GrainA micro-segment of audio, typically 1–100ms. The fundamental unit of granular synthesis — a quantum of sound.
Grain CloudThe aggregate texture produced by many overlapping grains. The cloud is the composition; no single grain is audible in isolation.
Grain SizeDuration of each grain in milliseconds. Determines whether the source is recognizable — small grains produce texture, large grains preserve identity.
Grain DensityNumber of grains triggered per second. High density produces continuous texture; low density creates sparse, isolated sonic events.
ScatterRandomization of grain position within the source file. High scatter reads from random points; low scatter reads linearly, preserving the original sequence.
Pitch RandomizationRandom transposition of individual grains, measured in semitones. Spreads the cloud vertically — from unison to wide chromatic smear.
PositionThe read-head location within the source file, from 0.0 (start) to 1.0 (end). Can be automated to create granular scrubbing through a recording.
MicrosoundCurtis Roads's term for sound at the sub-note time scale (1ms–500ms). The domain in which individual grains exist and human pitch/rhythm perception breaks down.
StochasticRandom processes governed by probability distributions. Xenakis's key contribution: replacing compositional decisions with statistical functions that generate controlled randomness.
Probability DistributionA mathematical function describing the likelihood of each possible value. Gaussian (bell curve), uniform, and exponential distributions all produce different grain cloud characters.
Time-StretchingChanging the duration of audio without affecting pitch. Granular time-stretch achieves this by repeating or omitting grains — the most artifact-free method at extreme ratios.
Phase VocoderAn FFT-based time-stretch method that maintains phase coherence between frequency bins. Smoother than granular but prone to characteristic "phasiness" on transient-rich sources.
Acoustic QuantaGabor's 1946 term for the elementary units of acoustic information, analogous to photons in light. Each quantum has a location in time and frequency — but not both with arbitrary precision.
Gabor AtomA Gaussian-windowed sinusoid — the mathematically optimal grain shape for representing sound. Minimizes the time-frequency uncertainty product as defined by the Heisenberg limit.
DissolutionThe gradual breakdown of granular parameters — increasing scatter, randomizing size, reducing density — until the source is unrecognizable. Basinski's tapes dissolved literally; granular synthesis can simulate this entropy.
Freeze EffectSetting position scatter to maximum and position automation to zero, creating a static cloud from a single moment in the source audio. A sonic photograph — an instant held forever in granular suspension.
GranulationThe process of breaking audio into grains. All granular synthesis begins with granulation of a source — live input, recorded sample, or synthesized waveform.
TextureIn granular synthesis, texture refers to a sonic mass without clear pitch or rhythmic identity — the emergent quality of thousands of micro-events heard as a single homogeneous sound.
MurmurationThe visual analog of granular synthesis: a flock of starlings moving as a unified cloud. Each bird follows simple rules; the group produces complex, unpredictable, beautiful patterns.
Async. GranularAsynchronous granular synthesis: grains are triggered at random intervals within a density range, rather than at fixed periodic intervals. Produces more organic, less mechanical cloud textures.