The basic premise behind granular synthesis is that a sound can be broken down into tiny particles, or grains. These sampled grains—usually no more than 10 to 50 milliseconds long—can be reorganized, or combined with grains from other sounds, to create new timbres.
In many respects, granular synthesis is similar to wavetable synthesis, but it works on a much finer scale. As you might expect, this method is ideal for creating constantly evolving sounds and truly unique tones.
The downside is that granular synthesis is extremely processor-intensive, and it wasn’t possible to do in real time until relatively recently. For this reason, it has remained largely ignored by all but a few in academic institutions. Today’s computers, however, have sufficient processing power to make this synthesis method a practicality, so there are a number of commercial products now available.