GGrain breaks down the incoming signal into grains and then reassembles it. A grain is just a very short section of the sound, shaped to avoid clicks.
You can specify the number and size of the grains taken from the incoming signal. The level and pitch of each grain can then be altered before the sound is reconstructed, giving a variety of interesting effects.
The grain-size, pitch and level can all be randomized using the SizeVar, PitchVar and GainVar parameters. These specify the largest variation from the associated centre value, set by the Size, Pitch and Gain parameters.
For example, if Size is set to 35ms and SizeVar is set to 5ms, then the grain-size used will vary randomly from 30ms to 40ms. If all of the Var parameters are 0 then no randomization will be applied.
The Mix parameter allows the original and resynthesized sounds to be mixed together.
There is also a quality switch. High-quality mode adds filtering to minimize aliasing when changing the pitch of grains. It also uses improved interpolation. GGrain uses significantly more CPU power in high-quality mode.