What does it take to transform a composer's sample-based demo into a form that real musicians can read and play?
The job of the professional orchestrator is one of the most misunderstood in today's music industry because, reflecting that of the composer, it has changed so profoundly in recent years. Only a couple of decades ago, a composer like John Williams would sit with Steven Spielberg and play the cues through on the piano, occasionally explaining that "the French horns will come in there," or "the strings will play that motif." The composer would then prepare a sketch, which would be fleshed out by an orchestrator — the director wouldn't hear the final piece until the recording session.
That method of a composer showing the client their work is now obsolete. Instead, they're expected to use their DAW and sample-based instruments to conjure up super-realistic mockups of their scores. My aim in this article is to explain how this change has impacted on the role of the professional orchestrator working in the film, TV and games industries, and to offer some tips for anyone looking to start out in this line of work.
First, it's worth considering why this approach has become the norm. While the sophisticated sample libraries that enable the mockup approach are a relatively recent development — orchestral libraries have existed since the '90s but we've had to wait longer for the computing power and affordable data storage that allow them to be everyday tools of the trade — the change in approach wasn't driven purely by new technology. Mockups are now the norm for a far more predictable reason: money.
The budgets of Hollywood movies can often run to several hundred million dollars, but even for something like a prime–time BBC drama, there's considerable money at stake and a lot of people are involved. On the music side alone, several minutes of overtime with a large orchestra can easily run into four figures — it's not hard to burn £100,000 in a few days with a full orchestra at somewhere like Abbey Road or AIR. So delays and errors at any stage of the project cost significant sums. One of the worst nightmares for a director, then, is to turn up at Abbey Road Studio 1 to hear the first cue and hate it. In days gone by this happened; whole scores for some very famous films were jettisoned at the recording stage, later to be replaced by the music we're familiar with today. Just imagine how terrifying that would've been for all involved.
A desire to minimise the financial risk in large projects means that composers' clients want any aspects of the music they're not keen on fully addressed long before the recording sessions. Convincing MIDI mockups make this possible. The director will hear each cue hundreds of times against the picture and come to know every hook and melody. Every conceivable musical direction can be discussed before the project gets anywhere near the studio. Occasionally, cues do still get rewritten on the stand, but I've been to hundreds of recording sessions and at none has the composer had to start rewriting a cue, with the orchestra waiting. In fact, I know of an extremely successful TV composer who informs his clients that the only changes he's willing to make at the studio are dynamic markings, and I think he's right to do that. After all, what's the point in the new approval process if the client will go back on their word at the last minute?
Overtime with a large orchestra can easily run into four figures — it's not hard to burn £100,000 in a few days with a full orchestra at somewhere like Abbey Road or AIR.
In the past, the orchestrator's job would typically involve taking a composer's sketch and expanding it, 'changing' parts of the composer's work or adding new melodies, counter melodies and melodic fragments; there was plenty of scope for imagination and creativity. But now, the details of the instrumentation and arrangement in the composer's mockup will already have been signed off before the orchestrator is involved, which means there's much less scope to embellish things. So, essentially, the main focus of our job is to 'realise' the MIDI performance as a score that real musicians can play.
That sounds simple, but MIDI mockups are generally not hugely idiomatic when displayed in notation form. For one thing, the sessions comprise so much more than notes: in particular, there's a lot of extra MIDI information used to switch articulations and inject believable expression into the sampled instruments. You can't just switch to your DAW's score view and hit 'print'! It's our duty to interpret these sessions, and generate a score that's as faithful as possible to the sound of the demo, rather than to the data used to create it.
So... why are we still called 'orchestrators' and not, say, 'MIDI transcribers'? Well, our job may have changed, but the skills required to do it remain the same. A deep and powerful knowledge of orchestration is essential when laying the MIDI parts down in notation form in scoring software such as Sibelius, Finale or Dorico. We still need to understand how instruments work, and know what is and isn't physically possible for an instrument or musician. We must spot any hiccups before they hit the stand at the session.
We also need to be able to work quickly: the whole project's production schedule and workflow have a huge bearing on how much time an orchestrator has to complete their job. The time from a composer first seeing the picture to the very last recording session ranges from a mere few weeks to a few months, so the speed at which orchestrators and copyists are expected to work can be genuinely challenging.
Arguably, it's more important today than ever for a professional composer writing to picture to have an interesting musical voice, and to capture the mood and vibe of a scene, than for them to be particularly diligent in terms of orchestration and harmony. That's not to say that composers don't have a good handle on orchestration — many are excellent orchestrators in their own right — but simply that their primary focus must be the picture they're writing against. Unfortunately, the speed at which they work and the sampled instruments themselves can sometimes mean losing sight of real-world constraints, by which I mean, what musicians playing real instruments are capable of. The note ranges of virtual instruments may be appropriate to real-world norms (they're sampled, after all), but they can still be played loud and fast for as long as your mood dictates. A composer can easily be too wrapped up in the pressure of delivering great-sounding music on time and within budget to consider whether flautists can actually play for 12 bars at 120bpm in 4/4 without taking a breath. Indeed, I frequently find that composers allow insufficient time for the breaths of wind and brass players. This is not a criticism by any means, just an observation and something to be aware of if you're new to orchestration.
Some composers use breath controllers to avoid this problem — they finger the notes on the keyboard and literally blow MIDI data into their mockups. But a simpler tactic I use when assessing mockups is to expel air as I play the part back; this costs nothing, and helps me to determine, broadly, if the line will actually be playable. It's important to remember, though, that playing louder requires more air to be blown through the instrument, particularly on lower-pitched instruments such as the tuba; when faced with a series of repetitive notes in the winds or brass, it's often wise to leave quite a few rests in the lowest instrument, rather than have them play on exactly the same beats as the others. (Having instruments like the tuba play only on the strong beats can also provide natural accents.)
Another trap that composers who've grown up with sampled instruments sometimes fall into is writing melodies in a range that doesn't project particularly well on a specific instrument, whilst scoring a dense texture around that part. This can be made to sound great in a MIDI mockup, but won't work particularly well with real instruments. For example, if a composer has written a melody on a flute around the middle of the treble clef stave, with a roaring French horn counter melody right in its sweet spot, and with high strings at a loud dynamic, there's no way you'll ever hear that melody on the flute in real life; the flute doesn't project well in the lower notes of its range and cannot be played particularly loudly.
A good orchestrator will flag up such issues to the composer long before the project reaches the recording studio. In this particular scenario, I'd probably mention to the composer that they won't hear the flute and ask if it would be OK to try moving the melody up an octave — or perhaps to rewrite it so that it falls above the treble clef stave, where the flute begins to sing and project clearly. Alternatively, depending on the piece in question, I might suggest that we instead overdub that melody as a separate entity in isolation, to provide full control over the flute's volume at the mixing stage. Another option would be to replace that instrument with a different one that will cut through better in the same register. In short, it's the job of the orchestrator to suggest solutions that will make the part work in the context of the cue.
Note, though, that we're talking about options to be discussed: newbie orchestrators must understand that they do not have permission to start changing instruments which are playing the melody without first consulting the composer. I'll come back to this, but for now remember, again, that at the orchestration phase the cues you're working on have more than likely been 'signed off'. For that reason alone most composers are normally reluctant to allow radical changes.
Precisely how you, an orchestrator, should work with the composer varies from job to job and person to person, but the worst thing you can do is fail to communicate, and it's hugely important that you establish their expectations early on. As a starting point, you need to know how faithful you should be to their mockup. I've mentioned the rule of thumb that the sound of the demo, not its MIDI data, is king, but you'll still need to know whether it's to be a slavish transcription of what you hear, if they'd prefer you to exercise some judgement when it comes to things like doubling parts, or if they'd like you try to 'fill out' the score where appropriate, to make it sound 'larger' than the demo.
Here's an example of making things sound larger: if I can hear in their demo that the composer wants a very grunty, deep, violent low end, and can see that the basses are playing but the celli are doing nothing, it's an obvious decision to me to double the celli and basses in octaves. But if the part is less thunderous and has been written with the celli doubling the basses, perhaps it would work better without the celli — context is everything, and you must trust your judgement.
You must also consider the size of the orchestra that will be used for the recording session. If a composer has had to work quickly, they might have opted not to double lines that would normally be doubled — perhaps they have the first violins playing on their own rather than playing in unison with the seconds. For a smaller orchestra, it's standard to double in this way to help everything sound larger, whereas on bigger sessions, with a full symphonic orchestra, you might want the first violins playing on their own, and then to bring in the second violins gradually, for more of a textural effect.
Second-guessing a composer's intentions is risky, though, so it's really important to get clarity at the outset. Naturally, on occasion, you'll need to clarify something with the composer mid-project, and it's best to communicate and get this sorted quickly so that everyone's on the same page. But there's a big difference between asking pertinent questions and haranguing a composer with constant emails and phone calls when they're up to their eyeballs in notes and rewrites — so don't make it an hourly habit! If they, like me, use Sibelius, I'll often write text notes about less urgent matters in the software, so they can see what I've changed in each cue without the distraction of a million emails going back and forth. The bottom line is that the more information you get from a composer at the start, the easier, more efficient and more satisfying you'll find your job.
I'm a stickler for ironing out anything at the start of the job which will save me time and money, and to that end I highly recommend setting up a template score — a template file for your scoring software that you'll always use to start orchestrating each cue from scratch. This will be your rock. Your lifeline. You'll set up everything, from fonts to your instrumentation, and it will make your job easier, more cost-effective and more time-efficient.
It's amazing how quickly parts can be tidied via a template, and if the parts always look semi-tidy before they've even been sent off to be tidied, you'll put a smile on the copyist's face. The copyist could be dealing with over 50 cue notation files, and since you've already spent 15 minutes sorting out things like multi rests, bar numbers under every bar and so forth, they won't have to waste time on it. Importantly, it also reduces the chances of human error creeping into the project: adding bar numbers under every bar, and making stave sizes thicker in 40 different instrument parts for each cue becomes mind-numbingly dull, and can easily lead a copyist to introduce mistakes that you could have prevented.
With this in mind, it really is worth taking the time to get everything perfect in your template file. You'll appear 10 times more professional, and have the pleasure of contributing to the whole team around you being happy and motivated — which is just nice!
At some point, you'll encounter a composer's DAW session that's as messy as a hoarder's garage. It'll be your job to make sense of it — and this ability to explore a composer's DAW session, and figure out what will and what won't be necessary in a live recording session, is one of the most crucial skills in modern-day orchestration. In practical terms, it's simply a matter of listening to the demo and working out how you're going to achieve that sound. But that can be much harder than it sounds; it really is an art in itself, and one that can take several years to master.
Because there's such a strong emphasis on the mockup, composers inevitably write for the sample instruments they have available to them at the time, and this doesn't always allow for the kind of writing that translates well to paper. I've touched on some examples above, but it can get a whole lot less practical than that.
For example, a composer may find that the string samples they're using just don't sound dense and thick enough. So, to achieve the sound they want, they double the arco string ensemble patch with a flautando string ensemble patch, a harmonics string ensemble patch and a con sordino string ensemble patch. Potentially, each patch could come from a different sample library. So the mockup contains four enormous string sections, played in several different acoustic spaces, in several different octaves, by hundreds of different session musicians. It may well yield a really warm and sweet sound, but it's impossible to achieve with a real orchestra. Even with a limitless budget, good luck finding a studio that can host four large strings ensembles as well as the rest of an orchestra (even Abbey Road Studio 1 would struggle).
As an orchestrator, you'll have licence to delete notes where necessary, as long as it leads to a sound that's the same as — or preferably better than — the demo. At this level, nobody really cares about the detail of what you did or didn't do, as long as you deliver what's asked and everything runs smoothly. To achieve the desired sound, you'll need to consider things like player ranges, whether instruments will 'cut through' when played in a particular range, how idiomatic the part is, and how the parts balance with each other. For example, in a string chord, you generally won't want too many thirds or fifths — otherwise, the chord will sound dreary, dull and unbalanced.
There are, of course, many different ways to voice and balance a chord, but Diagram 1 shows a few ways in which you can voice a simple G major chord. The specific voicings shown may or may not be appropriate to the style in which you're orchestrating; these are just examples to show you how a balanced chord looks on paper. In the 'bad' examples, doubling the third and fifth of the chord is overkill; there are more thirds and fifths than root notes. You might consider using the voicings at bars 4 and 5 when there's a full symphonic orchestra at your disposal, but a chamber string orchestra would lose its power with too much divisi. If you're going to divide the section in the low end, I'd recommend dividing the celli — they have a lot of depth and weight, so you don't seem to lose too much power when dividing them. There are many more possible voicings, though; you must use your initiative and balance the chord in whatever way best suits the style of the cue. Sometimes, I've even seen a fairly uninventive unison chord work just fine as the very final chord.
So far, I've written about chord voicing as a way to solve problems, but if you've established a long-standing working relationship with a composer, they might give you more flexibility, and encourage you to use them more creatively. (Others might not; it's their work, so you must respect their preference.) By revoicing a chord you can create an incredible, more sonorous effect which pleases the listener in a way that they might not consciously notice. The effect is very subtle, but you know good voicings when you hear them. Whenever I hear any of Thomas Newman's string writing, I'm instantly hooked — and a lot of that has to do with the way the parts are voiced.
Let's consider an example of how an orchestrator given such licence might look to use chord voicings. The first picture (left) in Diagram 2 shows a very simple string ensemble chord patch, moving from F major to A minor in Logic Pro's piano–roll view. The second (middle) shows what this would look like on a score if you dissected the chord in its simplest form, with the first violins taking the top note, the basses taking the bottom note, and the second violins, violas and celli playing the notes in between.
Now, I know from experience that some of London's finest session musicians can make this simple voicing sound good. But a composer working at speed and focusing on the bigger musical picture may well have played the chord like that despite having a slightly different sound in mind. So, imagine you've received the latest cue but then remember, a few days ago, the composer mentioned that "in the cue 1:20, the music needs to be quite dark sounding, so feel free to mess around with the voicings"? This is where you really earn your money.
On the third (right-hand) image in Diagram 2, while the notes remain the same the violins now aren't playing at all — instead, the viola and celli take the weight of the chord. (The basses sound an octave lower than written so, despite appearances, they are below the celli). The close voicing lends itself well to a darker, more mellow sound, but what's really making this chord sound darker is the divisi of the viola and celli. You inevitably lose a little power as the section splits in half (unless instructed otherwise), but unless there was a particularly small section to start with, I've personally never noticed a big difference when the section has divided.
If we compare this last version with the original voicing (the way it was played into Logic Pro), you'll see that the viola, celli and basses are in a much better register than the violins. They're playing on their lowest string (yes, the G string... no jokes please!), which sounds rather rich and dark. It fits the sound we're after, but it doesn't have much 'carrying power'. That's not a problem at these dynamics, but note that at a louder dynamic the low end of the chord would cut through more than the high end.