Skip to main content
The AudioManager class provides centralized control over audio session configuration, device management, and audio processing for LiveKit.

​
Singleton Access

let audioManager = AudioManager.shared

​
Audio Device Management (macOS)

​
outputDevices
[AudioDevice]
List of available output devices (macOS only).
​
inputDevices
[AudioDevice]
List of available input devices (macOS only).
​
outputDevice
AudioDevice
Currently selected output device.On macOS, you can set this to change the output device.
​
inputDevice
AudioDevice
Currently selected input device.On macOS, you can set this to change the input device.
​
defaultOutputDevice
AudioDevice
The default output device for the system.
​
defaultInputDevice
AudioDevice
The default input device for the system.

​
Audio Session Configuration (iOS/tvOS/visionOS)

​
isSpeakerOutputPreferred
Bool
default:"true"
Determines whether the device’s built-in speaker or receiver is preferred for audio output.
  • true: Speaker is preferred
  • false: Receiver is preferred
Only applies when audio output is routed to the built-in speaker or receiver. Ignored if customConfigureAudioSessionFunc is set.
​
sessionConfiguration
AudioSessionConfiguration?
Specifies a fixed configuration for the audio session, overriding dynamic adjustments.When set, takes precedence over dynamic configuration logic including isSpeakerOutputPreferred.
Ignored if customConfigureAudioSessionFunc is set.

​
Voice Processing

​
isVoiceProcessingEnabled
Bool
default:"true"
The main flag that determines whether to enable Voice-Processing I/O of the internal AVAudioEngine.
Toggling this requires restarting the AudioEngine. It is recommended to keep this true and toggle isVoiceProcessingBypassed when possible.
Setting this to false prevents voice-processing initialization and muted talker detection will not work.
​
isVoiceProcessingBypassed
Bool
default:"false"
Bypass Voice-Processing I/O of internal AVAudioEngine.Can be toggled at runtime without restarting the AudioEngine.
​
isVoiceProcessingAGCEnabled
Bool
Enable or bypass the Auto Gain Control of internal AVAudioEngine.Can be toggled at runtime.

​
Audio Processing

​
capturePostProcessingDelegate
AudioCustomProcessingDelegate?
Delegate to modify the local audio buffer before it is sent to the network.
Only one delegate can be set at a time. If you only need to observe (not modify) the buffer, use add(localAudioRenderer:) instead.
​
renderPreProcessingDelegate
AudioCustomProcessingDelegate?
Delegate to modify the combined remote audio buffer (all tracks) before it is played to the user.
Only one delegate can be set at a time. If you only need to observe the buffer, use add(remoteAudioRenderer:) instead.

​
Ducking Control

​
isAdvancedDuckingEnabled
Bool
default:"false"
Enables “advanced ducking” of other audio while using Apple’s voice processing APIs.When enabled, the system dynamically adjusts ducking based on voice activity:
  • More ducking when someone is speaking
  • Less ducking when neither side is speaking (SharePlay/FaceTime-like behavior)
Defaults to false, which keeps a fixed ducking behavior with minimal ducking to keep other audio as loud as possible.
​
duckingLevel
AudioDuckingLevel
default:".min"
Controls how much other audio is reduced (“ducked”) while using Apple’s voice processing APIs.Available on iOS 17+, macOS 14.0+, visionOS 1.0+.
  • .min: Keep other audio as loud as possible (SDK default)
  • .default: Apple’s historical fixed ducking amount
  • .max: Better voice intelligibility, more ducking

​
Recording

​
isRecordingAlwaysPreparedMode
Bool
Whether recording is kept initialized for low-latency publish.
​
isMicrophoneMuted
Bool
The mute state of the internal audio engine using Voice Processing I/O mute API.Normally handled automatically, but can be set manually if needed.
​
isManualRenderingMode
Bool
Whether manual rendering (no-device) mode is enabled.
​
isEngineRunning
Bool
Whether the internal AVAudioEngine is currently running.
​
engineAvailability
AudioEngineAvailability
The current availability state of the audio engine.

​
Callbacks

​
onDeviceUpdate
OnDevicesDidUpdate?
Callback invoked when audio devices change.
AudioManager.shared.onDeviceUpdate = { audioManager in
    print("Audio devices updated")
}
​
onMutedSpeechActivity
OnSpeechActivity?
Detect voice activity even if the mic is muted.
Internal audio engine must be initialized by calling prepareRecording() or connecting to a room.
AudioManager.shared.onMutedSpeechActivity = { audioManager, event in
    print("Speech activity: \(event)")
}

​
Methods

​
setVoiceProcessingEnabled(_:)

Set whether voice processing is enabled. 
try AudioManager.shared.setVoiceProcessingEnabled(true)
Throws: Error if the setting cannot be applied.

​
setManualRenderingMode(_:)

Enable manual rendering mode where you provide audio buffers. 
try AudioManager.shared.setManualRenderingMode(true)
In this mode:
  • Provide audio buffers via AudioManager.shared.mixer.capture(appAudio:)
  • Remote audio will not play automatically
  • Get remote audio with add(remoteAudioRenderer:) or per-track renderers

​
setRecordingAlwaysPreparedMode(_:)

Prepare the microphone capture pipeline for low-latency publishing. 
try await AudioManager.shared.setRecordingAlwaysPreparedMode(true)
When enabled:
  • Audio engine starts configured for mic input in muted state
  • Keeps recording initialized and pre-warms voice processing
  • Persists across Room lifecycles until disabled
Note: Microphone permission is required. iOS may prompt if not granted.

​
startLocalRecording()

Start mic input to the SDK even without a Room or connection. 
try AudioManager.shared.startLocalRecording()
Audio buffers will flow into LocalAudioTrack renderers and capturePostProcessingDelegate.

​
stopLocalRecording()

Stop mic input started with startLocalRecording(). 
try AudioManager.shared.stopLocalRecording()

​
setEngineAvailability(_:)

Set whether the internal AVAudioEngine is allowed to run. 
try AudioManager.shared.setEngineAvailability(.enabled)
This flag has highest priority over any API that may start the engine. Useful for CallKit flows where you need to set up connections without touching the audio device yet.

​
add(localAudioRenderer:)

Add a renderer to receive PCM buffers from local input (mic). 
AudioManager.shared.add(localAudioRenderer: myRenderer)

​
add(remoteAudioRenderer:)

Add a renderer to receive PCM buffers from combined remote audio. 
AudioManager.shared.add(remoteAudioRenderer: myRenderer)
For individual track buffers, use RemoteAudioTrack.add(audioRenderer:) instead.

​
set(engineObservers:)

Set a chain of AudioEngineObservers. 
AudioManager.shared.set(engineObservers: [observer1, observer2])
Defaults to having a single AudioSessionEngineObserver initially.

​
Example Usage

// Configure audio session for speaker output
AudioManager.shared.isSpeakerOutputPreferred = true

// Enable low-latency recording
try await AudioManager.shared.setRecordingAlwaysPreparedMode(true)

// Monitor speech activity while muted
AudioManager.shared.onMutedSpeechActivity = { _, event in
    print("Speech detected: \(event.isSpeaking)")
}

// Add local audio renderer
AudioManager.shared.add(localAudioRenderer: myAudioProcessor)

// On macOS, change audio device
#if os(macOS)
if let device = AudioManager.shared.inputDevices.first(where: { $0.name == "My Mic" }) {
    AudioManager.shared.inputDevice = device
}
#endif

Build docs developers (and LLMs) love

Get started for freeTalk to us