Video game soundtracks, and a format for adaptive long music

17.58, Tuesday 28 Apr 2020

I mentioned Red Dead Redemption the other day. Here’s the original trailer from 2009 – it’s a gorgeous game. It looks and plays like a spaghetti western, which admittedly I’m a sucker for anyway, but you spend a ton of time just riding around dry plains and watching sunsets which is kinda perfect. And the soundtrack is :chef’s-kiss-emoji:.

Actually, maybe play the soundtrack now here on Apple Music or here on Spotify and listen while you read this post. For the atmosphere.

Reading about the making of the RDR soundtrack is fascinating. There is apparently 14 hours of music, and it has to work in loops rather than tracks…

The obvious difference is that film music is written to fit a finite scene, whereas with the video game, we’re working in five-minute loops. It’s really wide open, but also very hard, because there are all sorts of things happening with layers. If the player shoots someone, suddenly the music changes, so we have to think, ‘okay, does this work over the top of that?’. Also the big thing with a game is, you don’t know how long you’re going to be staying in that mood – you can’t state too much, it’s kind of like implying a mood. It’s a balance between having it interesting, but not so much that you get sick of it, because you could be riding that horse for 15 minutes…

And so:

Although this gives the impression of a formless improvisational process, because of the way the music reacts in real-time to the player’s actions, the underlying structure had to be meticulously planned. If a dramatic sequence suddenly kicks off, the soundtrack switches to something with greater intensity, while a more foreboding sound is required during moments of suspense. All of these loops have to segue into each other as events evolve on screen. … the whole score is composed in A minor and at 130 beats per minute.

It would be neat to be able to listen to all of this nuance and dynamic shifting without the need to actually play the game at the same time.


  • The Wetland Project is the sound recording of an endangered marsh in British Columbia, Canada, broadcast for Earth Day: The sounds of birds, frogs, airplanes and more, take over the airwaves for this twenty-four-hour experience in “slow radio”. (You can stream the wildlife sounds from that website.)
  • Chris Watson’s wildlife and landscape recordings – a musician and artist, he’s also a sound recordist for BBC nature documentaries. His album The Weather Project (you can find it at that link) includes recordings of dawn in the Masai Mara, and the creaking of an Icelandic glacier.
  • This old iPhone app Pocket Storm (there are examples on that website): A storm that approaches, surrounds you, and then drifts away from chirping crickets to the sound of an absolute monster of a downpour, and it’s all generated. You choose the length of the storm you want to the minute and then hit play.

I wonder, when I listen to these soundscapes, whether it would be possible to make an album that is intended to be listened to over a full 24 hours, as a kind of live soundtrack to your life?

  • We live real lives, which means the soundtrack would need to follow you around, and adapt dynamically to fit in, so rather than having it as an album somewhere, you’d probably deliver it as an online radio stream that you could play through your phone and Apple Music, or Sonos at home, or your TV – only it would be a private, personalised stream, made just for you.
  • To adapt, the service would need some minimal inputs: the robot DJ would need to know whether you were asleep, or rushing for a train (remember that?), or perhaps could bring the soundtrack to a moment of quiet if you opened a video call app. So it would require some kind of permission to watch what you were doing.
  • The artist would create the music as loop and layers, not tracks, and set the rules about how the composition should come together and behave given the inputs, and over 24 hours - as the listener - you would get to experience the whole piece.

The temptation would be to make this functional somehow, like it have the sound of birdsong when you had a meeting incoming, distant barking when an email in your inbox contains urgent-sounding words, but that’s not really what I mean. (Although that’s interesting too.)

There’s an apocryphal story that the compact disc runs 74 minutes because that’s the length of the longest recording of Beethoven’s 9th symphony.

(Ignoring the 24 hour version, 9 Beet Stretch, I’m guessing… which is wonderful by the way).

It turns out the size of the CD is more to do with data compression, fierce corporate strategy, and German manufacturing capabilities – but it’s a reminder that the album “format” isn’t inevitable. What we’ve got is contingent on the market, path dependent on its own history.

And so streaming music apps, royalty calculations, ID3 tags, charts, indie upload sites, etc, all still perpetuate the album, even - as they are now - untethered from the physicality of vinyl or plastic disc peppered with microscopic dots to be read by lasers – all still propping up the old system just because that’s what everyone else does, like the end of The Good, the Bad, and the Ugly, nobody able to drop their gun and walk away.

So I wonder what it would take to break that Mexican standoff?


Amazon could team up with video game publishers via the relationships established with Twitch, together defining a “soundscape” format for long music, to be broadcast through Amazon Echoes and wherever the Amazon app is to be found, initially as a way to publish game soundtracks in a more authentic form than 74 minute static albums, but really open to any artist, the format eventually finding its own Brian Eno or Ennio Morricone but actually it’s a 14 year old kid somewhere in the sticks?

How change happens. How new formats are born. I wonder.

Update 29 Apr

I’ve been chatting on Twitter with designer + musician Matthew Irvine Brown about this, and I think it’s worth saying that explicitly what I don’t mean is a video game music engine as an app, or procedurally generated soundscapes… like, there’s a lot of that and it’s great, but I’m into the idea of a format which is closer to existing albums but REALLY LONG, with just as much adaption as required to make it so you can listen all day.

And by far the closest I’ve seen to what I mean is Matt’s own 2011-2016 project Music for Shuffle. e.g. from the sleeve notes on his first composition:

I wrote a series of short, interlocking phrases (each formatted as an individual MP3) that can be played in any order and still (sort of) make musical sense.

…which also includes lovely nuggets like: Doing this experiment meant thinking about what happens when the listener presses skip – in a sense, they become a performer.

Matt’s made 22 of these! Whoa.

So… each one is a shuffle cassette maybe? The format is a ZIP file hosted on Dropbox plus a metadata file and an image cover, or maybe a podcast feed which lists all the individual MP3s, so anyone can make and host them – but so that each cassette is still recognisably a contained thing? And then perhaps a player app which can download and play these cassettes?

And perhaps v1 just does shuffle, but then v2 has a couple of simple triggers like “only play this track if the listener running”.

Maybe for the triggers (and this is slightly absurd but, perhaps?) the “language” could be the generative grammar Tracery which is used by Cheap Bots Done Quick for a trillion amazing Twitter bots, being exactly the right combination of accessible yet expressive?

I’m into this.

More posts tagged:

If you enjoyed this post, please consider sharing it by email or on social media. Here’s the link. Thanks, —Matt.