ACE-Step 1.5 Prompt Guide: Write Better Song Prompts and Lyrics

ACE-Step 1.5 takes two inputs: a style prompt that describes the sound, and optional lyrics that describe what is sung. Good results come from being specific in the first and well-structured in the second.

1. The style prompt

Describe the music the way a producer would brief a session band. A useful order is:

genre → mood → vocals → instruments → tempo/key → production

Weak:

a happy song

Strong:

Upbeat synth-pop, euphoric summer mood, bright female lead vocal with airy harmonies,
punchy four-on-the-floor drums, plucky synth arps, deep sub bass, 124 BPM, A minor,
polished radio mix with wide stereo reverb

Tips:

Name the vocal: male or female, breathy, raspy, choir, auto-tuned, rap
Name 3–5 instruments: a long list blurs the arrangement
Give a BPM when the groove matters (you can also leave it to the model)
Use production words: lo-fi, vinyl crackle, live room, wall of sound, 80s gated reverb
Keep it under 512 characters

2. Lyrics and structure tags

Put each section on its own block with a tag in square brackets:

[Intro]

[Verse 1]
Streetlights flicker on the avenue
Every window glowing something new

[Pre-Chorus]
And I can feel it rising

[Chorus]
We are the summer, we are the light
Running through the city all night

[Bridge]
...

[Outro]

Tips:

Common tags: [Intro], [Verse], [Pre-Chorus], [Chorus], [Bridge], [Outro], [Instrumental]
Keep lines short, about 6–10 syllables, so they fit the beat
Repeat the chorus word for word, which helps the model build a hook
For a song in a non-English language, select it as the vocal language so the pronunciation is correct
Turn on Instrumental (or write only [Instrumental]) for music without vocals

3. Choosing a length

The model fits your lyrics into the length you choose. Too many lyrics for 30 seconds will sound rushed, and too few for 3 minutes leaves long instrumental gaps. A verse plus a chorus fits about 60 seconds, and a full song with a bridge needs 2–3 minutes.

4. Iterate

Generate 2–3 versions of the same prompt. Variation between runs is normal, and the best take is often the second or third one. Change one thing at a time: tempo, vocal or instrumentation.

Ready? Try your prompt now.