Start with one song idea
A coherent song begins with one central pressure, promise, question or transformation. The concept should tell you who is speaking, what changed and why the listener should remain until the final section.
“A song about trust” is a topic. “A night-shift worker notices that every promise arrives after the damage is done” is a usable song concept. The second version suggests setting, detail, vocal restraint, lyrical conflict and a direction for the hook.
Write the one-sentence record brief
A guarded Atlanta melodic-trap record in which late-night work, receipts and broken promises push the narrator from suspicion toward a firm boundary.
This brief becomes the test for every later choice. Cosmic sound design can support the isolation. A three-note underlick can remain unresolved during doubt and rise during resolve. Those choices belong because they serve the same story.
Build the Song DNA
Song DNA is the small set of decisions that must remain recognizable across every section. It prevents the second verse, bridge or breakdown from sounding like an unrelated prompt.
Keep the DNA stable while allowing its expression to change. A hook may widen, the drums may become denser and the underlick may gain new notes, but the song should still feel governed by the same identity.
Design section jobs
A section name is not a musical instruction. “Verse, hook, bridge” states order but does not explain why each part exists. Give every section a job.
Intro: establish the world
Introduce the identity with the smallest convincing amount of information. A vocal pickup, signature underlick, room tone, instrument answer or lyric fragment can establish the record faster than a full arrangement.
Verse: create evidence
Verses carry scenes, actions, receipts and perspective. Keep enough space for detail. The groove can remain stable while selected fills or instrumental answers underline key words.
Pre-chorus: change the pressure
Shift harmony, register, phrase length, drum density or breath spacing. The pre-chorus should make the hook feel necessary.
Hook: deliver the central truth
The hook needs a clear title phrase, memorable rhythmic or melodic shape and a response the listener can predict. Supporting vocals should strengthen the lead rather than restate every word.
Bridge or breakdown: reveal another angle
Remove a familiar support, change the point of view or reinterpret the motif. Contrast should deepen the same song instead of introducing a separate concept.
Final return: prove that something changed
Bring back the hook with altered harmony, response, ad-libs, instrumental weight or lyrical meaning. A final hook should feel earned rather than copied.
Write lyrics for movement
Build the lyric around concrete progression. The first verse introduces the pressure. The second verse adds evidence or consequence. The bridge changes understanding. The final hook lands differently because the listener now knows more.
- Use objects, places, times and actions instead of mood words alone.
- Give the title phrase one stable meaning, then let context deepen it.
- Place important words where the vocal behavior can emphasize them.
- Leave room for breath, response and instrumental punctuation.
- Do not force every line to explain the theme; let selected details carry it.
Verse one: the unanswered call. Verse two: the timestamp and receipt. Bridge: the narrator stops asking for an explanation. Final hook: the same title becomes a boundary instead of a question.
Direct vocal and rhythm behavior
Voice and rhythm should behave as one performance system. A behind-the-beat singer changes where the drums can fill. A dense rap cadence needs cleaner gaps from supporting parts. A choir response needs the instrumental bed to yield enough space for its words.
Vocal behavior
Describe range, weight, tonal core, timing, phrase connection, breath recovery, articulation and endings. Then direct how those traits change: restrained verse delivery can open into a chest-mix hook, or clipped verse phrases can stretch at the emotional turn.
Drum behavior
Explain the relationship among kick, backbeat, hats and bass. Let hat density, pickups, fills and dropouts mark sections. In trap, keep kick and 808 from masking each other; the kick may chase selected 808 movements instead of mirroring every attack.
Ducking and musical space
Use interaction when parts compete. Lower choir support under solo declarations, ease instruments during choir answers and create brief low-end space when kick and 808 collide. The purpose is clarity, not an obvious pumping effect.
Develop the underlick with the story
The underlick is a recurring musical response that helps identify the song. It may be a synth figure, guitar turn, piano answer, horn phrase, choir fragment or rhythmic cell. Its notes, rhythm, register and instrumentation should express the theme.
A lonely three-note analog signal ends unresolved after distrust lines. It rises by one scale degree when the narrator chooses resolve, gains sustained strings before the final hook and returns at full ensemble weight near the scene peak.
This is stronger than “add a cosmic synth.” The motif now has a dramatic job, a return pattern and a reason to change. Do not place it after every phrase. Repetition creates identity; selective absence preserves meaning.
Generate, select and diagnose
Judge generations by the record brief, not by novelty alone. A surprising voice or instrumental texture can be exciting while still weakening the intended song.
- Identity: Does the generation sound like the requested record?
- Performance: Does the lead voice carry the correct weight, pocket and emotional progression?
- Hook: Is the title phrase memorable after one listen?
- Section contrast: Do later sections develop rather than merely become louder?
- Musical response: Do drums, underlick and supporting parts react to the lead?
- Technical condition: Are artifacts, masking or structural problems repairable without losing the performance?
If several generations share the same failure, revise the relevant prompt layer. Thin vocals require stronger vocal-body and register direction. A flat hook requires clearer lift, response and contrast. Generic production requires more specific rhythmic and instrumental behavior.
Repair, arrange and finish the record
Once a performance is worth preserving, stop treating regeneration as the only tool. Repair isolated problems, tighten the arrangement and decide whether stems are needed.
Frequently asked questions
Should the full song structure be written in the prompt?
Include the sections that define the song's journey. A familiar structure does not need excessive explanation, but unusual transitions, vamps, breakdowns and final resolutions should be directed clearly.
How many generations should I make before changing the prompt?
Generate enough to identify a repeated problem. If several results fail in the same layer, revise that layer instead of changing the entire concept.
What makes one section feel different from another?
Contrast can come from register, harmony, drum density, instrumental weight, vocal timing, response behavior, space or lyrical perspective. Use more than volume alone.
When should stems enter the workflow?
Use stems after choosing a performance worth preserving. Stems help repair, rebalance and arrange a strong generation; they do not replace a clear song identity.