← Guide / Learn / Vocal delivery and performance

vocals · lesson

Vocal delivery and performance

Delivery is what the singer does with the line — whispered, belted, spoken, chanted — and it can change across a song by section.

Delivery is a verb

Register says where the voice lives and texture says what it is made of; delivery says what it is doing right now. It is the only vocal dimension that naturally varies within one song, which makes it the best tool for building dynamics.

  • whispered — barely voiced, maximum intimacy, needs a sparse bed.
  • spoken word — no melody, rhythmic phrasing.
  • half-sung — between speech and melody; indie and alt-R&B staple.
  • belted — full chest voice at volume, the standard chorus lift.
  • falsetto — light head voice above the natural break.
  • chanted — repetitive, rhythmic, often group-voiced.
  • shouted / screamed — punk, hardcore, metal.
  • melismatic — many notes per syllable; gospel and R&B runs.

Vary it by section

The most common reason a generated song feels flat is that the vocal performs identically from start to finish. Assigning delivery per section fixes this faster than any production change: whispered verse, half-sung pre-chorus, belted chorus, falsetto final chorus.

Section tags are the place to put these cues — they are read as local instructions rather than as global style, so a delivery note inside a chorus tag does not bleed into the verses.

Do not ask for belted and whispered in the same breath at the global level. Split them across sections or the model will pick one and ignore the other.

Support the delivery with arrangement

  • A whisper needs the band to drop out; name the reduction.
  • A belt needs headroom; thin the mid-range or it fights the guitars.
  • Chants want doubling and width; say group vocal, wide.
  • Spoken word wants a stable, looping bed so the rhythm of the speech carries.

Example

[Verse 1] (whispered, close, band almost silent) I counted the streetlights to keep myself awake [Pre-Chorus] (half-sung, drums enter, rising) and every one of them was burning out [Chorus] (belted, full band, wide) so let it burn, let it burn, let it go

Practice

Take a song you have already generated that feels monotonous and rewrite only its vocal delivery, section by section, leaving every other word of the prompt untouched. Generate once. The improvement is usually larger than anything you would have achieved by re-tuning genre or instrumentation, and it costs you four short parenthetical cues.

vocalsdeliverydynamics

FAQ

Where do delivery cues go — style prompt or lyrics?
Section-level cues go in the lyrics next to the section tag. A single overall delivery term can live in the style prompt.
Will parenthetical cues get sung as lyrics?
Sometimes, if they read like lyric text. Keep them short, descriptive and clearly performance-directions rather than sentences.

Keep going