Audio Device Optimization

allow listen-only mode in AudioUnit, adjust when category changes (#2)

release mic when category changes (#5)

Change defaults to iOS defaults (#7)

Sync audio session config (#8)

feat: support bypass voice processing for iOS. (#15)

Remove MacBookPro audio pan right code (#22)

fix: Fix can't open mic alone when built-in AEC is enabled. (#29)

feat: add audio device changes detect for windows. (#41)

fix Linux compile (#47)

AudioUnit: Don't rely on category switch for mic indicator to turn off (#52)

Stop recording on mute (turn off mic indicator) (#55)

Cherry pick audio selection from m97 release (#35)

[Mac] Allow audio device selection (#21)

RTCAudioDeviceModule.outputDevice / inputDevice getter and setter (#80)

Co-authored-by: Hiroshi Horie <548776+hiroshihorie@users.noreply.github.com>
Co-authored-by: David Zhao <dz@livekit.io>

Author: 3 people

audio/audio_send_stream.cc

@@ -424,6 +424,11 @@ void AudioSendStream::SetMuted(bool muted) {
   channel_send_->SetInputMute(muted);
 }
 
+bool AudioSendStream::GetMuted() {
+  RTC_DCHECK_RUN_ON(&worker_thread_checker_);
+  return channel_send_->InputMute();
+}
+
 webrtc::AudioSendStream::Stats AudioSendStream::GetStats() const {
   return GetStats(true);
 }

audio/audio_send_stream.h

@@ -96,6 +96,7 @@ class AudioSendStream final : public webrtc::AudioSendStream,
                           int payload_frequency,
                           int event,
                           int duration_ms) override;
+  bool GetMuted() override;
   void SetMuted(bool muted) override;
   webrtc::AudioSendStream::Stats GetStats() const override;
   webrtc::AudioSendStream::Stats GetStats(

audio/audio_state.cc

@@ -98,14 +98,26 @@ void AudioState::AddSendingStream(webrtc::AudioSendStream* stream,
   UpdateAudioTransportWithSendingStreams();
 
   // Make sure recording is initialized; start recording if enabled.
-  auto* adm = config_.audio_device_module.get();
-  if (!adm->Recording()) {
-    if (adm->InitRecording() == 0) {
-      if (recording_enabled_) {
-        adm->StartRecording();
+  if (ShouldRecord()) {
+    auto* adm = config_.audio_device_module.get();
+    if (!adm->Recording()) {
+      if (adm->InitRecording() == 0) {
+        if (recording_enabled_) {
+
+          // TODO: Verify if the following windows only logic is still required.
+#if defined(WEBRTC_WIN)
+          if (adm->BuiltInAECIsAvailable() && !adm->Playing()) {
+            if (!adm->PlayoutIsInitialized()) {
+              adm->InitPlayout();
+            }
+            adm->StartPlayout();
+          }
+#endif
+          adm->StartRecording();
+        }
+      } else {
+        RTC_DLOG_F(LS_ERROR) << "Failed to initialize recording.";
       }
-    } else {
-      RTC_DLOG_F(LS_ERROR) << "Failed to initialize recording.";
     }
   }
 }
@@ -115,7 +127,8 @@ void AudioState::RemoveSendingStream(webrtc::AudioSendStream* stream) {
   auto count = sending_streams_.erase(stream);
   RTC_DCHECK_EQ(1, count);
   UpdateAudioTransportWithSendingStreams();
-  if (sending_streams_.empty()) {
+
+  if (!ShouldRecord()) {
     config_.audio_device_module->StopRecording();
   }
 }
@@ -143,7 +156,7 @@ void AudioState::SetRecording(bool enabled) {
   if (recording_enabled_ != enabled) {
     recording_enabled_ = enabled;
     if (enabled) {
-      if (!sending_streams_.empty()) {
+      if (ShouldRecord()) {
         config_.audio_device_module->StartRecording();
       }
     } else {
@@ -203,6 +216,39 @@ void AudioState::UpdateNullAudioPollerState() {
     null_audio_poller_.Stop();
   }
 }
+
+void AudioState::OnMuteStreamChanged() {
+
+  auto* adm = config_.audio_device_module.get();
+  bool should_record = ShouldRecord();
+
+  if (should_record && !adm->Recording()) {
+    if (adm->InitRecording() == 0) {
+      adm->StartRecording();
+    }
+  } else if (!should_record && adm->Recording()) {
+    adm->StopRecording();
+  }
+}
+
+bool AudioState::ShouldRecord() {
+  // no streams to send
+  if (sending_streams_.empty()) {
+    return false;
+  }
+
+  int stream_count = sending_streams_.size();
+
+  int muted_count = 0;
+  for (const auto& kv : sending_streams_) {
+    if (kv.first->GetMuted()) {
+      muted_count++;
+    }
+  }
+
+  return muted_count != stream_count;
+}
+
 }  // namespace internal
 
 rtc::scoped_refptr<AudioState> AudioState::Create(

audio/audio_state.h

@@ -47,6 +47,8 @@ class AudioState : public webrtc::AudioState {
 
   void SetStereoChannelSwapping(bool enable) override;
 
+  void OnMuteStreamChanged() override;
+
   AudioDeviceModule* audio_device_module() {
     RTC_DCHECK(config_.audio_device_module);
     return config_.audio_device_module.get();
@@ -64,6 +66,9 @@ class AudioState : public webrtc::AudioState {
   void UpdateAudioTransportWithSendingStreams();
   void UpdateNullAudioPollerState() RTC_RUN_ON(&thread_checker_);
 
+  // Returns true when at least 1 stream exists and all streams are not muted.
+  bool ShouldRecord();
+
   SequenceChecker thread_checker_;
   SequenceChecker process_thread_checker_{SequenceChecker::kDetached};
   const webrtc::AudioState::Config config_;

audio/channel_send.cc

@@ -98,6 +98,8 @@ class ChannelSend : public ChannelSendInterface,
   // Muting, Volume and Level.
   void SetInputMute(bool enable) override;
 
+  bool InputMute() const override;
+
   // Stats.
   ANAStats GetANAStatistics() const override;
 
@@ -161,8 +163,6 @@ class ChannelSend : public ChannelSendInterface,
                    size_t payloadSize,
                    int64_t absolute_capture_timestamp_ms) override;
 
-  bool InputMute() const;
-
   int32_t SendRtpAudio(AudioFrameType frameType,
                        uint8_t payloadType,
                        uint32_t rtp_timestamp,

audio/channel_send.h

@@ -95,6 +95,8 @@ class ChannelSendInterface {
   virtual bool SendTelephoneEventOutband(int event, int duration_ms) = 0;
   virtual void OnBitrateAllocation(BitrateAllocationUpdate update) = 0;
   virtual int GetTargetBitrate() const = 0;
+
+  virtual bool InputMute() const = 0;
   virtual void SetInputMute(bool muted) = 0;
 
   virtual void ProcessAndEncodeAudio(

call/audio_send_stream.h

@@ -190,6 +190,7 @@ class AudioSendStream : public AudioSender {
                                   int event,
                                   int duration_ms) = 0;
 
+  virtual bool GetMuted() = 0;
   virtual void SetMuted(bool muted) = 0;
 
   virtual Stats GetStats() const = 0;

call/audio_state.h

@@ -59,6 +59,9 @@ class AudioState : public rtc::RefCountInterface {
 
   virtual void SetStereoChannelSwapping(bool enable) = 0;
 
+  // Notify the AudioState that a stream updated it's mute state.
+  virtual void OnMuteStreamChanged() = 0;
+
   static rtc::scoped_refptr<AudioState> Create(
       const AudioState::Config& config);
 

media/engine/webrtc_voice_engine.cc

@@ -2285,6 +2285,9 @@ bool WebRtcVoiceMediaChannel::MuteStream(uint32_t ssrc, bool muted) {
     ap->set_output_will_be_muted(all_muted);
   }
 
+  // Notfy the AudioState that the mute state has updated.
+  engine_->audio_state()->OnMuteStreamChanged();
+
   return true;
 }
 

media/engine/webrtc_voice_engine.h

@@ -93,6 +93,8 @@ class WebRtcVoiceEngine final : public VoiceEngineInterface {
 
   absl::optional<webrtc::AudioDeviceModule::Stats> GetAudioDeviceStats()
       override;
+  // Moved to public so WebRtcVoiceMediaChannel can access it.
+  webrtc::AudioState* audio_state();
 
  private:
   // Every option that is "set" will be applied. Every option not "set" will be
@@ -105,7 +107,6 @@ class WebRtcVoiceEngine final : public VoiceEngineInterface {
 
   webrtc::AudioDeviceModule* adm();
   webrtc::AudioProcessing* apm() const;
-  webrtc::AudioState* audio_state();
 
   std::vector<AudioCodec> CollectCodecs(
       const std::vector<webrtc::AudioCodecSpec>& specs) const;

modules/audio_device/audio_device_data_observer.cc

@@ -307,6 +307,10 @@ class ADMWrapper : public AudioDeviceModule, public AudioTransport {
   }
 #endif  // WEBRTC_IOS
 
+  int32_t SetAudioDeviceSink(AudioDeviceSink* sink) const override {
+    return impl_->SetAudioDeviceSink(sink);
+  }
+
  protected:
   rtc::scoped_refptr<AudioDeviceModule> impl_;
   AudioDeviceDataObserver* legacy_observer_ = nullptr;

modules/audio_device/audio_device_generic.h

@@ -135,6 +135,10 @@ class AudioDeviceGeneric {
   virtual int GetRecordAudioParameters(AudioParameters* params) const;
 #endif  // WEBRTC_IOS
 
+  virtual int32_t SetAudioDeviceSink(AudioDeviceSink* sink) { return -1; }
+  virtual int32_t GetPlayoutDevice() const { return -1; }
+  virtual int32_t GetRecordingDevice() const { return -1; }
+
   virtual void AttachAudioBuffer(AudioDeviceBuffer* audioBuffer) = 0;
 
   virtual ~AudioDeviceGeneric() {}

modules/audio_device/audio_device_impl.cc

@@ -72,15 +72,17 @@ namespace webrtc {
 
 rtc::scoped_refptr<AudioDeviceModule> AudioDeviceModule::Create(
     AudioLayer audio_layer,
-    TaskQueueFactory* task_queue_factory) {
+    TaskQueueFactory* task_queue_factory,
+    bool bypass_voice_processing) {
   RTC_DLOG(LS_INFO) << __FUNCTION__;
-  return AudioDeviceModule::CreateForTest(audio_layer, task_queue_factory);
+  return AudioDeviceModule::CreateForTest(audio_layer, task_queue_factory, bypass_voice_processing);
 }
 
 // static
 rtc::scoped_refptr<AudioDeviceModuleForTest> AudioDeviceModule::CreateForTest(
     AudioLayer audio_layer,
-    TaskQueueFactory* task_queue_factory) {
+    TaskQueueFactory* task_queue_factory,
+    bool bypass_voice_processing) {
   RTC_DLOG(LS_INFO) << __FUNCTION__;
 
   // The "AudioDeviceModule::kWindowsCoreAudio2" audio layer has its own
@@ -93,7 +95,7 @@ rtc::scoped_refptr<AudioDeviceModuleForTest> AudioDeviceModule::CreateForTest(
 
   // Create the generic reference counted (platform independent) implementation.
   auto audio_device = rtc::make_ref_counted<AudioDeviceModuleImpl>(
-      audio_layer, task_queue_factory);
+      audio_layer, task_queue_factory, bypass_voice_processing);
 
   // Ensure that the current platform is supported.
   if (audio_device->CheckPlatform() == -1) {
@@ -116,8 +118,13 @@ rtc::scoped_refptr<AudioDeviceModuleForTest> AudioDeviceModule::CreateForTest(
 
 AudioDeviceModuleImpl::AudioDeviceModuleImpl(
     AudioLayer audio_layer,
-    TaskQueueFactory* task_queue_factory)
-    : audio_layer_(audio_layer), audio_device_buffer_(task_queue_factory) {
+    TaskQueueFactory* task_queue_factory,
+    bool bypass_voice_processing)
+    : audio_layer_(audio_layer),
+    #if defined(WEBRTC_IOS)
+    bypass_voice_processing_(bypass_voice_processing),
+    #endif
+    audio_device_buffer_(task_queue_factory) {
   RTC_DLOG(LS_INFO) << __FUNCTION__;
 }
 
@@ -280,7 +287,7 @@ int32_t AudioDeviceModuleImpl::CreatePlatformSpecificObjects() {
 #if defined(WEBRTC_IOS)
   if (audio_layer == kPlatformDefaultAudio) {
     audio_device_.reset(
-        new ios_adm::AudioDeviceIOS(/*bypass_voice_processing=*/false));
+        new ios_adm::AudioDeviceIOS(/*bypass_voice_processing=*/bypass_voice_processing_));
     RTC_LOG(LS_INFO) << "iPhone Audio APIs will be utilized.";
   }
 // END #if defined(WEBRTC_IOS)
@@ -937,6 +944,27 @@ int AudioDeviceModuleImpl::GetRecordAudioParameters(
 }
 #endif  // WEBRTC_IOS
 
+int32_t AudioDeviceModuleImpl::SetAudioDeviceSink(AudioDeviceSink* sink) const {
+  RTC_LOG(LS_INFO) << __FUNCTION__ << "(" << sink << ")";
+  int32_t ok = audio_device_->SetAudioDeviceSink(sink);
+  RTC_LOG(LS_INFO) << "output: " << ok;
+  return ok;
+}
+
+int32_t AudioDeviceModuleImpl::GetPlayoutDevice() const {
+  RTC_LOG(LS_INFO) << __FUNCTION__;
+  int32_t r = audio_device_->GetPlayoutDevice();
+  RTC_LOG(LS_INFO) << "output: " << r;
+  return r;
+}
+
+int32_t AudioDeviceModuleImpl::GetRecordingDevice() const {
+  RTC_LOG(LS_INFO) << __FUNCTION__;
+  int32_t r = audio_device_->GetRecordingDevice();
+  RTC_LOG(LS_INFO) << "output: " << r;
+  return r;
+}
+
 AudioDeviceModuleImpl::PlatformType AudioDeviceModuleImpl::Platform() const {
   RTC_LOG(LS_INFO) << __FUNCTION__;
   return platform_type_;

modules/audio_device/audio_device_impl.h

@@ -43,7 +43,8 @@ class AudioDeviceModuleImpl : public AudioDeviceModuleForTest {
   int32_t AttachAudioBuffer();
 
   AudioDeviceModuleImpl(AudioLayer audio_layer,
-                        TaskQueueFactory* task_queue_factory);
+                        TaskQueueFactory* task_queue_factory,
+                        bool bypass_voice_processing = false);
   ~AudioDeviceModuleImpl() override;
 
   // Retrieve the currently utilized audio layer
@@ -145,6 +146,10 @@ class AudioDeviceModuleImpl : public AudioDeviceModuleForTest {
   int GetRecordAudioParameters(AudioParameters* params) const override;
 #endif  // WEBRTC_IOS
 
+  int32_t SetAudioDeviceSink(AudioDeviceSink* sink) const override;
+  int32_t GetPlayoutDevice() const override;
+  int32_t GetRecordingDevice() const override;
+
 #if defined(WEBRTC_ANDROID)
   // Only use this acccessor for test purposes on Android.
   AudioManager* GetAndroidAudioManagerForTest() {
@@ -165,7 +170,9 @@ class AudioDeviceModuleImpl : public AudioDeviceModuleForTest {
   AudioLayer audio_layer_;
   PlatformType platform_type_ = kPlatformNotSupported;
   bool initialized_ = false;
-#if defined(WEBRTC_ANDROID)
+#if defined(WEBRTC_IOS)
+  bool bypass_voice_processing_;
+#elif defined(WEBRTC_ANDROID)
   // Should be declared first to ensure that it outlives other resources.
   std::unique_ptr<AudioManager> audio_manager_android_;
 #endif

modules/audio_device/include/audio_device.h

@@ -21,6 +21,15 @@ namespace webrtc {
 
 class AudioDeviceModuleForTest;
 
+// Sink for callbacks related to a audio device.
+class AudioDeviceSink {
+ public:
+  virtual ~AudioDeviceSink() = default;
+
+  // input/output devices updated or default device changed
+  virtual void OnDevicesUpdated() = 0;
+};
+
 class AudioDeviceModule : public rtc::RefCountInterface {
  public:
   enum AudioLayer {
@@ -56,12 +65,14 @@ class AudioDeviceModule : public rtc::RefCountInterface {
   // Creates a default ADM for usage in production code.
   static rtc::scoped_refptr<AudioDeviceModule> Create(
       AudioLayer audio_layer,
-      TaskQueueFactory* task_queue_factory);
+      TaskQueueFactory* task_queue_factory,
+      bool bypass_voice_processing = false);
   // Creates an ADM with support for extra test methods. Don't use this factory
   // in production code.
   static rtc::scoped_refptr<AudioDeviceModuleForTest> CreateForTest(
       AudioLayer audio_layer,
-      TaskQueueFactory* task_queue_factory);
+      TaskQueueFactory* task_queue_factory,
+      bool bypass_voice_processing = false);
 
   // Retrieve the currently utilized audio layer
   virtual int32_t ActiveAudioLayer(AudioLayer* audioLayer) const = 0;
@@ -171,6 +182,10 @@ class AudioDeviceModule : public rtc::RefCountInterface {
   virtual int GetRecordAudioParameters(AudioParameters* params) const = 0;
 #endif  // WEBRTC_IOS
 
+  virtual int32_t SetAudioDeviceSink(AudioDeviceSink* sink) const { return -1; }
+  virtual int32_t GetPlayoutDevice() const { return -1; }
+  virtual int32_t GetRecordingDevice() const { return -1; }
+
  protected:
   ~AudioDeviceModule() override {}
 };