Granular Synthesis

What a Grain Is

A grain is a micro-sample. Take any audio — a voice, a violin recording, a field recording of rain — and cut a piece between 5 and 50 milliseconds long. That piece is a grain.

On its own, a single grain is barely audible. It is too short to register as a recognizable sound. Play it once and you get a click. But play hundreds or thousands of grains per second, overlapping each other, and something emerges. The character of what emerges depends entirely on which grains you pick, how fast you play them, and how much randomness you introduce.

A useful working definition: a grain is “a unit of microtime no greater than about 50 ms but no less than about 5 ms.” Below that lower bound, you lose pitch information. Above the upper bound, you start hearing the source material as a loop rather than a texture.

Grain Parameters: Size, Density, Position, Spray

Four parameters define what a granular synthesizer does with its source material.

Grain size controls how long each grain lasts. Small grains (5-10 ms) produce buzzy, abstract textures. Larger grains (30-50 ms) preserve more of the original character. Grain size is the single biggest factor in whether the output sounds like the source or like something alien.

Density is how many grains play per second. Low density gives you isolated clicks and stutters — individual grains poking through silence. High density fills the space with a continuous texture. When density gets high enough, the output becomes a solid, evolving tone.

Position determines where in the source audio the grains are pulled from. If your source is a ten-second vocal recording, position controls whether you are grabbing grains from the beginning, the middle, or the end. Automating position slowly scrubs through the source material. Freezing position on a single point loops the same tiny fragment over and over.

Spray is the wild card. It modulates the position randomly within a defined range. Zero spray means every grain comes from the exact same point — a tight, repetitive loop. High spray means grains are scattered across a wider area of the source material. Spray is what gives granular synthesis its characteristic unpredictability.

In VCV Rack, you can implement spray by routing a noise source through a sample-and-hold module into the position control of a sampler. Each time the sampler grabs a new grain, the noise source gives it a slightly different read position. That is all spray is at the circuit level: controlled randomness on the playback head.

Windowing and Overlap

If you play raw grains with hard start and stop points, you get clicks at every grain boundary. The solution is windowing — applying a tiny amplitude envelope to each grain so it fades in at the start and fades out at the end. The shape of that window matters.

A Gaussian window (bell curve) produces the smoothest result. The grain peaks in the middle and tapers symmetrically on both sides. A triangular window is sharper — faster attack, faster decay, more present in the mix. A rectangular window (no fade at all) produces the harshest, clickiest texture, which is sometimes exactly what you want for glitch and noise work.

Overlap happens when grains are dense enough that a new grain starts before the previous one finishes. With high overlap, you get a smooth, blended texture because multiple grains are always sounding simultaneously. The overlap amount is a function of grain size and density — if grains are 20 ms long and you trigger a new one every 5 ms, four grains overlap at any given moment.

Time Stretching as Granular Synthesis

The most common real-world application of granular synthesis is one most people never think about: time stretching.

When your DAW stretches a vocal take to fit a different tempo without changing pitch, it is doing granular synthesis. The algorithm chops the audio into grains, spaces them out (to slow down) or overlaps them more (to speed up), and crossfades the boundaries. The pitch stays the same because each grain is played at its original speed — only the spacing between grains changes.

This is why extreme time stretching sounds weird. Slow an audio clip down by 50% and the grains are spaced so far apart that artifacts appear — warbling, phasing, metallic shimmer. Those artifacts are the sound of the granular algorithm struggling to fill gaps between grains. Some producers treat those artifacts as a feature, not a bug. Slowing a recording to 10% or 1% of its original speed produces long, evolving drones where the original source is barely recognizable. That is granular synthesis working at the extreme end of its range.

Freeze

Freezing is what happens when you stop the position parameter from moving. The granular engine keeps generating grains, but they all come from the same tiny region of the source audio.

The result is a sustained texture derived from a single moment in the recording. Freeze a vocal on a vowel and you get a continuous, shimmering tone with the formant character of that vowel. Freeze a drum hit and you get a buzzy, noisy sustain. Freeze an orchestral chord and you get an evolving pad.

Spray becomes critical during freeze. With no spray, freeze produces a repetitive micro-loop — the same grain repeating identically, which gets stale fast. Add spray and the frozen point becomes a cloud: grains are pulled from a small area around the freeze point, introducing micro-variations that keep the texture alive and breathing.

In VCV Rack, you can build a freeze control by routing a gate signal to the CV input of the position knob’s attenuverter. When the gate is high, the position stops advancing. When the gate drops, the position resumes moving through the source. Combine this with a foot pedal or a keyboard gate and you can freeze and release in real time during a performance.

Turning Any Recording into a Synthesizer

This is the real power of granular synthesis. Unlike subtractive, FM, or additive methods — which all start from mathematically generated waveforms — granular starts from a recording. Any recording. Your voice. A field recording from a train station. A sample of a cello. The creak of a door.

The recording does not need to be musical. It does not need to be clean. It does not even need to be interesting on its own. Once it passes through a granular engine, the micro-structure of the recording becomes the timbral palette. A cello recording run through granular synthesis does not sound like a cello anymore. It sounds like a synthesizer whose raw material happens to be a cello.

Try building a granular synthesizer in VCV Rack using only a sampler module, a clock, a noise source for spray, and a VCA with an envelope for windowing. Load a vocal recording into the sampler. By adjusting grain size, density, position, and spray — all with external modulators patched in — the vocal transforms from recognizable speech to whistling drones to skittering noise textures. Same source material, radically different outputs.

Granular-as-effect is equally powerful: feed live audio into a granular processor and use it as a short glitch looper, or modulate the pitch and window size to create new melodic lines from incoming audio.

What to Practice

• Build a granular synthesizer from scratch in VCV Rack. Use a sampler module loaded with any audio file, a clock to trigger grains, a noise source routed through sample-and-hold for spray, and a VCA with a short envelope for windowing. No dedicated granular module — the point is to understand what is happening inside one.
• Load a vocal recording and experiment with grain size. Start at 50 ms and work down to 5 ms. Listen to where the voice disappears and texture takes over.
• Freeze on a single point in your source material. Try different spray amounts: zero spray (tight loop), moderate spray (subtle variation), maximum spray (wide cloud). Notice how spray determines whether the freeze sounds static or alive.
• Take a non-musical recording — traffic, a dishwasher, rain, a conversation — and try to make it sound pitched and tonal using only granular parameters. Position and grain size are your pitch controls.
• If your DAW has a granular plugin (Ableton’s Granulator II, Pigments, or a free VST), load the same source material you used in VCV Rack and compare the results. The parameters should be familiar now.
• Extreme time stretch: take a 30-second clip and slow it to 5% speed. Listen to what the granular algorithm reveals in the micro-structure of the sound.