This patch shows several techniques relevant to granular synthesis, playing a stream of short excerpts of recorded sound. (The patch uses one abstraction, called pan~, that's provided in the example titled "Constant-intensity panning subpatch". You'll need to download that abstraction and save it with the name pan~ somewhere in Max's file search path.)
This example allows a choice of four different modes of intensity panning, and two ways to specify the rate of panning change. The choice of four possible pannings is: static centered, left-to-right sudden switching, left-to-right gradual gliding, and random gliding. The rate of change can be controlled by sliders, either in Hertz (changes per second) or note values (based on the current transport tempo).
In classic granular synthesis, grains are very short windowed segments of sound, normally from 5 ms to 100 ms in length. A stream of sound grains be spaced at exactly the same time interval as the grain length, or at some greater interval of time. The spacing can even be randomized (some random time interval greater than or equal to the grain length). A single stream of grains is all you can do with a single groove~ object.
Preloading sound into random-access memory (RAM)—as opposed to reading it off of a hard disk—allows you more easily to access any point in the sound file instantaneously, read backwards as well as forwards, access different points in the sound simultaneously, etc.
The groove~ object plays sound from a buffer~, using two important pieces of information: a rate provided as a MSP signal and a starting point provided as a float or int message. For normal playback, the rate should be a signal with a constant value of 1. (The sig~ object is a good way to get a constant signal value.) A rate of 2 will play at double speed, a rate of 0.5 will play at half speed, and so on. Negative rate values will play backward.
The fact that groove~ can leap to any point in the buffer~ makes it a suitable object for certain kinds of algorithmic fragmented sound playback. In this example it periodically plays a small chunk of sound chosen at random (and with a randomly chosen rate of playback).
This example demonstrates creating a RAM buffer to hold a 10-second stereo recording, recording live audio into it (with input volume adjustment), and then playing randomly chosen backward clips of that sound, with a trapezoidal window to taper the beginning and ending of each clip to avoid clicks.
This shows how to record into a RAM (random-access memory) buffer, and how to play back the contents of the buffer at any rate (even backward by using a negative rate) starting at any point in the buffer. A timer is used to keep track of the duration of the recording. The example also demonstrates how one might use a quick fade-in and fade-out to avoid clicks when doing realtime capture during a performance.