Field reference
per_scene
One entry per scene row, in assembly order:
| Field | What it holds |
|---|---|
scene_index / start / end | Which scene and its time window |
music.present / music.mood | Whether music plays and its mood — tense, uplifting, dramatic, playful, dark, neutral, sad, epic, mysterious, energetic, or none |
music.genre | Specific — “cinematic-orchestral”, “lo-fi hip-hop”, “ambient-pad”, “synthwave”… |
music.bpm_estimate / music.instruments | Tempo and the 2–4 instruments or synths heard |
music.role | bed | lead | stinger | transition | silence |
sfx[] | Each effect: at, a specific type (“riser-buildup”, “sub-drop”, “tape-stop”, “UI-click”, “cash-register-ding”…), description, intensity (subtle | medium | punchy), purpose (transition | emphasis | emotional_beat | comedic | attention_grab | ambient) |
voice.present / voice.tone | Whether there’s voice and its delivery — “energetic-narration”, “calm-explainer”, “breathy-whisper”… |
summary | The scene’s overall audio vibe in one sentence |
music_segments
Consecutive scenes with the same musical mood are merged into continuous bands — this is the “Music · segmented cuts” list on the teardown page. Each segment: start, end, mood, genre, bpm_estimate, instruments, role (bed | lead | transition | drop | intro | outro).
sound_palette
The whole video’s audio fingerprint:
| Field | What it tells you |
|---|---|
total_sfx_events / unique_sfx_types / type_counts | How busy and how varied the SFX layer is |
dominant_purpose | What the SFX are mostly for — transitions, emphasis, emotional beats… |
music_dominance_pct | How much of the runtime sits under music |
silence_used_as_tool | Whether silence is deployed deliberately — a strong signal in ads |
sound_density | sparse | moderate | dense |
voice_style | The dominant voice tone across the video |
recreation_recipe and signature_moves
recreation_recipe is a 3–4 sentence spec you can hand to a sound designer — overall genre and BPM range, the dominant SFX palette, key transition cues, voice-over style. signature_moves lists the 2–4 audio techniques that make this video’s sound distinctive.
Stems
On demand, the original audio can be split into separated stems — vocals / drums / bass / other — plus two merged logical layers: music (drums + bass) and SFX (other). The vocals stem also powers the voice-isolation workflow for cloning setups. Stems are stored with the teardown and downloadable from the app.What to do with it
- Brief your music search with
music_segments, not adjectives. “Ambient pad, tense, 110 BPM, 0:00–0:37, then uplifting synths to the close” finds the right track in one pass. - Copy the SFX placement, choose your own sounds. If every scene transition carries a riser and the product reveal lands on silence, that architecture transfers — the specific samples shouldn’t.
- Check
silence_used_as_toolbefore adding more sound. High-performing ads often earn their loudest moment by going quiet first.
Related layers
The Cuts
The shots these audio scenes are synced to.
Script analysis
The voice-over element — tone, pace, pauses.
Recreation
The visual half of the rebuild recipe.
Scorecard
Where sound design shows up in the format score.