Spatial audio and outputs

AudioEmitter

A virtual speaker in the 3D world. It has one Input pin, and any number of wires can feed it. Its position comes from its parent (Attachment, Camera or PVInstance), or from PositionInstance when PositionType = Instance. Under any other parent it is silent.

Property	Type	Notes
AudioInteractionGroup	string	Only listeners with the same group hear it. Default "".
DistanceAttenuationMode	DistanceAttenuationMode	Custom (default) uses your curve. Inverse, InverseTapered, Linear and LinearSquared are presets.
DistanceAttenuationBounds	NumberRange	Full volume inside the minimum and silent past the maximum. Presets only. Default 4–10,000 studs.
PositionType	EmitterPositionType	Parent (default) or Instance
PositionInstance	Instance	Used when PositionType = Instance
AcousticSimulationEnabled	bool	Default true. Also needs the listener and SoundService switches on.
OcclusionEnabled, DiffractionEnabled, ReverbEnabled	SimulationMode	Per-feature override: Default, Enabled or Disabled
SimulationFidelity		Deprecated. Use AcousticSimulationEnabled.

Method	Notes
SetDistanceAttenuation({[studs]=vol})	Custom curve: up to 400 points, volumes 0–1, linear between points, flat past the ends. Only used in Custom mode. An empty table or nil means inverse-square.
GetDistanceAttenuation()	Returns the custom curve, even while a preset is active
SetAngleAttenuation({[deg]=vol})	0–180° off the emitter's LookVector. Makes voices project forward, or speakers directional.
GetAngleAttenuation()	Empty table when the curve is the default (flat)
GetAudibilityFor(listener) → 0–1	Combined distance × angle volume for that listener
GetInteractingListeners()	Listeners sharing this emitter's interaction group
GetWorldCFrame()	Where the emitter actually is

Presets

Inverse falls off naturally from the lower bound. InverseTapered follows inverse up close and fades to exactly zero at the upper bound. Linear and LinearSquared ramp between the bounds. Voice chat's Legacy curve is LinearSquared from 7 to 80 studs.

AudioListener

A virtual microphone. It hears every emitter in its interaction group and outputs the spatialized mix on one Output pin. Positioning works the same way as for emitters. The listener's orientation comes from its parent's CFrame.

Property	Notes
AudioInteractionGroup	Which emitters it can hear. Default "".
PositionType, PositionInstance	As for AudioEmitter (ListenerPositionType)
AcousticSimulationEnabled, OcclusionEnabled, DiffractionEnabled, ReverbEnabled	As for AudioEmitter

Method	Notes
SetDistanceAttenuation(curve)	Better or worse hearing for this listener only. The default is flat (1 everywhere).
SetAngleAttenuation(curve)	Directional hearing, such as a shotgun mic
GetAudibilityFor(emitter)	Same number as the emitter-side call
GetInteractingEmitters()	Emitters it can hear
GetWorldCFrame()
Reset()	Clears the listener's internal state

A listener does nothing on its own. It must be wired to something, usually an AudioDeviceOutput.

AudioDeviceOutput

The player's speakers or headphones. It has one Input pin, and anything wired to it is heard.

Property	Notes
Player	Optional. If set, only that player hears it. If empty, everyone does. This lets the server author 2D audio aimed at one player.

You can have as many outputs as you like. They all mix into the same hardware. SoundService output-device selection (SetOutputDevice) is for the player's chosen hardware, not routing.

Acoustic simulation checklist

1. SoundService.AcousticSimulationEnabled = true.
2. Emitter and listener AcousticSimulationEnabled stay true (the default).
3. Walls that should block sound keep BasePart.AudioCanCollide = true. Turn it off for foliage, glass panes or anything that shouldn't muffle.
4. Tune materials with PhysicalProperties.AcousticAbsorption (echo) and Density (muffle).
5. Set SoundService.CharacterSoundsUseNewApi = Enabled if you want footsteps and other default character sounds simulated too.
6. Voice chat is simulated automatically when VoiceChatService.UseAudioApi = Enabled.

Simulation is tied to graphics quality. On low-end devices reverb, occlusion or diffraction may lose accuracy or turn off.