The granular synth I have designed (someone suggest a name!) offers unique timbral controls via a system of 'probability set based grain selection'. A probability set itself is a gaussian distribution in time around a single reference point. Moving the reference point through the source sample and only using this probability set gives classic granular synthesis performance.
The user is not limited to one, or indeed any number, of probability sets. If they define two sets, for example, located near the start, and the end of a source sample, then over the course of a note the grains can be selected initially from only the first, then gradually introduce more and more grains from the second set, finishing with grains selected entirely from the second set.
A second example may clarify : Setting probability sets to the ADSR controls, could slowly fade between selecting grains from different parts of the source file, or multiple files, based on the position in the ADSR envelope. Most grains come from source sample A during the attack phase, but as it transitions to Decay, the probability of choosing a grain from another set is increased, blending the overal timbre to a granularly resynthesized version of source sample B, and so on through the ADSR envelope.
Probability set changes are also connectable to pitch, duration, and external controllers, allowing the sets to be switched or 'cross faded' over longer periods. The user is not limited to choosing either ADSR or pitch, but again a weighting system can be selected to say how much one controller influences the probability set to choose the next grain from.