mcpField
B
bevy-audio

bevy-audio

github.com/chrisgliddonDevelopment2663 installsUpdated 9w ago
Repository

About bevy-audio

bevy-audio is a skill for developers building audio in Bevy 0.19 applications. It explains playback, spatial audio, mixing policies, browser restrictions, and when to use an audio-graph library.

Key features

  1. 01Control clips, loops, and playback sinks
  2. 02Manage audio categories and transitions
  3. 03Configure simple spatial stereo audio
  4. 04Handle browser audio unlock requirements
  5. 05Understand the Bevy Seedling boundary

Use cases

  1. 01Add music and ambience transitions to a game
  2. 02Control live playback volume, pause, or seeking
  3. 03Place audio sources relative to a listener
  4. 04Plan audio support for browser builds

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/chrisgliddon/bevy-skills --skill bevy-audio

After installing, type/bevy-audioin Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

references/
built-in-audio.md
mixing-lifecycle-and-transitions.md
spatial-web-and-seedling.md

Details

From the project README

Bevy 0.19 — audio

When to use this skill

  • Spawn or control clips/loops through AudioPlayer and sink components.
  • Build music/ambience transitions, mix categories, ducking, or pause policy.
  • Add spatial sources/listeners or define the built-in HRTF/effects boundary.
  • Ship audio on browsers or evaluate bevy_seedling 0.8 for a graph backend.

Use Bevy's built-in entity-based audio for straightforward clips, loops, playback control, and simple spatial stereo. Add an application mixer policy for categories and transitions. Move to an audio-graph library only when the measured/design requirement exceeds that boundary.

Canonical pattern

use bevy::{audio::Volume, prelude::*};

#[derive(Component)]
struct Music;

fn start_music(mut commands: Commands, assets: Res<AssetServer>) {
    commands.spawn((
        Name::new("Exploration music"),
        Music,
        AudioPlayer::new(assets.load("audio/exploration.ogg")),
        PlaybackSettings::LOOP.with_volume(Volume::Decibels(-9.0)),
    ));
}

fn mute_music(mut sinks: Query<&mut AudioSink, With<Music>>) {
    for mut sink in &mut sinks {
        sink.mute();
    }
}

Bevy inserts AudioSink after non-spatial playback begins and SpatialAudioSink for spatial playback. Query the sink for live volume, mute, pause, speed, position, seek, and stop. Changing PlaybackSettings or GlobalVolume does not update audio that is already playing.

PlaybackSettings::{ONCE, LOOP, DESPAWN, REMOVE} define completion ownership. An ONCE player cannot simply replay after its sink drains; remove/reinsert the audio components or spawn a new player. Removing a sink while AudioPlayer remains causes the unchanged source to start again.

Gotchas

  • PlaybackSettings and GlobalVolume affect initial playback, not existing sinks.
  • A sink appears only after playback starts; query it optionally.
  • Drained ONCE playback is not replayable without replacing audio components.
  • Removing a sink while its unchanged player remains restarts the source.
  • Built-in spatial audio is stereo panning, not HRTF or a general effects graph.

Spatial audio boundary

fn spawn_spatial_loop(mut commands: Commands, assets: Res<AssetServer>) {
    commands.spawn((
        SpatialListener::new(0.2),
        Transform::default(),
    ));
    commands.spawn((
        AudioPlayer::new(assets.load("audio/generator.ogg")),
        PlaybackSettings::LOOP.with_spatial(true),
        Transform::from_xyz(4.0, 0.0, 0.0),
    ));
}

Keep exactly one SpatialListener; attach it to the authoritative listener/camera pose. Configure SpatialScale when world units do not match audio distance. Built-in spatialisation is simple left/right stereo panning: it does not provide HRTF or a full 3D acoustic model.

Mix groups and transitions

Built-in audio has no general bus/effects graph. Model categories (Master, Music, Sfx, Dialogue, Ambience, Ui) as application state and tag player entities. Compute effective gain from master × group × transition/ducking × source gain, then write sinks. Store sliders in decibels or a perceptually useful curve, not a raw linear UI that puts most useful range in a few pixels.

For music/ambience changes, overlap old and new loops and crossfade using real/virtual time according to pause policy. Do not despawn the old entity until its sink reaches silence. Preload assets before a latency-critical transition, and define what happens when a new track is not ready.

Browser rule

Browsers commonly block audible playback until a user activation. Treat audio unlock as an explicit app state: let the first intentional input resume/start audio, keep the request pending if blocked, and present a usable muted/unlock control. Test refresh, tab background/foreground, device change, and mobile browsers; startup playback is not evidence that every browser will allow it.

Built-in versus Seedling

Choose built-in Bevy audio for simple playback/control/spatial panning. For Bevy 0.19, bevy_seedling 0.8 is the compatible graph-based option when you need buses/effects, pooled emitters, sample-accurate scheduling, richer routing, or HRTF through its Firewheel stack. HRTF is not a default feature; enable Seedling's hrtf feature (and hrtf_subjects only when all embedded subjects are needed). Disable Bevy's built-in audio feature when using Seedling so the two backends/types do not compete. Pin and test the exact library release and platform.

Choose the relevant deep dive

ProblemRead
Players, playback modes, sinks, replay and cleanupBuilt-in audio
Mix model, crossfades, ducking, pause and asset lifetimeMixing, lifecycle, and transitions
Spatial limits, browser unlock, accessibility and Seedling boundarySpatial, web, and Seedling

Review checklist

  • Playback entity ownership and completion mode are intentional.
  • Runtime changes target the sink, not PlaybackSettings.
  • Group/master changes update already-playing sinks without losing source gain.
  • Transitions handle delayed assets, rapid state changes, pause, and entity cleanup.
  • Exactly one listener exists and scale/pose follow the intended viewpoint.
  • Browser audio unlock/failure is visible and recoverable.
  • Dialogue has captions and important audio cues have non-audio equivalents.
  • Headless tests exercise mixer/state logic without requiring an audio device.

See also

Comments

Markdown supported