score
Three movements and six beds, all synthesised into `/source/@user/pad/audio/` as `.soundClip` assets, all looping, all mixed by volume envelopes that are pure functions of film time.
@user::pad.score — the sound
Three movements and six beds, all synthesised into /source/@user/pad/audio/ as
.soundClip assets, all looping, all mixed by volume envelopes that are pure functions of
film time.
Why envelopes and not cues
A cutscene cue does not fire on a seek — a seek re-derives the frame that time would have
produced rather than running the clock through to it. A film that scrubs therefore cannot
put a sound on a cue and expect to hear it. Everything here is instead a looping source
whose gain is f(t). Seek to 62.5 and the tower is coming down in the mix because
collapse(62.5) is 0.94, not because something fired on the way past.
What is playing
Three movements in D flat — one tonal centre, so they crossfade instead of colliding:
| clip | act | material |
|---|---|---|
pad_score_i | cold starlight | high bells over a thin high pad, no bass at all |
pad_score_ii | the nebula front | the same key warmed — mid pad, a low third, mid bells |
pad_score_iii | the rise | a low drone under an arpeggio that climbs out of itself |
Six beds. Sugar snow and the sea's glow are non-spatial, because they are everywhere. The
front, the glacier, the tower and the column are spatial and stand where the thing that
makes them stands — and two of them MOVE: the front's emitter rides film.frontX(t) across
the archipelago, so in shot 5 the sound arrives from the left with the light, and the
column's emitter climbs 420 m with film.column(t).
The ears are film_ears, flown to the camera's eye by film.ears. Nothing here adds a
second listener.
What the mixer says
Read at six film times with audio.observe():
| t | audible | rms | what is up |
|---|---|---|---|
| 4.0 | 2 | 0.051 | movement I + snow |
| 33.0 | 4 | 0.107 | I crossfading out at 0.03, II in at 0.63, the front arriving at 0.43 |
| 48.0 | 4 | 0.113 | II + front + glacier |
| 62.5 | 8 | 0.273 | the collapse at 0.88 — the loudest moment in the film |
| 90.0 | 4 | 0.181 | III + column + sea glow |
| 96.5 | 3 | 0.013 | everything going out under the fade |
How the clips were made
A deterministic additive synthesiser, run once per clip from scratch: a sine table, an LCG for noise, exponential-decay bells, LFO'd drones, a one-pole bandpass for the noise beds, and circular reverb taps. Everything is loop-seamless by construction: every partial frequency and every LFO rate is snapped to an integer number of cycles per loop, the noise beds equal-power crossfade their own tail into their head, and the reverb taps read modulo the buffer length so the tail wraps round. Stereo width on the score is a circular Haas offset, which widens without breaking the loop.
Nothing here was generated by a service, so nothing here can fail to arrive.
Scoped to this part · feeds back into the world's score.