add sf2 soundfonts to sound registry

[felixroos]

instead of (or in addition to) .soundfont, soundfont presets should be usable via .s, integrating with .out. When loadSoundfont is called, all presets in the loaded soundfont are added to the available sounds of s. Some type of sound registry has to be implemented to be able to add sounds from outside (webaudio package should not depend on soundfont package). Also remove the old non sf2 file based soundfonts

[felixroos]

idea: add a new function that takes in a callback for creating the source web audio node:

note("c a f e").sourceNode((hap, deadline, duration, cps) => {
  /* webaudio code that uses above params to create a webaudio node * /
  return node;
}).lpf(1000)

the function is similar to onTrigger, but instead of being just a general purpose callback, the function is expected to return a web audio node, which is used by the webaudio output to apply the standard chain. Advantages:

• allows to add custom sound sources from the outside. the existing oscillators, the sampler and soundfonts could all just implement a source.
• the standard chain is preserved, meaning a custom source will still work with all the effects (like lpf in the example)
• it can also be used in the user code to create custom instruments with webaudio code

the same approach could potentially also work for custom effects, e.g. effectNode

[felixroos]

this works really well! it opens up some interesting possibilities:

const lfo = (freq, type = 'sine', range = 1) => {
  const ctx = getAudioContext();
  const osc = ctx.createOscillator();
  osc.type = type;
  osc.frequency.value = freq;
  osc.start();
  const g = new GainNode(ctx, { gain: range/2 });
  osc.connect(g); // -range, range
  return g;
}
const pitchLFO = lfo(.1, 'sine', 20);

note("c eb a*4 c4".add("12,12.05")).legato(.5 )
  .source((t, value, duration) => {
    const { release = 0.01 } = value;
    const ctx = getAudioContext();
    const o = ctx.createOscillator();
    o.type = 'sawtooth';
    const freq = getFrequency({value});
    o.frequency.value = Number(freq/2);
    o.start(t);
    o.stop(t + duration + release)
    pitchLFO.connect(o.frequency)
    return o;
  })
  .room(.5).lpf(800)
  .gain(.5)

this cries for a thin web audio abstraction layer..

[felixroos]

PR for sound registry https://github.com/tidalcycles/strudel/pull/516

[felixroos]

note to avoid confusion: https://github.com/tidalcycles/strudel/pull/516 now has integrated gm soundfonts based on webaudiofontdata, but the soundfonts loaded from sf2 files are still not part of the registry, so I'll leave this open