mp3 | vorbis | 44 seconds

You can check out the code after the jump:

TempoClock.default.tempo = 2; // 120 BPM

SynthDef(\audioinfilt, { |freq = 300, sustain = 1|
  var sig;

  sig = MoogFF.ar(SoundIn.ar(0), freq, 2.5) *
    EnvGen.kr(Env.linen(0.01, sustain, 0.01, 0.7), doneAction: 2);

  Out.ar(0, sig ! 2); // sig ! 2 is the same as [sig, sig]
}).send(s);

a = Task({
  var delta = 0.25;

  loop {
    [100, 800, 8000, 600, 3000, 1000, 150, 300].do({ |cutoff|
      Synth(\audioinfilt, [freq: cutoff, sustain: delta * 0.5]);

      delta.wait;
    });
  }
});

a.play;

The SynthDef defines a synth routine that plays a chunk of audio from the audio inputs, run through a filter, while the Task section defines the pattern, looping through an array of values and kicking off a synth for each value.

I tried a few approaches, and did actually get this working in a couple of ways, but none exactly to my liking yet:

• JACK Rack can control plugin parameters using MIDI, so I set it up with a Calf Filter, configured the cutoff for MIDI control, and created a sequence of parameter changes in Qtractor. This was quick to set up, and it kinda worked, but the timing wasn’t stable — I’m not sure if JACK Rack’s MIDI support is meant to be used this way.
• The next idea was to use a modular synth. I tried AlsaModularSynth, but I’m not sure that its step sequencer is generic enough to use for modulation, instead of playing notes — I had it all hooked up but just couldn’t get it working. Ingen looked a bit more promising, but it’s still very alpha, and crashed too often for me to be bothered dealing with.
• Following that line of thought, I fired up Pure Data (“Pd” to its friends) and tried to get to grips with it. Pd is one of the many programmatic audio/music synthesis environments available for Linux, and I’ve been looking for an excuse to check one out; this seemed to fit the bill pretty well. Pd is a bit like a modular synth, but much lower-level — it doesn’t have a step-sequencer, but I was able to build one! I got a working setup, but the resonant bandpass filter I used didn’t sound ideal to me, and I had hassles trying to pull in a better one, either as a Pd add-on or through LADSPA. I could’ve actually built a better filter, but that wasn’t something I was interested in doing today.
• Moving on from Pd, I had another crack with SuperCollider, another synthesis environment. Unlike Pd, which is graphical, SuperCollider is based around writing code, which should mean that it’s more to my liking. I can get sound out of it, and it has some good filters, but the sequencing stuff is gonna take a bit of work to get my brain around.