This was a short paper I wrote about the titular song for a class on music technology, which I said at one point I might post. Here it is!
I’ve made two changes: the transitioning of my citations from a “available in my notes but not visible otherwise” to “accessible by all,” and the addition of this note.
I chose to partially ignore the “no YouTube videos” part of this assignment, because I felt that the video was an important part of the song. They were created together, after all, as part of an even larger multimedia project: Portal 2, one of the top-selling games made by one of the world’s most famous video game companies. The compositional arc of the game as a whole is fascinating: Valve’s in-house composer, Mike Morasky, wrote almost the entire soundtrack for the game1 while working closely with their programming teams. Though the soundtrack was eventually rendered down to a still form for the release of Songs to Test By, within the game they’re procedurally-generated MIDI, with pre-set starting points that are then algorithmically developed to match the gameplay in a way that’s almost guaranteed to be unique to the player. (Morasky once stated that one of the pieces only repeats itself every 76,911.3 years, roughly.)
“Cara Mia Addio” was not a procedurally generated song, though the exact method by which it was made did rely on MIDI audio. In the studio, Morasky gave McClain2 the music he’d written for the turrets to sing and a melodic line for her, and asked her to improvise the words. The resulting melody, based on what she referred to as “my terrible Italian” became the Turret Opera. Morasky edited that recording to ensure that it didn’t sound too human – the ‘singer’ within the game being a robotic gun-turret – and then fed the backing sounds into the game engine itself.
That’s what I found most interesting about this – though the scene was rendered as a video file, not running live on the game engine,3 it was built within the same game engine that ran Portal 2, Source. Valve’s in-house animating tool, now released to the public as Source Filmmaker,4 provides deep control of every aspect of the game engine. Morasky (and, presumably, some of Valve’s animators) used sounds that had already been implemented in the game engine to provide all the voices save the melodic line. If I had to guess, I’d say that the system running animation queues was based on MIDI, as that’d be the easiest way to sync the visuals with the triggered sounds.
And a final note on those triggered sounds: all of the ‘turret voice’ effects were based on McClain’s voice, meaning that she sang the full chorus and solo of the song. Quite an impressive range.