Dance of the Dead

In addition to creativity, producing music for games requires a keen sense of balance in negotiating aesthetic decisions, budget constraints, and technical

In addition to creativity, producing music for games requires a keen sense of balance in negotiating aesthetic decisions, budget constraints, and technical issues while working within a fast-paced and ever- shifting schedule. Now that PC and console game technologies allow for high-resolution music recorded live, the stakes have been raised higher than ever. Given the impact that live performance adds to a soundtrack, and the vast musical territory that must be covered with a limited budget and short schedule, key parts must be chosen and recorded with great efficiency. Every take must count, and never was this more true than when I worked on the score for LucasArts' Grim Fandango.

Grim Fandango is a graphic adventure game, which means that it's like watching a movie while you control the actions of the main character as he or she interacts with the environment and the other characters in the game. Grim Fandango's leading man, Manny Calavera, sells travel packages to souls who journey in the Land of the Dead. But something is awry in this film-noir-like world. The travel tickets are going to souls who don't deserve them, leaving the good souls without a way to the Ninth Underworld, or Land of Eternal Rest.

One of these victims is the lovely Mercedes Colomar, for whom Manny has developed a more-than- professional interest. Manny, along with his demon driver and sidekick, Glottis, must find Mercedes and get her the tickets she deserves. Thus the scene is set for a four-year journey through a fantastic land with cities of sin, corrupt port towns full of casinos and shady characters, wild forests, an underwater slave city, and a majestic Mayan temple in the mountains. Bringing this diverse world to life with music was a huge task that yielded three hours of finished score-as much music as you would typically hear in three feature films.

Manny Calavera and his sidekick, Glottis, share a rare casual moment in Manny's cafe. In this scene, we hear Glottis noodling absent-mindedly on the piano. As Glottis leaves, muffled jazz can be heard from the casino next door. Such use of "source" music (rather than underscoring) was made more believable by employing live performers.

PLAN AHEADWith all this music to produce on a limited budget, how do you make the most of your live-recording time? Well, for one thing, you begin early and plan ahead, keeping in mind a complete picture of the production process. From its inception, a game like Grim Fandango may take more than two years to produce. Most production activity takes place over an 18-month span, with the heaviest emphasis at first on the art. The emphasis later shifts to programming. Because much of the art and programming must be finished before the composer can decide how to score the game, the bulk of music production occurs during the last nine months.

Nonetheless, the initial phase of my work begins at least nine months before the music production starts. As soon as I know that I will be working on a project, I obtain all the preliminary documents and learn about the game and the story. I keep a handheld cassette recorder with me at all times and sing into it whenever an idea strikes. When I'm near a keyboard, I record simple harmonized versions of these themes. Because I'm a big theme guy-characters first, then places-I like to see at least rough pictures of what these elements will look like in the game. I constantly run my themes past the game designer by just singing them live with piano accompaniment. At this point, I don't even bother sequencing a MIDI version.

A THEME TO YOUR LOOKEven though this initial phase is a background task while I'm working on other projects, hitting the mark with these themes as quickly and frequently as possible is still important. Coming up with the right themes early on inspires the confidence of the game designer and the project team.

As we go over the themes, we often talk about the style and direction of the music. I think about what the orchestration will likely be and which parts I would want to record live. The diversity of the Grim Fandango universe suggested a number of disparate musical styles: swing jazz from the noir era, jazz from other periods, and mariachi, Peruvian, and even Indian music.

I kept the score cohesive by building it on a backbone of classic Max Steiner-style orchestral underscoring. To internalize Steiner's sound, I studied full scores from sections of The Treasure of the Sierra Madre, Casablanca, and The Big Sleep. I watched those films about a hundred times, along with the Adolph Deutsch-scored film The Maltese Falcon. I read scores to several Debussy works (for the ocean music), listened to plenty of Duke Ellington, and made trips to San Francisco clubs for swing homework.

ENSEMBLE CASTINGWhile the first phase is still underway, I start to assemble a synth "ensemble" for the next phase of music production. I firmly believe in having one master template of sounds that never changes. In that respect, Grim Fandango presented a real challenge, because it had so many types of environments. I finally decided on a sort of virtual orchestra set up on an imaginary stage in such a way that there was a space for small ensembles of (mostly swing) musicians in front of the stage.

This setup was created using two E-mu E4s maxed out with 250 MB of memory, along with an E-mu Procussion and a Roland Sound Canvas. Good sample libraries were essential. I set up a template in MOTU Digital Performer, so that each virtual instrument was accessible on one of 68 MIDI channels (64 E4 channels and two channels each for the Sound Canvas and Procussion modules). External processing was provided by a Lexicon PCM-70 and an Alesis Q2.

With most of the major themes written and approved and a MIDI/synth configuration established, I was ready for the killer nine months of heavy production. At this point, it's worth taking a little detour to discuss the roles that music plays in an adventure game such as Grim Fandango. Each role has different implications with regard to live recording.

ON A ROLEThis type of adventure game has essentially three kinds of musical pieces: ambient pieces, event-triggered episodes, and underscoring for noninteractive movies (called "cut scenes" at LucasArts). In planning the production of these kinds of music, you must keep several things in mind: How does the player experience each piece in the game? How often will a piece be heard? Does the music need to loop or repeat, and if so, for how long? Ambient pieces provide the background for most of the action in the game. Event-triggered episodes and cut scenes, however, are likely to be experienced only once by the player and therefore may not be worth the time and expense of producing several live parts. On the other hand, these types of pieces serve as rewards to the player and showpieces for the game, so you must strike a balance.

The schedule further complicates the situation because cut scenes and one-time animations usually are not ready until late in the production process, and even then they're on an unpredictable timetable. This creates difficulties when scheduling recording sessions-writing the music in the allotted time is hard enough, let alone recording it. There were weeks during which I had to compose as much as two minutes of finished music a day, and that's a heavy regimen.

Nevertheless, I still wanted to keep the quality high by adding live parts where possible. One way I accomplished this was to emphasize parts that we could play directly into Digital Performer, such as violin and acoustic guitar parts. In that way, we avoided having to schedule sessions with outside musicians. Another tactic was to adapt the compositional style to include the editing of "wild" or improvised recorded parts (more on this later).

STATE OF THE STATEAmbient music takes up the lion's share of the production process and is by far the trickiest to write. In the LucasArts sound department, we also call this "state" music, because each piece of this type accompanies an open-ended "state" or condition at which the game has arrived. Such states are often tied to location in the game but may also relate to the player solving a puzzle or reaching a goal or pivotal point in the story.

Because there is no set time as to how long a player will take to change from one state to another, writing state music is the least like any other kind of composition. State music must be out of the way without being boring, it must intrigue and encourage the player without getting obnoxious, and it must withstand repeated listenings. These pieces are usually loops, and part of the trick is to make the loop happen in an unexpected place, so the listener is well past the loop point before realizing that the music is repeating.

Another compositional approach to writing state music is to start with a short flourish, or "stinger," and then end or fade the music and let the ambient sound effects take over. I used this second approach heavily in Grim Fandango.

Composition of state music begins when the main themes of the game have been established and the synths are configured. Because there was so much ground to cover (roughly 110 state pieces in Grim Fandango), every compositional shot had to count. My approach was to create a short sketch for each state-maybe 15 seconds long-and then move on as soon as I'd hit the target once and gotten the gist of the mood. You'd be surprised how much a piece is determined by what happens in the first 15 seconds. Besides, coming back later and completing an idea is much easier than coming up with a fresh idea under even greater pressure.

SESSION CONCESSIONSHaving paved a good portion of the game with 15-second stingers, it was time to start scheduling recording sessions so they could start at least two months before the score had to be finished. This is a bit of a gamble because you typically need as much as a month of lead time to obtain certain players and because there is usually disagreement about what the actual schedule will be. The upshot is that you're booking sessions for pieces that you haven't written yet, within a time frame that makes you really nervous. To be on the safe side, I also schedule the sessions in groups. I do the parts that positively must be live first and the less critical parts later because I realize that I might never get to some of the parts.

Because the sessions were booked before the music was written, I had to come up with some guidelines for what to record first and what kinds of parts I was likely to need. I set clear priorities, based on the heavy jazz element in the score, as follows: (1) With the exception of a few scattered cymbals, not a single fake drum sound would make it into the final product. (2) Nearly all of the bass parts would be real. (3) Next, sampled horns were to be replaced by real ones whenever possible. (4) The reeds were next, with saxophones taking precedence over clarinets. (5) The stringed instruments, including solo violin and guitars, were to be replaced last (see Fig. 1).

As the sessions approached, leaving enough time to print out and prepare parts for the musicians became important. For a number of reasons, it takes two or three solid days to prepare parts for a week of recording. On Grim Fandango, I decided it would be most efficient to use my sequencer's built-in notation section to print parts. That way, I could avoid getting bogged down with the zillions of extra features found in most high-end dedicated notation programs. However, I ran into two big headaches: pitch and rhythm. In any given piece, it may make the most sense to spell a note as F# at one point and as Gb somewhere else. Having just one of those two for the whole piece can be pretty hard on the players, so I ended up using lots of correction fluid.

As for rhythm, I dislike quantization in anything but techno music (which I didn't use in Grim Fandango). The jazz, in particular, had to be rhythmically very loose to be convincing. That may be good for the musicians to play along with, but it's terrible for printing scores. In addition, standard notation doesn't always represent the way the music is actually played. ("Straight" eighths and "swing" eighths, for example, have different values but look the same on paper.) So I usually had to make a separate copy of each sequence for playback purposes only, and then move note positions and lengths by hand to get them to make sense in the notation section.

STUDIO STRATEGIESWith the instrumental parts in hand, it was time for the best part of the whole effort: the recording sessions. To ensure that the sessions go smoothly, it's important to start with the right studio setup.

At LucasArts, we have an unusual studio configuration. The composers' offices are linked to the main control room and studio via audio and control tie lines. This allows us to keep sequencer sessions on a Mac in the composer's office (which contains the full sampler/synth configuration mentioned earlier) and sync them up with a 32-channel Mac-based Pro Tools system in the main studio. That's where all the good mixing gear is. In addition, we can use the tie lines to monitor either room from the other.

This setup has two advantages: first, you don't have to tax one machine with the burden of running both Pro Tools and a sequencer. Second, it allows composers to sequence off-line in their offices and even to record simple live parts directly into Digital Performer without tying up the control room. You could call this technique "guerrilla production"; it's quick and dirty but allows you to knock off lots of live parts with minimal effort. Then when the main studio is really needed, you can bring it into the picture with only minor reconfiguration. To streamline the operation even more, we use Farallon's Timbuktu software to command the composer's machine from the control room. That way, there's no running back and forth. Everything can be done in the control room.

Recording in LucasArts' main studio is the most rewarding part of the production process. I get to take off my composer's hat, put on a producer's hat, and collaborate with the musicians and a great engineer to bring the score to life. Of course, it can be tense, too; the clock is always ticking, and the budget is always dwindling. This is where it helps to have an ace engineer-always a luxury in our industry. Jeff Kliment was the man behind the faders on Grim Fandango, and he made sure that everything sounded great and that the sessions ran smoothly. Kliment was also lead sound designer for the game, so he was responsible for the final mix of music, dialogue, and sound effects-in other words, the overall sound of the game.

Another important way to keep the recording process fun and efficient is to be clear about what you want as a producer and to discern what each musician can bring to the music. One player may be a phenomenal reader, for example; another may give you something amazing if you just provide a few melodic and stylistic guidelines and cut him or her loose. On Grim Fandango, I soon learned that with music based so heavily on swing jazz, getting the best results often meant not being too attached to the written score-no matter how good I thought my original ideas were.

I took this approach to a new extreme as the recording sessions progressed, and it actually changed my compositional technique. As explained earlier, I often have to schedule sessions in groups, before the music is written. I realized that live playing was really giving life to the score. In fact, the playing mattered almost more than the notes themselves. What's more, I could benefit doubly by taking advantage of improvised sections of music to cover more ground in the game than I could if I painstakingly recorded multitracked parts. In other words, I could hit more musical targets in less time.

With this in mind, Kliment and I started recording musicians in twos and threes so that they could play off each other. We then used digital editing in order to shape these "wild" sessions into more structured pieces. In some cases, I even had one of the musicians, Hans Christian (a producer in his own right), do this editing himself. Then I added synth orchestra parts to fill out the textures and provide more harmonic context. That was exactly opposite from the order in which we did the more "standard" pieces in the score. The result was a truly alive feeling borne of the spontaneity and collaborative nature of the recording process-a huge benefit to the overall musical score.

The mixing session usually follows the recording sessions. In this case, however, we had to be superefficient, so we merged the two stages by mixing some pieces before others were recorded. Everything about scoring for interactive media revolves around one basic fact: you don't know how much you will get to finish until you're near the end. You can't record three hours of multitrack music, mix it down, then dump it in one big truckload on the project, and expect it all to be programmed into the game during the week before final testing. Instead you have to mix early and often, to get the flow of finished pieces coming into the project as soon as possible.

Remember that for the engineer, mixing involves not just the music but the integration of the ambient sound-effects bed, which is a major part of the final sound. In the case of cut scenes, there are also dialog tracks and foreground sound effects. And there is also an important phase in which levels are adjusted so that the cut scenes and mixed ambient pieces all match each other.

Finally, the entire soundtrack is slowly scrutinized in one last meticulous pass that takes a couple of weeks. During this time, the engineer tweaks the levels of all the sonic elements until they balance each other and the music and sound programming work the way they should. This process alone is fodder for another complete article; suffice it to say that, because it accounts for perhaps a third of the perceived overall sound quality, we take it very seriously.

FINAL TAKEAs Clint Eastwood once said, "A man's gotta know his limitations." Every project has its constraints. Indeed, I believe that good art thrives under constraints. While working on a big interactive project like Grim Fandango, it's imperative that you keep your sense of perspective. You want the greatest part of the experience to have the best quality possible. That always means cutting some corners somewhere. Looking back, for example, I wish I had been able to find a better piano sample, but I'm not sorry that I didn't record the piano live. Too many other things would have suffered.

Unlike Manny Calavera and his friends, the art and craft of scoring to interactive picture exist in the real world. So does the payoff-playing a finished game with a rich sonic landscape, and knowing that thousands of other adventurers can do the same.

When Peter McConnell isn't composing and editing music at LucasArts Entertainment, he sings and plays electric violin in the Bay Area band Spinray.