AI Music Prompting Guide: Get Better Results From Text-to-Music

Song Creator Pro is powered by ACE Step 1.5, an open-source AI music generation model that runs entirely on your PC. It produces 48kHz stereo audio that ranks competitively with cloud-based services like Suno.

This guide covers everything you need to write prompts that produce fantastic music in Song Creator Pro. Here's the short version to get you started — keep reading for the full deep dive.

The Quick Version

If you just want to start making music, here's what you need to know:

Turn on AI Enhance — it's the toggle above the description field. It reformats your prompt so the model understands it better.
Be specific in your description — don't write "a sad song." Write "melancholic piano ballad, soft female vocals, gentle strings, slow tempo." Include genre, instruments, mood, and vocal style.
Break your lyrics into sections — use tags like [Verse], [Chorus], [Bridge]. For instrumentals, just leave the lyrics blank.
Leave settings blank — BPM, key, and time signature are auto-detected from your description. Only set them manually if you need precise control.
Generate multiple versions — batch generate 2-4 variations and pick the best one.

That's it. Those six things will get you 80% of the way to great results. The rest of this guide goes deeper into each area if you want to fine-tune your workflow.

How Song Creator Pro Interprets Your Input

When you create a track in Song Creator Pro, your input is split into three channels that the underlying ACE Step 1.5 model processes together:

Description — The main text field where you define the overall sound: genre, mood, instruments, production style, vocal character.
Lyrics — Your song's words. You can paste lyrics as-is, but results are better when you break them into sections like [Verse], [Chorus], and [Bridge].
Settings — Numerical controls: BPM, key/scale, time signature, duration, and vocal language. You can leave these empty — the model is smart enough to infer them from your description and lyrics.

If you do set BPM, key, or time signature manually, make sure they're consistent with your description. Setting BPM to 160 while your description says "slow piano ballad" confuses the model and quality drops. When in doubt, leave the settings blank and let the model decide.

Writing Better Descriptions

The single biggest factor in output quality is how much detail you put in your description. The more the model knows about what you want, the better it can deliver. Vague prompts produce vague results.

The Key: Include As Much Detail As Possible

Think of it like briefing a session musician. "Play something sad" gives them almost nothing to work with. "Play a slow piano ballad with soft strings, breathy female vocals, and a reverb-heavy intimate feel" gives them everything they need.

Here are the dimensions to cover in your description:

Genre and era — pop, rock, jazz, "80s synth-pop", "90s grunge"
Instruments — acoustic guitar, piano, synth pads, 808 drums, strings
Mood — melancholic, uplifting, energetic, dreamy, aggressive
Timbre and texture — warm, bright, crisp, lo-fi, polished, raw
Vocal character — female vocal, male vocal, breathy, powerful, raspy
Production style — lo-fi, live recording, bedroom pop, orchestral

You don't need to hit every single one, but the more you cover, the more control you have over the output. The difference between a one-line description and a detailed one is dramatic.

Vague (unpredictable results):

a sad song

Detailed (much better results):

slow piano ballad with soft female vocals, gentle string accompaniment, intimate and heartbreaking atmosphere, reverb-heavy mix

AI Enhance

Song Creator Pro has an AI Enhance toggle above the description field. When enabled, it takes your description and automatically reformats it into the format that the underlying model understands best.

This means you don't need to worry about syntax or formatting — just focus on including as much detail as you can about the sound you want. Write however comes naturally, and AI Enhance will handle the rest.

Use Positive Language Only

Describe what IS present, never what should be absent. "Without drums" or "no bass" does not work — the model processes additive descriptions only. If you want a track without drums, list only the instruments you do want.

Structuring Lyrics

The lyrics field is the temporal blueprint of your song. Bracketed section tags tell the model how the song's energy and arrangement should evolve over time.

Section Tags Reference

Tag	Purpose
`[Intro]`	Opening section, establishes atmosphere (usually instrumental)
`[Verse]` or `[Verse 1]`	Narrative progression, moderate energy
`[Pre-Chorus]`	Builds tension toward the chorus
`[Chorus]`	Emotional peak, highest energy
`[Bridge]`	Shift in melody or feeling
`[Instrumental]`	Pure instrumental, no vocals
`[Guitar Solo]`	Named instrument feature
`[Build]`	Gradual energy increase
`[Drop]`	Energy release (electronic music)
`[Breakdown]`	Reduced instrumentation
`[Outro]`	Conclusion, wind-down
`[Fade Out]`	Fading ending

You can also add modifiers with hyphens for additional nuance:

[Chorus - anthemic]
[Verse - whispered]
[Bridge - building]

Special Formatting

UPPERCASE — signals emphasis or shouted delivery
(Parentheses) — processed as background vocals or harmonies
[Instrumental] between sections — creates dynamic variety and breathing room

A Complete Example

Here's a well-structured prompt that covers all three input channels in Song Creator Pro:

Description:

indie rock, electric guitar, driving drums, male vocal, energetic, bright, garage band feel, slightly distorted, 130 BPM

Lyrics:

[Intro]

[Verse 1]
Burning daylight on the road
Nothing left and nothing owed
Windshield cracking in the heat
Tapping rhythms with my feet

[Pre-Chorus]
Tell me is it worth the drive
Tell me are we still alive

[Chorus]
We're running out of highway
Running out of time
But I don't want to slow down
This reckless state of mind

[Instrumental]

[Verse 2]
Radio is fading out
Static hums replace the sound
Desert stretching either way
Nothing left for us to say

[Chorus]
We're running out of highway
Running out of time
But I don't want to slow down
This reckless state of mind

[Outro]

Generating Instrumentals

For instrumental tracks, just leave the lyrics field blank and make sure your description focuses on the music — genre, instruments, mood, production style — without mentioning vocals. That's it. The model will produce an instrumental track.

For instrumental breaks within a vocal song, insert [Instrumental] between lyric sections to create solos or breathing room.

Technical Settings That Matter

Song Creator Pro exposes several settings from the underlying ACE Step 1.5 model that significantly affect output quality.

Inference Steps

Inference steps control how many passes the model makes to refine your audio. Each step removes a layer of noise and adds detail — instruments come into focus, vocals emerge, drums tighten up, and the mix clears. More steps means higher quality but slower generation.

Model	Recommended Steps	Best For
Turbo	8 (default, optimal)	Remixes and style transfers
SFT	32-50	Custom songs and new compositions
Base	50-100	Final production output, maximum quality

Guidance Scale (CFG)

Guidance scale controls how strictly the model follows your description versus taking creative liberties. A higher value means the output sticks closer to what you asked for. A lower value gives the model more freedom to interpret and improvise.

This setting only applies to non-turbo models (base and SFT):

3.0-7.0 — the recommended sweet spot, balancing prompt adherence with natural-sounding results
Below 1.5 — the model tends to ignore your prompt and do its own thing
Above 12 — audio artifacts and distortion become common as the model over-constrains itself

In our testing, 5.0 is a reliable default for the SFT model.

BPM and Tempo

BPM ranges from 30-300, but the practical sweet spot is 60-180. Extreme values at either end are unstable because the model has less training data in those ranges. Slow jazz at very low BPMs is a known weak area.

Duration

The supported range is 10-600 seconds. For tracks longer than 120 seconds, we strongly recommend explicitly setting BPM.

The Batch Generation Workflow

This is the workflow advice we'd give anyone starting with Song Creator Pro: never generate a single version of anything.

AI music generation involves inherent randomness. Even with identical prompts, each generation produces different results. The practical workflow is:

Generate 2-4 variations of the same prompt using Song Creator Pro's batch generation
Pick the best one (or the best sections from each)
Iterate on the prompt — adjust the description (tempo/mood/instruments)

Quick Reference: Prompt Checklist

Before you hit generate in Song Creator Pro:

Pack your description with detail — genre, instruments, mood, vocal style, production feel. Keep AI Enhance enabled.
Leave settings blank — reset them between generations so the model can pick appropriate values. If you do set them manually, make sure they don't contradict your description.
Set batch size to at least 2 — multiple generations means faster iteration and better odds of getting a great result.

What's Coming Next

Song Creator Pro currently ships with ACE Step 1.5, but we're not stopping there. We're actively working on adding more models as the open-source AI music space evolves, so you'll always have access to the best available options.

We're also building features like the ability to modify specific parts of a song — so if the verse is perfect but the chorus needs work, you'll be able to regenerate just that section without starting over.

All future updates are included with your one-time purchase. As the tools get better, so does your copy of Song Creator Pro.

Ready to start generating? Get Song Creator Pro on the Microsoft Store — $49.99 one-time, unlimited AI music generation, runs entirely on your PC.