Goodbye, Steam Audio

Steam Audio is a spatial audio plugin that simulates how sound behaves in the real world. It's extremely powerful and an incredible free plugin that you can use in your own projects. Instead of simple left/right panning and volume fallout, it models the following:

• Binaural (HRTF) spatialization for more realistic 3D positioning over headphones

• Occlusion (objects blocking sounds)

• Transmission (sound passing through walls w/ material calculations)

• Reflections and reverb (sound bouncing around the environment)

• Propagation (how sound travels through space)

Steam Audio primarily uses the CPU for heavy lifting by applying ray tracing to audio. Instead of tracing rays for light, it traces acoustic rays for sound. These calculations can become very heavy on the CPU when you have many active audio sources in a real-time reflected environment with dynamic geometry.

Honcho is a game where players horde truck components, repair parts, spray paint, trade goods and vending machines all in a run-time generated dynamic world. This has lead to some serious CPU heavy calculations in the game, and the need to remove the plugin.

Honcho has a unique audio challenge: the player's position inside and around their truck dramatically changes how sounds should behave. Sitting inside the cabin with the windows and doors closed should sound completely different from lying underneath the truck. Exhaust placement matters. Missing exhaust parts matter. A smashed window changes the acoustics. Leaning your head out of the window changes what you hear. Rain hitting the roof, high-speed travel without a windscreen - all these small details affect the soundscape.

Steam Audio was a great fit because it could simulate these interactions automatically.

Unfortunately, now at version 0.4, the performance cost has become too high. In its place, I've implemented a state-based audio system that aims to recreate much of the same effect in a much more performant way.

Instead of constantly recalculating the environment, audio states are updated when something changes: entering the truck, opening a window, leaning out of the cabin, removing a component, and so on. Sounds outside the truck now use dynamic low-pass filtering, while environmental acoustics rely on Unity's reverb zones rather than physically simulating every material in the world.

CPU performance has increased significantly, which was the main bottleneck here and I'm fairly happy with the simulated ambisonics - even if they are now "faked".

Narrative elements, and truck storage are still being worked on. Full patch notes will be written up once I'm happy with the whole 0.4 content package.

Thanks,

Greg