Abbey Road's famous Studio One has played host to innumerable big-budget orchestral recordings for film, but the final session before its refurbishment in 2001 saw it being used for a new purpose: to record the soundtrack for a video game.
Richard Jacques (left) and Matt Howe (right) flanking Headhunter's virtual hero Jack Wade."This is probably the most expensive computer game soundtrack yet," says Richard Jacques of his music for Headhunter. "The orchestra, who have done loads of films, had never done a game before. There are some games that use orchestras now, but to my knowledge this is the first one that's been done on this scale. A couple of PC games have used orchestras in Moscow and Prague, and done it comparatively on the cheap. There's still a lot of orchestral soundtracks done on synths in games, because the publishers still don't quite understand how important it is to commit the financial resources, and also they don't really understand the business model where you make the money back on releasing the soundtrack. It's huge in Japan — I've had nearly every game I've done come out as a soundtrack in Japan, and nowhere else!"
Although he's recently left to become a freelance composer and producer, Jacques was in-house composer at Sega UK for many years. For his last project as a staffer, he was given the job of writing the soundtrack for a crime-fighting action game funded by Sega but developed by Swedish programming house Amuze. "The developers of the game were very keen to make the whole game as cinematic as possible," explains Richard. "They got a scriptwriter in who'd worked on X-Files and Millennium and sci-fi shows like that, and there's a lot of cut scenes in the game, some involving live actors, so the whole ethos of the game was to make it very cinematic. The director of the game always said he wanted an orchestral soundtrack, and I said 'Well, I'd like to try and get the budget to do it properly.' I had to persuade my ex-boss at Sega to put up the money. I did a few costings and the cheapest would have been about £60,000, that would have been to go over to Prague or somewhere like that, and this recording cost about £100,000."
The Planning Stage
Although £100,000 might sound like a lot of money, orchestral recording is expensive, and bringing the project in to budget required hard work, some compromises, and a lot of careful planning: "We got three days of studio time with a 64-piece orchestra. That included all the personnel, a bit of equipment, Pro Tools hire, and score copying. I actually put the team together myself, mainly out of people I've worked with before. I've done two games with Matt [Howe, who engineered the recording] before, there's a guy called Steve Lloyd who conducted the orchestra, who's a friend of mine, so I selected people that I wanted to work with, that I knew were good and could come up with the goods on time. Abbey Road probably gave me a bit of a discount because I block-booked three days, and I booked it very early, which made a big difference to getting the people I wanted. Then the mixing was done up at Sega. I did most of that, and Matt helped out."
An already ambitious project was complicated by the fact that on the day after the Headhunter sessions were scheduled to finish, builders were moving in to completely refurbish the studio. "The 31st January was the last day before they started tearing the control room apart, and a lot of the musicians were doing Captain Corelli's Mandolin the day before we came in, so it was pretty tight. Also, the budget was £100,000 and the three days were costing us about £99,600, so there was no leeway with time or budget! We were pushing the orchestra right up until five o'clock on the last day. There was sweat pouring off them!"
"The whole thing was, I think, an incredible achievement," says Matt Howe. "It was the first time Richard had had an orchestra record his work, and we crammed in almost 70 minutes of music in three days. None of the orchestra members were familiar with Richard or his music, and Steve was new on the scene to a lot of the players, so everybody had to get acquainted with everybody else, and we had to get a sound balance that we liked."
"The players were used to sight-reading, but they hadn't played any of my music before, so they didn't quite know how I write," adds Richard. "I'd done a lot of orchestral writing at college, but not that had particularly been recorded, and then since I've been doing games for the past seven or eight years, it's the first time I've had the opportunity to work with an orchestra. So I certainly put my all into it, and made sure that nothing could go wrong, and I think that was they key, to be as prepared as possible."
Streaming Or Surround?
Because Matt and Richard used surround-compatible miking techniques to record the orchestral soundtrack to Headhunter, they could have created a 5.1 mix. Not all games consoles would have been able to take advantage of this, however, and on those that could, working in surround would have made impossible some of the tricks Richard used to make the music seem interactive. "Originally, Headhunter was only going to be released on Dreamcast, which is stereo," says Richard. "The Playstation 2 can do 5.1 surround, but the PS2 version isn't going to be in surround, because it ties up the drive and we couldn't have done the cross-streaming thing. You have to pull six streams of audio off the disc to do 5.1, and there's one point where we're already using six streams and a string note held in sample memory, and that was more important.
"I'm going to think about it next time, but I think I still might choose playing multiple music streams over just playing music in surround. I'm much happier doing real-time in-game sound design in 5.1, which you can do on PS2. If the console is capable of decoding surround in real time, like PS2 or X-Box, then I would much rather work with the sound designer to say 'Right, you have these three extra speakers, and I'll have another four streams coming off the CD!'"
Composition & Preparation
The most important element of the preparation was, of course, writing the music in the first place. The soundtrack had to be tightly integrated into the game, which required a lot of precise timing and working at strict tempos, and Richard mocked up the entire score in detail on his array of Akai samplers. (You can hear MP3 excerpts from these demos in the righthand sidebar of this article.)
"I spent about six months actually writing the music," he says. "It was all written and sequenced in Cubase and programmed up on Akais. I write in the sequencer rather than on paper, especially when I'm doing this kind of music. It's much better having the whole palette of sounds set up. I've spent five years tweaking my orchestral setup, and it's now taking up nine Akais — one S6000, four S5000s, two 3200s and two 3000s — as well as an MPC2000. So it's about 3Gb worth of sample memory in total, and a lot of expensive libraries as well."
The choice of Abbey Road was made when Richard played his sample-based demos to Matt Howe. "I knew straight away I wanted to be there," says Matt. "It's a very familiar sound, particularly with the French horns and the brass. You can instantly recognise some of the Jerry Goldsmith or the Harry Gregson-Williams stuff where they've been there, which typifies the symphonic sound that Richard was looking for. I wanted to keep that classical open recording approach, ideally recording it with a main pickup, but also to have close mics as well to cover us if we needed to bring out a solo line. But we needed control as well. There are one or two other venues that could have coped with the control aspect, but it wouldn't have been so well recorded.
"I approached this exactly like a film score orchestral session. The thing is that Richard had tempo-mapped everything so he could do his edits and looping and whatever, so it was all tempo-mapped, and everything was accompanied by drum loops, so it was metronomic. Everyone in the orchestra was hearing a click on headphones."
"I made sure I put all the tempo maps from Cubase into Pro Tools so that I could check that it was working before we came to Abbey Road," explains Richard, "so every cue for the game had a Pro Tools Session with all the tempo information in there already, and I had eight submixed tracks with the sampled parts — drum loops, basic strings, brass, wind, percussion — so that we could hear any of it if we wanted. And then we basically just recorded the orchestra on spare channels. I literally just took two removeable Rourke hard drives back to the office, loaded up the Sessions and they were ready to go."
The Headhunter soundtrack was a large Pro Tools mix, incorporating as it did both ambient and spot mics for the orchestra, plus programmed material, percussion and other overdubs.
Through The Prism
Matt's decision to undertake the engineering exactly as he would have done for a film project extended to the technique used to mic up the hall: "Everybody uses the Decca principle with three mics above the conductor, you'll see that on so many sessions. It's very suitable for surround-sound film mixing, the main overhead pickup is a left-centre-right, and you can actually use it on a discrete system as L-C-R and feed each mic into the system, so it's very compatible with surround sound technique. So if Richard needs to, he could actually do a 5.1 surround mix."
As the plan was for Richard and Matt to mix the entire project themselves in Sega's in-house studio, which was based around Pro Tools, this was the obvious choice of recording medium. A Sony 48-track DASH machine served as a backup system, and the multitracks were actually recorded to two Pro Tools systems. As well as the standard rig the pair had hired, which used four of Digidesign's own 888I/O interfaces, Matt was in the process of evaluating another system with Prism converters and mic amps, and brought this to the sessions as well. "Matt just said 'Look, I've got this stuff, can I bring it along and run it in the background and try it out?'," says Richard. "I said 'Yeah, as long as it doesn't get in the way of my system!' I was recording at 16-bit, 44.1kHz on my system, and Matt took everything at 24-bit, 48kHz through the Prisms. We ended up using all the recordings made on the Prism, and they sound absolutely amazing. You really notice it on the top end of the strings, especially, there's a massive difference."
"I ran all the mics directly into Prism outboard mic preamps, so we didn't record through the console, even though it's a very adequate console," says Matt. "I was being as purist as I possibly could, because I was able to take advantage of the equipment. I didn't expect the comparison to be so wide. It was almost as though we were comparing between Minidisc and CD on stereo content, but we were comparing multitrack content, and the difference in spatial width and clarity made it sound like it was a different take. I'm still using other people's systems that are 888-equipped, though, and for a lot of work they're absolutely fine. It didn't do Ricky Martin any harm! The discrepancies we're talking about are noticeable to recording engineers, producers, and composers, but I'm sure the general public will be happy to accept the results."
Richard agrees: "I personally have never had a problem with the 888s, but on orchestral material, it's more noticeable. A lot of my work is 16-bit, 44kHz, and when it does go onto a games console, it often loses a bit of top end due to compression — if there's any clever streaming going on then it has to come down a little bit. So it was even better using the Prism, because there was that extra top end — by the time you did run it through the converter and get it working in the game, it sounded as good as CD."
Interactive Music
Writing music for video games shares some of the constraints of film work, but adds a few of its own. Whereas the scenes in a film have a fixed length and order, game music has to evolve and adapt to reflect the player's actions. This is one reason why games composers are often willing to sacrifice the superior sound of streamed, pre-recorded audio in favour of the flexibility that comes from using MIDI files to trigger samples from a console's internal sound chip.
"There are a lot of interactive soundtracks in games now, but they have to be done largely on the actual sound chip and held within memory and sample RAM, they're effectively MIDI sequences," says Richard. "The reason why we didn't do that is because we didn't originally know it was going to be out on Playstation 2, and the Dreamcast has only got 2Mb of sample memory, which has also got to hold all the sound design. So we weren't keen, and I knew it wouldn't sound that good. We decided to make it slightly more linear, but work with the CD, and do some clever streaming and cueing of the actual music. What we've done is that we've got various kind of 'moods', if you like — you get some very sneaky walking music for various characters, then as soon as a battle takes place, full-on action music that kicks in immediately.
"Although the music is pre-recorded, it's always cued in an interactive way. It was always written with that in mind, so it could loop if need be, so the pieces end up in the right key and so forth. A lot of them had to be pretty much the same all the way through — not necessarily melody- or harmony-wise, but it had to be 'this big' all the way through, because you're providing a minute's worth or three minute's worth of this kind of mood, so you can't go off on a tangent and change it too much. It proved to be quite hard to keep the momentum up and not make it boring, especially with some of the bigger action sequences. It's quite a thematic score — there's quite a few themes for the main characters of the game, so that provided a lot of material to make it easier to rework some of that.
"There's only about two pieces that loop, and they loop for the whole piece rather than just a section. Of course, I had to write starts and ends for a potential soundtrack release as well, so I had to finish them off properly. On the soundtrack CD they have proper ends, and on the game they loop, so we had to do different edits, and we had to make sure when we were mixing that it was at the right level when it comes back to the beginning. There's a lot of considerations to think about right from the early stages."
After one battle scene, an ingenious scheme was concocted to create an impression of interactivity: "We've done some clever streaming when it comes out of that scene. There's a look-up table held in the sound memory containing all the tonality information for that particular cue, so wherever you are in that piece, when the gun battle's finished, it says 'OK, we're now in F sharp minor' or whatever, and it holds to a string note which is held in sample memory, then it fades out the music track over a short while, and then it brings in a percussion track. It seems to work really well. It's all very seamless, the player doesn't notice that it's playing different tracks. It's like you're in the middle of this gunfight, and all the drums are going off and all the brass are going crazy, and then you gradually go back to just walking along."
Division Of Labour
When recording commenced, it soon became clear that because of the nature of the music, a single-take live recording of the whole orchestra was not going to give them the control needed for best results at the mixing stage. "After the first cue or two, we realised that we had to alter the balance. From the templates that Richard was sending me, I had had a suspicion that we might have to alter our approach as we went along," admits Matt. "I had wanted to be completely traditional and purist and just record everything openly and together, but the meter was very tight. A lot of the music is really highly charged, very energetic for the musicians, and even the close mics are extremely live because the studio's got the decay of a church, very long and trailing. When you've got fast rhythmic writing, that can sometimes work against you.
"Next time I think we'll do what I suspect a lot of contemporary film composers do, which is do some extensive overdubbing in the percussion department. The flavour of the soundtrack is pretty military, and we really needed the focus to be on the writing on the snare, with the snare playing sitting inbetween the first and second violins, quite up-front. The other thing I noticed was that the weighting between the strings and the other instruments was a bit out. If the string section had been a little bit larger, that would have helped the power ratio. The usual numbers in a symphony orchestra, when you're talking about desks of musicians with two players per desk, would be 16 first violins, which would be eight desks of firsts, plus eight desks of seconds and six or seven desks of violas."
"I would have liked to have had a bigger string section, but the budget wasn't really there," agrees Richard. "We had 14 firsts, 12 seconds, 10 violas, eight 'cellos, and six basses. After about the first morning's worth of recording, we took strings and harp, and then wind, brass and percussion, just completely separately. I think we more or less did the strings first, and then did the rest, which meant that when we wanted more control or a bigger string sound, when we rode up the string lines we weren't also riding up the snare and all the really loud percussive stuff in the background. And that was the way we mixed the whole lot. When we realised that this was what we wanted to do, we'd sit down, they'd have a bash through sight-reading it for 10 or 15 minutes, we'd send the wind and brass off for a tea while we do the strings, swap 'em over, and job done. It worked very smoothly after that."
"We made sure that every time we did the brass and percussion separately from the strings, I was also capturing the room every time, so it wasn't just the close-miked signal, I was also recording the main array," adds Matt.
As well as the programmed rhythm loops, other percussive elements were added later from sample libraries, as were piano parts and choral pads. "We use live timpani and snare," says Richard, "but stuff like suspended cymbal piatti and triangle, they're off libraries. I thought 'Well, I'd rather have two extra first violins than a suspended cymbal roll, which is always going to sound the same.' A crash cymbal's a crash cymbal."
"In a room like that, as well, even a triangle can really ricochet all the way around the room, and in some of the fast-tempo pieces that was really what we didn't need," adds Matt.
Hardware vs Software Samplers
Richard Jacques has spent a considerable amount of time and money investing in the technology used in creating this type of music. His studio includes not only a Pro Tools system with seven Mix Farm cards, but no fewer than nine Akai samplers, on which Richard has built up an impressive suite of orchestral sounds from the various libraries available: and despite the quantum leap in sample memory offered by software systems, he seems happy to stick with hardware.
Some of Richard Jacques' Akai samplers: from top, two S3200XLs, two S5000s and two S3000s."I'm going to have a serious look at Gigasampler, but at the moment I don't think I need it — I'll just buy another two S5000s!" he laughs. "You're still limited with polyphony and memory, admittedly more memory, and I don't know what the latency's going to be like. The only thing that does tempt me at the moment is that there are libraries starting to come out that are not getting ported to Akai format. I haven't really used Gigasampler, and as far as I know the only real operational advantage that it has is that you've got control keys that are used to switch sample sets, so you can hold down bottom C, and that defines that you're playing legato, or you hold down the bottom D and that defines that you're playing staccato, and it's actually switching sample sets in real time as you're playing it in. That might save me a bit of time. Some of the demos on the Gigasampler web site are just not good, though — I'm sure the libraries are great, but they're horrendously programmed demos. Give me the Dan Dean stuff and I'll show you what can be done with it!
"It's not really what you've got, it's how good it sounds, although it might take you a bit longer to get there doing it on Akais! In my setup in Cubase, I've got over 200 tracks. In my first violin, I've got 35 sequencer tracks triggering the array of sounds I've got in there. My setup is from various different libraries, and I set it up how I want to work with it. If I want to do legato strings pianissimo in the high register, I know that library X is going to sound better than library Y, but if I want to do fast, spicato, fortissimo, I know that library Z's going to sound better, so I'll set it up accordingly, and once I've played the line in, I work with it until it sounds how I want it. I'm lucky because I was originally an orchestral player, and I know how orchestras play, and a lot of people don't, which may be why there's a lot of bad programming out there. It takes a bit of getting used to reading what people write in the sleeve notes. The Miroslav Virtous library has disgraceful packaging for the amount of money it costs — they haven't even named the volumes or anything, and it's three grand's worth of library, whereas the Peter Siedlazcek library is really well explained. Because these are the people who designed the libraries and they know how they recorded them, they can say 'When you play a scale on the first violin, if you add 100mS to the length of the note it will sound legato,' or whatever."
At The Mix
The mix in Pro Tools was no trivial matter, involving as it did the blending of both close-miked and ambient orchestral tracks with extensive programmed parts. "We used reverb on the close mics, because they can have too much definition," says Matt. "Otherwise it would be too obvious when you ride the faders up to lift the violins that they would become very dry. If you want them louder, I think you have to design a reverb to the dimensions of the room."
"Primarily we used Lexiverb on Pro Tools, which is quite an under-used plug-in," continues Richard. "TC Megaverb is far too bright and it's not very realistic, D-Verb's OK as a budget reverb, but Lexiverb is absolutely superb, and a decent large hall setting on that is ideal for any kind of orchestral music. On a lot of my sampled demos all the stuff is done with Lexiverb, and we used that on the overdubs. I think I actually separated it into string reverbs and woodwind reverbs, they're both Lexiverb hall settings, and then we had one room setting for some of the people further back in the hall to get that sort of slap-back on the woodwind and brass. I might have used a TC Megaverb for some of the sampled or electronic percussion, because it's perfect for that kind of thing.
"I set up one Waves L1 compressor master setting for the whole game, and left it at that, and there was only a very slight bit of compression on the mix. We were quite careful with the EQing, though. I was running a Vac-Rac, which are 3U things you can have four modules in. In mine I've got a mic preamp, a limiter and two EQs, and we were using the EQs on that. They're probably not as good as an Avalon or something like that, but that was the best analogue EQ I had available, and it actually worked really well. We just set it up and left it over the orchestral mix. I had one stereo buss with just the live orchestra, and that EQ was set when Matt came in and set up the first mix, and then I left it throughout the whole project. As for the rest, because there is a lot of synth stuff, I had to work with it a bit, and there I normally used plug-in EQs just to sort out a bit of bass and add a bit more clarity on the top end of drum loops, and that sort of thing. It's a relatively bright mix, but we liked it that way."
One of the big differences between recording for film and recording for computer games is that in the former case, the principal playback medium will be a high-quality cinema surround system with a dynamic range of 100dB or more. The games composer is not so lucky, and has to face the fact that many players will not even be listening in stereo. "I think most games composers and people working in the medium realise that you have to find a common balance, which is normally a pair of decent stereo speakers," says Richard. "If someone's going to have a 14-inch black-and-white TV from Tandy they should expect it to sound crap, and it will sound crap. But I still try and make the mixes sound good on a good pair of monitors, and then you have to be aware that people will come down from that. The only real heinous crime I've made in the industry was on about my second project, when I did this drum & bass track, but when it came to playing the game it was only drums! It was one of those deep sine-wave basses, and I mixed it so low it just wasn't there!
"I mainly mix on Dynaudio M2s, constantly checking on NS10s, and I've got some Tannoy PMB 6.5s, and I've got some little Dynaudio BM5s as well. I check on a horrible half-inch speaker built into a Sony monitor, and I check on Yamaha PC speakers, and check the mono button quite frequently, and then I always check it at home on my knackered old hi-fi speakers, and Matt checks all the stuff on his system at home. So basically, the rule of thumb is to check it on as many systems as possible. You can never play it on enough systems. It should sound good and beefy on a decent pair of monitors, and the Dreamcast and PS2 have both got relatively good capabilities, but you're still only talking about a chip that cost £1 that's going to be decoding an audio signal. It's still not going to be as good as a £100 CD player, so you have to compensate."
Their quest for clarity also led Richard and Matt to 'cheat' slightly on one or two cues, by thickening the real orchestral recordings with the sampled parts Richard had created for demo purposes. "I think on about two tracks I did that just to thicken it up," says Richard. "It was minus 25dB or something that the sampled orchestra was set at, and the rest were up at minus 0.8 or something. It was more for the definition on the faster cues than anything else. There are some fast, scaley-type runs, and that was where I needed to bring the strings out a couple of times. If I'd had a bigger string section and perhaps a bit more rehearsal time, I probably wouldn't have needed to do that.
"Not many people publicise it, but I reckon it goes on a lot more than people admit. It's a bit taboo for some people, but any Hans Zimmer or Harry Gregson-Williams film, I can tell that they've got a bit of Miroslav Virtous in there as well as an orchestral recording!"
Headhunter is available on Dreamcast and PS2 now.
Jack's Theme
Here are two demos from Richard Jacques' soundtrack to Headhunter, put together in Cubase with the aid of his Akai samplers and orchestral library. Richard is at pains to point out that these are only rough demos and were never intended for release in this form — as they were only ever intended for demo purposes, they weren't mixed with the same care and attention he normally devotes to his sampled orchestral work — but even so, they're pretty impressive... If you want to compare them with the full versions recorded with the London Session Orchestra at Abbey Road, buy the game, buy the soundtrack CD, or visit www.music4games.net.
headhuntertrack01.mp3 Track 1
headhuntertrack05.mp3 Track 2