Sample-based vs Procedural: Its not quite as dramatic as an all out death match between these two approaches and philosophies, even though the temptation is to see things in either/or black/white terms.

One thought is that, procedural audio, even though it has been around for a while now, is still fledgling and even though there are inherent ‘cost’ savings to using this method for sound generation and propagation (particularly in games with huge amounts of content), finding a home in a largely risk-averse entertainment software industry is a big ask as the applicable approaches still feel fundamentally ‘experimental’. The thing I’ve come to realize, perhaps somewhat later than everyone else (and perhaps because of the ‘either/or’ polemics), is that a lot of the techniques and tools we are using are already in transition to a more procedural status.

This is just a quick categorization attempt that I wanted to get down before it evaporates with the rest of my thoughts and doodles on a Friday morning…

The Sample-Based Approach.

Relying entirely on streaming or preloaded sample based assets sitting on a disc.

(Most games of the PS2/PS3 generation and some mobile games today)

Re-triggering of pre-recorded material, usually wave file assets.

The Procedural Approach

Moving the sound generation effort from the disc (and the streaming throughput bandwidth) to the processor.

Synthesis-based sound objects, acoustic models, grain players, noise-shaping and DSP intensive – in-essence everything is generated at run-time, based on (hopefully) elegant, efficient and simple real-time models.

(currently fringe aesthetic games, some music based games)

For me, the process of just writing these two (admittedly loose) definitions down, made me realize that any proposal to exclusively use either of these models would need to be either a) aesthetically niche or b) technically or artistically challenged in some way. And, even though I tried to say definitively which games used these approaches, I think I’m on unsafe ground in my generalizations. It also made me realize that, of course, there is already a ton of crossover in these categories in most proprietary sound engines, and certainly inside middleware audio solutions. A purely sample-based approach is probably getting quite rare these days. So, are we in the midst of a hybrid approach without even really realizing it?

Hybrid Procedural Approach

(Most console games today)

A fundamentally sample-based approach, but one that goes much further towards the implementation side of things than simple triggers. Breaking down sounds into constituent molecules (granular) or even small recognizable chunks (automatic weapons). Parametrization of sound. Sound ‘shaping’ in the form of Procedural DSP used for ‘additional layers’ like reverbs, filters and flutter. Some soundseed or air implementation in wwise, but just as a subtly mixed in ‘layer’, rather than to supply the overall effect.

We are using procedural techniques and technologies more and more in the form of reverbs and DSP effects. But also in our implementation, we are thinking more procedurally about sound, even if still using sample-based playback material as the starting point and raw material. My feeling is that we have moved towards this often without even realizing the big picture. Could this slow-bleed approach eventually end up with interactive sound designers working completely with acoustic models and unique sound object based propagation? Perhaps for certain genres and platforms. But it is difficult to imagine a move away from a hybrid position into exclusivity. I can though, see certain projects leaning one way or the other.

Chances are if you work in game audio, you are already working in a hybrid procedural audio world.