How video game music is created

From those primitive little bloops and bleeps to the cinematic sound tracks of this generation of video games, the creation of the music of games is an interesting and technically challenging aspect of the industry.

  • Old Game Systems

  • The sound hardware available on older systems actually resembled the hardware found in analog synthesizers.

    These sound chips usually consisted of 1 to 3 simple oscillators, capable of producing a handful of wave shapes. Sine waves, square waves, triangle waves and white noise generators were common.

    The main technique used to shape these sounds was to use envelope generators. The envelope of a sound is like a graph of its amplitude or volume over time.

    The ADSR model for envelope generation was quite common and could actually vary the sound these simple chips could produce quite a bit.

    Composing music with such limited hardware was, to say the least, challenging. Especially when you consider that the music had to share the limited number of oscilators with the game's sound effects.

    Usually the composer would create a melody and simple accompaniment. The composer would then assign a higher priority to the melody. The game could then 'steal' less important oscilators or channels to use for sound effects, without harming the melody.

  • Modern Game Systems
  • Fortunatly for us, a few advances in sound chips paved the way for the much richer experience we now enjoy.

    The first advance was the introduction of DACS. These chips used digital sampling to reproduce recorded sounds. These chips also had their own memory, which was used to store the digital representation of the sound. Some of these chips were actually separate processors that ran their own programs, concurrent with the game system's main processor.

    The second advance was the dramatic increase in the number of available oscillators or hardware channels, as they are now called. While the SNES could hardly be called a 'modern' game system, it had 8 hardware channels available, and the original Playstation had 24 channels. That may seem like a lot, but the Playstation 2 has 48 channels!

    The third advance was the increase in the amount of sound RAM or memory that the systems now have available and the ability to play compressed sounds, which can take up 1/4th to 1/10th the size of an uncompressed sound.

So what did all this new technology do for the composer working on the game? Well, it became possible to create realistic instrument sounds, using sampling of actual musical instruments, and it gave the composer the ability to create much more complex music in the MIDI environment that most modern musicians are very comfortable with. The final result of moving the music over to the game very closely matches the MIDI music the composer creates using their computer and MIDI synthesizers.The modern game system has enough hadware channels for the music and the sound effects to co-exist peacefully, without the sound effects stealing channels from the music.

One other technique exists for the ambitous composer, one that allows the ultimate flexiblility. It's a technique called streaming, and it involvs taking a complete recording of a song and breaking it up into little chunks that can be loaded into small buffers or areas in sound RAM continuously while the game is being played. Since the music is a recording, the artist is not limited in any way. It's possible, using this technique, to produce a musical score and have a symphony orchestra playing it, for a truly cinematic experience.

Games that features some good examples of modern scoring:

Ultima Online - Whether you like the game or not, the music has to be looked from a musical standpoint and appreciated. The period pieces (Trinsic, Britain, Buc's Den) are extremely faithful to a medieval style, and the more modern pieces for the magic cities (Occlo, Magincia) are sometimes pleasant and sometimes spooky. These pieces are all midi with no sampling and its how midi composition should be done: the finished product sounds good on any system.

Unreal Tournament - The reason I mention this music is because it never gets old. It is perfectly incorporated into the media its backing up, and as thus functions like any good film score should. The midi scoring is never so thick its distracting, but its never too thin. The orchestration employed (whether the composer intended this or not) is almost perfect; you always hear what's going on and what the composer wants you to hear.

Vampire The Masquerade: Redemption - Not only is this game visually stunning, the soundtrack was beautifully recorded by the Seattle Symphony Orchestra and conducted by the composer who's name unfortunately slips my mind. This music uses an mp3 delivery system rather than midi and I seriously hope that this is the wave of the future, both as a composer and a gamer. The music is beautiful, chilling, emotional, and very very well written.

In my experience, it's more difficult to write 45 2 minute pieces of music than one piece that's an hour and a half long, because with the 2 minutes pieces you have to get in and say what you mean and then get the hell out. There's no time to develop, no time to screw around with acoustic staging or the big reveal. In short, the music for this game was better than most film scores I hear on a daily basis.

I would like to hear which music from which games my fellow noders have enjoyed. I'll be adding to my w/u as time goes on.

Tips on composing chiptunes

I have composed music for several NES games and demos. Some useful tricks I've found:
  • Most of the time, the percussion is simulated using white noise. (Listen to the Super Mario Brothers NSF to hear how.) This sounds okay, but a fellow can do better by stealing from the bass line (which is a triangle wave on the NES). Play an E-3 (MIDI note 42) with a sharp downward pitch bend; this becomes your bass drum. Put soft, short notes in the white noise channel for the hi-hat. A triangle B-3 with a slightly louder noise serves as your snare drum.
  • Arpeggios. Many tracker systems can rapidly change the period of an oscillator to two or three different notes in sequence. Use octaves and fifths with square waves to add texture to your compositions.
  • Play sound effects through your system's sample channel. The NES supports a decent 6-to-1-bit predictive sample compression method called DMC (delta modulation coding) and can play sounds in the background. I use DMC for sound effects (a la Super Mario Brothers 2) and the oscillators for music; it's a lot easier on MMC3 where you can bankswitch the sample area at $C000-$DFFF. Some developers like to use DMC for drum hits (Super Mario Brothers 3; Contra), but I do just as well with the triangle wave.

Hear what I've done: http://nesdev.parodius.com/2A03/

the countdown begins...
Some observations on video game music, and why it sounds like it does. (Work in Progress)

The first music generated on a home computer was a rendition of "Fool on the Hill" (Lennon/McCartney) created on the Altair (as the machine had no sound capability, the ingenious method of throwing it into differently timed loops and placing a detuned radio next to it was used) - this was described as the only useful thing that was ever done with an Altair. Jeff Minter (The Yak) later reproduced this feat with a Vic 20. I won't go into Sinclair Spectrum and Commodore 64 (or earlier) music here, as it was generally unlistenable beeping. Someone else can wax lyrical about Manic Miner and Ghostbusters...

Sega Master System / Game Gear: If I recall correctly, does not have any hardware capability to reproduce samples, just simple waves and white noise (as described above). However it was possible to get the machine to produce a very fuzzy rendition of speech (see American Baseball), usually pausing all other operations while this was going on. It was best suited to producing a rough approximation of a ghostly flute sound (best used in Castle of Illusion), and those old-style oriental instruments (as in Shinobi). (Sorry, I forget the exact names, but they always appear on even the simplest multi-instrument synth). Bass was practically nonexistent.

Nintendo Game Boy / NeoGeo Pocket Color: both of these machines used the same battery-friendly sound system. It was capable of producing stereo, and seemingly was a bit easier to do sampled sounds on (witness Faselei! on the NGPC, which stored its entire intro music as a 30 second sample, complete with vocals!). Drums are generally handled by white noise, but the processing power of the NGPC at least allows for fairly complex instrumentation to be synthesised (able to do a slightly squelchy, but listenable rendition of the soundtracks of the arcade games that made up most of its software library). The "fluty" noise that is the hallmark of PCM(?) style sound is slightly meatier than on the Master System (maybe an extra channel or two in there?).

Sega Mega Drive: Aah, now this machine started to give developers some room to play with. A slightly unusual* Yamaha chip was given sound duties on the MD, at last providing some FM synthesis. It was, of course, down to the ingenuity of the developer to make these resources sound like interesting. Rob Hubbard at EA and also the sound guys at Technosoft both discovered that the machine could produce a slightly odd but versatile guitar sound (See, any John Madden game, Thunderforce IV, Mutant League Football, Road Rash), even if it did sometimes sound like ukelele speed metal.

Yuzo Koshiro on the other hand was going through his big techno phase, and used the sample channel for kick drums and vocal stabs while using the PCM channels to create an uncanny rendition of a DX7, almost anything by Roland, or a fantastic synth piano (Streets of Rage, Streets of Rage 2, Story of Thor, Revenge of Shinobi). JVP Productions (now Toejam & Earl Productions) took a little from both columns to create funk tracks using the sample channel as a drum machine (with some handclaps and vinyl scratching in there too) and the PCM channels as a funky bass guitar (Toejam & Earl). The MD traded off overall sound quality for versatility - as a result, most games sound very different, but there are no attempts at orchesteral clarity... which brings us nicely to...

Super Nintendo: If the Mega Drive and SNES were "a bit different" in the graphics department, they were completely at odds when it came to sound. The SNES's sound hardware has the inbuilt capability to handle a bank of compressed samples (you can in fact rip these samples directly from the ROM using modern emulator tools). This of course allowed farty, low-bitrate cries of "Hadoken" in Street Fighter 2, but more importantly allowed composers to allocate as much or as little memory as was necessary to build unique sets of instruments. The result, a rich, broad sound (although with rather ropey sound effects backing it up, IMHO).

Square took this to insane levels, with Final Fantasy 6 and Chrono Trigger both having truly awesome soundtracks (check out the opera scene in FF6 ... genius!). Yuzo Koshiro popped up again to score the memorable Actraiser. Many developers seemed to use the same approach and exactly the same instruments however ... for example Konami, whose Vangelis-like synth strings graced dozens of titles (of which I liked Pop & Twinbee best, but that's just me).

Amiga and Atari ST: Amiga owners were always bragging that their games had better music, which was sadly often true, even though the humble ST was powering most of the chart music of the time, thanks to the wonders of MIDI.

The 32-bit consoles brought in more channels, and started to close the gap between in-game music and prerecorded music (which was at last an option, thanks to the advent of CD drives). However not all game music went the way of CD. Nights on the Sega Saturn is notable for its generative musical score, which has hundreds of variations.

Around this time "proper" musicians started to get in on the act (although those with long memories will remember Frankie Goes To Hollywood, Journey, Bomb The Bass and god knows who else all tried their hands at game music in the 1980's), with Sony's Wipeout having a showcase soundtrack by various popular beat combos. Gran Turismo brought in Ash (for a ridiculous fee as the story goes), EA started tacking the latest pub-singalong drossfest on each new FIFA title... the list was endless.

But Quake was the first** game to actually have an original soundtrack made by a recording artist (Trent Reznor), previous games being content to simply plunder the artists' back catalogues, usually not resulting in music that fits the game very well.

(I've not covered PC games in much depth here, due to the fact it would take a large book to cover 20 years of sound technology and hackery, and the title of this node probably means videogames as in console games anyway... check out oldskool.org for some interesting info though.)

So, as we begin the third millennium, game music is seen as a more legitimate art form than ever before (there's even a game music Grammy) - hopefully soon to be seen in the same (or better) light as movie soundtracks. Although of course a game soundtrack has to deliver much more than the fairly straightforward process of aurally illustrating 90 minutes of linear, passive entertainment.

(Please note that I've probably got some or all of the technical terms wrong in this -- I don't know all that much about old sound hardware I'm afraid. Sorry. Since writing this I've discovered that the distinction between what I've called "samples" and "waves" is less clear cut than I thought.)

* This is why DOS Megadrive emulators, which generally map the sound to the Soundblaster 16 mode of the PC soundcard, never sound quite right.

**AFAIK

Actually, Quake may be the first American game with a soundtrack made by a recording artist, but there are at least two British games predating it in which professional musicians were involved.

The first computer game I remember having a soundtrack specifically done for it by "mainstream" artists is the Amiga version of Xenon 2: Megablast, a Bitmap Brothers vertical scrolling shoot'em-up from 1989, whose musical theme was composed by Bomb The Bass.

Also, in 1995, Ocean's 3D shooter Inferno had original music by Alien Sex Fiend.

Nintendo 64 audio

Nintendo 64 had an interesting way of doing sound. It had a MIPS R4300 CPU, a "Reality Co-Processor" (mainly a graphics chip, but I think it did more than just graphics), 4 megabytes of unified Rambus RAM, and game cartriges with 4 to 64 MB of ROM (most used 8 to 16). No sound chip, no sound RAM, not much ROM, so for N64 to make sound, the developer had to cut into the fairly plentiful non-sound resources.

Developers without imagination just let their games sound bad. But those who had made their N64 games sound great. 1080 Snowboarding had very fitting metal, rap, techno and narration. Perfect Dark used MP3 compression and several megabytes of ROM to give its characters a movie's worth of dialogue. The Mario games had hum-able tunes and the Zelda games had sound effects for every occasion. Factor 5 created and leased the MORT codec so N64 games could play voice samples compressed 12:1 without using much CPU time. They used it in Rogue Squadron, Battle for Naboo, Pokemon Stadium 2 and I believe Indiana Jones and the Infernal Machine. And despite taking clock cycles and bytes out of the system, the above mentioned are some of N64's best playing and best looking games.

So N64 could do great sound. But, in order to accomodate everyone, Nintendo has built their latest console Gamecube with a 1.5 GB optical drive, 43 MB of RAM, a dedicated sound chip and a 485 MHz PowerPC CPU. With all that, developers can record music for games just as they do for CD players, and store it all in RAM so the lens can be free to load new game data during play.

Log in or registerto write something here or to contact authors.