What is layering?
If you have spent some time working with orchestral libraries, you have probably noticed that sometimes sections don’t feel quite big or lush enough, especially in those cases when you’re going for that massive Hollywood-style sound. Or maybe they lack a particular timbre or character that you’re looking for, even though they’re pretty good in all other ways that matter (sound quality, articulations etc).
One way of dealing with this problem is simply shopping around until you find a sample library that sounds just like what you’re after. And if money’s not an object, this is certainly the most fun option. We all love new musical toys, right? On the other hand, good orchestral libraries are expensive and for most people this approach is simply far too cost-prohibitive.
Thankfully there’s a different solution to this problem that, with a little work, might yield just as good results as finding that “perfect” sample library (if there is such a thing), and that is layering. By simply doubling sections with other similar samples — e.g. 1st violins from library X combined with 1st violins from library Y — it might be possible to add that missing or extra dimension to your current samples.
Subtlety is the key to successful layering. The most important thing to keep in mind is that the layer samples should be sensed rather the outright heard. Think of it as adding a complementary nuance to the sound rather than transforming it into something completely different. If the foreground samples are the ingredients of your musical dish, the layer samples are the spices that help bring out the flavor. Like adding a bit of attack to a staccato articulation that is a little too wimpy, or some additional size to a section that seems to have too few players.
With this in mind, it will come as no surprise that layer samples should in most cases have very modest volume levels relative to the foreground samples. Best approach is loading a layer sample that you think will go nicely with the foreground one, and bring it’s level down to zero. Then, while playing the sound in question, slowly raise the level until you start to just barely hear it.
So how do you know when the levels are good? With a recorded part playing back, mute the layer samples. If you can’t hear any difference, the levels are too low. If the difference is extreme, the levels are too high. You need to find the sweet spot where the layer blends enough with the foreground to add its character to the sound, but not so much that it actually sounds like it’s a double.
Choosing layer samples
The catch with the whole layering thing is that you will need a fair amount of different stuff too choose from. If you have just, say, EWQLSO Silver and Miroslav Philharmonik and nothing else, there’s no guarantee that the two libraries will go well together in all thinkable scenarios (especially since they’re two libraries with distinctively different characters).
On the upside though, layer samples don’t have to be as complex and detailed as the foreground samples. They’re there to add a subtle touch of something, they’re not meant to be in focus and provide the meat of the section sound. That’s the foreground samples’ job. So basically, for layering you can get away with using stuff that you would never dream of choosing for your foreground samples. Old freeware soundfonts, synth patches, and so on.
There’s also a few cheap orchestral samples that you can pick up for peanuts and use for layering, such as the DSF soundfonts, Zero-G Strings and Brass & Woodwinds. Perhaps not great on their own but blended with other stuff, they can work wonders.
This might be a no-brainer but it’s worth mentioning anyway: you will need to use the same, or at least very similar, articulations for both foreground and layer. Doubling fast staccatos with a slow sustain sound doesn’t make any sense, for example. If the foreground has vibrato, the layer should preferably have vibrato too, and vice versa.
Layer samples doesn’t have to have velocity layers or round robins though. The layer will be masked by the foreground anyway, so if the foreground has round robins and the levels are set correctly, you will not be able to tell that the layer is repeating the same sample over and over. As for velocity layers, you can (if necessary) set up velocity-triggered low pass filters in your sampler, creating the illusion of dynamics in the layer samples.
When to layer?
So, should you just habitually layer each and every part of your virtual orchestra? Nope, just where you find that it adds something that’s lacking. Not all orchestral instrument families take kindly to layering and if you just stack things on top of each other willy-nilly you will end up with something that resembles a giant “symphonic” synth sound rather than a believable orchestra. Let’s have a quick look at the different instrument families:
Strings are easy to layer as they already have a very diffused sound and adding even more of the same will just make them bigger and warmer. You can take a lot of liberties with strings as even tuning problems, as long as they’re fairly small, only adds to the lushness. Only thing to watch out for is frequency buildup (rumble in the low register, harshness in the mid and high registers). Be prepared to use your sampler’s filters or an EQ to tweak the layer where necessary.
Brass is more difficult. These are smaller sections where even minor tuning discrepancies will be immediately apparent, and will turn your blockbuster sound into an under-rehearsed high school band. Brass instruments also have a wider range of dynamic timbres than the other instrument families. Combining pp strings and ff strings can work, but pp brass and ff brass sound nothing alike and layering the two will give you a strange, ambiguous sound.
Woodwinds? I wouldn’t layer them at all, unless we’re talking about layering solo samples to create an ensemble sound. Just like the smeary, diffused sound is a large part of a string section’s character, the lack of diffusion is an equally large part of the ensemble woodwind sound. Too many woodwind instruments sandwiched together will sound more like an accordion than a wind section, so tread carefully.
Percussion — it depends. Big, lower-pitched percussion (like bass drums, timpani, taikos and toms) can usually be layered without problem if additional fatness or a percussion ensemble sound is desired. High or non-pitched percussion (cymbals, glockenspiel, triangle etc) are better left alone, as layering will lead to weird, chorusy artifacts. Simple rule of thumb: is the instrument in question meant to sound huge? If yes, layer away and see where it gets you. If no, don’t bother.
How to layer?
Here’s where things get more hairy. Depending on your DAW of choice, its routing capabilities, your understanding thereof, and your preferred way of composing, there’s a number of different ways you can go about this.
The most basic way is simply adding another instance of your sampler, loaded with the layer samples you need, and doubling each part manually. Record a part with a sound from sampler A. Then re-record the same part with the corresponding sound from sampler B. This is mindlessly redundant and time-consuming, as you will need to record everything TWICE. And it leads to project clutter. But if you’re reluctant to learn more efficient ways of doing it, or if you’re after the (possibly beneficial) slight timing and velocity differences this method will yield, you can definitely do it that way.
Second option is the same as the first one, only that you record a part with sampler A then copy and paste it to the corresponding track on sampler B. But if you’re going to do this, you might as well bring out your DAW manual and figure out how to use…
Third option — which I prefer, for the most part — is sending one midi channel to multiple sources. For example, 1st violins sustain vibrato goes to both sampler A and B — or different channels on the same sampler, depending on how you have everything set up. Alternatively, you can also set up two sampler tracks to receive on the same midi channel, one for foreground and one for layer. Which way is easier depends on whether you’re starting on a new orchestral project or if you’re going to add layers to an existing one.
As for outputs, you can either route the layers to the same output as the foreground, or keep them on their own separate channels. Having them routed to the same outputs ensures that panning and reverb levels are matched for both foreground and layer. On the other hand, sometimes you will want to have finer control over the layer samples (EQ, lower reverb level if the samples have a lot of recorded ambience), in which case routing them to a completely separate output is the better option.
And yes, it’s perfectly fine to send all layers of an instrument group to a single stereo output and setting the panning within the sampler. Having the same reverb level on (e.g) all layer strings isn’t that big of a deal as you won’t be able to here minute differences in distance in the layers anyway. The foreground will take care of that.
As you can see, sample layering is not an exact science and it takes some patience, experimentation and a set of decent ears to get good results. But when you get it right — when you hear that so-so section suddenly come to life — it’s totally worth the trouble.