如何静音WebRTC中OpenAI实时API的远程音频输出

我开发了一个用WebRTC连接OpenAI Realtime API的应用，目前麦克风静音功能正常，但想实现静音Realtime API的输出音频。

相关代码如下：

internal class WebRTCClient (
    onIncomingEvent: (OpenAiEvent) -> Unit,
    context: Context,
) {
    private val peerConnectionFactory: PeerConnectionFactory
    private val peerConnection: PeerConnection
    private val negotiateJob = AtomicReference<Job?>(null)
    private val audioSource: AudioSource
    private val localAudioTrack: AudioTrack
    private val dataChannel: DataChannel

    init {
        peerConnectionFactory = createPeerConnectionFactory(context)
        peerConnection = createPeerConnection()
        createAudioSourceAndTrack().let {
            audioSource = it.first
            localAudioTrack = it.second
        }

        val sender = peerConnection.addTrack(localAudioTrack)
    }

    private fun createAudioSourceAndTrack(): Pair<AudioSource, AudioTrack> {
        Napier.d(tag = TAG) { "Audio track: creating and adding" }
        val audioSource: AudioSource = peerConnectionFactory.createAudioSource(MediaConstraints())
        val localAudioTrack = peerConnectionFactory.createAudioTrack("mic", audioSource)
        localAudioTrack.setEnabled(true)

        return Pair(audioSource, localAudioTrack)
    }
}

我尝试了以下代码，但没有任何效果，AI的输出音频仍在播放：

fun setAudioTrackEnabled(enabled: Boolean) {
        localAudioTrack.setEnabled(enabled)
    }

解决方案 #

你当前的setAudioTrackEnabled方法操作的是本地音频轨道（麦克风输入），只会影响你传给AI的声音，无法控制远程（AI输出）的音频。要静音远程音频，需要操作收到的远程音频轨道，步骤如下：

1. 监听PeerConnection的onTrack回调，保存远程音频轨道
   修改createPeerConnection方法，设置PeerConnection.Observer并重写onTrack方法，获取并保存远程音频轨道引用：

   private var remoteAudioTrack: AudioTrack? = null
   
   private fun createPeerConnection(): PeerConnection {
       val observer = object : PeerConnection.Observer {
           // 按需实现其他必要的回调方法（如onIceCandidate、onConnectionChange等）
           override fun onTrack(transceiver: RtpTransceiver?) {
               transceiver?.receiver?.track?.let { track ->
                   // 判断是否为音频轨道
                   if (track.kind() == MediaStreamTrack.AUDIO_TRACK) {
                       remoteAudioTrack = track as AudioTrack
                   }
               }
           }
   
           // 示例：实现其他必填回调（空实现即可，按需补充逻辑）
           override fun onIceCandidate(candidate: IceCandidate?) {}
           override fun onConnectionChange(newState: PeerConnection.PeerConnectionState?) {}
           override fun onSignalingChange(newState: PeerConnection.SignalingState?) {}
           // 其他未实现的Observer方法按需补充
       }
   
       // 传入你的PeerConnection配置和observer创建实例
       return peerConnectionFactory.createPeerConnection(/*你的RTCConfiguration*/, observer)
   }
   

2. 实现静音远程音频的方法
   在WebRTCClient类中添加控制远程音频的方法：

   fun muteRemoteAudio(mute: Boolean) {
       // enabled设为false时静音，true时恢复
       remoteAudioTrack?.setEnabled(!mute)
   }
   

调用muteRemoteAudio(true)即可静音AI的输出音频，调用muteRemoteAudio(false)恢复播放。

内容的提问来源于stack exchange，提问作者Sorry