FunProductivity

How to Write AI Music Prompts That Get the Song You Imagined

An AI music generator can only follow what you give it. Most disappointing songs come from prompts that name a genre and a feeling and stop there. Here is what to add, and how to say it without music jargon.

A frosted-glass speech bubble shape floating above a chrome tuning fork and three pastel sliders on a small sculptural mixing block.

Type “a sad song” into a music generator and you will get a sad song. It will be slow, it will be in a minor key, and you will have forgotten it by tomorrow. Nothing went wrong: the generator did exactly what it was asked. The prompt simply didn’t ask for anything worth remembering.

A music prompt is a brief. It is the note you would hand to a session band who has never met you. This guide breaks that brief into parts, shows how each one changes the result, and ends with examples you can adapt in ChatUp’s AI music generator.

The six parts of a music prompt

You don’t need all six every time. But when a song comes back wrong, the fix is almost always in one of these.

1. The subject

What the song is about. This matters most when there are lyrics, because the words are built from it.

  • Weak: “a song about love”
  • Strong: “a song about the first winter in our cold flat, wearing coats indoors and laughing about it”

Names, places and objects give the lyrics something concrete to say. If the song is for someone, put their name in.

2. The feeling, and how it changes

Adjectives on their own are flat. Describe the emotional movement instead.

  • Weak: “emotional”
  • Strong: “starts lonely and ends hopeful, like the moment you decide to call someone back”

Movement is what makes a song feel like it goes somewhere.

3. The energy

How busy and how fast the music should feel, in everyday words.

Instead of…Try…
“upbeat”“fast enough to run to, with handclaps”
“chill”“slow, steady, nothing sudden, room to think”
“intense”“builds for a long time, then everything hits at once”
“calm”“sparse, quiet, lots of space between the notes”

4. The instruments

Naming two or three instruments gives the arrangement a clear direction. Say what each one does, not just that it’s there.

  • “Soft piano carrying the melody, warm bass underneath, brushed drums.”
  • “Crunchy electric guitar riff, pounding drums, a big bass line.”
  • “A lone cello at the start, then full strings.”

5. The vocals

If the song has singing, describe the delivery: “a soft, close vocal”, “a big, belting chorus”, “half-spoken verses, sung chorus”, “a choir in the last chorus”. ChatUp has no voice picker, so treat this as direction, not a guarantee. If you don’t want singing at all, don’t describe it away; switch to Instrumental.

6. The structure

Most people skip this, and it’s often the difference between a loop and a song.

  • “A quiet verse, a big chorus, and a stripped-back bridge before the final chorus.”
  • “A short intro, two verses and choruses, and an outro that fades.”
  • “No chorus, one long build from start to finish” (useful for cinematic pieces).

Settings do some of the work

In ChatUp, three things are settings rather than words, and it’s best not to fight them in the prompt:

  • Vocals or Instrumental. A switch, not a sentence.
  • Style. Pop, indie pop, hip-hop, R&B, electronic, rock, acoustic or classical. Your prompt can then narrow it (“garage rock”, “lo-fi”, “neo-soul”).
  • Lyric mood. Serious, funny or heartfelt, when ChatUp writes the lyrics.

Use the prompt for everything the settings can’t express: the story, the scene, the arc, the instruments.

Before and after

A birthday song

Before

Birthday song for my sister.

After

A bright, bouncy birthday song for my sister Nadia, turning 25: she runs every morning at 6, owns too many plants and gives the best advice in the family. Big singalong chorus with her name in it.

Try it with the birthday song generator, mood set to funny or heartfelt.

Background music for a video

Before

Chill music for my vlog.

After

Light, warm instrumental for a cooking vlog: plucked acoustic guitar, soft shaker, a gentle bass line. Steady and cheerful, no big changes, easy to talk over.

Use Instrumental, acoustic style, or start from the background music generator.

A beat

Before

Hip-hop beat.

After

A mid-tempo boom-bap beat: punchy kick, dusty snare, a looping jazzy piano sample, deep warm bass. Head-nodding and relaxed, with a short break every eight bars.

More in how to make beats with AI.

A cinematic piece

Before

Epic orchestral music.

After

A slow cinematic piece that opens with a single piano line, adds strings halfway through, and builds to a full orchestral swell before ending quietly on the opening melody.

The classical music generator starts here.

Common mistakes

Stacking contradictions. “Calm but energetic, sad but uplifting, minimal but epic.” The generator has to pick something, and you won’t like which. Choose one direction, or describe a change from one to the other.

Naming artists instead of describing them. A name tells a person a lot and a generator very little you could not say yourself. Describe the era, the instruments and the vocal style you associate with that artist. You’ll get closer to the sound and keep the song your own.

Writing the lyrics into the prompt. If you have lyrics, use the lyrics box (“Use my lyrics”), not the description. The description is for how it sounds; the lyrics box is for what it says.

Changing everything between takes. If the second version is better, you won’t know why. Change one part of the prompt at a time.

A template to copy

Template

A [energy] song about [subject, with a name or a detail]. It starts [feeling] and ends [feeling]. [Two or three instruments and what they do]. [Vocal delivery, if any]. [Structure: e.g. quiet verses, a big repeated chorus, a short bridge].

Fill in only the parts you care about, pick a style and a mood, and press Create. If you are making your first song, how to make a song with AI covers the whole process from start to finish.

Try it in ChatUp

Turn this guide into a workflow.

Run the prompts above against the model that suits the task, keep the useful context across chats, and pick it back up on any device.

Try for Free