Chapter 6 — Song Structure
A melody and a chord progression are components; a song is the assembly. This chapter treats a song as an ordered list of named sections, each with its own register, energy, and role — like routes in an application, composed into a predictable flow. By the end you will read a full song form as a data structure and map it onto MidiSketch's structure presets.
Sections: the building blocks
Up to now every score has been a phrase in isolation. Real songs concatenate phrases into labelled blocks, and those blocks repeat. The labels matter because each one carries a convention: a listener expects a verse to be calm and a chorus to be loud, the same way a reader expects a function body to follow its signature.
Section
A section is a contiguous span of a song with a single structural role — a verse, a chorus, an intro, and so on. It is the unit of arrangement: melodies, dynamics, and instrumentation are decided per section. Think of a song as Section[], where each element names its role and the engine fills in the notes.
MidiSketch generates one melody and accompaniment per section, then stitches them together. Because sections are typed, the engine can apply different rules to each: a verse generator favours small steps and a low register, a chorus generator pushes to the top of the vocal range. The rest of this chapter walks through each section type.
A-melo: the verse
The first sung section is the verse. Its job is exposition: establish the key, the groove, and the lyrical setup without spending the song's emotional budget too early. Melodies here stay low and stepwise, leaving headroom so a later chorus can feel like a climb. In the score above, notice the narrow range and the relaxed quarter-note rhythm — deliberately understated.
A-melo (verse)
The verse — Japanese A-melo (Aメロ) — is the song's opening sung section, typically calm and conversational. It carries the narrative and sits in a low register. MidiSketch and J-pop convention use both names interchangeably; this course writes "A-melo (verse)" so you recognise either.
The naming split is worth flagging once: the Japanese pop tradition labels sections by letter (A, B, then the hook), while the Western tradition labels them by function (verse, pre-chorus, chorus). The engine speaks both dialects. Wherever you see A it means the verse.
B-melo: the pre-chorus
Between verse and chorus most pop songs insert a ramp — a section whose only purpose is to raise tension so the chorus lands harder. The melody climbs, the harmony leans on the dominant, and the section deliberately refuses to resolve. The unresolved ending is a feature: it hands the release to the chorus that follows.
B-melo (pre-chorus / bridge)
The pre-chorus — Japanese B-melo (Bメロ) — is the transitional section between verse and chorus that builds tension toward the hook. It often ends on the dominant (V) without resolving. In some forms the same slot hosts a contrasting bridge later in the song; the engine treats both as the B role.
In the score, the bracketed stepwise climb and the closing chord on V do exactly this. Read the V ending as a pending promise: the chorus is the resolve.
Sabi: the chorus
The chorus is the payoff. It occupies the highest register, the biggest dynamics, and the most repetition in the song — and crucially, it contrasts with everything around it. A chorus feels like a chorus mostly because the verse stayed low; the contrast is the effect. Compare the chorus score below with the verse example: same key, same tempo, but a full register higher and far more insistent.
Sabi (chorus)
The chorus — Japanese sabi (サビ) — is the song's high point and most memorable section, carrying the main hook. It sits in the highest register with the greatest energy and the most repetition. In MidiSketch, hook intensity and the loudest arrangement concentrate here.
The chorus here runs over an F–G–Em–Am loop and peaks at the top of the vocal range. That peak is not accidental — it is the structural reason every preceding section held back.
Intro, interlude, and outro
Three non-sung (or lightly sung) sections frame and pace the song. The intro opens it, the interlude breaks it up, and the outro closes it. They carry no new lyrics but do real structural work: setting expectations, giving the listener a rest, and landing the ending.
Intro
The intro is the opening instrumental section, before the first verse. It establishes key, tempo, and often the main motif so the listener is oriented before the vocal enters. Keep it short; its job is setup, not payoff.
Interlude (間奏)
An interlude — Japanese kansō (間奏) — is an instrumental section between sung sections, often after a chorus. It gives the arrangement room to breathe and can reprise the intro motif or feature a solo. Functionally it is a controlled pause in the vocal line.
Outro
The outro is the closing section that ends the song. It frequently quotes earlier material — often the intro motif, slowed or thinned — so the song feels framed. Think of it as the finally block: the place the energy settles and stops.
The score plays the intro motif, then the same material as an outro echo. Reusing one idea at both ends frames the song like bookends — a cheap, reliable way to make a sketch feel intentional.
Song form: the full assembly
Stack these sections in order and you have a song form. A form is just a sequence — Intro → A → B → Sabi → ... — and pop relies on a small catalogue of well-worn sequences rather than inventing one per song.
Song form
A song form is the ordered sequence of sections that makes up a complete song, e.g. Intro–A–B–Sabi–A–B–Sabi–Outro. It is the top-level structure: a template that says which section types appear, in what order, and how often. MidiSketch exposes forms as numbered presets.
The schematic above is not a melody — each whole note stands for one section, drawn at the register that section typically occupies, so you can see the shape of a song at a glance: a low verse, a rising pre-chorus, a high chorus, a settling outro. MidiSketch ships 18 such structure presets (formId 0–17) that assemble sections into complete songs.
Common pitfall — targetDurationSeconds overrides formId
If targetDurationSeconds is greater than 0, the engine auto-builds a structure to fit that length and ignores formId's section layout. Leave it at 0 to honour formId. Likewise formExplicit: false lets the engine adapt your chosen form — set it true to lock the form exactly as numbered.
The final-chorus key change: modulation
One structural device earns its own section because it is almost a cliché of the form: lifting the last chorus into a higher key. It adds no new melody or harmony — it transposes a section the listener already knows, and that familiarity-plus-lift is the whole effect.
Modulation (key change / 転調)
Modulation is a shift of the entire song to a new key partway through. In pop it is overwhelmingly a late, upward move — the final chorus jumps up a step or two — used as a last-minute energy injection. Because every part transposes together, the harmony's function is unchanged; only the absolute pitch rises, which the ear hears as a fresh surge.
MidiSketch controls this with two fields. modulationTiming chooses when the key change happens: LastChorus (the classic final-chorus lift), AfterBridge, EachChorus (rare and aggressive), Random, or None. modulationSemitones chooses how far: +1 to +4, with +2 (a whole step) the most common pop choice. A +1 lift is subtle; +3–+4 is dramatic but pushes the chorus higher into the vocal range, so pick the amount with vocalHigh in mind.
MidiSketch mapping
| Concept | Config field | Range / notes |
|---|---|---|
| Song form preset | formId | 0–17 (18 structure presets) |
| Use the form exactly as given | formExplicit | true = honour formId; false = engine may adapt |
| Build a form to a target length | targetDurationSeconds | seconds; 0 = use formId's structure |
| Final-chorus key change | modulationTiming | None / LastChorus / AfterBridge / EachChorus / Random |
| Key-change amount | modulationSemitones | +1–+4 semitones (+2 most common) |
| Section roles | (internal) | Intro / Verse (A) / Bridge (B) / Chorus / Interlude / Outro |
When targetDurationSeconds is 0, the engine uses the structure from formId; set a positive value to auto-build a structure of roughly that length instead.
See the Presets reference and Option relationships for the full structure catalogue and how formId, formExplicit, and targetDurationSeconds interact.
Continue with Chapter 7 — Mapping Concepts to Config.