Generation Pipeline

When you call generator.generate(config), the engine executes a six-step pipeline that transforms a configuration object into a complete Bach-style composition.

Step 1: Configuration Resolution

The engine resolves all configuration options, filling in defaults for unspecified fields. Each form defines a default instrument, voice count, and tempo.

How Defaults Cascade

When you specify form: 'fugue' without other options, the engine applies: instrument = organ, voices = 4, BPM = 85. Any field you explicitly set overrides the form default. See Option Relationships for the complete dependency graph.

Field	Resolution Order
form	Required (default: 0 = Fugue)
instrument	User value → form default
numVoices	User value → form default
bpm	User value → form default (0 uses form default)
key	User value → 0 (C)
isMinor	User value → false (major)
seed	User value → 0 (random)
scale	User value → 1 (medium)
targetBars	User value → determined by scale
character	User value → 0 (balanced)

Step 2: Subject/Theme Generation

The engine generates the primary melodic material that will serve as the compositional seed for the entire piece. The approach varies by form family.

What is a Fugue Subject?

In Baroque music, the fugue subject (Latin: subjectum) is a short melodic theme — typically 1 to 4 bars — that serves as the entire fugue's foundation. Every voice will eventually state this subject, and all episodes and development sections derive from it. The quality of the subject largely determines the quality of the entire fugue.

Character Influence

The character parameter shapes the melodic profile of the generated theme:

Character	Effect on Subject
0 (Balanced)	Mix of steps and leaps, moderate rhythm
1 (Lyrical)	Predominantly stepwise motion, smooth contour
2 (Energetic)	Wider intervals, more rhythmic variety
3 (Dramatic)	Bold leaps, dotted rhythms, chromatic elements

Step 3: Formal Structure

The engine plans the macro-level structure of the piece. This is where different form families diverge most significantly.

Fugal Forms (Fugue, Prelude and Fugue, Toccata and Fugue, Fantasia and Fugue)

The fugue structure follows a well-defined plan:

• Exposition: Each voice enters in turn with the subject. The first answer is typically at the fifth (dominant).
• Episodes: Free counterpoint that modulates to related keys, often derived from the subject.
• Middle entries: The subject appears in new keys (relative major/minor, subdominant, etc.).
• Stretto: Voices enter with the subject in increasingly close overlap, building tension.
• Final entry: The subject returns in the home key for a conclusive statement.

For compound forms like Toccata and Fugue or Fantasia and Fugue, a free section precedes the fugue:

Variation Forms (Passacaglia, Chaconne)

• The bass theme (passacaglia) or harmonic progression (chaconne) repeats throughout
• Each variation adds new melodic and rhythmic material
• Texture and complexity build progressively
• The scale parameter controls the number of variations

Cantus Firmus Forms (Chorale Prelude)

• The chorale melody appears in long notes in one voice (typically soprano)
• Other voices weave independent contrapuntal lines around it
• Each phrase of the chorale is treated as a unit

Trio Sonata / Cello Prelude

These forms follow their own structural conventions — the trio sonata with two interacting upper voices over a bass, and the cello prelude with continuous arpeggiated figuration.

TIP

The scale parameter controls how many sections or variations are generated. Use scale: 'short' for compact pieces or scale: 'full' for maximum development.

Step 4: Voice Generation

Each voice is generated independently within its assigned register. For a detailed breakdown of voice roles and ranges, see Voice Architecture.

Key aspects of voice generation:

• Register assignment: Each voice is assigned a pitch range based on its role (soprano, alto, tenor, bass) and the instrument
• Melodic contour: Voices follow natural melodic shapes — arch contours, descending lines, sequence patterns
• Rhythmic identity: Each voice maintains its own rhythmic character to ensure independence
• Motivic development: Thematic material from Step 2 is developed differently in each voice

Step 5: Counterpoint Rules

The counterpoint module validates and adjusts the voice interactions. For the complete rule set, see Counterpoint & Voice Leading.

Key rules applied:

• Parallel motion: No parallel fifths or octaves between any pair of voices
• Dissonance treatment: All dissonances must be properly prepared and resolved (suspensions, passing tones, neighbor tones)
• Cadential formulas: Standard cadences close phrases — authentic (V→I), half (→V), deceptive (V→vi)
• Voice crossing: Voices should not cross into another voice's register without musical justification
• Leap resolution: Large intervals are followed by stepwise motion in the opposite direction

Step 6: MIDI Encoding

The final step maps the internal representation to Standard MIDI:

Internal Concept	MIDI Representation
Voice	MIDI Track
Instrument preset	General MIDI Program Change
Tempo	MIDI Tempo meta-event
Key	Key Signature meta-event
Note	Note On / Note Off events
Bar structure	Time Signature meta-event

• Each voice becomes a separate MIDI track (Type 1 MIDI file)
• Instrument presets map to General MIDI program numbers (e.g., organ = program 19, harpsichord = program 6)
• The output is a valid Standard MIDI File that can be opened in any DAW

// Access the output
const midi = generator.getMidi()       // Uint8Array (MIDI file)
const events = generator.getEvents()   // Structured event data

TIP

The structured event data from getEvents() provides programmatic access to every note, track, and metadata field. See the JavaScript API for the complete type definitions.