Granular Synthesis

Every method so far starts from a wave that repeats. Granular synthesis starts from a recording, and shatters it. It slices the audio into tiny grains, gives every grain its own little fade-in and fade-out, and fires them off like a confetti cannon: dozens per second, overlapping, from wherever in the recording you point it. The result is sound as texture: clouds, smears, frozen moments, shimmering walls.

How a granular engine works

A grain is typically 10 to 200 milliseconds of audio, wrapped in a smooth volume envelope (the "window") so it starts and ends silently instead of clicking. One grain alone sounds like a blip. The engine's real instrument is the stream: a scheduler that launches grains at some density, say thirty per second, each reading from a playhead position in the source file. Keep the grains long, dense and overlapping and they fuse into a continuous, organ-like tone. Make them short and sparse and the same recording turns into rhythmic, glitchy confetti. Add random offsets to where each grain reads from (spray) and the texture blurs like a photo out of focus.

Power the lab on. It renders its own 2.4 second source phrase, then starts chewing:

Granular Lab · grain cloud
Waveform · time
Spectrum · frequency
25 %
Position
120 ms
Size
25/s
Density
1.00×
Pitch
10 %
Spray

The source is a 2.4-second phrase this page rendered for itself. Sweep Position to scrub through it frozen in time; raise Spray to smear it into a texture.

Try this: sweep Position very slowly from 0 to 100%: you are scrubbing through the recording in slow motion, hearing each frozen instant. Then crank Spray and Density: the phrase dissolves into an ambient wash. Drop Pitch to 0.5x for instant mystery.

Time and pitch, finally separated

On a tape machine or a classic sampler, speed and pitch are welded together: play it faster and it gets higher, chipmunk-style. Granular breaks the weld. The playhead position decides when in the recording you are, and the per-grain playback rate decides what pitch comes out, and the two are completely independent. Freeze the position and time stops while pitch still works. Walk the position at half speed without touching pitch and you have invented time-stretching, which is literally how most time-stretch algorithms work under the hood. Stretch a three second recording over ten minutes and you get the cavernous ambient style known as paulstretch.

A short history: from acoustic quanta to modular cult classics

The idea is older than synthesizers. Physicist Dennis Gabor proposed in the 1940s that all sound could be described as tiny "acoustic quanta". Composer Iannis Xenakis realised it musically in the late 1950s by splicing tape into hundreds of fragments. Curtis Roads spent the 1970s coaxing granular clouds out of room-sized computers, one overnight render at a time, and in 1986 Barry Truax achieved real-time granular synthesis, turning it from a theory into an instrument. Today it is everywhere: Ableton's Granulator, Arturia Pigments' granular engine, and the Mutable Instruments Clouds module, whose grain-smeared haze became so beloved in Eurorack that clones of it outlived the company.

The five controls that matter

  • Position: where in the recording the grains read from. Your time-travel knob.
  • Size: grain length. Long grains stay recognisable; tiny grains become pure texture.
  • Density: grains per second. Sparse is rhythmic; dense is a continuous cloud.
  • Pitch: per-grain playback rate, gloriously independent of time.
  • Spray: random position jitter. The blur slider for audio.

Where you will hear it

Ambient music is soaked in granular processing: it is the standard way to turn any source into a wash of atmosphere. Film and game sound designers use it for risers, drones and the texture of unreal spaces. Producers freeze a single vowel of a vocal into a pad that holds forever, or granularise drum loops into stuttering fills. Once you know the sound, half of modern "how did they make that" moments turn out to be a recording, a window function, and a confetti cannon.

Granular vs sampling

A sampler treats a recording as a performance to replay: press a key, hear the file, maybe pitched up or down. Granular treats the same recording as raw material, a quarry to mine one pebble at a time. That difference changes what you listen for. Choosing a source for a sampler, you want a great complete sound. Choosing one for granular, you want great moments: a breathy consonant, a bow scrape, one shimmering piano decay, because any single instant can be frozen and stretched into a minute of texture. Some of the best granular sources are recordings that sound unremarkable played straight.

Checkpoint
A 'grain' in granular synthesis is…
Challenge · Match the cloud

Same source, hidden grain settings. Is the cloud smooth or stuttery? Is it pitched up or down?

15/s
Density
1.00×
Pitch