AUD · Audio tools

Grain Delay

Continue editing Open in workflow
Often followed by + Filter + Reverb + Limiter

Shareable link

Settings are written to the URL as you change them. Nothing differs from the defaults yet.

Repeat grains, not echoes

A grain delay keeps a rolling capture of the last two seconds and replays short windows of it. Each window, or grain, has its own delay offset, its own place in the stereo field, and a transposition applied on the way out. Grains overlap two at a time and are faded in and out with a raised cosine, so nothing steps at the edges.

The controls set the shape of that: Delay is where the grains are cut from, Grain size is how long each one runs, Spray is how far they scatter behind the delay point, Pitch transposes them, and Feedback decides how much of the output goes back into the capture.

Grain size

Grain size is the single control that changes what the effect is.

  • 20 to 40 ms: too short to carry pitch, so the output reads as a rough texture on top of the source.
  • 60 to 120 ms: long enough to hold a note. Repeats are recognisable and transposition sounds like transposition.
  • 150 to 200 ms: whole syllables and drum hits survive intact, which turns the effect into a scattered delay rather than a texture.

Density stays the same at every setting: a new grain starts every half grain, so two are always sounding and their windows sum flat. Short grains therefore arrive more often rather than leaving gaps.

Spray and the stereo field

Spray adds a per grain offset of up to 500 ms behind the delay point, and every grain also gets its own equal power pan position. The two together are what separate this from a delay with a wide setting: a delay places one repeat somewhere in the image, while this places every grain somewhere different, so the field fills in rather than widening.

Both values come from a hash of the grain’s position in the file. That is what keeps the effect reproducible: rendering the same file twice gives the same grains, and seeking in the preview lands on the same grid the export uses.

Pitch with feedback

Pitch transposes each grain by up to an octave either way. With Feedback at 0 that is a single transposition, since every grain is cut from the source.

Raise Feedback and grains are written back into the capture, so the next grain is cut from one that has already been transposed. Each pass stacks another interval. At +12 and high feedback the repeats climb an octave at a time and thin out, which is the classic granular ascent; at -12 they descend into a low blur. A gentle low pass in the feedback path takes a little off the top on every pass, which is what stops the repeats from accumulating into a whistle.

Common settings

A wide ambient bed: Delay 400, Grain 120, Pitch 0, Spray 70, Feedback 50, Mix 40. A tight granular double: Delay 80, Grain 60, Spray 10, Feedback 20, Mix 35. A rising tail on the end of a phrase: Delay 250, Grain 90, Pitch +12, Feedback 65, Mix 50.

Where it sits

Before a reverb, so the reverb has grains to work with rather than the other way round. After compression, since the grains inherit whatever dynamics the source arrives with. A filter after it tames the top end when feedback is high, and the limiter belongs last as usual.

Frequently Asked Questions

It cuts short windows out of the recent past, transposes them, scatters them in time and across the stereo field, and plays them back. A normal delay repeats the signal itself; this repeats fragments of it, so the output is a texture built from the source rather than a copy of it.

An echo has one repeat interval and returns the signal unchanged apart from level. Here every repeat is a grain of 20 to 200 ms with its own delay offset and stereo position, and the Pitch control transposes each one. At small Spray settings and unity pitch it behaves much like a short echo; at high settings it stops sounding like a delay at all.

How far each grain is allowed to sit behind the delay time, up to 500 ms. At 0 every grain comes from the same point in the past and the output is tight. Raised, the grains scatter, and because each one also lands somewhere different in the stereo field the result widens as it loosens.

The feedback path writes grains back into the same capture buffer, so the next grain is cut from an already transposed one. With Pitch at +12 the first repeat is an octave up, the second two octaves up, and so on. Setting Pitch to 0 keeps the repeats where the source was.

They are deterministic. Each grain's offset and stereo position are a hash of its position in the file, so the same file always produces the same result and the preview and the exported file agree. A generator seeded at playback time would give a different answer after seeking.

Explore Our Tools

Browse all tools