No matter what technology you use, from high-resolution samplers to FM synthesis, an essential part of orchestral simulation lies in giving passionate performances, and that requires understanding the orchestra and using subtle gestures. Here, I'll describe how to use MIDI continuous controllers to effect those subtle gestures. Unless you chant MIDI parameters in your sleep, you might want to read the sidebar “Take Control” for a quick refresher on MIDI continuous controllers.
FIG. 1: Expression Controller is used here to create a natural-sounding dynamic contour for a bowed string instrument.
Violin samples or patches are usually programmed to start with a relatively hard attack, a drop in loudness, a sustain with a little vibrato, and an eventual decay as simulated by the MIDI Expression Controller (CC 11) curve in Fig. 1. But, violinists have many different ways to play a note — they might hit the strings hard, reverse bow direction, and slowly decay, or they might strike softly and slowly swell in a long crescendo that slurs together several notes. Your synth or sampler may not have those articulations, but you can also use Expression to simulate those.
In the opening notes of the score for Alfred Hitchcock's “Psycho,” composer Bernard Herrmann calls for the strings to play aggressively and swiftly. To create that effect, start the note with maximum Expression, and then quickly move to a much lower level followed by a slight swell and a final decay. That effect is best applied to a marcato patch or sample, which has a heavy attack. If your instrument responds to Aftertouch or Note Off Velocity, you can further differentiate bow strokes using those messages. Alternatively, you can use the Expression Controller to soften the attack of a string by starting at a low level, then ramping up slowly. That produces a soft bowing effect. Applied tastefully, CC 11 will help your music flow, hiding the seams in your sequences.
BOWED BUT NOT BROKEN
A violinist will often slur together several notes with a single bow stroke. If your synth or sampler supports MIDI Legato Footswitch (CC 68) or has some other form of legato control, use that to simulate slurred notes. The trick is to ensure that the envelopes don't retrigger for each note. For MIDI instruments that have poor-sounding legato, try the following workaround.
If your instrument allows you to edit the attack portion of its amplitude envelopes, you can create a string patch that has almost no attack. First copy your normal violin sound to a new memory location, then edit its amplitude envelope to deemphasize the attack. You may also want to experiment with reducing the filter-envelope amount. When you want to play legato notes, use the new patch on a separate MIDI channel, overlapping the notes slightly to ensure continuity.
ONE FROM COLUMN A
Of course, a violinist can't go on bowing in one direction forever. Eventually, he or she must either lift and replace the bow or change direction. For long smooth passages, the player can softly reverse the bow to avoid a sharp transition, but that still entails a subtle shift in timbre. When the violinist pulls the bow toward the body, it's called up-bowing, whereas pushing away is called a down-bowing. Accented notes are generally played with down-bowing.
To simulate real bowing, place two distinct violin sounds — one with a harder attack — on separate MIDI channels. After recording a violin part, drag your down-bowed notes to the channel with the harder attack, and put your up-bowed notes on the other channel (see Web Clip 1).
YOUR OWN VIBE
To create vibrato (an undulating change of pitch), string players move one finger back and forth while holding the string against the fingerboard. Vibrato can be performed rhythmically or offbeat, fast or slow. Synthesized vibratos that fluctuate at a fixed rate sound phony and should be avoided, or at least mixed with custom vibratos. The problem with using sampled vibratos is that they are usually provided at only one speed and depth. You can add organic dimension to string and woodwind parts by performing your own vibratos.
Vibratos start soon after an attack, often speeding up and growing shallower as they progress. To emulate that, start with a string patch or sample that has little or no vibrato, and wiggle your finger back and forth on the Pitch Bend controller while sustaining a note. If your Pitch Bend controller is too sensitive, just focus on the rhythm and scale down the Pitch Bend values afterward. In Digital Performer, I use the Change Continuous Data command to limit the amount of bend, but you can also set the Pitch Bend Sensitivity to a comfortable range in advance. Web Clip 2 illustrates an accelerating vibrato performed individually by several members of the string family.
FIG.2: Bowed tremolos are played by bowing rapidly on a single string. Here, a cellist starts by bowing in triplets, bowing irregularly by the end. Note that down-bows are generally shorter than up-bows.
Bowed tremolos are played by rapidly bowing down and up on a single note. Like vibratos, they can be played rhythmically or not. But the technique is tiring, causing players to inevitably slow down or bow sloppily if the tremolo must be sustained. Bear in mind that a tremolo is not the same as a trill, which uses only one pitch and is much faster. As with vibrato, tremolos are often included in sample libraries, but they are usually recorded at only one rhythm and tempo. By performing your own, you can, for example, construct bowed tremolos in triplet 16ths that will automatically match the passage's tempo.
To avoid a monotonous machine-gun sound, use separate up- and down-bow patches. Fig. 2 shows a tremolo constructed using two MIDI channels — one with a marcato cello patch for down-bows, the other with a softer cello patch for up-bows. Both the initial attack and final release notes are doubled (played on both channels) to produce a heavily accented sound. (When stacking notes, keep the note Velocity on one of the channels low to avoid its sounding like two cellos.) During the tremolo, emphasize downbeats by using higher Velocities, and keep note durations short, with a little overlap between down-up pairs. You can color the tone by adding tiny amounts of Pitch Bend, as well as by shortening and lengthening accented and weak notes, respectively (see Web Clip 3).
Winds of Inspiration
Woodwinds have a pronounced attack, rising sustain, and controlled release, which are physically produced by tongue movement, airflow, and the cessation of air. To sound convincing, woodwinds need even more dynamic contour than strings do. Use Expression to heighten or deemphasize attacks, create continuous swells, and sculpt natural releases. You can also add Pitch Bend during the attack to simulate a slow surge of air. To do that, start the note about a quartertone low, then quickly release the Pitch Bend controller after you strike the key.
During sustain, airflow is regulated by the player's lungs, and pitch can be modulated to produce vibrato in three ways: with the diaphragm, with jaw movement, and by changing the orientation of the mouthpiece to the lips. The third type, called mechanical vibrato, is easy to produce but considered inferior on most woodwinds. Flutists are more likely to use diaphragmatic vibrato, whereas sax players use jaw vibrato.
A woodwind soloist can play legato, nonlegato, or staccato, which you can emulate by varying note durations and Velocities as well as using MIDI legato (CC 68) if available (see Web Clip 4). Legato phrases contain smoothly connected notes with only one clear attack at the start of the phrase. Nonlegato phrases give each note a soft attack, controlled with the tongue; the flow of air is lightly interrupted for each attack. Staccato notes have strong attacks and are clearly detached from each other.
In your MIDI sequence, legato notes should overlap slightly for continuity and to mask the attacks. Whether you use CC 68 or another approach to legato, the result should be a silky, continuous sound with no apparent note transitions.
Nonlegato notes should come close to joining, but should be separate. After the first note, notes in a nonlegato phrase should have relatively low Velocities, gradual releases, and only small Expression peaks to mimic soft-tongued attacks. Often, it's most convincing to play connected notes in legato mode, while using Expression to sculpt the soft attacks and quick releases.
Staccato notes require higher Velocities and Ex-pression Controller values. Use short note durations with rapidly falling Expression curves to effect a sharp release.
Due to their pronounced attacks, chiffy flute patches often sound weird in legato and nonlegato phrases. For realistic flute solos, use chiff samples to start a musical phrase, but switch to flute sounds with little attack to slur successive notes. You should vary playing with legato, nonlegato, and staccato techniques as melody demands, but be mindful of how much air each approach requires.
When playing legato or nonlegato, there should be a quick interruption or musical break every 8 to 16 seconds; otherwise, the listener will start to feel asphyxiated. Performing with a wind or breath controller will force you to take breaths naturally. To simulate a breath, create a drop in Expression, release any note you're sustaining, and begin the next note with a definite attack. Releasing a note slightly early is the best way to suggest a breath. If a new musical phrases begins immediately, play its first note a touch late, too.
FIG. 3: The Pitch Bend curve in blue, below the left note, is mirrored by the Expression Controller automation in purple. On the right note, the Expression Controller curve has been further altered to simulate an attack followed by a slow crescendo.
Woodwind vibrato is similar to string vibrato in that it involves Pitch Bend, but it also entails noticeable changes in dynamics as airflow is restricted and resumed. You can create a dynamic curve that matches Pitch Bend in most sequencers by first creating Pitch Bend data, then copying and transforming that data to Expression. To achieve the result in Fig. 3, I used Digital Performer's Reassign Continuous Data (copy) command to duplicate Pitch Bend as CC 11. Because Pitch Bend messages are centered around zero, I used DP's Change Continuous Data command to add an offset to the Expression curve. Finally, I used DP's Reshape tool to emphasize the attack, draw a slight crescendo, and finish with a diminuendo.
Glissandos are achieved by altering the embouchure and air column while playing a chromatic scale or by rotating or moving the instrument in or out of the mouth for a smoothly bending portamento. There are limits, however, as to how far a natural pitch bend can go. For example, the opening clarinet solo for Gershwin's “Rhapsody in Blue” starts with a trill, followed by a seemingly wide pitch bend. In reality, few clarinet players can bend that far. They typically play a chromatic scale that ends with a glide, giving the impression of a full bend. Long clarinet glissandos should be performed the same way — start with discrete notes and end with about a four-semitone pitch bend.
Rip It Up
With brass, you can use legato, nonlegato, and staccato, as described for the woodwinds, but vibrato is rare. A more characteristic technique on trumpet is the rip, played by rapidly varying lip tension and air pressure to produce natural overtones at the release of a note. Trumpet overtones follow the fixed series shown in Fig. 4. Most trumpet players can play only the first eight overtones, and rips spanning four overtones are most common. If you use the seventh overtone in a rip, remember that it is a little flat, so use Pitch Bend to emulate that.
FIG. 4: The natural overtone series of a Bb trumpet shown here is used in playing rips. Black note heads indicate tones that don''t coincide perfectly with chromatic steps and sound flat.
In addition to rips that accompany the release of a note, you can run up the overtone series before striking a note. You can also fall off (run down the overtone series) when the tongue is relaxed after playing a high note. All rips should be played quickly and quietly — like a single grace note — overlapping each tone slightly (see Web Clip 5).
As an example of a brass instrument's broadly shifting timbre, you can add punch (or “brassed tone”) to any loud note by doubling it with a low-Velocity note an octave higher. The listener will perceive the added note as upper harmonics of the fundamental, rather than as a distinct note. That technique is best used briefly during the attack of a note or at the tail end of a crescendo.
Shake It Up
Shakes are produced by moving the lips, tongue, jaw, or brass instrument to shift between two different partials while sustaining a note. Fig. 5 shows a trumpet shake between pitches D and F. You should hear an initial attack, followed by rapid, smoothly alternating pitches. You create that effect by starting two pitches simultaneously on separate MIDI channels and crossfading between them using the Expression Controller. The two pitches alternately grow louder and softer, shown as a mirrored pattern of purple Expression curves in Fig. 5.
FIG. 5: Trumpet shakes require two MIDI channels that have symmetrically crossfading notes. Adding Pitch Bends extends realism.
The upper partial in a shake should be significantly quieter, approximately one-third the Velocity of the fundamental. Always start with the fundamental note at maximum Expression and with the upper partial at zero. If the melody continues afterward, end with the fundamental note at maximum Expression.
Because there's a lot of random motion during a shake, small pitch bends also occur (shown as blue lines in Fig. 5). The amount of bend is discretionary, but I generally keep it within a third of a semitone in either direction. To produce a more organic feel, don't synchronize the Pitch Bend and Expression curves (see Web Clip 6).
Any brass or woodwind instrument can be muted, though the average synth provides only muted trumpet sounds, and those are usually staccato. You can create a more flexible trumpet mute by rendering your trumpet part as an audio track, then applying an automated filter or EQ effect. That way you can mute notes of any length and at any speed, try effects such as a muted rip, and simulate different styles of mutes.
Straight mutes allow frequencies above 1.8 kHz to pass, whereas a cup mute acts as a bandpass filter whose center frequency lies somewhere between 800 Hz and 1.2 kHz. The Harmon mute passes frequencies between 1.5 and 2 kHz (see Web Clip 7). Mute action should always be enhanced by a simultaneous boost of around 1.5 to 3 kHz. To avoid sounding like a filter sweep, thin your EQ automation to just a few nodes — no more than one message every 100 ticks or so.
Like string players, trombonists find pitches by feel, adjusting the slide and noting its relationship to the bell. The tuning is usually not equal tempered, and the pitches are generally less stable than other instruments. Bends occur when it's necessary to move the slide in a legato or a nonlegato phrase. Use small amounts of Pitch Bend to simulate those transitions and to keep the trombone sounding loose and natural.
Trombonists often use their tongue to quiet the slurs that accompany slide movement, so except when the slurs are a desired effect, use the Expression Controller to soften slides, as if you were trying to hide them. Knowing when and how far to bend the pitch requires either direct experience or a chart detailing the notes playable on the seven trombone slide positions. When in doubt, keep an orchestration textbook handy.
Colors of Woodwinds and Brass
Woodwinds and brass grow brighter as the volume swells, and darker as the player runs out of breath. You can emulate continuous timbral shifts on many synths by varying CC 74, which is often assigned as a filter-cutoff or brightness controller. Other instruments often provide custom brightness controls.
FIG. 6: Flutter tonguing uses a quiet dissonant partial to create a buzzing effect. Soft notes shifted an octave higher at the attack and release simulate brassy upper harmonics.
Another technique common to brass and woodwinds is flutter tonguing, produced by rolling the tongue as if saying “r-r-r” while playing. Airflow is never shut off completely between tongued accents, and a pretty faithful simulation can be made using sustained notes and Pitch Bend, as shown in Fig. 6. To emulate the rapid alternation, try drawing a triangle-shaped wave with a very short period. Infuse variety by avoiding completely uniform amplitude. It sounds more natural if the rolling slows a bit and the Expression curve drops near the end.
Timbre shifts rapidly during a flutter, producing a distinctive buzzing quality. To simulate that, add a dissonant partial a semitone above the tonic, and have it start 20 to 30 ticks after the fundamental note. To avoid its sounding like a chord, make sure that the added note has no more than a third of the fundamental note's Velocity. You can also add a short, parallel dissonant note, shifted up an octave, a few ticks before the note to brass up the onset of the flutter.
Cymbal crashes and timpani rolls are less about keeping time than generating excitement, so accents should be treated with care. Vary note Velocity more than you might think is necessary in order to clearly emphasize rhythms and allow timing to slip when you want to create suspense. Because percussion instruments can't express themselves through melody, gestures are especially important.
If you have several timpani patches, use darker samples for heavy hits or attacks with harder mallets. You can also try using separate timpani patches to represent two physical drums. If a drum is played heavily during a passage, let a tiny but growing amount of downward Pitch Bend creep in over time, lowering the pitch to simulate natural detuning. Just be sure to fix it before the next movement starts, the way a real percussionist would.
Also use Pitch Bend to vary the pitch depending on where and how the drums are struck. Add Pitch Bend to the strongest beats, but rarely go off more than a quartertone unless you're actually detuning the drumheads while playing. For that effect, apply a lurching Pitch Bend (timpani skins are adjusted with a foot pedal), don't use an interval greater than a major third, and avoid chromatic steps.
Timpani players usually stop a drum with their hand on a final staccato note to avoid its ringing awkwardly after other players have stopped. You can damp the decay of your drum by quickly ramping down the Expression Controller at the end of a piece.
Triangles can resonate at many indefinite pitches depending on how and where on their three sides they're struck. Single strokes are played at the bottom of the triangle, where pitches are relatively low. Tremolos are played by quickly beating the two upper sides of the triangle, where the harmonics can be as high as 20 kHz. In addition to a muted hit, there should be at least three distinct open-hit pitches available for triangle performance. If you have only a single open-hit triangle sound, use Pitch Bend over a narrow range, with one shift per hit.
Be sure to vary Velocity widely with each strike to properly accentuate rhythm as well as to help fool the ear into believing that the triangle is being struck in many different places. The first stroke in a tremolo is often the loudest and can be accentuated by doubling it with a muted triangle note. Tremolo dynamics should swell up and down and sound most convincing if you perform the part with two fingers on separate keys or pads of your MIDI controller. Even if you have only the triangle sound assigned to one key or pad, play the part using two, then move all notes to the correct key or pad. That is much better than performing with two fingers on one key, which cuts off the decay of each note.
Synths are also often underequipped in the gong department. Many have only a single sample or patch dedicated to this sonorously rich instrument. A gong's timbre shifts slowly as it evolves, and the initial impact influences how many overtones are heard. If you have only one gong sound, try applying a gentle lowpass filter sweep to produce some variations. A gong's natural decay carries on for a long time, but for a damped-gong effect, apply a quickly falling Expression curve.
Crash or Slide
The piatti or crash cymbals have indefinite pitch and can be played in more than one way. When a cymbal is struck near its rim the tone is low, and when it's hit up further, the sound starts to bloom. Use more than one cymbal sound to represent different strike zones, and use Pitch Bend to approximate small variations within each strike zone. Layering several cymbal sounds can enhance a climactic crash. Piatti can measure as much as 30 inches in diameter and sound quite low, so you needn't be shy with Pitch Bend.
FIG. 7: The cymbal tremolo (left) was constructed using Pitch Bend, and the cymbal roll (right) was constructed from crossfaded reverse and forward samples.
Piatti can also slide. Sliding involves continuous changes in pitch as undulating cymbals meet at different points. A piatti roll involves continuous circular sliding, which is most easily simulated through digital audio editing. Start by reversing a sample of a sustained cymbal crash. That becomes the first crescendo of the roll. Just before the sample reaches its peak, crossfade to the original cymbal hit without its initial attack. The roll should swell smoothly up and down, with no breaks, as shown in Fig. 7. If you want another iteration, crop the attack and tail from another copy of the crescendo and repeat before the first surge has a chance to decay. You can stop on either a crescendo or a diminuendo or finish with a big crash.
A tremolo begins with a loud crash, followed by a soft Doppler effect, as the performer moves the resounding cymbals toward and away from the audience. Using a long sustaining cymbal sound, wait at least one second after the attack, then start rocking the Pitch Bend control to achieve the Doppler effect.
Build Sections with Soloists
Instrumental section patches with names like “Brass Ensemble” or “Three Flutes” might sound full and detailed, but they lack the feel of live performance because they're homophonic. When eight violas or three trumpets play in unison, the result is the sound of that many individuals following the conductor. Although playing 20 or 30 solo violin parts would be overkill, it's worth at least separately performing the First Chair Violin, Violins I, and Violins II parts on three separate MIDI channels, using different samples or patches that contrast and blend in a realistic way.
In a section built from solo woodwinds or brass, have the players alternate breaths. For example, a lead trombonist might take breaths on the fourth beat, while other players breathe on the second to allow continuity.
Don't copy and paste. Instead, perform each line, allowing slight differences in the dynamics of each musician, the start and stop times of each note, the onset and speed of vibratos, and the tuning of each instrument. At the end of a measure, some musicians will stop ahead of others. To simulate normal variations in string tuning, add very slight pitch bends to one or more tracks.
When playing to a metronome, be sure that the pace varies organically, the way a conductor would lead, with scaled ritardandos and accelerandos. It's more important that parts stay together than that they follow a rigid pulse. Let the parts diverge, but ensure that each section follows the most rhythmic part of the orchestra. You can accomplish that by loosely quantizing the main voices in each section, then letting the other voices contrast to a measured degree.
If you must quantize a passage, I recommend groove quantizing or quantizing by a percentage, because strict quantizing can remove those little hesitations or anticipatory strokes that give life to a performance. And only quantize when you truly need it. Many parts will already be drifting a bit, so it only makes sense for a part to line up numerically if most of the orchestra does.
Real is Better
If you play notes far outside an instrument's physical range, it will sound fake. The same goes for trying to jump from pizzicato to triple stop in less than two seconds. The best way to ensure realism is to empathize with performers. Are they bowing or plucking? Must they stop and breathe? Are they switching mallets?
Leafing through an orchestration handbook will expose the limits of each instrument as well as suggest many exciting ways to perform. And, when you need that whisper-through-a-trumpet effect that just can't be simulated, there are plenty of soloists willing to come to your aid.
Many low-budget projects still use live performers, and as a rule, the more skilled the players you use, the better the overall sound. Even if it only means recording a real tambourine part for an otherwise synthesized scherzo, it's amazing how that one instrument will improve your perception of the whole recording.
Adding subtle gestures to your sequences demands patience, so it pays to store them as clippings or reusable components. You'll also discover the point of diminishing returns. In a tutti section details get buried, so you can use fewer gestures when the orchestra plays in unison. But don't skimp. To remind yourself of the goal, go to concerts and listen to recordings. Instrumental solos and chamber works are often best for revealing a particular instrument's character; you'll notice bent pitches, slow attacks, loud harmonics, and other organic variations. As much as possible, make your music sound like the real thing. The gestures might be subtle, but the improvement will be vast.
Most of the techniques mentioned in this article involve automation using MIDI continuous controllers, which include widely known MIDI messages such as Pitch Bend, Aftertouch, Note On Velocity, and Control Change (CC) messages 7 (Master Volume) and 11 (Expression Controller). I also make use of the lesser-known CC message 68 (Legato Footswitch), but because that is often poorly implemented or not implemented at all, I offer a work-around that, while not perfect, will generally get the job done. Another, also imperfect, alternative is to use the MIDI Channel Mode messages CC 126 (Mono Mode On) and CC 127 (Poly Mode On) — with Mono Mode acting as a kind of forced legato. In short, if it works, use it.
My use of the Expression Controller needs some further explanation. The routing of the Expression Controller varies from instrument to instrument and often from patch to patch within an instrument. I use it as a pre-Master Volume control, which is the most common implementation. But it is often simultaneously routed to affect filter-cutoff frequency and sometimes to also control envelope amount or attack time. Affecting filter-cutoff frequency is probably not a problem in most cases, but affecting the filter or amplitude envelope will distort the uses for which I intend it. The rule of thumb here is to know your synth.
Finally, with today's highly programmable synths and samplers, you can route virtually any MIDI continuous controller to any synth parameter. Some instruments will have consistent assignments for some parameters from patch to patch, whereas others — typically software instruments with patches from a variety of sound designers — will be all over the map. Using what's provided as well as implementing your own automation routings can add a lot to your toolkit of gestures.
David Pedergnana (www.moodpaint.com) creates film scores at Mood Paint, his San Francisco Bay Area company. You can hear samples atwww.moodpaint.com/gestures. Thanks to Rich Busby for his trombone knowledge.