Shop Talk

Ben moderated the session and Kevin runs the Discord community that hosted it. Their introduction sets up what the room is for: writing string parts that sound natural, balancing and doubling across registers, blending articulations, and getting a convincing sound out of sample libraries in a DAW.

My own introduction is short. I’ve been orchestrating and writing orchestral music for a little over ten years, and writing music far longer than that. What I wanted to share is how I write for strings, and how to push them toward the realistic side.

Before I write a note, there are three things I’m thinking about, and together they get you 90% of the way to a good string sound.

Writing. The writing itself has to be good. Nothing downstream rescues it.

Dynamics. Simulating the performance - how hard the players are leaning their bows into the strings.

Balance. How the sections sit against each other in volume: first violins, seconds, violas, cellos.

The workshop is mostly about the first two, because the third largely solves itself. I use one library for every section - Cinematic Studio Strings - and inside a single library the sections come pre-balanced, so there’s nothing for me to fix.

None of this is a rule. They’re three tools out of many, and they happen to be the three most of my writing comes through.

Open voicing is refusing to play the chord tight. Take a C major in close position and move the middle note up an octave - same chord, open. Same with an F: lift the A up, and up again. You can go further and take the low C up an octave too. It’s the difference between a chord that sits and a chord with air in it.

Inversions mean putting a chord tone other than the root in the bass. A C chord over E, or over G. They work especially well with strings - with any section of the orchestra, really.

Suspensions are the third: instead of landing on the chord tone, hang the note next to it and let it resolve.

All three get demonstrated at the piano rather than in the strings, and that’s deliberate. The strings only sound right if the writing underneath them is right, so the foundation gets built at the keyboard first.

A simple piano sketch, written with those three concepts in it, orchestrated live into a string section.

The sketch came out in one shot - melody and chords together. People ask whether I start with the melody or the chords, and the honest answer is both at once, most of the time. From there I read the sketch top down: the top line is the violins, then violas, then cellos, then basses, and I record each section in place.

No click for this kind of writing. When I’m sketching something lyrical I don’t record to a click, and often not even to a time signature. I want the tempo to breathe - up a little, down a little - because it sounds more natural. Tight rhythmic material is the exception: percussion, ostinatos, anything in fast succession gets a click.

On EQ, which came up as a question here: what you’re hearing is EQ’d, but the out-of-the-box CSS sound is fine. I cut a lot of the mids and left some lows and highs. How real it sounds has far less to do with EQ than with the orchestration and the dynamics. EQ and effects are the last 10%. I spent weeks, months, years tweaking before I worked out that the 10% wasn’t where the time belonged.

The thing to watch in this section is the CC1 curve, not the notes.

I record dynamics in waves - up and down, up and down - never flat. A flat curve sounds lifeless. On a single line the difference is subtle; once every section is moving that way, it’s large.

The basses are never recorded alone. I play them in unison with the cellos, basses at one octave and cellos an octave above. That’s what gives the bass its body.

The sketch stays basic on purpose. I could go back and write in every embellishment, and there’d be nothing wrong with that, but I want room to improvise while I’m performing each section. The second violins here got a counter melody that isn’t in the sketch at all.

The sections don’t peak together. Compare the violins to the violas, and the seconds to the firsts - the curves crest at slightly different moments. Whichever line I want heard most gets the higher dynamics. It’s rare in a real orchestra for every section to play identical dynamics at the same instant unless someone designed it that way.

There’s no significance to starting with the violins, by the way. Working from a sketch, I could start anywhere, and sometimes I start with the bassline.

A second sketch, same exercise, with more embellishment invented during the take - a counter melody in the cellos that wasn’t in the sketch, and the basses copied from the cello line.

Latency is normal in CSS, and I don’t quantize it away. If a note lands badly I move it by hand.

On headroom: most of my writing starts at lower dynamics than it ends up at. If I know something is going to soar, I’ll start higher, but generally I leave myself room and raise later - or lower, just as often. What I record on the pass is almost never final. As sections stack up, the balance between them keeps shifting and I go back and re-tweak. It’s a constant push and pull.

On where CSS lives: I favor the lower dynamics, and even on epic material I’m almost never in the top 20% of the mod wheel. Under 80%, nearly always. The loudest sample layers tend to feel rough; leaning softer reads as more realistic.

A line gets buried once the track fills up - how do you make it stand out? Two choices, and only one of them is good. I can push the dynamics, but if loudness is the only reason I’m pushing them, it sounds odd - like the violins are trying too hard. So instead I double the line: flute first, then French horn, then a divisi viola section, until it carries.

Do you reach for CC11 to go louder in epic writing? Almost never in CSS. In Cinematic Studio Strings the loudness is baked into the modulation - as CC1 climbs, the volume ramps with it. Other libraries don’t ramp as much, which is why people end up on CC11. I’ll tweak it plus or minus three to five and no further, because the moment I use volume to fix a balance problem I’ve broken the balance, and balance is one of the three things holding the realism up.

Why are the basses so big? They’re never alone. I double the basses with the cellos in octaves, essentially always - the bassline by itself doesn’t sound like that.

How was the cello divisi done? Right-click, duplicate the track. Nothing more.

The CSS staccato patch holds four different lengths - staccatissimo, spiccato, staccato and sforzando - and the mod wheel moves between them. I use all four, especially on exposed lines.

Played on one articulation, a short-note passage sounds mechanical, because no real player makes every note the same length. So I move the mod wheel as I play and let the length change underneath my hands. On an exposed melody I’m quite intricate about it; the cellos, violas and basses in a supporting role get much less, maybe a little at the end of a phrase.

Doubling shorts with other instruments works, but it’s context-sensitive - it depends on how fast the line is. A quick violin figure might take flutes happily and French horns not at all.

The same progression played twice - identical chords both times, inversions the only difference. All of it is one sustain patch, not two; the second track exists only to play the piano.

What made the second pass sound so much fuller was almost nothing: I took one note down an octave, then another. That’s it. Sometimes the really simple stuff is what makes the difference, which is the whole argument for caring about voicings.

Then suspensions, which is where I have the least discipline - I throw them all over the place, and that’s genuinely how I like to write. A question, then an answer.

For long notes CSS gives you sustains, tremolos, harmonics, trills and marcato, and using the full range is one of the cheapest ways to make a passage sound richer.

This example is deliberately over-orchestrated, just to hear everything at once:

• first violins and violas on straight legato
• first violins divisi on harmonics - barely perceptible, but there
• second violins on trills
• violas divisi on tremolo
• cellos on tremolo
• basses on pizzicato at the start, legato by the end

Stacked up it’s more than I’d normally write, but it’s the right kind of clutter: the trills and tremolos add suspense and flavor that straight legato doesn’t have, and the whole thing takes on the liveliness of a real section.

Adding a flute an octave above the violin line, and hearing the difference it makes, is the start of the broader point: from here on, the answer to “how do I make the strings better” keeps turning out to involve instruments other than strings.

Nothing here is quantized. Because I play free-form, the lines never meet perfectly - look at the flutes against the violins and the difference is sometimes stark. That’s by design. A real orchestra isn’t quantized either. Go back to any orchestral recording that isn’t heavily produced and you’ll hear the violins a fraction late into a legato where the cellos weren’t, or the horns entering a hair behind. If something is off badly enough to bother me I move it by hand, and that’s the only correction it gets.

EQ for CSS. The chain I built years ago is more complicated than it needs to be, and I wouldn’t do it again. Starting over, I’d cut the low mids slightly - somewhere around 330, up toward 500 or 600 - by maybe a decibel and a half, and stop there. A decibel and a half is nothing on one instrument, but across five instruments it’s additive, and it keeps mud from accumulating across the score.

Mud is usually a writing problem, not an EQ problem. Before reaching for the EQ, look at the orchestration: how many instruments are down in the low register, and how close together are their notes? Up high, notes can sit tightly against each other and still be clear. The lower you go, the wider you need to space them, or it turns into a mess. Fix that first; if a little muddiness survives, then cut the frequency.

Mic mixes. An anecdote that answers this better than a principle: I thought I’d dialed in the perfect mic blend for the string patches, then wrote an entire album over four months. Mixed, mastered, released, on Spotify. Opening the project afterwards, I found I’d knocked the close mics up five or six decibels on every string section at some point and never noticed. It matters, but not as much as you’d think.

When to toss in a sus chord. I don’t have a rule for this one, and I couldn’t invent one on the spot. What I have is experiments, repeated. A few days before the workshop I sat at the keyboard for an hour and a half doing nothing but looking for chords I liked. I used to think that wasn’t productive; it’s how the intuition gets built.

He didn’t invent it and plenty of people used it before him, but this is the stack I’d call the John Williams sound:

• first violins and violas in unison
• second violins an octave up
• cellos an octave down

In the demo I’m getting there with MIDI modifiers, transposing in real time - seconds +12, cellos -12 - so one performance feeds all four.

The useful part is mixing and matching rather than always using all four. If the first violins lack power or definition, add the seconds an octave up. First violins and violas in unison alone is already fuller than either, partly because the violins sit on the left of the stage and the violas on the right, so you get width out of it. Firsts and seconds, or firsts, seconds and violas - all valid. I rarely use all four at once, because the cellos can crowd a register that’s already busy.

One of my favorite sounds: basses with the cellos an octave above, and the cellos con sordino. Muted, it turns smoother and darker - low and seething, more menacing, an under-the-hood sound. Without the mute the same line is plainer.

This example started with no sketch at all, just a bassline I liked. Everything else was added in a deliberately supportive role, nothing competing for attention: violins on harmonics, almost below the threshold of hearing, violas on tremolo. Then a little woodwind and a glockenspiel on top, which is the point of the example - making strings sound realistic by putting other instruments around them, even when those instruments aren’t doubling anything.

Everything in the workshop is Cinematic Studio - Strings, Brass and Woodwinds. Percussion is Spitfire, an old library that still does what I need.

A different way to sketch, learned from Chopin’s Prelude in E minor: start on a chord you know, then move individual lines - sometimes together, sometimes at staggered moments - and don’t worry about what the resulting chords are called. Past the first one, I usually don’t know. It takes a lot of experimentation before the ear starts telling you which moves work.

This example uses tight voicing, which contradicts what I said at the top about open voicing - and that’s the point. Open voicing is a tool, not a rule. Here I wanted the lines close, then opened the sound up a different way: splitting the three lines across sections, doubling the top line with first violins divisi an octave above, doubling the main line with violas, putting a lower line in violas divisi and the bottom line in the cellos.

Keep the ensemble patch underneath. After splitting lines into sections I leave the ensemble patch playing, and it adds a lot of body. I used to think ensemble patches were pointless; I changed my mind a long time ago. A/B it and the version without sounds thinner.

On divisi and volume: no, I don’t lower the volume to compensate. The one exception is a divisi where the upper half sits an octave higher - that I’ll pull down slightly, so it doesn’t get too prominent. Strictly realistic? Probably not, but it’s a modern balance rather than a classical one, and it’s the only time I touch volume.

The related worry - “a four-horn patch playing three notes means twelve horn players, so I should use solo patches” - may hold in a real orchestra, but with samples it’s the last thing worth your attention. Same for divisi: keep the volume, keep the balance, move on.

Everything so far, applied to one piece: sketched on the ensemble patch, split into sections, then built out with the rest of the orchestra.

How the brass goes in. I look at what the strings are doing and fit the brass into the spaces they leave. It doubles on some notes, but it also plays the notes the strings don’t, and that’s what makes the sound full.

The mistake I used to make was trying to do too much with the strings. Something felt thin, so I’d add more strings. What I eventually learned is to write the strings so they leave space, and let other instruments fill it. The woodwinds do a lot of doubling here, and also fill gaps. If someone asks how to make strings sound good, and everything else has been done properly, the answer is the instruments you put around them.

Where this one started. Not a sketch - a melody, then chords found to fit under it. I start with strings maybe 90% of the time and both hands at once most of the time, but there are no rules. The only useful advice is to try all of it and keep what suits your personality.

On closing up an open voicing by filling the middle with other instruments: it depends on register. High enough and the textural difference keeps it clear; low enough and it turns into the same garbled mess as any tight low voicing.

On the template: it’s my normal one, balanced mostly by being one library family. I trim a flute half a decibel here or there, and it’s insignificant.

On the click: the short-note and ostinato material is played to a click, always. Quick notes in succession make a metronome of themselves, so any drift is audible immediately. Slow legato writing tolerates far more looseness without sounding sloppy.

Written from a single line, with the rest of the parts already in my head: violins on the melody, cellos on a counter melody, violas on a counter-counter melody.

My default woodwind pairings for doubling: flutes on the first violins, bassoons on the cellos, clarinets on the violas. Not a hard rule, but it’s where I start, and it gives each line definition.

Unison, then divergence. The woodwinds start locked to the strings and drift apart toward the end of the phrase. Tight at the opening, fuller as it goes. On anything long I’ll write that divergence in deliberately rather than double all the way through.

Doubling for presence. Where the melody wasn’t cutting through, nothing about the volume changed: seconds doubled the firsts an octave up, then violas joined the firsts - the John Williams stack, stopped short of the cellos - and finally the woodwinds took the melody line too: flutes, oboes and clarinets, everything but the bassoons, which were busy elsewhere.

Copy-paste or play it in? Case by case. String-section doubling like that was a straight copy-paste. Woodwinds I play in, because their dynamics behave differently from the strings. If I want to be more intricate, I’ll re-record the second violins too.

How high are the dynamics there? Under 50%. The guesses in the room were around 99%, which is the whole point - CSS sounds like that well below the top of the wheel, and pushing it higher would be far too much. The brass in that passage sits around 41%. I go by ear rather than by the curve: getting balance right took training my ear, mostly by listening to a lot of music with one question held in mind - how does the brass sound in the context of the strings, and the strings in the context of the brass?

Brass curves here follow the strings, in and out, in waves. For epic writing they invert: start high and come down, more marcato, for the harder attack.

When I started, I pictured short notes as something you play with power. They don’t have to be: very quiet shorts sound great - so quiet you can barely hear them - and they can come up a little from there when the music needs it.

The shorts in this example are CSS doubled with Performance Samples’ Vista Chamber Strings Light Ensemble, which had just come out, though CSS carries it on its own too.

The staccatissimo patch also does things I didn’t expect at first - quick-succession figures that sound genuinely playable. Most of what’s here came from messing around and keeping what worked.

Three released tracks, each played in full and then with the strings soloed.

Short strings under a synth pulse, from my album Survivor. They sit low in the mix and aren’t prominent, but soloed they’re the groove of the track. The relationship with the pulse is indirect: the pulse is locked to the beat, the strings are locked to the beat, and they play against it in different ways rather than mirroring each other. That wasn’t planned - it’s intuition.

String runs, which I’m still learning to write well. CSS handles them unusually well: choose the marcato patch, turn legato on, and the runs become playable.

A full track traced back to its sketch. The same process as the small examples, at full length - sketch, extract each line into its own section, embellish, then build the rest of the orchestra on top. Everything in it is sustain legato, which is worth sitting with: no articulation gymnastics, just writing and dynamics.

On libraries: CSS is what I’d recommend if you can get it. Ark 5, where you can buy individual patches, is a decent budget alternative with playable runs of its own. And there’s a real workflow argument for staying inside one family - the balance is already there, and the woodwinds respond like the brass, which responds like the strings, so learning one teaches you the others.

Severely underrated. My music comes and goes in waves, and part of that is simply letting notes end rather than connecting everything. A line stops, there’s a gap, and it doesn’t read as a mistake - it’s a singer taking a breath.

It isn’t only the melody. I give every instrument room to come in and go out, and that’s where a lot of the dynamic range in a track actually comes from. A continuous, unbroken series of notes gets tiring to listen to.

Do you record CC afterwards or play it in? Both hands at once - notes and modulation together - and then I fuss with the curves afterwards until they’re exactly right. Live in the workshop I took the first pass and moved on, but on my own I almost never keep it. The mod wheel gets moved with my pinky, even under big chords.

How do you sync free-time playing back to a click later? The last track demonstrates it: the opening is played free, the click starts partway in, and the seam is inaudible. When I play freely I’m not swinging wildly - I’m orbiting a tempo I hear in my head, a little above and a little below. So there’s an average, and when I need a click I find that number (80 BPM here) after the fact rather than deciding it up front. Playing free also buys you things a click won’t: the strings swell before the beat arrives and slow right down there, then pick back up, all by feel. You could draw that into a tempo map instead. I’d rather play it.

How do you decide a song’s tempo? I just play, and whatever it is, it is. If you don’t play a controller: hum the idea, then use your DAW’s tap-tempo button - ten or fifteen seconds of tapping gives you the number.

Is practicing from downloaded MIDI files a good idea? If it works for you, yes. There’s no one-size-fits-all here - a friend of mine works in a way I don’t resonate with at all and it serves him well. Examine a file, try to replicate it, see what it teaches you. Having tried something already puts you ahead.

Reverb. One plugin: AIR Studios Reverb Essentials. Everything goes to it through sends, all at the same amount - strings -6, brass -6, woodwinds -6. That’s the whole mixing section. I’ve watched a lot of tutorials walk through compression, saturation, resonance control and EQ chains; it’s all good, and it’s all the last 10%. The 90% is writing, dynamics and balance.

Mastering. No engineer. I export out of the DAW and master outside it, and barely do anything - what you hear is close to straight out of the DAW. On some tracks I use LANDR, which is AI mastering. A good human will beat it every time, but my own mastering is weak and I’d rather have the AI’s version than mine. If I could hire someone, I would.

Lessons. Yes, I teach privately - Discord DMs are open. And the broader lesson: once you have a tool set, learn it, and stop buying plugins.

Kevin closed the room on behalf of the community - thanks for the preparation, and for the snippets built to demonstrate each technique, with an open invitation to come back and do it again.

There was also a request for a brass workshop. We’ll see.