Ok, I'm off to the US for a couple of weeks but I thought I'd leave with a quick blog entry on mocking up string sections.
Again before I start, I don't pretend to be an expert on this subject but it is something that I've spent time working on and I'd like to trade ideas with the people on this forum.
As everyone probably knows, mocking up string sections can be very difficult and the results can be somewhat disappointing. They manifest a number of problems from sounding 'synthy' to having a 'sucking' or backwards sound. Often they can sound too perfect or just a little lifeless when compared to a real string section.
To begin with, I will list out the string samples I'm currently using: -
Sonic Implants Symphonic Strings
VSL Appassionata Strings
VSL Special Edition (Chamber Strings, Orchestral Strings, Solo Strings)
Dan Dean Solo Strings Advanced
VR Sound Kontakt Strings
East West Quantum Leap Gold XP Pro
Garritan Stradivarii 2.0
I'd also like to mention Atmosphere's Hybrid String sounds which are extremely usable and end up in my mixes surprisingly often.
These include a mixture of pre-panned strings with room sound and dry strings taking up a full stereo spread.
Let's say we're aiming for some kind of orchestral string section anywhere between 30 and 60 string players.
Firstly I would ensure that all libraries are panned and sitting in a similar space. For me, this means collapsing stereo samples using Waves S1 Stereo Panner and panning them to the normal orchestral seating. I've not experimented with the 2nd violins to the right although I can imagine that it works very well for film music (evening up the high frequencies and moving the low frequencies closer to the middle).
Once my dry samples are correctly panned then I use Altiverb to move them backwards into the virtual room. I've been using the Todd AO samples recently and I like the sound. I use separate Altiverb ER instances for left, right and centre strings. I also ensure that these are insert effects and that all direct sound is coloured by the Altiverb room. This is particularly effective in moving samples back into the room, especially with very bright and 'up front' samples like VSL.
Ok, so let's say we have a fairly good mixture of samples now and they all sound though they are in the same space more or less.
Let's pick the actual samples to use. I use a variety of criteria here. Are the samples ever going to be exposed clearly? Are they going to be playing lyrical lines or holding occasional pedal notes? This is very useful for reducing the amount of work your DAW has to do. If your double basses are only used to emphasise a couple of notes then there's no point layering 3 top quality legato samples on them. A lower quality sample will suffice. On the other hand, the 1st violin lines might require 3 or 4 high quality samples to really bring them to life.
For an average section I will always use two sets of samples as a minimum. This allows me to stagger note changes and endings, adding some slop and looseness to the sound. It also gives me a natural ability to thin out the sound for divisi sections.
I usually load at least a staccato and a sustain version of every section. For most of the samples I use K3 and create instrument banks, obviously the Vienna Instruments players allows me to create an instrument matrix of all articulations required.
Staccato strings are generally quite easy to create so I'll deal with them first. It's important to have some kind of round-robin in there (like the Dan Dean cloaking device) to avoid a 'machine gun' effect. Also ensure that velocities reflect the rhythm being played so that prominent notes are slightly louder. Sometimes it works well to quantise this kind of idea, sometimes it's better to leave it sloppy. Don't forget that some samples come in slightly late and might need to be played ahead of the beat to keep in time. Other than that, staccato strings should work well and just need to be balanced well (and watch out for too much reverb). Pizzicato is similar in requirements.
Legato and sustained string passages are really quite difficult to create. Very few stand up to any kind of scrutiny next to real strings. Still, I urge all digital orchestrators to take some solace. We're not trying to sound like real strings, we're trying to sound like a recording of real strings and believe me, I've heard some orchestral recordings that sounded very synthy so don't forget that we're overly critical of our mockups.
Firstly, the better orchestrated your strings are, the more realistic they will sound. Now I really don't pretend to be a great string orchestrator but I know a few things. Try and split the lines down to the separate sections and make sure that the lines balance so they're in the correct range for their part. Often strings are arranged in open voicings of fifths and sixths. The cellos and basses are usually arranged in octaves, basses or tenths. Spend some time learning how to write for strings and make sure that the lines make sense horizontally. Learning counterpoint or 4-part harmony is good practice for this kind of writing.
Secondly, be very careful about applying EQ to your string samples. This is a very quick way to make your strings sound synthy (and is probably what happens to those poorly recorded strings I mentioned previously). I've been told that if I have to EQ strings then use a Finite Impulse Response EQ such as PSP Neon and only cut frequencies. Having said that I found that the Voxengo GlissEQ can add a great sheen to strings if used to boost the 'air' frequencies (10-15KHz). Unfortunately it is only available for PC so I can't use it on my Mac...
So for a smooth homophonic part or melody I would probably use a legato/portamento sample plus a sustained sample and maybe an espressivo sample as well. They will all be crossfade samples controllable via Expression or Modwheel. Varying the ratio of the layers can help add interest on long sustained notes. This is often where a quiet layer of Atmosphere hybrid strings sometimes works as the sound tends to evolve naturally.
For extra elements to the mix, I might consider adding in solo strings or chamber strings, particularly in smaller ensembles. These can often give more of an upfront or close-miked sound. Dan Dean has some great small groupings in his Solo Strings Advanced and VR Sound strings can also work well in these situations.
For huge ensembles I often add a quiet layer of East West to give that massive sound. I'll often just use an ensemble patch and a cellos patch to really add the shimmering top end and bass.
I also sometimes add in a quiet layer of either tremelo or muted strings to add more interest. For passages of extreme intensity I will sometimes double the main melody note quietly an octave higher. In this way it sounds as if the players are really digging in and the first harmonic is coming out stronger.
Once the part is really mocked up I make a huge effort to go through the piece and listen out for 'aural cues' that sound unrealistic and then try and minimise them. These can be unnatural attacks, machine gun sounds, unrealistic timing or whatever. Generally this involves carefully drawing velocity curves in the Expression/Modwheel layer. Also I'll sometimes make the pitching less than accurate with various techniques (pitchwheel, Kontakt scripting, etc).
If the strings still have a couple of noticeable problems I can't get around then I'll use the last resort of most digital orchestrators: hide it behind something else, usually percussion or piano. Not the best thing to do but it certainly works!
Well I'm away from noon tomorrow but hopefully this will generate more discussion and I'd be really happy to pick up advice from anyone else who reads this.