Procedural Sound: The Split Between Cost & Creativity
- Callum Steele
- Oct 8, 2025
- 8 min read
"The creation of sound in real time, using synthesis techniques such as physical modelling, with deep links into game systems" (Paul Weir, 2017)
The above quote is a succinct way to define procedural audio, given by the man behind the audio of No Man's Sky at the Game Developers Conference in 2017. But what does this mean?
Often, when playing a game, the player is hearing playback of loops or one-shots that interact with gameplay. Procedural sounds differ in that they are created during runtime as opposed to coming from a pre-rendered source. In other words, procedurally generated audio creates the sound effects that the player hears on the go, based on a set of pre-defined behaviours (Ronny Mraz, 2021).
What does this mean for the developer and the player? It can mean a much smaller RAM usage, a wider range of audio, a more unique game experience (especially when we talk about roguelikes or the aforementioned No Man's Sky that have other forms of procedurally generated content) to name just a few (Ronny Mraz, 2021). This post aims to look at some examples of games utilising procedural audio for such creative benefits versus how they can be used to save on costs.
No Man's Sky
Of course we had to start with the posterchild for procedurally generated worlds. No Man's Sky is a survival and space exploration game featuring an endlessly growing universe. A landscape entirely procedurally generated needs procedural vocalization; all creatures would need sound, and the chosen method has to work for everything from eldritch beasts to little birds (Paul Weir, 2017). For this the team created a plugin for Wwise, called VocAlien, to closely tie the vocalisation of each creature with its characteristics. The instrument itself functions like a pipe, or four pipes stuck together, and some noise generators which function as frequency modulators. The sound was then affected by such qualities as size, body-to-head ratio and time of day to deliver alien vocals that matched each creatures physical attributes and circadian rhythms (Paul Weir, 2017).
Below you can hear the result of Vocalien in action, compared with the audio of sentient creatures in the game, which is not procedurally generated but chosen at random from sound banks.
Spore
As games continue to grow there is an argument to be made for procedurally generated music being the way forward to avoid fatigue. In 2012, Tomas Dvorak (composer of the soundtrack to Machinarium) was quoted as saying “Soundtrack music has to be more abstract to give space for the image and also to not be annoying if it repeats”. He is referring to the large open worlds and open-ended narratives of modern games, for which a simple loop can be anything from tedious to inappropriate (Plans & Morelli, 2012).
An example of this might be Spore, a game that follows a species designed by the player from microscopic, to land mammal, to space-travelling society. The music in Spore is actually created by a generative algorithm that constantly shuffles the composition. The entire game was built on the back of procedural generation (Wright, 2005) so why not have the soundtrack follow suit? Composer Brian Eno used a software called 'The Shuffler' which was created using PureData, a visual programming language. 'Shufflers' (now called 'Elements') allowed the game to combine various samples following only a few simple rules to generate a unique piece of music for each scene. This allowed every scene to have 'a sonic character -- but it wouldn't be exactly the same each time you visited' (Eno, 2012). The player ends up with a soundtrack that is different every time they play through. Should they design a creature with more aggressive features and play accordingly the music will be more aggressive to match. If, on their next play through, their creature is less aggressive and perhaps the society that evolves a pacifist society, the music will reflect this. As there is so much middle ground between these two extremes there are practically limitless possibilities for musical variation.
Other games have since utilised this technique, including Rain World, in which the player takes control of a "slugcat" and explores an industrial, post-apocalyptic world. Each region of the world contains layers of sound that are triggered selectively based on levels of danger, itself impacted by the complex ecosystem of enemy AI. In 2023, James Primate, the composer for Rain World said of the enemy AI, "It has its own system of weights and desires... You can encounter a creature and it might just not be into you. It might be into doing something else, in which case the threat music would not build up to the degree that it otherwise would." (Laced, 2023)
You can now see the same technique being deployed in AAA games. In 2019, Woody Jackson explained to The Telegraph how the music in Red Dead Redemption 2 comes from many stems which react to the player. He says of the process of making the music react to the player, “It is usually based on 11 Stem, 4 to 5-minute loops. This is not your usual song stem session where you just split up the parts into drum/bass/guitars/melody and everything can stack on top of itself but rather it caters to what will or could happen during gameplay." (Hoggins, 2019).
Two Dots
Two Dots is a mobile puzzle game developed by Playdots, inc in which players must connect two dots of the same colour in order to make a move and progress through to the next level. Playdots regularly release expansions which, this being a mobile game, runs the risk of taking up a huge amount of disc space and RAM on its user’s device. One workaround would be for Two Dots to compress its audio files as it releases more themes but this would ultimately mean the sound quality would get worse and worse until it detracts from the game. Instead, Playdots uses PureData, the same visual programming language used by Spore, within Unity to synthesise their sounds in real time. Audio behaviour is scripted inside PureData to generate audio through ‘patches’, with parameters that change the audio output based on gameplay as it runs. (Ronny Mraz & Splice, 2021)
Peggle Blast
Just like Two Dots, the team behind Peggle Blast were working within the restrictions of a mobile game but had to contend with a small budget on top of this. While they did use sampled audio files it was sparingly as the bulk of sound effects were procedurally generated. They managed to do all of this using only Wwise. (Popcap Games, 2017)
They used only the tools available to them in Wwise for real-time synthesis and digital processing to create sounds whilst using very little RAM. This also led to a much faster design process as the team were not worried about game capture or spending any time exporting assets (Shumate, 2016).
The result was sounds that were adaptable and in sync with gameplay, all while remaining within the 5MB limit set for the audio team; Peggle Blast ultimately shipped with 1.3MB of sound effects and 3.5MB of music (Shumate, 2016). However, a side effect of the process meant that the layers of each sound were not all available in one sample (as you would find when exporting a track from a DAW) but now existed as individual voices. (Ronny Mraz & Splice, 2021)
The Last of Us
Speaking of creating a lot from a little, a challenge for much larger games (games that can give up hundreds of hours of gameplay) is keeping sound effects varied. In The Last of Us, a game in which players must fight their way through a post-apocalyptic, zombie-infested USA, players will hear countless gunshots, explosions, footsteps, to name just a few. To keep this from getting tedious, or even fatiguing (without requiring an untold number of SFX to be rendered) tricks like repitching, gain and envelopes can be utilised at random to manipulate many variations from a tiny number of sound samples (Stevens, 2016). The Last of Us does exactly this (Crawford, 2018). The result of this added randomness is greater verisimilitude within the game which increases the chance of player immersion as, in the real world, this randomness is exactly what we expect (Brown & Cairns, 2004).
Procedural then clearly has many uses, from grand creative endeavours such as never ending musical variations, to tiny sonic details contributing to an immersive world. Game developers have been repeatedly stretching the uses of procedural audio to their limits, and constantly finding out how far those limits are. It is important for a developer to understand when a procedural audio solution may be best suited to very specific problems, or projects; the music of Spore being ambient allow for a lot more flexibility with structure than the spaghetti Western inspired score of a game like Red Dead Redemption 2, in which people might expect to hear a theme, or for creating a franchise narrative. While it can be a win for saving resources on creating more raw sound samples (or the availability of storage) this must be balanced with the time that is can be spent creating new software to manage demands or the workarounds required in the available software. Overall, procedural audio is a versatile and useful tool for sound design and music making and this group of audio engineers look forward to learning just how much further it can be pushed.
References
How procedural audio brings sounds to life in video games - Blog | Splice. (n.d.). Blog | Splice. https://splice.com/blog/procedural-audio-video-games/ [Accessed 2 October, 2025]
Game Developers Conference. (2017b, July 11). The Sound of No Man's Sky [Video]. YouTube. https://www.youtube.com/watch?v=zKJ_XuQjjiw [Accessed 2 October, 2025]
Game Developers Conference. (2017a, June 11). Peggle Blast: Big Concepts, Small Project [Video]. YouTube. https://www.youtube.com/watch?v=Nsn3E3eGPhk [Accessed 2 October, 2025]
Steadman, I. (2012, September 28). Brian Eno on music that thinks for itself. WIRED. https://www.wired.com/story/brian-eno-peter-chilvers-scape/ [Accessed 3 October, 2025]
Purchese, R. (2007, January 16). Eno's unique Spore soundtrack. Eurogamer.net. https://www.eurogamer.net/enos-unique-spore-soundtrack [Accessed 3 October, 2025]
Shumate, J. (2016, May 30) Real-Time Synthesis for Sound Creation in Peggle Blast! | Audiokinetic Blog. (n.d.). Audiokinetic. https://www.audiokinetic.com/en/blog/real-time-synthesis-for-sound-creation-in-peggle-blast/ [Accessed 2 October, 2025]
Rain World audio design stories: Part 1 - Maze Runner. (n.d.). Laced Records. https://www.lacedrecords.co/blogs/blog/rain-world-audio-design-stories-part-1-maze-runner [Accessed 3 October, 2025]
Wright W, (2005) Zeusmonomon. (2020, August 27). Will Wright GDC 2005 Spore (The Future of Content) Remastered [Video]. YouTube. https://www.youtube.com/watch?v=ofA6YWVTURU [Accessed 2 October, 2025]
Plans, D., & Morelli, D. (2012). Experience-Driven Procedural Music Generation for Games. IEEE Transactions on Computational Intelligence and AI in Games, 4(3), 192–198. https://doi.org/10.1109/tciaig.2012.2212899 [Accessed 6 October, 2025]
Hoggins, T. (2019, November 6). Vintage guitars, sneezing sheep and artificial intelligence: How the shifting soundtrack of Red Dead Redemption 2 was built. The Telegraph. https://www.telegraph.co.uk/gaming/features/red-dead-redemption-2-shifting-soundtrack-built/ [Accessed 7 October, 2025]
Journey Into Starfield's Epic Game Audio - with Bethesda Game Studios' Sound Team | A Sound Effect. (n.d.). A Sound Effect. https://www.asoundeffect.com/starfield-game-audio/ [Accessed 2 October, 2025]
Crawford, D (2018, July 26) Unique Challenges of Interactive Audio in The Last Of Us — Dara Crawford. (n.d.). Dara Crawford. https://daracrawford.com/audio-blog/unique-challenges-of-interactive-audio-in-the-last-of-us [Accessed 7 October, 2025]. [Accessed 7 October, 2025]
Richard Stevens (2016) Why Procedural Game Sound Design Is so Useful - Demonstrated in the Unreal Engine. A Sound Effect, 18 January [Online blog]. Available from: <https://www.asoundeffect.com/procedural-game-sound-design/> [Accessed 7 October, 2025]
Brown, E. & Cairns, P. (2004) A Grounded Investigation of Game Immersion. In: Proc. Chi 2004, ACM Press, 2004. Press, pp. 1297–1300. [Accessed 7 October, 2025]
Comments