Merge pull request #15534 from kencooke/audio-PTT-output-ducking

Reduce echo when push-to-talk is used

interface/src/scripting/Audio.cpp

@@ -366,8 +366,10 @@ void Audio::onContextChanged() {
 void Audio::handlePushedToTalk(bool enabled) {
     if (getPTT()) {
         if (enabled) {
+            DependencyManager::get<AudioClient>()->setOutputGain(0.1f); // duck the output by 20dB
             setMuted(false);
         } else {
+            DependencyManager::get<AudioClient>()->setOutputGain(1.0f);
             setMuted(true);
         }
     }

libraries/audio-client/src/AudioClient.cpp

@@ -221,6 +221,35 @@ static float computeLoudness(int16_t* samples, int numSamples, int numChannels,
     return (float)loudness * scale;
 }
 
+template <int NUM_CHANNELS>
+static void applyGainSmoothing(float* buffer, int numFrames, float gain0, float gain1) {
+
+    // fast path for unity gain
+    if (gain0 == 1.0f && gain1 == 1.0f) {
+        return;
+    }
+
+    // cubic poly from gain0 to gain1
+    float c3 = -2.0f * (gain1 - gain0);
+    float c2 = 3.0f * (gain1 - gain0);
+    float c0 = gain0;
+
+    float t = 0.0f;
+    float tStep = 1.0f / numFrames;
+
+    for (int i = 0; i < numFrames; i++) {
+
+        // evaluate poly over t=[0,1)
+        float gain = (c3 * t + c2) * t * t + c0;
+        t += tStep;
+
+        // apply gain to all channels
+        for (int ch = 0; ch < NUM_CHANNELS; ch++) {
+            buffer[NUM_CHANNELS*i + ch] *= gain;
+        }
+    }
+}
+
 static inline float convertToFloat(int16_t sample) {
     return (float)sample * (1 / 32768.0f);
 }
@@ -2109,6 +2138,14 @@ qint64 AudioClient::AudioOutputIODevice::readData(char * data, qint64 maxSize) {
     int framesPopped = samplesPopped / AudioConstants::STEREO;
     int bytesWritten;
     if (samplesPopped > 0) {
+
+        // apply output gain
+        float newGain = _audio->_outputGain.load(std::memory_order_acquire);
+        float oldGain = _audio->_lastOutputGain;
+        _audio->_lastOutputGain = newGain;
+
+        applyGainSmoothing<OUTPUT_CHANNEL_COUNT>(mixBuffer, framesPopped, oldGain, newGain);
+
         if (deviceChannelCount == OUTPUT_CHANNEL_COUNT) {
             // limit the audio
             _audio->_audioLimiter.render(mixBuffer, (int16_t*)data, framesPopped);

libraries/audio-client/src/AudioClient.h

@@ -241,8 +241,10 @@ public slots:
     void setInputVolume(float volume, bool emitSignal = true);
     void setReverb(bool reverb);
     void setReverbOptions(const AudioEffectOptions* options);
+
     void setLocalInjectorGain(float gain) { _localInjectorGain = gain; };
     void setSystemInjectorGain(float gain) { _systemInjectorGain = gain; };
+    void setOutputGain(float gain) { _outputGain = gain; };
 
     void outputNotify();
 
@@ -395,6 +397,8 @@ private:
     int _outputPeriod { 0 };
     float* _outputMixBuffer { NULL };
     int16_t* _outputScratchBuffer { NULL };
+    std::atomic<float> _outputGain { 1.0f };
+    float _lastOutputGain { 1.0f };
 
     // for local audio (used by audio injectors thread)
     std::atomic<float> _localInjectorGain { 1.0f };