Music Intelligence System

Drop a song. Get a cost-gated
multi-format music video.

A three-layer stack — keyframe, motion, assembly — turns one finished track into a 4K master and five platform cuts. Every run is priced and approved on a pre-flight card before a second of motion is bought.

See a pre-flight card The stack

The three-layer stack

Keyframe → motion → assembly.

Each layer has one job and routes to the engine that wins for it. The boundaries are the point — settle the look before you pay for motion, and never pay twice for assembly.

L1Keyframe

One locking still per shot — the design decision the whole video inherits. Identity, palette, and composition are settled here before a second of motion is bought.

nano-banana-proGoogle
nano-bananaGoogle
higgsfield-soul-idHiggsfield

L2Motion

Each keyframe becomes 8 seconds of motion. Body shots run cheap; the 2–3 hero shots that carry the video upgrade to the native-audio engine.

veo-3.1-standardGoogle
kling-3.0Kuaishou
seedance-2.0ByteDance
luma-ray3Luma
runway-gen-4.5Runway

L3Assembly

Beat-synced cuts, burned-in captions, audio-reactive overlays, and the six-format export — all local, deterministic, $0 marginal per render.

hyperframesHeyGen (open source)
remotionRemotion
ffmpegFFmpeg

The five style lanes

Pick a lane and the engines pick themselves.

Each lane is a reasoned routing of keyframe, motion, and assembly engines for one look. The lane decides the spend posture before the first prompt is written.

Realistic Cinematic

cinematic

Photoreal, film-grade. Real locations, real light, real faces. The flagship release look.

Keyframe: nano-banana-pro
Motion: kling-3.0
Hero: veo-3.1-standard
Assembly: hyperframes

Character Narrative

character

A recurring artist/avatar carries a story across every shot. Identity must hold across 10-12 cuts.

Keyframe: nano-banana-pro
Motion: kling-3.0
Hero: veo-3.1-standard
Assembly: hyperframes

Anime / Animated

anime

Stylized 2D/2.5D or anime-inflected. Expressive, saturated, motion-forward.

Keyframe: nano-banana
Motion: seedance-2.0
Hero: kling-3.0
Assembly: hyperframes

Abstract / Visualizer

abstract

Non-representational. Light, particles, fluid, shader. Audio-reactive. Mood over narrative.

Keyframe: nano-banana
Motion: luma-ray3
Hero: runway-gen-4.5
Assembly: hyperframes

Lyric Typography

lyric

Type-driven. Synced lyrics over atmospheric b-roll or solid/gradient fields. No video-gen required for the type itself.

Keyframe: nano-banana
Motion: — compositor only
Hero: —
Assembly: hyperframes

The six formats

Master once in 16:9 4K. Derive the rest.

The full release ships every platform from a single master. Five of the six are reframes or cuts — cheap by design.

Format	Aspect	Resolution	Max duration	Platforms
YouTube — Full Music Videoyoutube-full · master	16:9	3840×2160	unbounded	youtube
Shorts / TikTok / Reelsshort	9:16	1080×1920	180s	youtube-shorts, tiktok, instagram
Spotify Canvascanvas	9:16	1080×1920	8s	spotify
Apple Music — Motion Artworkapple-artwork	1:1 + 3:4	3840×3840	35s	apple-music
Visualizer Loopvisualizer · master	16:9 + 9:16	1920×1080	unbounded	youtube, instagram, spotify
Instagram Feed (1:1)square	1:1	1080×1080	90s	instagram

The pre-flight card

What you approve before a run.

No render starts without this. Design brief, shot summary, per-engine cost, the total in dollars and credits, and the ROI it has to clear — one surface, one decision.

Pre-flight · approve before render

The Awakening

Realistic Cinematic · 17 shots (2 hero) · 117s · 1080p · 104 BPM

Ready to render

Design brief

Audience: Late-night listeners who find the track on Discover Weekly and decide in the first two seconds whether to stay.
Emotional arc: A held, beautiful frame resolves into a single committed camera move that lands on the downbeat — wide establishing, then intimate close on the hook.
Hook doctrine: Open on stillness; the first motion is the push-in timed to beat one. No cut before the listener is already in.

YouTube — Full Music VideoShorts / TikTok / ReelsSpotify CanvasApple Music — Motion ArtworkInstagram Feed (1:1)

Cost — 2× selects baked in

Line item	Qty	USD
Nano Banana Pro (Gemini 3 Pro Image) keyframesL1 Keyframe	17 image	$2.28
Soul ID character lockL1 Keyframe · 40 cr	1 character	$2.50
Kling 3.0 — body (15 shots)L2 Motion · 130 cr	206.3 second	$6.32
Veo 3.1 Standard — hero (2 shots)L2 Motion	27.7 second	$11.08
HyperFrames — assemble + 5 formatsL3 Assembly	5 clip	$0
Total	170 cr	$22.18

170 Higgsfield credits ≈ $8.33 at the Plus rate, already counted in the total.

ROI — Spotify breakeven: ~5,545 streams
ROI — sync breakeven: 1 sync deal
Resonance: unscored — run virality_predictor

Sample: cinematic lane, full-release format set, 2× selects. Numbers are live from the rate card — verdict and totals recompute when the plan changes.

The production backlog

21 songs released, 21 awaiting video.

Released tracks with no full video on file. The next run starts here.

The Awakening
franks-vibes
1:57
Vibe O S
franks-vibes
4:00
Golden Age of Intelligence
nona
2:35
Lumina
arcanea
4:13
Starlight Delight (Remastered)
nona
2:59
Trust in Yourself
nona
2:19
Arcanea (light me up) (Remastered)
arcanea
3:08
I Feel the Vibe
franks-vibes
2:34
Art Of Soulful Living
nona
3:44
Magical Times
frank-riemer
3:21
Golden Frequencies v4
frank-riemer
2:39
Golden Frequency Choir
frank-riemer
3:26

+ 9 more in the queue.

The command reference

Five commands drive the whole pipeline.

Run from Claude Code. Each plans or renders against the same typed substrate — nothing publishes without the pre-flight gate.

/music-video
Plan a full release. Reads the song, builds the shot plan, renders the pre-flight card, waits for approval.
/mv-render
Execute an approved plan — keyframes, motion, assembly, then all six formats from the one master.
/mv-canvas
Spotify Canvas only. A 3–8s seamless loop under 8MB, derived from the master, no text or CTAs.
/mv-artwork
Apple Music motion artwork — ProRes 1:1 + 3:4, first frame locked to the static cover.
/mv-visualizer
Audio-reactive visualizer loop in HyperFrames. No video-gen spend — waveform/spectrum driven.

Drop a song. Get a cost-gatedmulti-format music video.