Positional audio

since 0.1.0 Simulation, audio and XR

Open the live demo · Read the source · View on GitHub

AudioScene decides what is heard and how loudly, from the position of every sound and the listener's position and facing. The geometry is computed in Dart rather than left to the backend, because a listener that turns every frame needs the pan to follow it immediately.

Step 1: A scene with one sound #

SilentBackend makes no sound and records what it is asked for, which is what lets a page like this one run with nothing behind it.

final backend = SilentBackend();
final scene = AudioScene(backend: backend);
const bell = SoundDef(name: 'bell', asset: 'bell.wav');

Step 2: Play it and measure it #

play starts an emitter at a position. update recomputes every emitter's gain and pan against a listener, once a frame.

final emitter = scene.play(bell, Vector3(4.0, 0.0, 0.0));
final listener = AudioListener()..aimAt(Vector3.zero(), 0.0);
scene.update(listener);

With the source four metres to the side, the listener hears it panned hard to that side.

Step 3: Move the listener #

// The listener turns to face the source directly and steps closer.
listener.aimAt(Vector3(2.0, 0.0, 0.0), -1.5707963267948966);
scene.update(listener);

Turning to face the source directly centres the pan, and stepping closer raises the gain: InverseRolloff, the default attenuation curve, gets louder the nearer the reference distance a sound is heard from.

Step 4: Walk round the bell #

The same two calls, made every frame. The blue block is the listener and its white nose is the way it faces; it walks round the bell, or wherever the sliders put it. The bell swells with how loud it is heard, the yellow bar is that gain, and the knob on the rail is the pan, from the left ear to the right. Turn the listener away and the pan swings, though the distance has not changed.

// The same scene as above, asked again every frame: the listener moves
// and turns, and what it would hear is read back off the emitter.
_audio = AudioScene(backend: SilentBackend());
_emitter = _audio.play(
  const SoundDef(name: 'bell', asset: 'bell.wav'),
  Vector3(0.0, 0.0, 0.0),
);
_listener = AudioListener();

_listener.aimAt(at, yaw);
_audio.update(_listener);