Write structured song lyrics and style prompts for AI music generation

How to use

You don't need every song section — some great songs are just one section that evolves, and structure serves the emotion, not the other way around.

When rewriting an existing song with new lyrics, map the original's skeleton first, then fit new words to it.

On Suno V4.5 and above, the Style field supports up to 1,000 characters — use them.

The Suno lyrics field holds about 3,000 characters, roughly 40-60 lines.

For local, GPU-based generation instead of Suno, two optional skills cover this; because of their heavy dependencies they are not installed by default.

The lyric-writing and prompting craft here applies to heartmula as well, since its input format is lyrics with bracketed structure tags plus comma-separated style tags.

Everything in this guide is a guideline, not a rule: art breaks rules on purpose, so use what serves the song and ignore what doesn't.

Pick a common song skeleton, or invent your own.

Common skeletons include ABABCB (verse/chorus/verse/chorus/bridge/chorus, most pop and rock), AABA (verse/verse/bridge/verse, jazz standards and ballads), ABAB (alternating verse/chorus, simple and direct) and AAA (strophic verse-only, folk and storytelling). Mix, modify, or throw them out as needed.

ABABCB  Verse/Chorus/Verse/Chorus/Bridge/Chorus    (most pop/rock)
AABA    Verse/Verse/Bridge/Verse (refrain-based)    (jazz standards, ballads)
ABAB    Verse/Chorus alternating                    (simple, direct)
AAA     Verse/Verse/Verse (strophic, no chorus)     (folk, storytelling)

Say the lyrics out loud — if you stumble, the meter needs work.

Avoid clichés on autopilot, like "heart of gold" without earning it.

Avoid forcing word order to hit a rhyme ("Yoda-speak").

Avoid the same energy in every section, which flattens dynamics.

Don't treat your first draft as sacred — revision is creation.

Map the original's structure first: count syllables per line, mark the rhyme scheme (ABAB, AABB, etc.), identify which syllables are stressed, and note where held or sustained notes fall.

Counting syllables per line reveals the shape each new line must fit.

Mark the rhyme scheme, such as ABAB or AABB.

Identify which syllables are stressed in the original.

Note where held or sustained notes fall.

Match stressed syllables to the same beats as the original.

On long held notes, try to match the vowel sound of the original — if it holds "LOOOVE" with an "oo" vowel, "FOOOD" fits better than "LIFE".

Sing your new words over the original — if you stumble, revise.

For original concept work, pick a concept strong enough to sustain the whole song and start from the title or hook, building outward.

Start from the title or hook and build outward.

Generate lots of raw material first — puns, phrases, images — then fit the best ones into the structure.

If you need a specific line somewhere, reverse-engineer the rhyme scheme backward to set it up.

Use the Style/Genre Description Field with this formula, adapted as needed: Genre + Mood + Era + Instruments + Vocal Style + Production + Dynamics. For example, instead of the bad "sad rock song", write the good version: "Cinematic orchestral spy thriller, 1960s Cold War era, smoky sultry female vocalist, big band jazz, brass section with trumpets and french horns, sweeping strings, minor key, vintage analog warmth".

BAD:  "sad rock song"
GOOD: "Cinematic orchestral spy thriller, 1960s Cold War era, smoky
       sultry female vocalist, big band jazz, brass section with
       trumpets and french horns, sweeping strings, minor key,
       vintage analog warmth"

"Begins as a haunting whisper over sparse piano. Gradually layers
 in muted brass. Builds through the chorus with full orchestra.
 Second verse erupts with raw belting intensity. Outro strips back
 to a lone piano and a fragile whisper fading to silence."

Use no artist names or trademarks in the style description — describe the sound instead: "1960s Cold War spy thriller brass", not "James Bond style"; "90s grunge", not "Nirvana-style".

Specify BPM and key when you have a preference.

Use the Exclude Styles field for what you don't want.

Put metatags in both the style field and the lyrics for reinforcement, keep to 5-8 tags per section maximum because too many confuses the AI, and don't contradict yourself (for example [Calm] plus [Aggressive] in the same section).

Always use Custom Mode for serious work, which gives separate Style and Lyrics fields.

Always add structural tags — without them Suno defaults to flat verse/chorus/verse with no emotional arc.

Without structural tags, Suno defaults to flat verse/chorus/verse with no emotional arc.

Proper nouns have the highest pronunciation-failure rate for AI vocalists, so test them early.

Test proper nouns and unusual words in a short 30-second clip first.

Once generated, pronunciation is baked in, so fix spelling in the lyrics before generating.

Write the concept or hook first — what's the emotional core?

If adapting, map the original structure: syllables, rhyme and stress.

Generate raw material — brainstorm freely before structuring.

Draft the lyrics into the structure.

Read or sing the lyrics aloud to catch stumbles and fix the meter.

Build the Suno style description, painting the dynamic journey.

Add metatags to the lyrics for performance direction.

Generate 3-5 variations minimum and treat them like recording takes.

Pick the best generation and use Extend/Continue to build on promising sections.

If something great happens by accident, keep it.

Expect about 3-5 generations per one good result — revision is normal — and restate genre and mood when extending, since style can drift in extensions.

The heartmula skill generates full songs with vocals from lyrics plus tags.

The audiocraft skill provides Meta's MusicGen for instrumental text-to-music and AudioGen for sound effects.