To simulate acoustic reverberation in a room, SoundSoup needs to take a “dry” sound recording (with no reverberation) and add the effect of reflections from the walls, ceiling and floor of the room, as well as reflections of the reflections; reflections of the reflections of the … etc. Each of these must be added to the “dry” sound with a delay that depends on how far the sound has travelled; and the strength of the reflection should depend on the sound absorption of the surface(s) that it was reflected from. Accurate modelling will need to include many thousands of these reflections.
There are computer models that will (in simplistic terms) follow the paths of thousands of sound “rays” as they are reflected from each surface in the room, calculate the time delay for each, apply a filter to simulate the effect of sound absorption by each surface, then add them to the “dry” sound. As you can imagine, this takes significant time, even for a very fast computer. It also requires precise knowledge of the position and sound absorption of everything in the room, so these models typically require a detailed computer model of the room before they can do their work.
SoundSoup uses a simper, but still relatively accurate, process. The “dry” sound is fed into ten digital “pipes” with different lengths, but whose average length represents the average distance between reflections in the room. At the end of each pipe is a filter which represents the average sound absorption by surfaces in the room. The sound that comes out of each pipe is then added to the “dry” sound, but also to the input of each pipe, so that each reflection is reflected again and again.
To calculate the average length between reflections, all we need to know is the room volume V and the total surface area A. In any room, the average reflection length is just 4V/S. Isn’t geometry wonderful?
Technically, that process gives an exactly accurate result if the room is a “Sabine” room, which is a mathematical idealisation of a real room. However, for most rooms the difference between SoundSoup’s result, which can be heard in real time, and the result of a more exact model which takes much longer to set up and run (and is much more expensive) is hard to detect. Don’t use SoundSoup to design a critical space like a concert hall or recording studio, but for school classrooms, offices, hospital spaces, foyers and all kinds of “ordinary” spaces, the sound simulation is very close to what you will hear when the space is built.
The attached article describes SoundSoup’s modelling process in much more detail.