Spatial Audio Functional Regression - different spatial scaling factor for each object

[johnny-smitherson]

What problem does this solve or what need does it fill?

I want to scale spatial audio differently for each effect. This is so I can have some "far away" effects (low volume, low distance scaling) mixed together with some "close up" effects (high volume, high distance scaling).

For example, explosions - when they happen far away, https://freesound.org/people/rhapsodize/sounds/255111/

When they happen close up, https://freesound.org/people/juskiddink/sounds/108641/

Both effects need to live at the same time with different spatial scales, because the camera can quickly move closer/further away from the audio source.

I would also want this to work on multiple listeners at the same time: one main camera for 80% volume, and a few extra cameras for 20% volume each.

This was trivial under Bevy 0.11 (in the old update positions system), but now is impossible with bevy_audio 0.12:

https://bevyengine.org/learn/migration-guides/0.11-0.12/#more-ergonomic-spatial-audio
If you were manually scaling emitter/listener positions, you can use the spatial_scale field of AudioPlugin instead.

But the AudioPlugin is a global, so all objects must share the same spatial scale.

What solution would you like?

Extra field spatial scale for each Spatial Audio Source

What alternative(s) have you considered?

• in a system, manually change the RelativeVolume component based on distance and some formula
  • problem: doesn't work with multiple listeners, because the volume is set per-effect not per-listener
• swap out bevy_audio for one of the external plugins

[rparrett]

"Spatial Scale" in Bevy 0.11 was also "global" -- it affects the position of the ears as well. At least how it was implemented in the examples.

I think another potential workaround would be adding a system that multiplies your emitter GlobalTransform translations after TransformSystem::TransformPropagate but before AudioPlaySet. If that works for you, I'm not sure if this needs to be treated as a regression.

Affecting a particular audio source's distance attenuation seems useful. I'd like to see examples of how other game engine audio APIs work for this use-case. Would a minimum and maximum attenuation distance be more useful? I'm not sure how much fancier we could get without ditching rodio. Your suggestion sounds reasonable enough to me though.

[rparrett]

Godot has the following on its emitters:

• Attentuation model
• Max distance
• "Unit Size" (scale)

And no configuration for the listener, only position. There can only be one active listener.

[johnny-smitherson]

"Spatial Scale" in Bevy 0.11 was also "global" -- it affects the position of the ears as well. At least how it was implemented in the examples.

Yes, what I did was scale the sink around the listener:

fn update_spatial_audio_locations(
    emitter_query: Query<(&GlobalTransform, &SpatialAudioSink, &SpatialAudioEffect)>,
    listener_query: Query<&GlobalTransform, With<SpatialAudioListener>>,
) {
    let Ok(listener) = listener_query.get_single() else {
        return;
    };
    let listener = listener.compute_transform();

    for (emitter, sink, effect_info) in &emitter_query {
        let emitter = emitter.compute_transform().translation;
        let emitter =
            listener.translation + (emitter - listener.translation) * effect_info.scale_range;
        sink.set_emitter_position(emitter);
        // don't touch the listener: sink.set_listener_position(listener, SPATIAL_AUDIO_EAR_GAP);
    }
}

---

I think another potential workaround would be adding a system that multiplies your emitter GlobalTransform translations after TransformSystem::TransformPropagate but before AudioPlaySet

I'll take a look, thanks!

---

I'd like to see examples of how other game engine audio APIs work for this use-case.

Godot

• support for single Audio Listener : Note: There may be more than one AudioListener3D marked as "current" in the scene tree, but only the one that was made current last will be used.
• each "audio source" is controlled by "audio stream player 3d" which has separate spatial controls, including linear damping, collisions, reverb
• support for multiple effect stacks (called Audio Busses), mixed together: https://docs.godotengine.org/en/3.2/tutorials/audio/audio_buses.html#doc-audio-buses -- each Emitter can choose the bus to send audio to

Unity

• "audio source" has a bunch of 3D Sound Settings - https://docs.unity3d.com/Manual/class-AudioSource.html
  • look at the graphs in the picture: volume, reverb, etc. as functions of distance
• multiple listeners NOT supported - Each scene can only have one Audio Listener.

Unreal

• extremely complex including volume attenuation, spatial collisions, focus, priority, etc.
• customizable for each separate audio source
• https://docs.unrealengine.com/5.2/en-US/sound-attenuation-in-unreal-engine/

---

I guess there are 2 separate discussions:

• different spatial audio settings for each emitter object
• meaningful support for multiple simultaneous listeners with different spatial settings and positions (like Godot Audio Busses) - should open a different feature request for this one

[rparrett]

• different spatial audio settings for each emitter object
  meaningful support for multiple simultaneous listeners with different spatial settings and positions (like Godot Audio Busses)
• should open a different feature request for this one

Agreed, thanks for doing that roundup. (and for the excellent issue writeup). On the first bullet point, I know that scale is obviously feasible but I'm guessing that other attenuation options might not be doable with rodio.

Just need to determine whether or not there's an acceptable workaround now.