arrival

The brief was one sentence:

an epic orchestral score, aliens coming to earth, not a battle but an emotional battle of will, lots of instruments, designed rather than chaotic, fear and hope at once.

What came back was 64 BPM in D minor, eight sections, 59 bars, 62 parts, 28 distinct instruments, and three minutes forty-one seconds. This is what a piece written in the language from the first post actually looks like, and this post takes it apart: the one idea that carries it, the numbers behind its balance, the mixing pass it just got, and what it cannot do.

The idea that carries it#

Two wills, side by side, that never merge.

The human side is a piano theme, with a harp and recorders behind it. It follows the harmony: the same scale degrees, re-rooted on whatever chord is underneath, so it moves when the music moves.

The alien side is handchimes and tubular bells. They do not follow. The chimes play the same three notes and the bells the same two, in the same rhythm, bar after bar, indifferent to the chords going past.

The entire difference in pitch between following the harmony and ignoring it is one boolean:

const theme  = { sound: 'piano',       follow: true,  notes: '0 ~ 2 4@2 ~ 5 4@2 0 ~ 2 4@2 ~ 7 6@2', ... };
const signal = { sound: 'handchimes',  follow: false, notes: '0 ~ ~ 6 ~ ~ 3 ~ ~ 0 ~ ~ 6 3 ~ ~', ... };

Hear it. Same four bars, same chords underneath, one part at a time:

The idea is implemented more than once. The human phrase is stretched over two bars and the alien phrase is one, so the human breathes slower. And in the fifth section, where the key turns to D phrygian, the human theme shrinks to a half-step sigh that bends with the chord, while the chimes sit exactly where they always were.

Since the mixing pass there is a third difference. The piano carries a humanize setting: every note lands up to twelve milliseconds late, by an amount that follows a slow curve over the phrase and a fixed lean per beat, the same way every play. The chimes carry nothing, and land exactly on the grid. The human side is played; the alien side is sequenced.

Where the idea breaks#

There is one section, plea, that modulates to F major. And follow: false turns out to mean something narrower than I thought.

It means “do not track the chords within a section”. It does not mean “do not transpose with the section’s key”. So when the key changed, the immovable signal moved. Its intervals, two stacked perfect fourths everywhere else in the piece, become a fourth and a tritone:

Nothing in the system reported this, because nothing in the system knows the chimes were supposed to be immovable. There is no way to express “these three absolute pitches, always” in the language; there is only “these degrees, in this key”. The composer’s intent and the language’s semantics come apart at exactly the one place the key changes.

I am not going to fix it. It is the most alien-sounding moment in the piece, and it got there by accident.

The walk#

void is one bar: an organ chord and a drone, and that is the whole section. signal is the alien alone, for eight bars, before anything human arrives. approach brings the theme in under a filter that opens across the section. contact is the first climax. will drops to D phrygian with the bass on a sustained didgeridoo. plea modulates to F major and is the one section the human side clearly wins. threshold is the peak, twelve bars, thirteen parts. after is six bars of leaving.

Who is winning#

The brief said neither side wins at the climaxes. That is close to true. The arc between them is the story.

The balance between the two sides across the eight sectionsMeasured this way the human side leads at both climaxes by around three to one, but the arc between them is the story: the alien side plays alone, wins the fifth section outright, loses the sixth by a factor of fifteen, and has the last word in the closing section.

Summed event gain per side, read 2026-09-19

Section	Human	Alien
void	0	0
signal	0	8.4
approach	21.3	8.4
contact	82.5	24.2
will	7.4	22.1
plea	78.5	5.1
threshold	150.5	53.8
after	3.2	7.7

Read per instrument rather than per side, the climaxes really are close: at the first, the piano leads the chimes and bells by about nine percent; at the peak, the piano and its vibraphone doubling lead them by about one percent, and once the piano’s velocity line is counted the alien side edges it. The human side wins on headcount, not on level, because by then it has four voices against two.

But the arc between them is better than “neither wins”. The alien opens alone. The human arrives. Parity. The alien takes the fifth section. The human takes the sixth. Parity again at the peak. And then the chimes get the last word.

Fear low, hope high#

The other half of the brief was a mixing instruction disguised as an image: fear in the low half, hope in the high half. It is in the numbers, as a clean separation between the two groups’ medians.

The separation is not luck, and it is not a hole in the pitch space either: plenty of individual notes span the middle, and the strings alone cover both halves. What is separated is the centre of gravity of each voice, because the register control moves a part by whole octaves rather than smoothly. The language pushed the two groups apart.

The best detail in the piece is the one voice that crosses. The bowed strings start in the fear half, climb into the hope half as the piece opens up, drop back down for the standoff in will, and drop again at the finale. One part moving between the halves section by section is the “both at once” idea living in a single instrument.

The mixing pass#

Everything above was true of the first finished version. What the file gained since is a mix, in the sense the previous post is about: not louder and quieter, but a scene.

The previous post opens with the measurements from before the pass: an impact hit as the loudest thing in the climax, a choir louder than the theme it was meant to sit behind, two parts too quiet to hear, nothing with a position, five separate reverbs.

After it: one hall on the whole song, with each part’s space value reduced to its send into that hall. A position for every part, so the piano is in front and a little right, the chimes and bells together on the right, the harp and recorders on the left, the doubling far left, the choir behind and to the right and darker, so the organ keeps the low mids. The impact set back. The pulse turned up to the top of its range because the timpani sample is quiet. The glockenspiel replaced by a vibraphone, because the glockenspiel sample was too quiet to place at any level. The two spiky bell strikes a bar given a compressor. And the piano and the pulse given the correlated humanize described above in place of the old dice-per-hit jitter.

The measure tool’s verdict went from five warnings to clean, and the mix’s peak fell by more than half, because the impact had been the peak. One number from the pass cuts against intuition: by the score’s own summed per-event gain the triangle is still the loudest voice in the finale, by a factor of two and a half over the harp, because the pulse is at the top of its level range and the triangle sits in the slot that fires most often. By the rendered measurement the whole pulse sits twenty-one decibels under the mix. The number the language exposes and the number the ear reports disagree, and the pass sided with the ear.

What it cannot do#

There is no counterpoint. The pads stack triads on the chord root and every voice moves in parallel, every time. The language now has inversions and a chord-tone follow mode; this piece uses none of them, and voice leading is not modelled anywhere. There are parallel fifths across the whole piece, and the fifth section, where organ and strings and drone and piano all step together, is parallel motion as the entire dramatic device.

The development is one line, transposed, mostly unchanged. One sixteen-step phrase carries the entire human side. There are three further hand-written versions of it: a truncated sigh in the phrygian section, a fragment in the closing bars, and the one that matters, the answer, which changes the second half so it descends and lands on the root instead of hanging. That last one is a single string edit, and it is the whole motivic argument the piece has.

The rhythm cannot really change. Density on a melody is a dice roll: below the midpoint it drops notes at random, above it doubles them. In the fifth section it is set low enough that two of the eight bars have no human theme in them at all. The section named for the human will asserting itself has two bars where the human is simply absent.

There is no tempo flexibility. 64 BPM, exactly, for 221 seconds. Film scores live on rubato. This has none.

What I actually think#

It is a well-voiced, convincingly orchestrated three and a half minutes with a real dramatic shape and one genuinely good idea about two incompatible musics coexisting, built from one melodic cell repeated without development over four chords repeated without variation, at one unwavering tempo. A very good sketch for a cue, played by a very good sample library, by someone who has not yet learned to write a second phrase.

I still think the good idea is good. Two wills, in the same key, over the same chords, and neither one gives way. It is expressed as a boolean, and it is audible to someone who cannot read music.

Here is the whole thing, and the instrument it came out of.

Open the editor on the peakHide the editor

The live editor needs a wider screen than this one. Open it in a new tab ↗