHT SHIFT: a grid of dark windows behind her, single cells lighting bubblegum pink
camera_motion: push in to a stop
action_visual: black glass tower, traffic reflections crossing the facade, fine rain sliding down it. She touches the sunglasses on her head without lowering them and raps: 'I own the night'
text: 'I OWN THE NIGHT' assembled from individual lit window cells across the tower grid behind her, one word per floor
exit: the lit windows switch off in a wave, one vertical pink column left burning
(MATCH CUT TO)
SHOT 2 (1.5 to 3s)
camera: medium, eye level, profile 90 turning three quarter. lens: 65 moderate wide
LIGHT SHIFT: hard pink projection sweeping a geometric mural
camera_motion: arc right around her, ending frontal
action_visual: concrete underpass, black and pink mural planes, ground fog sliding through. She shifts her weight, flicks one wrist and turns back into the lens rapping: 'Move to my rhythm'
text: 'MOVE', 'TO MY' and 'RHYTHM' in three oblique lines dropping onto the mural planes from three directions and locking together on the clap
exit: the projection sweeps diagonally off the lens
(ZOOM BURST TO)
SHOT 3 (3 to 6s) RESOLUTION
camera: close up, eye level, frontal 0. lens: 29 short telephoto
camera_motion: push in for two seconds, then settle and hold
action_visual: tunnel mouth opening onto a saturated pink horizon, tunnel lights flickering in sequence toward it, fog drawn forward past her, wet pavement carrying the reflection to the lens. She raps the full line into the lens: 'Watch me light the concrete'
text: 'CONCRETE' in monumental condensed capitals rising vertically out of the wet pavement behind her, 'WATCH ME LIGHT THE' small and crisp on the line above it
reaction: the last syllable releases, her shoulders drop, a controlled half smile arriving a beat late while her eyes stay on the lens
exit: the pavement reflection pulses once and settles, holding her gaze for the final eight frames


=== TECHNICAL CONSTRAINTS – paste at the bottom of all three runs ===

only the words listed appear in frame, spelled exactly as written, no captions, no subtitles, no misspelled or invented letters, no text over her face, no watermark, no logos, no generated music, no extra people, no costume changes, no identity changes

这是 Demo

当前是单页静态站:视频 + 指引 + 可复制提示词。后面可以做成多教程列表、账号、收藏提示词、一键复制、手机竖屏版等。

返回首页,看更多教程