While music from artists as diverse as Megadeth to Stevie Wonder and Muse immediately grab your attention when you talk about audio in a game like Guitar Hero, they are just a component of what makes the game fun to play.
There's a lot going on behind the scenes to help really put people up on that stage and to make sure it's fun, even as players clang a few notes. In this article I'll look at some of the things we did in the audio and how we did our part to take Guitar Hero up a notch.
We used FMOD as our sound engine, so some of our approaches might be more applicable to people who are familiar with the software, But even if you use a different engine, I hope that you might understand what we were aiming for, and why we did the things we did -- and find this info useful.
When we were given the honor of tackling the next Guitar Hero in 2007, we knew that we had to build on what the game offered to keep the series' momentum. It was also obvious that audio would be a major part of how we'd do it. That meant really sitting down and looking at the technology we had on hand, and how we were going to use it.
This led to a number of goals for our song audio player. It not only had to have great performance but also all a lot of new functionality to support the features we had planned, including: adding drums and vocals; letting players play the instruments they wanted, even if that meant doubling up; allowing a new drop-in and drop-out party mode and giving the player even more of a sense of location and space where their virtual band was playing.
We needed a system that would give us great audio streaming performance directly from a disc, a number of audio tracks with per-track digital signal processing, almost instant player feedback plus an audio visualizer, as well as atmospheric audio spatialization and on-the-fly reconfiguration for all of the above.
In early versions of Guitar Hero, there were, at most, three stereo tracks per song: guitar, bass or rhythm guitar and everything else. By adding drums and vocals, and giving the drums the space they required, we would have to triple the number of stereo tracks to nine. To ensure that this amount of data would smoothly stream off disk without performance issues we had to take some special measures. In this case it meant interleaving into three multichannel files.
Instead of having one massive file, three interleaving tracks of similar sonic complexity let us take advantage of variable bit rate space savings as well as the optimizations built into FMOD for handling 8, 6 and 4 channel input files.
Basically, it means instead of having one big file with the same bit rate for all instruments, we have three smaller files each with their own bit rate which changes based on the varying requirements of each group of instruments. Only sending bits down the stream when they are needed means more efficient streaming and disk space savings.
This is the layout of the stereo tracks:
1 - kick
2 - snare
3 - tom
4 - cymbal
5 - guitar
6 - bass
7 - misc
8 - vocal
9 - crowd sing-along