← 提示词库 在 GitHub 查看源文件 ↗
中文

Worked Example — Music Video / Performance (beat-synced)

Original worked example by jnMetaCode (MIT). Applies the 5-stage structure to a music video — and it's the deliberate counterpoint to movie-trailer.md: the trailer bans a music bed and designs sound; the MV is the one genre where the track IS the point. This template teaches when to want music and how to cut to a beat.

Concept: a single performer in a stylised space, cut to a track's rhythm. Swap the artist, look, and tempo; keep the beat-sync discipline.

Further reading (inspiration, not copied — all rewritten in our structure): performance / music categories in jax-explorer/awesome-veo3-videos (no license — inspiration only) and audio-sync notes in geekjourneyx/awesome-ai-video-prompts (MIT).

中文版 →


Variables you need to define first

VariableThis exampleSwap for…
{{artist}}a solo singera band / a dancer / an instrumentalist
{{genre_tempo}}dreamy synth-pop, ~100 BPM, mid-energytrap ~140 / ballad ~70 / hyperpop ~160
{{look}}neon-lit rain on a rooftop at nighta white cyclorama / a desert at golden hour
{{grade}}magenta + cyan neon over deep bluehigh-key pastel / warm film
{{hook_moment}}the chorus drop at ~4sthe beat the whole edit builds to

The complete prompt (copy-paste ready)

1 · Core theme

music video | beat-synced multi-shot | one bold grade | performance energy | no karaoke literalism, no random unmotivated cuts

2 · Character & scene

Performer: {{artist}}. Reference uploaded photo, features 100% preserved, no beautification. Real performance energy — sweat, breath, motion blur on fast moves, a stray hair stuck to the face. Imperfections keep it a real performance, not a polished CG idol.

Scene: {{look}}, built around {{grade}}. The space reacts to the music — lights pulse, rain catches the beat, reflections move.

3 · Atmosphere & quality

Shot on Sony Venice with anamorphic primes; bold stylised grade in {{grade}} (an MV can push colour harder than a narrative film). Film grain, real neon/practical light, motion blur allowed on fast moves. One committed look across the whole piece.

4 · Camera rules

Edited multi-shot cut to the beat — holds on the verse, faster cuts into {{hook_moment}}, a camera move that lands on the downbeat.

5 · Storyboard (beat-synced, ~12s)

Format header — lock the beat math first: 16:9 | 12s | ~100 BPM | 6 shots | beat-synced — stating tempo + shot count up front makes the model land cuts ON the beat instead of drifting. End on the hook: BGM cuts dead, frame freezes.

0–4s · Intro / verse (let it breathe)
Action: {{artist}} in {{look}}, mid-performance, {{grade}} establishing.
        Slower, longer holds matching the verse.
Camera: Slow push or a single slow arc. Breath-float.
Beat:   Cuts land on the downbeats, sparse.

4–8s · Hook / drop ({{hook_moment}} — the payoff)
Action: Energy lifts — the performer hits the chorus, the space reacts
        (lights pulse, rain bursts), motion sharpens.
Camera: Faster cuts, a move that lands hard on the drop.
Beat:   Cut rhythm doubles into the hook.

8–12s · Outro (restrained close)
Action: Energy settles back; one last held performance frame.
Camera: Hold, breath-float.
Close:  No lyric text on screen, no logo card, no freeze-frame wink. Just
        the performer and the space coming to rest as the phrase ends.

Negative prompt (Seedance / Kling — paste into the dedicated field)

blurry, low resolution, watermark, text overlay, lyric subtitles, logo, karaoke text, distorted face, asymmetric eyes, extra fingers, deformed hands, lip-sync mismatch, plastic CG skin, glossy idol render, video-game look, 3D cartoon, flat even studio lighting, grade shifting randomly between shots, unmotivated epileptic cuts, frame flicker, ghosting, lifeless locked-off camera

Why it's built this way

Usage: decide the track's structure first, then map cuts to it — generate the hook beat first since it's the payoff. If you have a real song, generate to temp/silence and sync the real track in post. Keep performers' faces consistent across cuts (generate the hook frame first to lock the look).

Model: Veo 3 and Sora 2 handle performance motion and (if you want it) native audio best; Kling 3.0 is strong on rhythmic action and multilingual dialogue. For real lip-sync to your own track, lay it in post rather than trusting generated vocals.