A Practical Create Music Workflow for Style and Lyrics

Start from a listening goal, not a keyword pile

Before you open Create Music, write one sentence about what you want to hear at the chorus: mood, tempo feel, and who is singing. That sentence is your brief. Style and Lyrics exist to serve it — not the other way around.

If you cannot say the chorus in plain language, the model will guess. A clear brief prevents random genre stacking and keeps later retries comparable.

• Example brief: “Mid-tempo indie pop, hopeful chorus, soft female lead, guitars in front, no trap drums.”
• Keep celebrity names out of Style; describe tone and delivery instead.

Treat Style and Lyrics as two lanes on Create Music

On Create Music, Style steers genre, instruments, vocal character, and production. Lyrics steers section jobs, rhyme, and singable line length. Mixing those jobs into one blob makes failures hard to diagnose.

Build Style from a short stack: primary genre → mood → tempo band → 2–4 instruments → vocal character → one production cue. Build Lyrics with section labels your track actually needs (Verse, Pre-Chorus, Chorus, Bridge).

• Style lane answers: what does the bed sound like?
• Lyrics lane answers: what is sung, and where does the hook land?
• Meta Tags can mark sections; keep them consistent with the Lyrics outline (see the Meta Tags section on the main Guide).

Generate a baseline before you optimize

Your first Create Music run should be intentionally simple. Save the exact Style string and Lyrics outline. That baseline is the control sample for every later change.

Listen once for the chorus hook, once for verse clarity, once for mix mud. Only then decide which lane failed.

Change one lane per retry

If the hook melody is weak, revise Lyrics or section contrast — leave Style alone. If the bed is muddy, thin Style instruments or production words — leave Lyrics alone. Parallel edits burn credits and erase cause-and-effect.

Use a stop rule: after three focused retries without progress, rewrite the listening goal instead of adding more adjectives.

When to leave the prompt and use product features

Not every problem is a Style rewrite. Extend continues a strong take past the current ending. Persona locks a voice character across songs. Reference Audio steers style transformation from a track you have rights to. Split Music separates vocals and instruments so you can hear masking.

Pick the tool that matches the failure: weak ending → Extend; inconsistent singer across takes → Persona; want a known production vibe → Reference; cannot tell what is crowding the vocal → Split Music.

• Browse ready-made Style ideas on /ai-music-prompts, then paste into Create Music.
• Draft Lyrics first in /tools/song-lyrics-generator if the words are the bottleneck.