A computer-readable medium stores a data structure that encodes an audio performance
for interactive playback. The data structure includes a virtual instrument pool,
which encodes an interactive part of the audio performance. Audio content of the
interactive part is encoded at least in a sequence of synthesizer control data.
Each datum in the synthesizer control data specifies a digital sample of the audio
content to be played back. The data structure also includes a global accompaniment
pool, which encodes non-interactive portions of the audio performance. The global
accompaniment pool includes timing information to synchronize the playback of the
audio performance.
