Although mostly famed for its near-photorealistic graphics, Tomas Neumann and Christian Schilling of Crytek, along with composer Inon Zur opened GDC 2008’s Audio track with a discussion of the decision to deliver the highest quality of audio to match the graphics of their recent release Crysis.
“We had to develop next-gen audio when most of our focus was on the graphics: that was a real challenge,” Neumann admitted, with Shilling joking that calling the session “the Crysis of Audio” was “just too easy.
They introduced the challenges that face audio designers -- not only competing within the game with the dominant requirements of the graphics, but also with the needs of simulating freedom in a real world, where players can easily break the intended sound design.
Using normal sound tools such as SoundForge initially, Crytek move on to a data-driven sound specification using a FMOD designer (“we can add and layer any behaviours to sounds that we wish,” said Neumann) and a central mixer, before the sound is integrated into their sandbox emulator for the game engine.
Crysis used 3,500 sound events, built from nearly 15,000 individual sound files, requiring over 30 gigs of uncompressed audio data, revealing that in the end about 1.7 gigs of the game DVD is audio.
Game Design and Audio
Schilling discussed the way in which the requirements of the game design lead the audio audio design using the example of weapons and vehicles : Crysis features 24 different weapons with different modes, requiring sound designers to deal with sounds for single shots, bursts of fire and guns indoors and outdoors.
Similarly, Crysis features both first-person and third-person viewpoints, requiring different sound design.
Shulling moved on to explain that the game featured 15 vehicles, and they quickly discovered they would have some of the most complex sound events, with each vehicle featuring many moving, sound creating parts, such as opening and closing doors, and could interact with ground materials, had burstable tires and other damage effects (requiring audio feedback.)
But “to get immersion,” Neumann argued, “the best thing to use is music.”
Introducting Inon Zur, Neumann explained their intentions for the dynamic music score: a cinematic feel that reflects the player’s gameplay (not only when the enemy AI are “alert”) by using a system of nodes to avoid repetition.
“We actually stole this tech from our animators,” Neumann admitted.
Zur introduced himself: “the way I see music is the emotional and dramatic tool to drive the game,” and explained that he had “quite a challenge in Crysis,” with three factions, the US, Koreans and aliens all requiring their own, quite clashing, scores.
He described his choices: a traditional military score for the US, a more traditional, eastern score for the Koreans, and with the aliens, rather than go for the obvious decision to use electronic music, he decided to use “weird string effects” to create an alien sound.
Zur explained he combined scores to create a backdrop for conflicts: with a battle between the US and the Koreans, for example a bombastic, military sound with touches of flutes.
In game, with battles moving between levels of intensity, Zur created three cues that the game could crossfade – soft music for exploring the levels, speeding for initial encounters, before exploding into a full action score in massive battles.
Though the demonstration felt seamless, Zur said that they chose to hide these transitions behind sound effects wherever possible so it would never jar the player.
Ambient sound effects were created by marking areas across the map for ambient sounds, with certain areas overlapping or being inside each other, with levels of priority based on the player’s location. “Nature should react to the player,” said Schilling, and so the ambiance also required dynamic behaviour, with bird sounds ending when gunshots are fired.
Neumann discussed the difficulties of getting sound design to work with a physics engine, with it being incredibly hard to judge what players will do and how the physical world will react. They created a matrix of all possible sound effect interactions and ended up with over 36,400 entries, but even in Crysis, where they strived for realisim, they had to scale down significantly to optimise their code.
Neumann talked shortly about working with dialogue, advising that using clean assets (i.e. straight, unaltered voice files) was very useful, with effects such as radio noise added in engine, as it allowed assets to be reused easily in a variety of situations and to deal with script or story changes (something they deal with throughout the development of Crysis.)
In conclusion, they noted the positive response that their work on the audio had in reviews, when it was naturally expected all reviewers would talk about was the graphics.
“We were humbled by this,” beamed Neumann.