Grain Delay: separate time scatter from pitch scatter
Instead of treating a grain cloud as one mysterious texture, separate when each grain reads the source, how often grains arrive, and how fast they play. That gives you a practical route from a recognisable pluck to a dispersed tail.
Leave a clear event at the front
The first original eight-second synthetic pluck has bright attacks followed by short decays. Dry/Wet at 40% keeps a direct reference while the delayed grains collect behind it. Begin by following the attack, not the busiest part of the tail. The aim is to make the empty space interesting without losing the start of the note.
Scatter the tail without losing the pluck
Interactive audio example
The panel's XY pad can change two selected parameters at once. Select Spray on one axis and hold the other axis still for this comparison. Bring Spray from 80 ms to zero, then toward 160 ms, keeping Pitch and Rand Pitch at zero. Listen for the delayed material spreading through time without deliberately scattering its transposition.
Spray chooses additional look-back into the delay buffer for each grain. It adds a random delay between zero and the selected amount, rather than placing grains equally early and late around the base delay. More Spray can therefore make the response feel later as well as less regular. If a tail collides with the next pluck, reduce Spray before assuming Feedback is the only cause.
Frequency is the grain rate, not a filter
Return Spray near 80 ms and compare Frequency at 12 Hz and 60 Hz. This sets how often grains are generated. Their window length also follows the grain period within limits, so you are changing the temporal texture rather than simply adding identical copies. Listen for separate fluctuations at the lower setting and a more connected layer at the higher setting; the exact impression depends on the source.
Now select Pitch and compare zero with +7 semitones, leaving Rand Pitch at zero. Each grain reads at a different speed to transpose it. This is varispeed inside grains, not a guarantee of transparent pitch shifting. If the original melody becomes ambiguous, reduce Dry/Wet before adding randomness. A clearly pitched source is useful here because it makes the transposition easy to recognise.
Build a cloud from something less tonal
The second original eight-second source is synthetic noise bursts with a low noise floor. Its preset uses a synced quarter-note base delay, lower grain Frequency and a downward Pitch shift. Hear the shaped noise rather than expecting a sung or played note from it. Rand Pitch at 12% introduces a small amount of grain-to-grain variation around the base transposition.
Turn short noise bursts into a grain cloud
Interactive audio example
Compare Rand Pitch at zero and 25% without moving Spray. Judge changes in the texture's colour and internal motion. Then hold Rand Pitch still and reduce Spray. These are different axes: scattered pitch does not require scattered delay. On a tonal upload, increasing Rand Pitch can blur the harmonic role much sooner than it does on these noise bursts.
Sync the distance, not the grain clock
Sync resolves the base delay from the chosen note division and tempo. It does not lock the grain scheduler to the transport grid: grains are free-running from node creation. With Sync off, Delay Time spans 1–128 ms; synced values can be longer, capped at two seconds by the processing path. Do not expect changing Frequency to select a musical delay division.
Finally, compare Feedback at 25% and 40% while leaving playback time for the result to accumulate. The granulated output returns to the buffer, so later generations can be scattered and transposed again. More feedback can obscure new bursts and raise the sustained level. Lower Feedback first when the cloud will not leave room; lower Dry/Wet when its timing is useful but its presence is excessive.