AI Music Video Prompt Breakdowns: Story, Cut and Style

An AI music video prompt is really three decisions per shot: what the shot shows, where the cut lands, and what the picture looks like. On VdoBloom the Music Video Maker does the timing for you — it analyses the track, finds the beats and sections, and plans a shot per phrase — so your prompts only have to answer story, camera and style. The breakdowns below show how to write that for every section of a song.

Updated October 2026.

How the shot plan is built

You start with a song: generate one in the tool, or upload an mp3, wav or m4a you own, up to 25 MB, and pick a 10 to 60 second window that usually lands on the hook or the drop. The analysis runs on the server and finds the tempo, the beats and the sections, then proposes a shot plan with a cut on every beat — roughly two bars apart inside a section, and a new shot whenever a new section starts. The plan is free to review, and you can merge, reorder or rewrite shots before anything is generated.

Then you choose a look, a 9:16 vertical or 16:9 widescreen frame, and one anchor: upload a photo of the performer and the same face is used across every shot, describe a character and a portrait is generated first, or skip the character for a scenery-led video. Shots render on the Seedance family — Mini, Fast, 2 and 2.5 — at 480p, 720p or 1080p depending on the model. Planning is free; each clip is charged at the model’s own price for its length, plus a flat 10-credit build fee that covers trimming, the beat-cut assembly and muxing your audio. Up to 20 shots per video.

Seven shot prompts, section by section

Write each prompt as one look plus one camera move. The bar notes tell you roughly where the planner will drop the shot.

1. Intro establishment (bars 1-4, Seedance 2 Fast, 720p): “Slow dolly push through a dark doorway into a hall lit by a single hanging bulb; dust in the air, a silhouette standing at the far end facing away. Steady forward move, no pans, deep shadow, photoreal, muted teal grade.”
2. Verse performance (bars 5-12): “Tight close-up on the singer’s face as they mouth the line, eyes closed, one side lit warm and the other falling into black. The camera creeps in slowly the whole shot; shallow depth of field, fine grain, no cut. Keep the face, hair and jacket from the reference photo.”
3. Pre-chorus build (last two bars before the chorus): “Handheld shot circling the singer once at chest height in an empty car park at dusk, lights flicking on behind them as the camera passes. Slight handheld sway, rising energy, cool ambient light, photoreal.”
4. Chorus wide (first bar of the chorus): “Wide shot of the whole band on a rooftop at night, city skyline behind, singer centre frame arms out. Camera cranes up slowly as the skyline lights flare. Bold contrast, saturated rim light, photoreal.”
5. Drop / peak (the loudest beat): “Hard backlit shot, the singer silhouetted as a wall of sparks bursts behind them on the downbeat. Camera locked, wide lens, high contrast with a warm flare through the frame, photoreal and punchy.”
6. Bridge (a quieter section): “Reflected shot in a rain-covered window: the singer on the far side, out of focus, one streetlight smeared across the glass. Camera holds still, soft desaturated blue, slow drifting rain, photoreal.”
7. Outro (final bars): “The singer walks away down a long corridor toward a bright exit; the camera stays put and lets them shrink into the light. Locked frame, even lighting, a long slow fade, photoreal, keep the wardrobe consistent with every other shot.”

Story: one idea per shot

A 10 to 60 second window holds a small number of beats, so each shot should carry exactly one image you could describe in a phrase — doorway, close-up, circle, rooftop. Do not write a mini-film into a single prompt; the planner has already given you a shot per phrase, and a crowded prompt fights the edit. Let the sequence tell the story and keep each prompt to one subject, one action and one camera.

Cut: writing with the beat in mind

You do not set the cuts by hand, but you can write for them. Because a new section always starts a new shot and interior cuts land on detected beats, give each shot a look that snaps cleanly: a clear change of scale or light between neighbours is what makes a beat-cut feel deliberate. A close-up following a wide, a night shot following a day shot, a locked frame following a moving one — those contrasts read on the beat. If a shot feels long, split it into two rather than slowing one down, and reorder the plan until the scale changes fall where the song lifts.

Style: lock the look across the video

Style is the one line you repeat. Decide a grade and a lighting phrase — “muted teal, practical lights, photoreal” — and reuse it in every prompt, so the shots look like one video instead of a playlist. For a performer-led video, the anchor photo already keeps the face consistent; add the same wardrobe phrase to every prompt and the character holds from the doorway shot to the outro. For a scenery-led video, drop the character and let one palette carry the whole piece. The same tool doubles as a lyric video: burn in lyric captions timed from the vocals, or use the standalone lyric video maker when the words are the point.

Frequently asked questions

Do I need Suno to use this?

No. VdoBloom is independent and not affiliated with Suno. The Music Video Maker works with any track you have the rights to use — a song you generated in the tool, or an mp3, wav or m4a file you exported from any app or DAW, up to 25 MB.

How do the cuts land on the beat?

The song is analysed server-side for tempo, beats and sections, and the planner places every interior cut on a detected beat, about two bars apart within a section, with a new section always starting a new shot. If you edit the plan, each new cut still has to sit on a beat.

How long can the video be?

Between 10 and 60 seconds of the song, with up to 20 shots. If your track is longer, slide the start control to the section you want, usually the chorus or the drop, and make separate videos for different parts.

How much does it cost?

Planning is free, and you only pay when you build: each shot at the model’s normal price for its length, plus a flat 10-credit build fee. A few real figures from the table: a 4-second Seedance 2 Mini shot is 7 credits at 480p and 14 at 720p, a 4-second Seedance 2 Fast shot at 720p is 40 credits, and a 4-second Seedance 2 shot at 1080p is 164 credits. The total for your plan is shown before anything renders, and failed clips are refunded.

Can I fix one bad shot without rebuilding the video?

Yes. Any single shot can be regenerated on its own, and only that clip is recharged. You can also reopen a finished video, edit a shot, and rebuild the cut.

Upload your song, review the free beat-synced shot plan and generate on the Music Video Maker, or begin from the Suno song to music video tool page. Generate a fresh track first with the AI music generator, and see our AI music video guide before you export. Videos are charged per shot with the price shown first, failed clips are refunded, and paid plans export watermark-free — plans are on pricing.