We explain how to create MIDI string arrangements that don’t sound like MIDI string arrangements!
A lot of musicians like using string ensembles in their productions, whether it’s for sweeping orchestral soundtrack cues, anthemic rock/pop choruses, or slippery disco‑funk countermelodies. Of course, most of us don’t have a real string section on speed‑dial; neither do we have a recording setup comprehensive enough to do it justice even if we did! So we ruefully fire up some sample library or virtual instrument and hope for the best — a tactic that rarely ends very well, to be honest. Indeed, the majority of ‘canned’ string‑ensemble parts I hear coming out of project studios only sound at all plausible if you bury them in the mix. The moment you try fading the tracks up as a feature, their manifest fakery just makes your whole record feel cheap.
But it doesn’t have to be this way, because there are ways to get remarkably realistic string‑ensemble sounds when working with MIDI strings, and in this article I’d like to suggest a few tips that can really help in practice.
Divide & Conquer
The first thing to realise is that you can’t just play your strings patch the same way you would a synth patch if you want best results. String ensembles don’t play like that, because each instrument section operates independently, tailoring the phrasing and tone to suit the nature of the line, whereas all the keyboard player’s fingers are being driven by the same musician!
There’s a world of difference between how Middle C sounds on double basses and how it sounds on violins, and part of the reason that canned strings sound a bit rubbish is that they don’t reflect this reality.
So one of the things I always recommend is splitting out each internal MIDI line as a separate track. For a start, this gives you much more flexibility to adjust the timbre of the ensemble sound to suit the musical context. You see, the ranges of the different string instruments overlap a great deal, but if you feed your MIDI data into an all‑in‑one ‘ensemble strings’ patch it’ll give you no choice about which instruments play which notes. There’s a world of difference between how Middle C sounds on double basses and how it sounds on violins, and part of the reason that canned strings sound a bit rubbish is that they don’t reflect this reality. In particular, melodic parts will sound a lot more convincing if they’re not hopping around between different instruments.
And speaking of melodic parts, the other big reason to split each MIDI line onto a separate track is that this makes it possible to take better advantage of the more advanced articulation and phrasing options provided by your choice of virtual instrument. Orchestral libraries now often provide special legato patches, for instance, which render overlapping MIDI notes as slurred phrases, something that can immediately improve the sense of musical realism. Things like accents, staccatos, and swells are also usually much better handled using dedicated articulation‑specific patches than by simple MIDI programming tweaks, and it’s useful to have the flexibility to switch between those articulations on a per‑part basis, rather than just switching the whole ensemble at once.
More advanced string‑ensemble sample libraries offer keyswitches that can select different articulations. Making use of these can help contribute to a more convincing‑sounding string arrangement.
Velocity Versus Volume
That said, there is plenty you can usually do to improve a string ensemble sound with your MIDI programming too. The most important consideration, in my view, is how you shape volume/expression and velocity data in conjunction with each other. It’s vital to realise that volume and expression data typically affect only the level of a string section (much like your DAW’s fader automation), but that velocity data will also change the string section’s timbre — typically by making higher‑velocity notes more strident, as you’d expect if the player were playing with greater intensity.
With this in mind, I usually prefer to adjust velocity data first to achieve a suitable overall timbre and believable musical phrasing, and only then balance the parts against each other (and indeed the rest of the mix) using volume/expression data or DAW fader automation. Mind you, if you’re using budget‑friendly virtual instruments, you may find that they only have two or three different samples available per note to cater for all 127 velocity values, so I do sometimes feel the need to boost/cut the level of some notes with volume data for phrasing purposes in order to avoid unmusically abrupt tonal switches between velocity levels.
Here you can see how one of the parts in a string‑ensemble arrangement has been split out, and then had some of its notes overlapped to take advantage of the more musical note transitions in a virtual instrument’s ‘legato’ mode.
Layering Live Parts
No matter how much of a MIDI ninja you are, or how flashy your strings software is, there is nevertheless still a limit to how realistic programmed strings can sound. Yes, you can put together a pretty convincing chord‑per‑bar string pad with samples these days, but the moment you start including faster runs of notes or more lyrical melodic lines you’ll quickly find things sounding more mechanical. Indeed, one of the simplest tricks to getting decent samples‑only strings is just to avoid those kinds of arrangement moves in the first place! When you can’t, though, it’s time to turn to a stock‑in‑trade trick that media composers have been using for decades: conceal the fakeness of your string patches by layering a single live player over them — a much easier recording job than tracking a whole ensemble! This is something I’ve had great success with on a number of occasions, so I have plenty of advice to pass on here as well.
To start with, it’s important to direct your efforts efficiently. You don’t need to track up live parts for every single line in the arrangement — just focus on those parts that most clearly reveal the shortcomings of your MIDI strings. In practice, the first violin is often a prime candidate for overdubbing, but if that part’s just playing some high held note while important countermelodies are happening in the second violin, viola, or cello parts, then a live first violin will hardly help the sense of realism at all. Similarly, there’s no rule that says you have to record the whole of the arrangement for a given part — if most of the song just has block chords, then just overdub on the section that doesn’t.
Now, if you’re on a budget, you may well be taking advantage of a friend who used to be a string player at school, but who now only picks up the instrument once in a blue moon — there are a lot of those players around, me included! In this case, it’s really important to realise that orchestral string instruments take a while to ‘wake up’ when they’ve not been played for a few months. I can only speculate as to exactly why this is, but I imagine that the wood simply becomes a bit stiff when it’s not been resonating in a long time. And this isn’t me being petty: high‑end string‑instrument dealers actually employ people to play their instruments regularly so that they remain on top form for potential customers. So your first job before overdubbing is just to make sure the musician has actually played the instrument at a decent volume for a good couple of hours, focusing particularly on the lower strings and any specific notes that feel at all dead‑sounding, thin, or reedy. Believe me, if you try to record anything before doing this, your captured sound will definitely suffer.
Mic Technique
For the recording itself, it’s important not to mic the instrument too close, otherwise the close‑miked sound won’t match the more distantly miked sound typical of most string sample libraries. And it’s not just a question of reverb levels, because that’s not actually such a big issue. The main thing is that the balance of frequencies and instrument facets you pick up a foot away from the instrument is very different from what you get four to six feet back; the more distant placement just sounds a lot more like a real instrument. Now, that said, you should certainly take precautions to keep reverb levels as low as you can (such hanging up a few of those ubiquitous acoustic duvets), especially if you’re recording in a small room, simply so that the recording environment’s acoustic character doesn’t unnaturally colour the captured timbre or conflict unduly with the room ambience of the samples you’re using.
What also helps, in my view, is to record each pass with a spaced pair of mics. Not only does this introduce some image width to the sound (which generally seems to enhance the naturalness of acoustic instruments, for me), but it also stereo‑decorrelates any room ambience you do pick up, so that your overdub sounds less like it was recorded in a corridor! Personally, I like to keep the mic spacing quite close, around nine inches, so that there’s no risk of mono‑incompatibility, and I’m not concerned with trying to generate any kind of ‘realistic’ stereo image — just a little width and tonal complexity from the two distinct mic positions.
Out of preference, I prefer using omni mics in this application if I can, again because I find that they usually sound more believable and neutral than cardioids. However, I can get away with that because my own home studio is set up as a mix room, so it’s acoustically quite dry. If your chosen recording location is more ambient, then directional mics may be the only option. That said, do experiment with figure‑eight polar patterns if you have them, as they’ll likely pick up the off‑axis ambience with less tonal coloration than the cardioids. And, speaking of figure‑eight mics, ribbon designs naturally have that polar pattern and are also a great choice for strings, especially with any violin that has a tendency to shrillness — ie. any violin that costs less than your car!
People get very worked up about mic positioning when close‑miking string instruments, and with good reason, given how unpredictable the frequency dispersion of wooden‑bodied instruments is at close range. However, if you’re miking six feet away, the mic positioning tends to be a lot less critical in my experience, and for violin or viola, you can usually set the mics up somewhere in front of the player at chest height and it’ll work just fine right away. The same kind of miking position can also deliver good results with cellos, but sometimes moving the mic closer to the floor can help fill out the tone if you want more low‑end warmth.
When layering live string overdubs over a MIDI arrangement, recording in stereo from a distance of at least four feet usually helps the recordings fit better with whatever string‑ensemble samples you’re using.
Layering, Editing & Mixing
A crucial question when layering up overdubs of a live string player is how many layers to record per line. Personally, I almost always go for three. If you only do one layer, that single live part will probably feel a bit stranded over the ensemble sound. With two layers, any tuning or timing vagaries between the two parts will sound more obvious, and if you pitch‑correct the layers to avoid this, you may then get an artificial phasing‑like interaction between them instead. Three layers per line acceptably mitigates all these issues for me. But why not layer things up even more? Well, in my experience, further layers never seem to improve things enough to be worth the extra effort.
Again, if you’re working with a string player of the ‘weekend warrior’ variety, you’ll almost certainly want to edit the overdubs to match their takes with the more finessed sound of your samples. Besides, it’s much more important to concentrate on the phrasing, tone, and musicality of the live takes, as those aspects of the performance aren’t nearly as easily fixable as a bit of dodgy rhythm or pitching. I’m particularly careful to keep my ears peeled to avoid mechanical squeaks from the bow or discontinuities in the timbre of sustained notes, and I also do my best to remind the string player to take advantage of idiomatic performance techniques wherever possible — things like vibrato, bow‑slurring, left‑hand position shifts, and volume swells.
Once the overdubs are recorded and edited, the lion’s share of the job’s done, and the mixing process is usually pretty straightforward. Clearly, if there are any obvious tonal disparities between the overdubs and the samples, then some EQ may be called for, and I sometimes find that gently low‑pass filtering the overdubs at, say, 15kHz can help sit them a step further back in the mix if they’re feeling too dislocated from the rest of the virtual ensemble. (That said, you often need to brighten string ensembles overall within a pop/rock mix context to combat high‑frequency masking from things like cymbals, percussion and electric guitars.)
Compression can come in handy, as string instruments can be surprisingly dynamic when recorded comparatively close‑up. Even so, you’ll almost always need to supplement that with some fader automation at mixdown to refine the balance within the context of the full production.
Beyond that, a healthy dose of artificial room ambience added to the live overdubs will often help blend them better with the samples, especially if your recording room wasn’t very reflective sonically. After all, in most real‑world situations where you’d hear a string ensemble, a good deal of what reaches your ears will be room reflections rather than direct sound. Some hall reverb applied to the entire string ensemble usually provides a nice final touch, further gluing the layers together into a smooth and coherent sound.
Even Better Than The Real Thing
The suggestions I’ve presented in this article aren’t rocket science (and are a heck of a lot cheaper and easier than trying to record a real string orchestra), but I can tell you from real‑world experience that they really do work in practice! To demonstrate this, I’ve uploaded some audio examples to the SOS website at https://sosm.ag/realistic-strings that illustrate the kind of improvements that can be achieved. Plus, if you head over to www.cambridge-mt.com/sosstrings you can download a supporting DAW project where you can experiment yourself with the balance of the samples, live overdubs, reverbs, and backing stems in my demonstration arrangement.
How Low Should You Go?
Within most chart‑oriented commercial releases, there’s rarely much room to spare at the low end once the kick and bass are in the mix. Furthermore, heavy masking from guitars, keyboards and vocals can often make it seem as though the only part of a string ensemble you’re actually hearing is the highest line. As such, it’s very tempting to shortcut the job of string arranging by using nothing but violins. I’d advise against that, though, because a big part of the characteristic appeal of strings is the textural complexity contributed by those lower instruments, and that’s present even when the listener can’t consciously pick out their specific musical lines.
By all means high‑pass filter your string parts at mixdown to avoid low‑frequency clutter, but keep those violas and cellos in the arrangement nonetheless. Otherwise, your strings will likely sound thin and uninspiring.