Case Study 2: The Uncanny Valley of the Orchestral Mockup

The composer in this story is invented; the techniques are the documented, standard craft of working media composers, and the robotics concept that frames it all is real and citable. Read this even if you never plan to touch a string patch — the lesson generalizes to every "realistic" virtual instrument you'll ever program.

In 1970, the roboticist Masahiro Mori published a short essay with an idea that escaped robotics and colonized the entire culture: the uncanny valley. His observation: as a robot becomes more humanlike, our affinity for it rises — up to a point. Then, just short of convincing, affinity doesn't merely level off. It plunges. A clearly mechanical robot is charming; a nearly-human one is creepy. The failures stop reading as machine limitations and start reading as something wrong with a person. Almost-human is worse than not-human-at-all, because the brain switches judges — from "how impressive, for a machine" to "what's wrong with this human?" — and the second judge is merciless.

Programmed orchestras live and die in exactly this valley, and string patches are its most famous residents. A deliberately synthetic pad — the kind on a thousand synthwave records — sits happily on the near slope: nobody expects it to breathe, so nobody misses the breath. A real recorded section sits on the far side. And in between lies the territory where a beautiful, expensive sample library plays your part and listeners get that polite, slightly embarrassed look and say "the strings sound a bit… fake?" They're not wrong, and this case study is about exactly what they're hearing — and the craft working composers use to climb out of the valley, because climb out they do, daily, professionally: the modern film and TV scoring business runs on mockups, programmed orchestral demos so detailed that directors approve cues — and sometimes whole scores ship to picture — before a single player is hired.

Lena's Cue

A composition student I'll call Lena sends me ninety seconds of score for a student documentary — a wide-shot nature cue, strings carrying everything. Her harmony is lovely. Her library cost more than her interface. And the cue sounds like a church organ cosplaying as an orchestra. She knows it; she just can't name why, and the not-naming is the problem, because every fix in this business starts with the diagnosis (you'll meet that doctrine at full scale in Chapter 30).

So we open the MIDI and do what Case Study 1 did to the piano: read what's actually there. Seven findings — and I want you to notice that every one is a chapter concept wearing a tuxedo.

1. The dynamics are frozen. Every note enters at its velocity-chosen loudness and stays there until release. But a sustained tone in the physical world is never still — bow pressure swells and eases, breath leans and withdraws; a held note is a thing a musician does continuously, not a switch they flip. Lena's CC1 lane is empty. Her strings aren't playing notes; they're exhibiting them. This is finding number one because it's half the entire problem: on sustained instruments, the performance happens during the note, and "during the note" is controller data.

2. Velocity is doing CC's job. Where Lena did want dynamics, she wrote them as velocity — so each note picks a dynamic layer at the moment it starts and is stuck with it. A crescendo written as rising velocities becomes a staircase: each note a fixed loudness, stepping upward. Real sections pour. The library can pour too — its dynamic layers crossfade under CC1 — but nobody's hand is on the faucet.

3. Every note restarts the orchestra. Her lines are sequences of separate sustains: note off, note on, note off, note on. Real string lines connect — the bow carries one note into the next, and the transition itself has a sound (a tiny slur, a bow change, a finger shift). Her library ships a legato patch that contains recorded transitions, triggered by overlapping consecutive notes. Her notes don't overlap. She owns a thousand recorded transitions and has triggered none of them.

4. Twelve players with one body. Chords land sample-simultaneous, machine-perfect. A real section of twelve violinists is twelve human nervous systems agreeing approximately — entrances spread across tens of milliseconds, a soft blur of consensus your ear reads as ensemble. Perfect simultaneity reads as one enormous synthetic instrument. (Notice this is Case Study 1's chord roll, scaled from one hand to a stage.)

5. The parts are unplayable. Lena writes at a keyboard, so her string parts are piano parts: six-note block chords spanning three octaves in one "violin" track, lines that leap like a right hand. Violins play one line (double-stops aside); sections divide; phrases fit inside a bow's length and a lungful of intent. The idiom rule from the chapter — learn what the instrument can't do and stop doing it — is being violated in every bar, and listeners who couldn't name a viola can still smell it.

6. The repeats machine-gun. A repeated-note ostinato in the middle section retriggers the identical sample eight times. No round-robins engaged (her library has them; the patch she chose doesn't), no alternating velocities. Eight identical photographs of a bow stroke, stapled in a row.

7. Everything is equally close and equally loud. Every instrument sits at the same distance, the same prominence, all the time — no phrase rises out of the texture and recedes, no front-to-back stage. Part of this fix lives in mixing (depth and space are Chapter 24's whole subject), but part is programming: real sections take turns mattering, and the CC data is where they negotiate.

Hold the list up to Mori's valley and the diagnosis names itself. Each finding is small. None is "a wrong note." Together they add up to almost-human — close enough to invoke the merciless judge, not close enough to survive him.

What the Professionals Actually Do

Here's the working craft, in the order Lena and I applied it — which is the order of audible payoff:

Ride the lanes, on hardware, in real time. The center of mockup craft is the one this chapter promised: CC1 carries the dynamic story; CC11 breathes inside it. Working mockup artists perform these lanes — mod wheel or a fader under the left hand while the right hand plays, or a dedicated pass per line, recorded like a take. Why performed rather than drawn? Because a hand hesitates, leans, overshoots and corrects — structured deviation, arriving free, exactly as the chapter said about played velocities. Lena's first hour of repair was nothing but this: every sustained line got a performed CC1 pass — the hairpin rule, no exceptions: every held note is going somewhere, swelling in, blooming through, or dying away — with CC11 micro-rides layered after. The cue improved more in that hour than in the previous month.

Trigger the legato you paid for. We switched lines to the legato patches and overlapped each note a few ticks into the next, so the library played its recorded transitions. Lines began to pour. Where she wanted audible position shifts — the expressive slide of a phrase reaching upward — the patch's portamento responded to wider overlaps. (Every library's conventions differ; the principle — connected notes need connected MIDI — is universal.)

Unstack the entrances. Section chords got the ensemble treatment: entrances spread by structured handfuls of milliseconds — lower voices leaning fractionally early, upper voices blooming after, the same direction each time (pattern, not jitter — Clip C's ghost haunts orchestras too). Solo-ish exposed entries kept tighter spreads; big tutti arrivals, wider ones.

Rewrite for bodies. The six-note piano chords became orchestration: violins I on the melodic line, violins II a third below, violas filling the middle, cellos on the harmonic floor — each section one playable line with phrase lengths a bow could survive, marked rests where players would breathe. This step costs the most musicianship and buys the most realism, and there's no plugin for it. (Composers learn it the Chapter 4 way: listening to real sections, scores open.)

Engage the variation machinery. Round-robin patches on the ostinato, velocities alternating in a bow-stroke pattern — up-bow slightly lighter than down-bow. The machine gun became a section digging in.

Layer for focus. A standard professional move: under (or over) the section patch, a quiet solo-instrument layer on the melodic line — a "front desk" the ear can focus on, restoring the slight individuality that sampled sections smooth away. Related: doubling a line an octave up at lower CC1 for shimmer, or blending a second library whose room and character differ, so the composite stops sounding like one sampling session. Dose matters; the layer supports, it doesn't duet.

Stage the space — later. We left distances and depth for the mix (sends, pre-delay, the front-to-back map — Chapter 24 will hand you that toolkit), but the programming groundwork was done: phrases now rose and receded via CC, so the mix would have motion to stage rather than a wall to decorate.

Two evenings of this — no new notes, no new purchases — and Lena's cue crossed the valley. Not to "indistinguishable from the London sessions," but to musical: the documentary's audience would feel weather and wonder, not polite embarrassment. Her professor asked what library she'd upgraded to. You already know the answer, and so does every chapter of this book: same library, different MIDI. The realism was rented through the performance data all along.

The Other Exit from the Valley

One more lesson, because it's this chapter's title hiding inside a case study. Mori's curve has two safe slopes — and so does yours.

Lena climbed out the far side: more realism, earned with craft. But the near slope was always available too: stop pretending. The moment a string sound declares itself synthetic — a sawtooth ensemble pad, an 80s synth-strings patch, anything that isn't auditioning as an orchestra — the merciless judge clocks out and the charming-machine judge returns. Whole genres live happily there; a synthwave "string" pad with zero CC riding isn't lazy, it's idiomatic. The fatal zone is only the middle: realistic samples, machine programming, the almost-human.

So before any orchestral (or "realistic" anything) part, ask the commit question from the chapter, out loud: is this part pretending to be played? If yes — pay the toll: CC lanes, legato, ensemble spreads, idiomatic writing, the full kit above. If no — choose a sound that's honest about it, gridlock with pride, and spend the saved hours on the song. Either slope works. The valley is for parts that never decided.

And if mockups turn out to be your calling — film, games, trailers, the whole scoring economy runs on this craft — the chapter's further-reading list points to the standard texts and the communities where working mockup artists trade exactly these techniques, CC lane by CC lane. It's a deep, employable specialty, and everything in this case study is its first day of school.

And Lena? She kept enrolling. Two more student films, a game-jam score, a trailer pitch — same library, same laptop. When she sent me her latest cue I asked what she'd upgraded, and the answer was nothing: more hours on the lanes, that's all. That's what graduating from the valley sounds like — the conversation stops being about what you bought and starts being about what you played. The merciless judge never left. She just stopped handing him evidence.