Two separate sessions at this year's GDC illustrated one of the most interesting aspects of game design: the audio. The exacting details of how to make the bubbles of the Atlantic sound just right might not be everyone's cup of tea, but we found the audio post-mortem for BioShock and the presentation from sound mastermind Masafumi Takada fantastic. Read on for our impressions of two of the most niche but interesting sessions at this year's GDC.
Patrick Balthrop, 2K's senior sound designer for BioShock, presented the game's audio post-mortem. He addressed the specifics of how the team crafted BioShock's soundscape, which was just as important to its atmosphere as its visuals (point in fact, BioShock won the GDC award for Best Audio this year). The first challenge was with the audio logs and short-wave radios.
In order to create the "sound" of these powerful but archaic devices, the team applied a wealth of post-production. Distortion, down-sampling, gating, and mid-band EQs all helped create the "old-time radio" sound. From there, the team had to create a solution to ensure that the audio could be clearly heard at all times. The player had to be able to hear key information even in the midst of a firefight. Initially, after tagging the logs and radios, the team applied a low-pass filter. That resulted in muddy sound, so instead (using the same tags), the overall mix volume is lowered when a radio or log plays. You're probably already familiar with the effect from when the whole game suddenly gets quieter as you focus on a dead resident's confession.
The next challenge was the need for voiceover that brought the AI of Rapture to life. Eventually the team decided on a high limit of three simultaneous speech events, to avoid "cacophony." They also kept a five-second delay between queued speech events. A simple example of the organization of a speech events parameters (queue, priority, trigger chance, minimum time between events) might be the Little Sister and Big Daddy voiceover sequence.
In order to bring the ecology of the two iconic BioShock characters to our ears, the team created the following sound flow chart (one of many potential sets): summoned Little Sister (the Big Daddy hitting a wall); summoned Little Sister annoyed (the noise as one angrily arrives at a pipe); exitvent; thankful ("Thank you, Mr. B!"); idle; scream (as you approach too closely); recover; threaten; damage; melee (a sequence of the whale song noises from the Big Daddy); death; and cry ("Mister B! Missstteeer Beee!"). Balthrop pointed out that that the whole sequence's key factors are that it's relationship-focused and that it's dynamic, crafted as the two AI characters interact.
Noting that the processing of human voices is always done with trepidation due to its risks of sounding "really cheesy," Balthrop pointed out that the Little Sister and the final tracks of Atlas' voice were processed. Neither voiceover was bad, they just weren't perfect. The team took the audio tracks, processed them extensively, and then added the processed version to the unprocessed version for the final audio.
The next major point was the establishment of the aesthetic through various tools. The first was music, selected from the jazz era. The team actually had to use vinyl emulators to add "that vinyl warmth" to some tracks. Balthrop had nothing but praise for working with composer Garry Schyman for original music. Schyman was sent entire videos of scenes, and would craft the music explicitly for those moments. What made the process work, Balthrop said, was constant iteration between the studio and Schyman.
