The schools of tiny flying cloth strands were made from processed bats and pitched up cloth flaps that programmer Martin Middleton was able to trigger from every tiny change in direction. For every cloth strand you see there are always multiple little grains of sound trying to play.
The calls of the friendly cloth fish creatures introduced in the third level are pitched down and processed aviary birds meant to sound kind of like dolphins. At first they played all of their sounds at random and seemed pretty dumb and robotic.
So I divided them into 16 categories: short, medium, and long idle chatter, mutters, laughter, questions, waking up, responding to a player, "come here", "uh-oh", "found something!", "I'll pick you up", "there you go", trapped cries of help, freed celebrations, and warbled frozen whimpers.
Designer Bryan Singh painstakingly added them in across the game as needed through a combination of script and hard code, and suddenly they seemed halfway intelligent. The jellyfish sounds are related but lower and were meant to sound like they’re getting tickled.
The giant cloth creatures' happy calls and the flying robots' sad eerie moans were made from humpback whales. According to my friend Sara Kerosky at Scripps Institution of Oceanography, humpbacks have a wide range of sounds, and each population sings the same song, which can last up to several hours. The song is picked up by other populations so that this one song (or calling sequence, or pattern) is sung by humpback whales worldwide each mating season, and it changes slightly each year. They surface to breathe at certain points in the song, never missing a beat.
The large flying enemies, which we mostly came to refer to as "guardians" or "war machines", were meant to be ancient but advanced stone robots. At the end of the opening ascent of the last level, the player originally used to shout and break apart their stone shells with a giant singing hawk shout, revealing cloth creatures inside. The wandering guardian / war machine moans are processed humpback whale songs and leopards run through a Kaoss Pad named K.I.T.T. for glitchy treatment, as are the digital babbles that play from the broken random body pieces you free cloth from in the earlier levels.
Their attack vocalizations were made from various vocoded animal roars, each with near and distant versions. Those emit from their head, and from three points on their body they emit power throb sounds made from bass drones gated with a square wave. The visor lowering and opening sounds include a giant, old, and very dangerous office paper cutter that was about to be donated to charity. (Nope, still mine!) Their searchlight emits a sound made from the static hums of TV tubes. Their eating cloth sound is made from vacuums and power saws, and the scarf getting ripped has dry ice-on-metal squeals and a gunshot.
All of the levels have multiple streaming surround ambiences made from library sounds and original recordings. In the snowstorm mountain level alone there are 16 different ambiences for the various areas and wind intensities.
What is the sound of a super hot, very still desert? I don't know, but the ambience in the third level is made from processed hissy warehouse room tone.
What I like about the game, Austin's excellent talk at GDC and your article here is how music and sound interacted much more closely than they do in most games. Depending on the genre / style / mechanics, a lot more games should do the same.
I find inspiring how creative you were with what you recorded. Great ingenuity in the sounds recorded!
One point you make that I think is very important for games trying to breath life into their worlds; if ambient creatures are added to the world, they are perceived as dumb if they don't react to the world in any way. These reactions can be simple audio elements that are triggered at logical times to denote emotional responses from these creatures, and immediately give the game-world more believability. I think this is a trick that could be used to great affect in many games.
Thanks for this great article. I am experiencing problems in looping sounds like surface slides, because there is a certain amount of phasing going on, and whenever the the sample loops to the start, you can notice the discontinuity. I try to cope with that by duplicating the audio and reversing the duplicate in the time domain. Any suggestions would be deeply appreciated.
Printer-Friendly Version
One point you make that I think is very important for games trying to breath life into their worlds; if ambient creatures are added to the world, they are perceived as dumb if they don't react to the world in any way. These reactions can be simple audio elements that are triggered at logical times to denote emotional responses from these creatures, and immediately give the game-world more believability. I think this is a trick that could be used to great affect in many games.