Granular synthesis builds textures from clouds of short enveloped sound grains

Granular synthesis treats sound as a stream of microsonic particles called grains — brief events, typically 1–100 ms, near the threshold of auditory perception. Each grain has two essential parts: a waveform (synthetic or sampled) and an amplitude envelope; the envelope (e.g. Gaussian, Hanning, expodec) is required, because abrupt grain boundaries otherwise create audible clicks. Grains are not heard individually — they are layered into streams and clouds, and the audible macro-texture emerges from their collective behavior, controlled by grain density, duration, pitch/transposition, amplitude, and spatial position rather than by editing individual samples. Grains can be scheduled stochastically or deterministically; around 15–30 grains/sec produce smooth texture, while below ~10/sec the grains become audible as discrete rhythmic events. The method enables transformations difficult with other techniques, notably time-stretching without pitch change and pitch shift without time change. Theorized by Dennis Gabor (1940s) and Iannis Xenakis, and implemented computationally by Curtis Roads and Barry Truax.

Examples

A 30 ms grain from a piano sample sounds like a timbral click, not a note; thousands overlapping at varied positions and rates produce a shimmering cloud resembling the piano’s timbre. Roads’ Cloud Generator (up to 90 grains/sec) slices a speech recording and re-scatters it from recognizable words through smeared texture to a swirling cloud. Truax’s Riverrun (1986). Strudel: s("voice").chop(32).speed(rand.range(0.5,2)).

Assessment

Define the two structural components of a grain and explain why the amplitude envelope is required, not optional. State the characteristic grain-duration range. What grain density produces smooth texture versus discrete rhythmic events? Name two transformations granular synthesis makes easy that other methods do not.