Add Spatial/Positional Audio in Rust

[harudagondi]

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

• Other game engines have the capability for positional audio sources:
  • Godot
  • Unreal Engine
  • Unity
• Bevy currently do not have support for spatial audio.

What solution would you like?

Create two Components:

• AudioEmitter: Emits sound from the current position of the entity.
• AudioListener: Listens for sounds from AudioEmitter. Has its own (potentially same) position.

AudioEmitter's API should be similar to bevy_audio::Audio, while AudioListener works similar to AudioOutput internally, or computes the resulting sound wave and send it to the mixer.

These components should probably have its own thread, as the sample rate (the number of audio samples per second) is usually around 44100 Hz, so systems dealing with audio components should be able to run for at least 44100 times a second to avoid underruns.

What alternative(s) have you considered?

• None so far.

Additional context

See the discussion in the bevy discord in the #audio-dev channel for more info.

[bjorn3]

These components should probably have its own thread, as the sample rate (the number of audio samples per second) is usually around 44100 Hz, so systems dealing with audio components should be able to run for at least 44100 times a second to avoid underruns.

I believe samples are pretty much always batched, so you don't need to run 44100 times for a 44100 Hz sample rate. As for running in their own thread, rodio (which bevy currently uses for audio) already spawns a background thread for audio handling AFAIK.

[harudagondi]

samples are pretty much always batched

background thread for audio handling

Whoops, i forgot about that. Maybe AudioListener and AudioEmitter could just use mspc channels, as long as it is fast enough.

[mockersf]

I have a small demo showing rodio limited spatial audio capabilities: https://vleue.itch.io/rodio-spatial-audio. It's waiting on a rodio release to be upstreamed in Bevy

[basilefff]

This issue seems to be resolved, should it be closed?

[mockersf]

It currently only does attenuation based on the distance to left/right "ears", it's very far from spatial...