Music Composers Unite!
Creating Realistic Legato Strings With Samples
I get a lot of requests from people asking about how to get good legato strings whether they need to buy LASS, what reverbs they need, etc. In the end I thought it would be a good idea to simply write a blog about the principles behind getting good legato strings and let people try out the the techniques and see for themselves. I'm really focusing on slower legato here as creating fast string runs is an artform in itself! In this article I'm going to look at the problem facing us, the tools available and my various approaches for getting the results I want. This promises to be a big subject so let's dive in and see what we get...
Section 1: The Problem
Samples are wave files of recordings. We wish to create sections of music using legato (Italian for 'tied together') strings. We wish the individual samples to sound joined together and flowing rather than clearly detached wave files played consecutively. Unfortunately this can be quite difficult for various reasons.
Firstly, one of the main attractive features of strings is the soft and natural chorusing effect caused by so many strings players playing just slightly out of tune with each other. Unfortunately we often find that sample libraries are recorded unnaturally in tune. This makes the string sections sound smaller and they have less of the rich and lush sound of live strings.
Secondly, the ensemble nature means that attacks are slightly out of time with each other. A sample of 20 violins will have them all start playing at the same time. 20 live violins will have a far softer attack due to the slightly staggered nature of the players starting the note.
Thirdly, we come to the problem of the actual 'legato' bit itself. By this I mean the sound of playing from one note to another without the sound of a new attack. A real nice to have would be the element of portamento so that we can get a realistic sound of sliding from one note to another.
A final 'nice to have' is to be able to play this all using an ensemble patch. While they are not realistic, ensemble patches are a very fast way of working and often more inspiring than layering one line upon another if you are writing at the keyboard.
There are a lot of further issues which don't help as well. These include the overall envelope of each wave file (a lot have a small crescendo built in). Another factor includes the amount of vibrato used by the string players. Finally the amount of articulations required can be a further factor. I'll touch on these points and maybe I'll try and cover this in future columns.
Section 2: The Tools
The Available String Libraries
There are a lot of string libraries out there and they each have their own strengths and weaknesses. Whether you're reviewing your existing libraries or considering purchasing a new product, here are some of the features to look for: -
Some libraries just have a lovely sound straight away. Part of this is down to included ambience but part of is down to the engineering skills, the microphones, the players and all manner of magic. For example, Symphobia is reknowned for having a beautiful and natural string sound straight out of the box. You also might have a specific string sound in mind and a certain library might be closer to the sound you want.
Some libraries have a nice built-in ambience. This can be an advantage in getting a nicer sound. I'll discuss use of reverb later but already having a nice ambience can reduce the amount of work you need to get a good sound straight out of the box. East West Gold has a lot of ambience, whereas VSL was recorded in their infamous silent stage and has very little. Other libraries like LA Scoring Strings (LASS) and Sonic Implants Symphonic Strings (SISS) fall somewhere in the middle.
I believe VSL were the pioneers here in including a special 'legato' noise between note transitions. They went and recorded the sound of every note combination and this little sound gets included on transitions. They've gone on to record portamento as well. LASS has of course done this and there has been a lot of praise for the realism of the LASS legato sound. Upcoming libraries will need to have this feature and Hollywood Strings is going to have it. A lot of older libraries don't have this feature and there are ways around this. Certainly a lot of reverb can help to hide awkward transitions. There is also a script (called SIPS) which can be added to patches in Kontakt (full version, not Kontakt Player). This can add a legato sound to patches which don't have it and is generally well regarded. Download the script from http://www.nilsliberg.se/ksp/scripts/sips/sips.htm
A collection of different articulations can be important. Although we're really dealing with the legato articulation here, soft strings are definitely enhanced by use of multiple articulations such as crescendo, diminuendo, sus w/ vibrato, sus w/o vibrato, etc. Although not an articulation as such, it's extremely useful to have patches with mutes (con sordino).
Layers and Cross-fading:
If you want dynamic levels in your strings then you need sample libraries that have been recorded at multiple dynamic levels. You also need a way of seamlessly cross-fading between these levels using either Expression (CC11) or Modwheel (CC1). Personally I prefer the modwheel although if I buy an expression pedal I might change my mind there. How well cross-fades work tends to vary from library to library and personally I think LASS does it better than anyone else.
How much flexibility do you have in the section sizes? Do you have one large violins patch or do you have solo strings, small sections and large sections? VSL tends to sell these all separately (solo strings, chamber strings, orchestral strings and appassionata strings) whereas many others include them all (EW and LASS for instance). The string sections of SISS and Symphobia are actually quite small and that may be part of their appeal.
Tuning and Humanisation:
It's possible to add in slight tuning errors with the pitch wheel (or tuning tools in Kontakt). Similarly you can add in 'humanisation' timing errors manually or with scripts. It can be a big advantage to have this done for you in a library with a feature that you can activate and configure yourself.
Reverb, Placement and Space
When aiming for a beautiful orchestral string sound, reverb is usually a significant factor. When working with samples this is even more important. Good use of reverb helps to hide some of the giveaway 'cues' that the strings are samples. The soft release caused by the reverb tends to blur transitions which sounds more realistic.
Within this subject we come to the ever-present issue of placing the instruments in a virtual space. At this point I think the user needs to make a decision as to what sound they want. Are you looking for a huge symphonic sound of many strings or something very close and intimate? Recently I've been working on a full sound with a lot of rich reverb but some of the strings are mixed closer so that I can bring them out if I need to.
I strongly recommend listening to reference recordings of live strings at this point. There are a lot of very different recordings and they can all sound beautiful with the right treatment.
How to use reverb for virtual placement is a huge subject in itself but I highly recommend people learn it. There are a lot of very high quality reverbs available on the market and the choice can be bewildering. At the moment I'll just say that I've had a lot of success using convolution reverbs for room placement and then adding a final wash from an algorithmic reverb.
I won't go into the various reverbs available but I will say that you should have one capable of splitting out Early Reflections from Reverb Tail and hopefully with some very good Impulse Responses of various rooms and at different microphone distances. Personally I tend to use both Altiverb and Vienna Suite Convolution reverb with a Lexicon PCM 90 for my final reverb tail.
Another thing to consider is using EQ. It's commonly accepted that losing the higher frequencies tends to make things sound further away however I tend to prefer a shallow wide dip from 1kHz to 5kHz which seems to help push things back but retain the air. I'd like to give a brief word of caution though. The very act of using EQ on strings can often make them sound synthetic (even with live strings) so be aware of this. I'm told that Linear Phase EQs (such as PSP Neon) tend to sound better on strings but they have significant latency and cannot be used while actually playing.
Lastly, the actual ambience of the library is a big deal here. Many libraries are already recorded in place and so this job is done for you. I've found that a really quick way of achieving a spacious yet detailed sound can be achieved by mixing an ambient library (such as EW) with a close-miked library (like VSL). Simply mix levels to taste. It's not a perfect solution but often can be surprisingly effective.
Section 3: The Solutions
So given that I've covered what we want to achieve, why it's difficult and what we have available to achieve this, let's look at some solutions that I use.
Learn to Write for Strings
I wasn't going to include this but I guess it's important to mention (and people will harangue me if I omit it). If you want realistic string parts then first make sure you can write for strings and aren't just writing a keyboard part with a string patch. I'm the first to put my hand up and say I'm neither conservatory trained nor a player of an orchestral instrument however I have put in hours learning how to write four part harmony and counterpoint and I've also spent a lot of time studying good string writing from the masters. I don't consider myself an expert by any means and for me it's a constant learning experience but at least I know that I am writing realistic 'lines' for the various sections and I pay attention to that.
To me, the most important technique is to use multple layers of strings and often this works best when layering from multiple libraries. The slight tuning discrepancies and differences in the 'envelopes' of each library can help give a more natural sound. Of course this approach has been used extremely well within LASS where each section is represented by 4 different tracks.
The other ways that layering helps is by allowing for slight change in texture during a note where perhaps one layer might come out more than others. It can be worth looking at different "volume curves" in Kontakt so that louder levels actually emphasise different layers to softer levels. You can also use humanisation scripts or play each layer in separately to get timing discrepancies as well.
As I mentioned in the reverb section, layering can also be used to bring out detail with an ambient sound. Combine this with the trick above and you can have detail come out at either lower or higher level as well.
One thing to watch for with layering is just a large build of sound. Sometimes you need to EQ judiciously in the low mid areas. It's often worth layering in a couple of solo strings as well to get a sense of detail.
Given that we are hoping to 'cheat' and use ensemble patches, I often layer in sections over ensembles. That way I can play everything in as an ensemble and get the basic sound. If I need more detail then I enhance the various lines with the individual sections.
I also have found that layering in some element of a synth (such as the Transparent Adagio strings from Omnisphere) can be very effective. The evolving nature of the synthetic sustain patch can help to break up the often unnatural uniformity of string sample sustains.
Well there really isn't a way around this. If you want good legato I recommend you buy a string library with good legato features. If you can't afford that then I recommend you load up your library in Kontakt and use SIPS. With judicious use of layers and reverb that should sound fine.
Of course, right now legato patches are limited to individual sections playing single notes (this will not be the case soon). This means that you will need to break up parts into their sections. I've found that I can often get away with just using this for one or two prominent lines and the general layering and reverb hides the other transitions fine.
The more sophisticated libraries include portamento effects and I think a judicious use of this mixed in with the legato can be an excellent effect.
"Twelve Years Ago"
Here's an example of some very simple writing that includes legato strings. It's not the cleverest writing I've done but it is an example of the kind of thing that gets asked for in tv underscore. It was also quite a fast piece to complete by playing most of the strings parts in one pass using ensemble patches and then breaking it up as separate lines for a couple of sections.
For this example I've used a lot of layers of strings. The ensemble patches are: -
Omnisphere Transparent Adagio Strings Bright
Sonic Implants Symphonic Strings Ensemble Sustain con Sordino
Symphobia Soft Sustain DYNAMIC
Symphobia Solo Strings
LASS Ensemble Sus Con Sordino
I've then layered in the following section patches: -
LASS 1st violins divisi B legato and portamento light
LASS 1st violins First Chair legato and portamento light
LASS Cellos divisi B legato and portamento light
LASS Cellos First Chair legato and portamento light
VSL Chamber Strings II (sordino) violins
VSL Chamber Strings II (sordino) violas
VSL Chamber Strings II (sordino) cellos
I'm using VSL because I wanted a mostly muted sound and (as far as I'm aware) right now they are the only library of muted strings with true legato. LASS has beautiful sordino strings but not with legato at the moment.
Now I'm aware that this is a very quiet piece and the detail might be hard to pick up but it is there. The solo strings come out from time to time and I've aimed for a mix of space and intimacy. I was going for a string sound like James Newton Howard's recordings, particularly Treasure Planet.
For reverb I've used Altiverb for positioning. I've created two separate reverb busses for Strings (FAR) and Strings (NEAR). The LASS section patches go through the NEAR bus and this helps to bring out details.
All tracks are then going through an ambient reverb (again with Altiverb) which adds a slight wash (around 20% mix) of a distant microphone.
Finally everything is going through the Lexicon PCM 90.
I appreciate that this is an enormous subject here and despite the length of this article I've just scraped the surface of it. I also don't pretend to be an expert. I'm just trying stuff out and learning from other people. I'm sure some people reading this will outright disagree with some of what I've written. Again I'm hoping to help people with techniques and advice that has helped me. We're all still learning here.
If you have comments or questions then please go ahead and ask.