@netpro2k I've tried to cover as many audio update cases as I could think of but I might be missing some.

In updateAudioSettings, we check whether we are switching from panning to stereo modes, and fire an event accordingly:

    if (
      (audio.panner === undefined && settings.audioType === AudioType.PannerNode) ||
      (audio.panner !== undefined && settings.audioType === AudioType.Stereo)
    ) {
      el.emit("audio_type_changed");
    }

In the event handlers, we delete the three js Audio (or PositionalAudio) node and replace it with its counter part. And we do the same thing in swapAudioType of audio-emitter-system.ts.

When we control the web audio node graph directly, changing from Stereo to Panner does not require deleting resources -- just rewiring of the graph to go through (or not go through) a PannerNode.

This also allows us to delete swapObject3DComponent, which was a hack written specifically for this situation.

I thought this was one of the main motivations for making this change, so I am surprised to not see it in the diff. Are there plans to follow up with that change, or did I misunderstand the problem you're intending to solve?

Good catch! I totally overlooked that case.

I thought this was one of the main motivations for making this change, so I am surprised to not see it in the diff. Are there plans to follow up with that change, or did I misunderstand the problem you're intending to solve?

The main goal of this PR is:

• Simplify the scene and audio node graphs remove unnecessary intermediate nodes
• Improve the audio performance reducing the amount of panner nodes and audio listener transform updates

Swapping audios is not a common case case, it only happens when you disable L/R panning so it doesn't really happen most of the time but now we can simplify that so that's a good side win.