Behaviour, Structure and Causality in Procedural Audio - Game Sound Technology and Player Interaction: Concepts and Developments - page 322

we now need a computationally more expensive
grain/segment-based player (for the hybrid case)
or a control and DSP synthesis layer for a fully
procedural implementation.
A further developmental disadvantage in mov-
ing beyond the limitation of events to thinking
about actions and energy flows is that previously
neat boundaries become blurred. A discrete time
interpretation of a game makes simple, easy sense
to level designers or script writers. In practice, be-
havioural sound objects coupled to the underlying
physics engine (for example, see Mullan, 2009)
and game logic may need to be presented such that
the boundaries of sounds may remain, in surface
appearance at least, based on indivisible events.

if (change(angle(gate)))
playgrains(scrape(angle(gate)))

The “event” that the audio engine main loop
is waiting for is a change in the angle of the gate
because a player has moved it. From here, we
can dare to abandon entirely the notion of state
and pretend that all objects in the world, subject
to their position in a masking/priority table, are
reactive and continuously parameterised . Use-
ful mask topologies are functions for spatial and
temporal relevance (player focus), geometric
distance from known listener actors (machine
listening objects and human players), occlusion
based on raycasting, and unimpeded work (power
delta as a measure of how loud the sound could
be in free space).
The “sound” is now no longer simply a file
whose duration should be matched to the rotation
time of a fixed animation: It is a command to a
granular synthesiser that creates screeching sounds
by picking grains from a file and replaying them.
It is a sound made as a process or function. The
domain of the function comprises two variables,
time and angle, while the range is a time-variant
audio signal. Further, the function could have
hidden internal mappings, for example, the rate
of change of angle (angular acceleration) could
be used to select different grains and timings that
stimulate a stick-slip friction model to obtain an
impulse signature such as outlined by Rocchesso,
Avanzini, Rath, Bresin, and Serafin (2004) or in
Farnell (2008).
As sound code rather than audio file, advan-
tages are better cohesion and decoupling since
no subsystem need remain aware of the state of
any other (an advantage for network replication
too). A disadvantage is that more data must pass
between the underlying world model (geometry
and physics engine) and the audio system. In use,
the first obvious advantage is that the player can
exert continuous control over the object movement
and hear corresponding sounds. A disadvantage
is that instead of a linear, sample replay routine

bEHAVIOUrAL AUDIO

Another advantage of procedural sound is that it
can easily obviate the old problem of repetitive
sound sources and the need to fake variations. If
a procedural sound object is repeatedly triggered
with precisely the same parameters, will it not
make an identical sound? A comparison I have
been fond of is between sound as a film, which
can reveal behaviour, and sound as a photograph
which cannot. It accords with a dynamic interpreta-
tion of sound where causality is significant. One
phenomenon here is iterated complexity, familiar
to some as the “butterfly effect” (viz. Lorenz
(1993)). Natural variation can be introduced when
a function of several variables is sensitive to initial
conditions. Although of course all computations
are deterministic, with small deliberate variances
even systems of low complexity yield large varia-
tion of output in a short time.
A weakness in the film analogy is that a film
also has a script. Watching the same film twice
does not alter the story. Films are also snapshots
(singular experiences, regardless of warped time,
in the sense of Tarkovsky) in (and of) time, despite
an extra dimension to play with. The analogy of a
theatrical play with real actors is an improvement,