> ## Documentation Index
> Fetch the complete documentation index at: https://docs.ozu.studio/llms.txt
> Use this file to discover all available pages before exploring further.

# Music Prompts

> Writing music prompts that produce the mood, genre, and energy arc your scene needs.

The **Song Node** generates a full music track from a short text prompt. Good music prompts are shorter than you'd think — focused, emotionally specific, and anchored in a clear genre. This page covers what actually moves the needle.

## The structure that works

Every strong music prompt layers the same elements:

**Genre → Mood → Instrumentation → Tempo → Use case**

Example: "A slow, melancholic piano melody over ambient synth textures, suitable for a tragic film scene, 70 BPM, instrumental only."

That's one sentence, five layers. The model has everything it needs.

<Tip>
  **Focus beats verbosity.** "Rainy day jazz cafe" often produces more cohesive output than a paragraph of conflicting instructions. Start short, add detail only where it matters.
</Tip>

## Mood first

Mood descriptors are the most efficient words in a music prompt. "Melancholic," "triumphant," "unsettling," "foreboding," "dreamy," "tense," "serene" — each one bundles harmonic progression, tempo, and instrumentation choices the model makes for you.

You don't need music-theory vocabulary to prompt music. "Sense of building dread" gives the model more to work with than a chord chart.

## Genre sets the palette

Lead with genre. It anchors the sonic world the model draws from:

* **Cinematic orchestral** — strings, brass, percussion. Epic, dramatic.
* **Ambient electronic** — synth pads, drones, sparse rhythm. Atmospheric.
* **Indie folk** — acoustic guitar, soft vocals, banjo, warm room.
* **Jazz** — piano, upright bass, brushed drums, horns. Cafe, noir.
* **Hip-hop** — 808 drums, bass, sampled textures.
* **Rock / post-rock** — electric guitar, live drums, build-and-release dynamics.
* **Lo-fi** — warm analog textures, tape hiss, relaxed tempo.

Blending is fine: "Ambient electronic with jazz piano influence" works. "Lo-fi hip-hop with orchestral strings" works.

## Instrumentation and tempo

Name specific instruments when you want them. "Solo piano" means the piano leads. "808 drums and sub-bass" sets a specific low end. "String quartet" is not the same as "orchestral strings."

Tempo can be given as BPM or as feel:

* **Slow / 60–80 BPM** — ballads, reflective, tragic
* **Mid / 90–110 BPM** — conversational, warm, everyday
* **Upbeat / 120–140 BPM** — action, energetic, driving
* **Fast / 140+ BPM** — chase, tension, adrenaline

## Energy arc

Music in a scene rarely stays at one intensity. Direct the arc:

* "Builds from tense to explosive over thirty seconds"
* "Starts quiet and minimal, adds layers as it progresses"
* "Steady throughout, no dynamic shifts"
* "Resolves to silence in the final five seconds"

Arc descriptions work best for longer tracks. Short cues (under fifteen seconds) don't need one — the model won't have room to develop it.

## What to avoid

* **Contradictory moods.** "Energetic but calm" or "aggressive yet peaceful" forces the model to reconcile opposites. Pick one primary direction.
* **Artist imitation.** "Sound exactly like \[artist]" produces inconsistent results. Describe the qualities you want instead: "Warm analog synths, reverbed vocals, dreamy feel."
* **Keyword stacking.** "Best, high quality, studio grade, professional mastering" does nothing. The model already defaults to high fidelity.
* **Too much structure for a short cue.** A fifteen-second bed doesn't need a verse-chorus-bridge instruction. Save structural direction for longer pieces.

## Vocals

By default, music prompts generate instrumentally. If you want vocals:

* Write **"with vocals"** and describe the vocal quality ("warm female voice," "raspy male tenor," "choral").
* Or provide lyrics. The model structures vocals to match.

For instrumental scoring, explicitly write **"instrumental only"** — otherwise some models add vocal elements you didn't ask for.

## Match the music to the scene

Use the visual mood of your storyboard to guide the music prompt. A noir scene calls for different music than a warm family memory, even if the beats are the same length.

| Visual mood       | Music direction                                              |
| ----------------- | ------------------------------------------------------------ |
| Dark, moody, noir | Sparse piano, upright bass, brushed drums, low ambient drone |
| Bright, upbeat    | Major-key acoustic, steady rhythm, warm strings              |
| Tense, thriller   | Staccato strings, deep bass pulse, percussive stabs          |
| Dreamy, ethereal  | Reverbed synth pads, gentle chimes, slow movement            |
| Action, chase     | Driving percussion, orchestral brass, 140 BPM                |
| Reflective        | Solo piano or acoustic guitar, slow tempo, minor key         |

## Related

<Columns cols={2}>
  <Card title="Song Node" href="/user/audio-and-music/song-node">
    Generating and editing music for your animatic.
  </Card>

  <Card title="Audio Node" href="/user/audio-and-music/audio-node">
    Voice, dialogue, and sound effects alongside music.
  </Card>

  <Card title="Timeline Editor" href="/user/timeline/editor-overview">
    Placing music under your scenes and syncing to frames.
  </Card>

  <Card title="Voice Prompts" href="/user/prompt-techniques/voice-prompts">
    Writing dialogue that reads naturally when spoken.
  </Card>
</Columns>

## FAQ

**How long can a generated track be?**
Length depends on your tier and the model. For most scenes you'll generate thirty seconds to three minutes at a time. Chain tracks or loop a short cue in the Timeline for longer sequences.

**Can I specify BPM exactly?**
Yes. "120 BPM" will pull the model toward that tempo. It's an approximation — expect to land within five BPM of the number you give.

**Why did my prompt produce vocals when I didn't ask for them?**
Some models default to adding vocal elements. Explicitly write "instrumental only" in the prompt to force a pure instrumental.

**Can I generate music that sounds like a specific artist?**
Not reliably, and not for commercial release — imitation prompts are inconsistent and legally murky. Describe the qualities you want (instrumentation, feel, era, production style) instead of naming an artist.

**What if the music doesn't fit my scene?**
Regenerate with a tighter mood direction, or adjust the genre. Music is the element most often reshot — expect to try two or three versions before one lands.

**Can I edit the music after it's generated?**
You can trim, loop, and fade on the Timeline. For deeper edits (changing an instrument, shifting structure), regenerate with an adjusted prompt — editing after the fact is not currently supported.