Praat AudioTools · User Guide

Babbitt’s Combinatorial Arrays

A serial-array composition and synthesis environment inspired by the opening of Milton Babbitt’s Semi-Simple Variations (1956). Explore combinatorial arrays, independent rhythmic voices, timbral transformations and musical form — then shape the complete result with a global dynamic arc.

Version 0.5.4 · 2026Author: Shai CohenAffiliation: Department of Music, Bar-Ilan UniversityLicense: MITGitHub repository

Overview

The script constructs a four-line combinatorial pitch array and realizes it as a synthesized composition. Its Historical / Analytical model uses a restrained interpretation of documented structures, while its Creative Combinatorial model adds independent rhythmic voices, register and spatial organization, gesture development, complementary spectra, and large-scale form. Both feed a shared event table for sound, visualization, diagnostics, and MusicXML.

12 presets

Eight foundational and four Creative configurations, from the opening array to evolving multi-cycle textures.

Two ways to set duration

Use a fixed tempo, or request a target number of seconds and let the script compute tempo.

Voice-aware register

Protect the four SATB parts with standard or custom ranges, preserving pitch class by octave placement.

Dynamic shaping

Add an optional composition-wide arch, inverse arch, crescendo, or diminuendo.

MusicXML

Retain a score as a Praat Strings object or save it as a MusicXML file for notation software.

Built-in diagnostics

Check serial integrity, rhythmic structures, event ranges, duration, audio levels and MusicXML consistency.

Design principle: the compositional engines generate pitch and rhythmic events. SATB Voice Range Protection selects valid octaves without changing pitch classes, while the optional Global Dynamic Arc shapes the amplitude of the completed mix without changing the underlying events.

Quick start

  1. Open Praat and run Babbitt_s_Combinatorial_Arrays.praat.
  2. Choose one of the 12 presets. Start with Opening Model (Semi-Simple Variations) to explore the basic array, or Combinatorial Morphogenesis for a longer Creative realization.
  3. Under Composition duration, keep Tempo (default) and set Tempo BPM, or choose Target duration and specify seconds. The other duration field is then ignored.
  4. Set Register centre MIDI if desired. Standard SATB Voice Range Protection is already on by default (advanced settings).
  5. Choose a Dynamic arc if you want an overall crescendo/diminuendo; None preserves the original dynamics.
  6. Enable Write MusicXML file to save a score directly. Use Edit details to adjust the structural and realization controls or the in-memory Strings output.
  7. Keep Draw visualization on for the analysis figure and Play result on to hear the resulting Sound. Click OK.
Interface note: the main form contains both Tempo and Target duration fields, plus the Turning point field. Standard Praat forms do not hide fields dynamically. The selected duration mode decides which time field is active; Turning point applies only to Arch and Inverted arch. All of these controls are in the main form, not the two advanced Details dialogs.
Not a score transcription: even the Opening Model is a realization of an analyzed structural mechanism, not a reconstruction of Babbitt’s original performance or full score.

The 12 presets

Presets select structural, rhythmic, synthesis and/or form defaults. The main-form duration and Global Dynamic Arc settings remain under user control; presets do not overwrite them.

01 · Foundation

Opening Model (Semi-Simple Variations)

The encoded opening hexachord, four-line array and original partition order; a restrained Historical / Analytical realization.

02 · Generator

Type-A Generator

A new ordering with interval classes 1–5 exhausted once, a new exhaustive rhythmic order and Creative realization.

03 · Rhythm

Rhythmic Transformations

Retains opening pitch organization while using the retrograde-complement time-point transformation.

04 · Pitch-class families

Associative Trichords

Foregrounds related trichord classes with associative rearticulation, bright timbre and optional trichord halos.

05 · Extension

Secondary Sets

Adds a second array cycle that extends the corresponding voice lines into 12-tone secondary sets.

06 · Derivation

Trichord Derivation

Opening, secondary-set and four derived cycles; includes checked hybrid-cycle construction and trichord halos.

07 · Serial sound

Serial Synthesizer

Multi-cycle derived structures projected into structural partial-slot timbre, in the Creative realization model.

08 · Experimental

Free Serial Field

Open combinatorial hexachord selection, a valid array operation, variable entry order, fresh rhythm and percussive synthesis.

09 · Creative

Structural Counterpoint

Independent rhythmic voices, structural distribution, spatial organization and gesture-based counterpoint.

10 · Creative

Spectral Combinatoriality

Complementary spectral groups derived from each active aggregate, with independent voices and spectral-form development.

11 · Creative

Temporal Labyrinth

Independent voices with a changing time-point path, partition-half segmentation and a contrasting temporal plan.

12 · Creative

Combinatorial Morphogenesis

A multi-cycle form derived from the array’s structural roles, with independent rhythm, gestures, spectra, space and responsive relations.

Two realization models: the preset defaults are editable in Details 2/2. The Historical model neutralizes incompatible Creative choices (for example structural form or independent rhythmic texture). The Creative model is an original extension; it should not be confused with a historically documented Babbitt technique.

Composition duration

New in v0.5.4: the main form lets you choose between Tempo mode and Target duration mode. Each structural cycle occupies 16 quarter-note beats (four 4/4 measures). The number of cycles is determined by the structural settings, and the new duration feature does not insert or remove cycles automatically.

16 beatsPer structural cycle
1 + S + DTotal cycles (S = secondary sets 0/1, D = derivation cycles 0–7)
20–300 BPMSupported effective tempo
musical_duration_seconds = 16 × cycles × 60 / tempo_bpm

target_mode_tempo_bpm = 16 × cycles × 60 / target_duration_seconds
ModeUser controlsWhat happens
Tempo (default)Tempo BPM (default 60)Legacy behavior. The composition’s musical duration follows from tempo and cycle count; the Target duration field is ignored.
Target durationTarget duration (s) (default field value 120)The script computes the effective tempo without rounding the timing internally. Tempo BPM is ignored. The number of cycles and the events’ structural positions are unchanged.

Worked example

Combinatorial Morphogenesis defaults to one basic cycle, one Secondary Sets cycle and four derived cycles: 6 cycles × 16 = 96 beats.

1:36At 60 BPM (Tempo mode)
48 BPMComputed for a 2:00 target
120 sMusical duration in Target mode

If a requested duration requires a tempo outside 20–300 BPM, Praat shows a dialog with the reachable duration range and options to change the requested duration or the number of derived cycles, or to cancel.

Musical duration versus audio-file duration: the target controls the structural duration, not necessarily the final sample count. The rendered Sound can include a synthesis decay tail. In Target mode it also preserves any final structural rests; in legacy Tempo mode the audio may stop before the final beat if the score ends with rests. Both durations are reported in the Info window.

The Global Dynamic Arc is measured as a fraction of musical time. If you set 180 seconds and a turning point of 65%, the requested turning point is at approximately 117 seconds, irrespective of the resulting tempo.

SATB Voice Range Protection

New in v0.5.4: voice range protection is enabled by default. Each Soprano, Alto, Tenor and Bass note is fitted to an octave within that voice’s allowed range. This avoids extreme high Bass or low Soprano notes in both synthesis and MusicXML.

VoiceDefault rangeMIDI min–maxPurpose
Soprano (S)C4–A560–81Higher voice band
Alto (A)G3–E555–76Middle-high voice band
Tenor (T)C3–A448–69Middle-low voice band
Bass (B)E2–E440–64Lower voice band
Relative ranges across MIDI 36–84
Soprano
MIDI 60–81
Alto
MIDI 55–76
Tenor
MIDI 48–69
Bass
MIDI 40–64

Illustrative shared MIDI scale; overlapping SATB ranges are intentional.

What is protected?

Custom ranges or experimental register

Open Edit details → Details 2/2: realization. Use Voice range protection on/off and Voice range profile Standard SATB or Custom. Custom values are entered as four min–max MIDI pairs in S–A–T–B order:

60-81 55-76 48-69 40-64

Each range must lie within MIDI 21–108 and span at least 11 semitones so every pitch class has a possible octave. Invalid custom ranges stop the run with an explanation. When protection is off, the script reverts to the broad MIDI 21–108 behavior for instrumental or experimental realizations.

QC: the Info window prints the actual minimum and maximum generated MIDI pitch of each voice, plus the permitted limits, octave corrections, suppressed displacements and any lost exact registral mirror symmetry. Structural self-tests check the event ranges and their agreement with the written MusicXML part ranges.

Global Dynamic Arc

The optional Global Dynamic Arc adds an overarching loudness trajectory to the completed synthesized mix. It does not restart at aggregate or cycle boundaries, and it does not change the array, timing, voice allocation or original event-level articulations and dynamics.

Arch

Crescendo → climax → diminuendo

Inverted arch

Diminuendo → valley → crescendo

Rising

Long-range crescendo

Falling

Long-range diminuendo

Main-form fieldDefaultMeaning
Dynamic arcNoneNone, Arch, Inverted arch, Rising or Falling. With None the global envelope is not applied.
Arc range (dB)18Attenuation span (0–36 dB). The arc rises to 0 dB at its loudest point and attenuates other sections; it does not add gain above unity.
Turning point (%)62%Where Arch reaches its climax or Inverted arch reaches its minimum (5–95%). Ignored for Rising, Falling and None.

The transition curves use smooth raised-cosine shapes. A single envelope table in composition time drives both the sample-level audio gain and the MusicXML dynamic directions and playback values.

output(t) = synthesized_mix(t) × 10^(arc_dB(t) / 20)
Listening versus measurement: a prescribed loudness maximum is not guaranteed to be the perceived climax. Note density, spectral brightness and existing local dynamics also contribute to perceived intensity. The optional visualization panel I compares the requested envelope with measured gain and smoothed output RMS.

Serial architecture

Opening four-line array

S = P     10  6 11 |  8  7  9
A = RI     2  4  3 |  0  5  1
T = R      9  7  8 | 11  6 10
B = I      1  5  0 |  3  4  2      (I = I11)

The ordered P hexachord 10 6 11 8 7 9 has successive interval classes 4, 5, 3, 1 and 2. Its segmental trichords include the set classes 015, 025, 014 and 012. Each vertical aggregate covers all twelve pitch classes.

Pitch construction and derived cycles

Hexachord sources

Opening hexachord; Type-A generator with successive ICs 1–5 exhausted once; or any compatible combinatorial hexachord searched by the script.

Secondary sets

An optional second cycle transforms each voice line by retrograde-plus-combinatorial operation, extending the line to a 12-tone secondary set.

Trichord derivation

Additional cycles partition the aggregate using a segmental trichord and transposition/inversion forms. The hybrid join is checked for repeated pitch classes within voice lines.

Horizontal and vertical checks

Each aggregate is checked for chromatic completion; each voice line is checked for six distinct pitch classes where required. A failed structural gate prevents output.

Time-point rhythm

The 16 partition symbols denote rhythmic types, not binary numbers. The opening order is:

Z 5 4 X 3 T 9 Y | 7 1 E 6 2 8 W 0

Examples of the decoded patterns (four sixteenths):
5 = 1100     3 = 0010     T = 0011     8 = 0110

The full collection contains all 16 four-position patterns exactly once, with 32 attacks in the composite stream. Retrograde, complement, order reversal and within-quarter mirroring act on the decoded attack patterns, not on symbol numbers. The opening’s two eight-quarter halves contain 18 and 14 attacks respectively.

Composite versus independent voices

Composite stream: time points from a single composite series are assigned to voices, with repeated attacks added according to the selected rearticulation strategy. Independent rhythmic voices (Creative): each voice follows its own transformed partition series. The aggregate is constrained to an eight-quarter window; simultaneous attacks are possible and elaborations may restate only pitch classes already introduced by that voice.

The Creative gesture engine varies elaboration through neighbouring notes, echo, pedal, contour and imitation strategies. Its form engine works with roles such as exposition, relay monody, density, spectral texture, withheld completion and revelation. The responsive layer can continue, transform, develop or contrast a previous cycle; it remains an algorithmic compositional model, not a claim of historical authorship.

Synthesis and timbre

Available timbres include Pure, Additive, Bright, Percussive, Structural partial slots (the earlier model), Complementary spectral and Interval spectral. The complementary spectral model relates upper partials to the active aggregate’s combinatorial groups (S+T against A+B). Creative realization also allows structural dynamics, articulation, stereo spacing and trichord halos.

Advanced parameters & controls

Main form

ControlDefaultAction
PresetOpening ModelSelect one of the 12 preset configurations.
Duration modeTempoChoose user-specified tempo or target-duration calculation.
Tempo BPM60Active in Tempo mode only (20–300 BPM).
Target duration (s)120Active in Target mode only; the effective tempo is computed from cycle count.
Register centre MIDI60Reference centre for register strategies, before optional SATB octave fitting.
Dynamic arcNoneNone / Arch / Inverted arch / Rising / Falling.
Arc range (dB)18Global attenuation depth (0–36 dB when arc is active).
Turning point (%)62Only for Arch and Inverted arch (5–95%).
Write MusicXML fileOffPrompt to save an actual .musicxml file.
Edit detailsOffOpen the two advanced dialogs described below.
Draw visualizationOnDisplay the structural picture and QC report.
Play resultOnAudition the final Sound.

Details 1/2 — structure

ControlOptions / behavior
Hexachord sourceOpening / Type-A / Any combinatorial.
Basic set transposition; Array operationTranspose P; choose P–RI–R–I, P–I–R–RI, two transpositional partners, or a random valid operation.
Secondary sets; Derivation cyclesEnable a secondary cycle; add 0–7 derived cycles.
RhythmOriginal, time retrograde, complement, retrograde-complement, quarter-order retrograde, mirror, new exhaustive ordering or Gray path.
Aggregate segmentation16+16 attacks or partition halves (8+8 quarters).
Rearticulation; Associative targetStrict, Grouped or Associative; target 015+012, 025+014, segmental P or current generator.
Sample rate; Output peak; Random seedDefault 44,100 Hz; peak 0.95; seed 0 means unpredictable and a positive seed produces reproducible choices.
Keep MusicXML StringsOn by default: retain the generated MusicXML lines in a Praat Strings object.

Details 2/2 — realization

ControlOptions / behavior
RealizationHistorical / Analytical or Creative Combinatorial.
Form planStatic, Counterpoint, Spectral, Temporal or Morphogenesis.
Voice distributionRound robin, Shuffled, Structural.
TextureComposite stream or Independent rhythmic voices.
Register strategy; Register spreadFixed, Distributed, Rotating, Structural inversion-axis or From form plan; spread default 10 semitones (0–18).
Timbre; Trichord haloSeven synthesis choices; optional halo chords.
Dynamics; ArticulationFlat / Structural dynamics; Uniform / Structural articulation.
Space; ChannelsFixed / Complementary / Symmetric / Exchange / Crossing / From form plan; Mono or Stereo.
Decay scalePositive multiplier for sound decay (default 1.0).
Voice range protectionOn (default) / Off.
Voice range profileStandard SATB (default) / Custom.
Custom rangesFour MIDI min–max pairs, e.g. 60-81 55-76 48-69 40-64.
Preset defaults: these values are prefilled according to the chosen preset and only need modification if you enable Edit details. The core interface remains compact; the SATB controls are advanced because most users can keep the standard ranges.

MusicXML output & notation

The MusicXML exporter reads the same corrected event table as the synthesis and visualization. It writes a four-part score (Soprano, Alto, Tenor and Bass) in 4/4 with four subdivisions per quarter note.

Write MusicXML file

Turn on Write MusicXML file in the main form. Praat opens a save dialog for a .musicxml document. File export happens after structural and final audio QC. Opening the file later in MuseScore Studio or another notation program is a separate step.

Keep MusicXML Strings

Enabled by default in Details 1/2. Produces an in-memory Praat Strings object named like musicxml_babbitt_arrays_SpectralCombinatoriality. Select it in Praat and use Save as raw text file… with a .musicxml extension when a file is needed.

FeatureScore behavior
Pitch and octaveThe MIDI pitch from the shared event table, including SATB range corrections where active. No separate octave correction is applied during XML writing.
RhythmNotated values follow the realized event’s written duration; rests, ties, barline splitting and articulation markings are generated as needed.
Voice structureFour monophonic parts with 4/4 measures and cycle/line labels. A part may contain rests.
TempoQuarter-note metronome marking and MusicXML <sound tempo="…">. Target mode retains higher precision in playback tempo.
Local dynamicsDynamic directions and score playback levels are derived from event dynamics.
Global Dynamic ArcVisible crescendo/diminuendo <wedge> directions, dynamic indications and note playback levels use the shared envelope. The hairpin turning point is placed on the nearest sixteenth-note grid position.
Extra synthesized layersTrichord halos and timbral partials are sound-design layers and are not independently notated as vocal parts.
Important playback limitation: MusicXML expresses dynamic intentions through score directions and per-note playback hints, not a continuous sample-accurate MIDI CC11 lane. MuseScore Studio and other notation programs may interpret those hints differently. The Praat audio has the authoritative continuous envelope; the MusicXML is its coordinated score representation. The script does not automatically launch MuseScore.

Visualization & quality control

With Draw visualization enabled, the Praat Picture window shows a multi-panel structural analysis of the actual event stream. Some panels appear only when their corresponding feature is active.

A — Event score

Four-voice piano roll: note spans, dynamics, new entries, rearticulations, aggregates and actual register.

B — Time-point partitions / array plan

Decoded attack/rest patterns for the composite stream; plan and rhythmic structure for independent voices.

C — Array & aggregate derivation

Opening or derived structures, line forms and aggregate relationships.

D — Surface trichords

Distribution of successive surface trichord classes, with targets highlighted.

E — Space / dynamics

Stereo event panning or mono event-level dynamics.

F — Spectral centroid

Measured spectral centroid for each aggregate, from the rendered sound.

G — Structural partial slots

Shown for the older Structural partial-slot timbre.

H — Per-cycle measurements

Polyphony, simultaneity, rhythmic independence and spectral distance across cycles.

I — Global Dynamic Arc

Shown when an arc is active: requested curve, measured gain and smoothed output RMS.

QC — Structural checks

Pitch-class and aggregate integrity, corrected rhythm, within-voice ranges, pitch agreement with XML, audio limits and other validation results.

Info window diagnostics

The Info window records the full event table, musical and rendered durations, effective BPM, SATB generated ranges and octave corrections, the arc’s theoretical turning point and measured gain, plus structural and listening-oriented metrics (such as sounding polyphony and gesture diversity). A failed structural self-test stops the run; an aesthetic metric is reported as evidence, not treated as an objective musical-quality verdict.

Complete workflow & outputs

  1. Load the selected preset and user-specified main-form settings.
  2. Resolve composition duration from tempo or target seconds, without changing the chosen cycle structure.
  3. Build and verify serial structures: hexachord, partner forms, arrays, secondary and derived cycles as requested.
  4. Realize events using composite or independent rhythmic voices, with gesture, articulation, register, dynamics and space.
  5. Protect SATB ranges by octave-fitting each event’s pitch class before sound and score output.
  6. Validate the shared event table and MusicXML structure; halt before producing outputs when a structural constraint fails.
  7. Synthesize audio, optionally apply the Global Dynamic Arc and complete audio QC and normalization.
  8. Export MusicXML and optional Strings only after final QC; show Info and visualizations, then play the Sound if requested.
OutputDescription
SoundFinal synthesized mono or stereo realization, named for the selected preset/run.
Praat PictureEvent score and optional analytical panels including the arc view.
Info windowStructural provenance, event log, register checks, durations, dynamic-arc measurement and QC.
StringsMusicXML lines retained in memory when Keep MusicXML Strings is on.
.musicxmlOptional file saved through the main-form Write MusicXML file control.

Troubleshooting

SymptomWhat to check
Requested duration cannot be reachedThe required BPM may be outside 20–300. Follow the dialog to change target seconds or derived-cycle count.
Custom SATB range rejectedEnter eight numbers as S/A/T/B min–max pairs, each within MIDI 21–108 and spanning at least 11 semitones.
Unexpected register or voice crossingsSATB protects each individual range but does not prohibit crossing. Turn protection off only for experimental register freedoms.
Dynamic arc appears absentCheck that Dynamic arc is not None and Arc range is greater than 0 dB; inspect panel I and Info gain measurements.
MusicXML file is not writtenCheck that Write MusicXML file is on, the save dialog was completed, required write permission was granted, and QC passed. Alternatively save the Strings object later.
Different dynamics in MuseScoreContinuous audio gain and score-playback hints are different representations; playback interpretation depends on the notation software.

Scope & interpretation

Historical / analytical scope: the coded opening pitch array, the 16-symbol rhythmic order and its 32-attack total are drawn from the analysis adopted by the script. Other implementation details — several rhythm-symbol decoding choices, segmentation, rearticulation scheduling, register, voice allocation and 4/4 notation — are model decisions. The attack/rest complement is not Babbitt’s modular time-point inversion. The Creative mechanisms and Global Dynamic Arc are contemporary extensions, not claims about Babbitt’s own compositional practice.

Recommended use: analyze the serial constraints in the Historical model, then use the Creative model to explore contrasting interpretations of the same structures. Evaluate the result by listening as well as by the available structural and signal measurements; a high numerical score alone does not establish musical quality.