Prompting for music

AI music and audio · Lesson 2 / 20

Four parameters that decide

A music prompt is simpler than a text one: the generator understands the language of genre and mood descriptions because it was trained on labelled tracks. You need to set four things.

[GENRE]      lo-fi hip-hop, cinematic orchestra, synth-pop
[MOOD]       calm, tense, triumphant, contemplative
[INSTRUMENTS] soft electric piano, muted bass, brushed drums
[TEMPO/DENSITY] slow, around 80 BPM, sparse arrangement

What does the most

Instruments. Genre sets the general direction; instruments determine what it actually sounds like. "Calm background music" is an empty description; "soft electric piano, warm bass, no drums" is a specific sound.

State the density

The commonest problem with background music is that it's too busy and pushes forward. The phrases "sparse arrangement", "minimal instrumentation", "no sharp accents", "leaves room for a voice" solve it immediately.

If the music sits under speech, say so directly: the generator will keep the mid-range clearer.

A reference instead of a description

A fast technique is to name not a performer but the features: "in the spirit of Nordic detective soundtracks: minimal, low strings, sparse hits". That works more precisely than a genre label and raises no questions about others' rights.

What not to specify

Too many parameters at once. Five instruments, three mood changes and an exact structure in one prompt produce mush. Start with the four parameters and add one at a time.

Insight. Instruments do more than genre. "Calm music" is empty; "soft electric piano, warm bass, no drums" is a specific sound.
Common mistake. Not specifying arrangement density. Background music defaults to busy and pushes forward, competing with speech.
Pro tip. If the music sits under a voice, say so explicitly. The generator frees the mid-range and mixing becomes noticeably easier.

Cheat sheet

  • Four parameters: genre, mood, instruments, tempo and density.
  • Instruments are more specific than genre.
  • "Leaves room for a voice" is a working phrase.
  • Reference by features rather than a performer's name.
1. What sets the sound more precisely than genre?
2. Why does background music often fight speech?
3. How do you set a reference without legal questions?
Task — checked by AI

Generate a background track for a real task of yours, setting all four parameters. Then make three versions: without specifying instruments, without specifying density, and with an explicit "leaves room for a voice". Listen to all of them under your own speech or a real video. Paste the prompts and describe which version worked and why the others got in the way.

← Back

🔒 Answer the question correctly to move on to the next lesson.

Prompting for music — AI music and audio — Skilvy