Ray tracing for the ears: When sound stumbles through the voxel forest

Ray tracing. For most people, this has so far meant pretty reflections in puddles, soft shadows, blinding rays of sunlight – and a GPU fan that spontaneously switches to runway mode. But now it’s getting thick – this time for the ears. The developer Vercidium, previously more at home in the niche market, is focusing on a concept that starts where many engines traditionally leave off: with the realistic simulation of ambient noise. The working title of the whole thing? Audio ray tracing. Sounds spectacular, but it’s not. At least not in the show-and-shine sense. What is being considered here simply follows the logic that makes optical ray tracing so effective: rays (in this case virtual sound waves) are emitted, interact with the environment, are reflected, absorbed, damped or penetrate materials. Sounds like a physics lesson at first, but is actually a useful attempt to bring some order to the acoustic proliferation in current games.

Simulation instead of placebo: acoustics by geometry

In contrast to the usual, rather lazy reverberation algorithms – based on the motto “one room, one echo” – Vercidium’s system calculates the propagation of sound based on the actual level structure. This means that geometry, material properties, obstacles and distances are all taken into account. If you stand in an empty hall, you get a correspondingly cold echo. If you fill the same hall with boxes, you will hear the attenuation. This may not be a revolution, but it is the difference between reverberation and acoustic illusion. For technology fans: the whole thing is not based on polygonal geometry, but on a voxel-based spatial grid. This is less computationally intensive and is easily sufficient for sound propagation. The advantage: even older systems can work with it because the process runs entirely on the CPU – GPU-friendly and without RTX constraints. Vercidium itself speaks of the “also runs on the space station” principle. Irony included.

More than just noise: the sound becomes intelligible

1. The implementation is where it gets interesting. Vercidium defines four steps:
2. Sound beams are emitted spherically from the player’s position.
3. These bounce off walls, floors, ceilings and other objects – similar to an echo chamber with a system.
4. Parameters such as distance, material and reflection angle are saved for each interaction.

The final soundscape is created from the sum of this information – including reverberation, filter effects and directional information.

The result: a sound that adapts to its surroundings. Not groundbreaking, but at least consistent. The difference can be heard particularly in dynamic scenes – such as weather effects or changing room occupancy. When rain pours through an open window, it no longer comes from the middle, but from exactly where the window is open. If you play it through headphones, you’ll notice it. Maybe.

For the eye: accessibility meets visualization

There is a very useful side effect for deaf players: the system allows the visualization of sound sources. Acoustic events are represented by small, colored dots. Gunshots? Red. Footsteps? Green. Volume? Size of the dot. Everything live, directly in the game environment. What looks like a tech demo at first glance could actually be a benefit for many players. Provided that it is implemented sensibly and not as a garishly flashing balloon circus.

Efficiency instead of escalation: CPU instead of shader overkill

In terms of performance, the system remains down-to-earth. Vercidium relies on background threads to minimize the load on the main game performance. Although the initial calculation is time-consuming, it is sufficient for 32 rays to be updated per frame – enough for changes in the environment without bringing the computer to the boil. A dedicated ray tracing GPU is not required. No DLSS, no frame generation, no patch tracing – just a properly structured CPU job that runs on the side. And this is perhaps the greatest charm of the concept: no excessive technical madness, but a pragmatic addition.

Still alpha, but with potential

The plugin is currently at a very early stage. It is being tested on the basis of our own engine structures, and integration into Unreal Engine 5 and Godot is planned. Whether and when it will be released is still up in the air – Vercidium is keeping a low profile in this regard. Whether audio ray tracing can establish itself in the long term probably depends less on the technology than on the will of the development studios. Because realistic sound doesn’t sell on screenshots. And what doesn’t shine rarely gets a budget. Nevertheless, anyone who is serious about acoustic immersion will find a possible building block here. Not a panacea, but a start.

Source: Youtube