Fix pitch correction latency

apps/common-app/src/examples/AudioVisualizer/AudioVisualizer.tsx

@@ -131,16 +131,14 @@ const AudioVisualizer: React.FC = () => {
         }
       />
       <View
-        style={{ flex: 0.5, justifyContent: 'center', alignItems: 'center' }}
-      >
+        style={{ flex: 0.5, justifyContent: 'center', alignItems: 'center' }}>
         {isLoading && <ActivityIndicator color="#FFFFFF" />}
         <View
           style={{
             justifyContent: 'center',
             flexDirection: 'row',
             marginTop: layout.spacing * 2,
-          }}
-        >
+          }}>
           <Button
             onPress={handlePlayPause}
             title={isPlaying ? 'Pause' : 'Play'}

packages/audiodocs/docs/sources/audio-buffer-queue-source-node.mdx

@@ -39,8 +39,8 @@ function App() {
     const audioBufferQueue = audioContextRef.current.createBufferQueueSource();
     const buffer1 = ...; // Load your audio buffer here
     const buffer2 = ...; // Load another audio buffer if needed
-    audioBufferQueue.enqueueBuffer(buffer1, false);
-    audioBufferQueue.enqueueBuffer(buffer2, true); // Last buffer should be marked as is
+    audioBufferQueue.enqueueBuffer(buffer1);
+    audioBufferQueue.enqueueBuffer(buffer2);
     audioBufferQueue.connect(audioContextRef.current.destination);
     audioBufferQueue.start(audioContextRef.current.currentTime);
 }

packages/audiodocs/docs/sources/audio-buffer-source-node.mdx

@@ -12,7 +12,6 @@ import { useGainAdsrPlayground } from '@site/src/components/InteractivePlaygroun
 
 # AudioBufferSourceNode
 
-The `AudioBufferSourceNode` is an [`AudioBufferBaseSourceNode`](/docs/sources/audio-buffer-base-source-node) which represents audio source with in-memory audio data, stored in
 The `AudioBufferSourceNode` is an [`AudioBufferBaseSourceNode`](/docs/sources/audio-buffer-base-source-node) which represents audio source with in-memory audio data, stored in
 [`AudioBuffer`](/docs/sources/audio-buffer). You can use it for audio playback, including standard pause and resume functionalities.
 
@@ -40,6 +39,12 @@ interface AudioBufferBaseSourceNodeOptions {
 }
 ```
 
+:::caution
+The pitch correction algorithm introduces processing latency.
+As a result, when scheduling precise playback times, you should start input samples slightly ahead of the intended playback time.
+For more details, see [getLatency()](/docs/sources/audio-buffer-source-node#getlatency).
+:::
+
 ## Example
 
 ```tsx
@@ -100,6 +105,34 @@ Schedules the `AudioBufferSourceNode` to start playback of audio data contained
 
 #### Returns `undefined`.
 
+
+### `getLatency`
+
+Returns the playback latency introduced by the pitch correction algorithm, in seconds.
+When scheduling precise playback times, start input samples this many seconds earlier to compensate for processing delay.
+Typically around `0.06s` when pitch correction is enabled, and `0` otherwise.
+
+If you plan to play multiple buffers one after another, consider using [`AudioBufferQueueSourceNode`](/docs/sources/audio-buffer-queue-source-node)
+
+
+#### Returns `number`.
+
+<details>
+<summary>Example usage</summary>
+```tsx
+const source = audioContext.createBufferSource({ pitchCorrection: true });
+source.buffer = buffer;
+source.connect(audioContext.destination);
+
+const latency = source.getLatency();
+
+// Schedule playback slightly earlier to compensate for latency
+const startTime = audioContext.currentTime + 1.0; // play in 1 second
+source.start(startTime - latency);
+```
+</details>
+
+
 ## Events
 
 ### `onLoopEnded`

packages/react-native-audio-api/common/cpp/audioapi/HostObjects/sources/AudioBufferSourceNodeHostObject.cpp

@@ -27,7 +27,9 @@ AudioBufferSourceNodeHostObject::AudioBufferSourceNodeHostObject(
 
   addFunctions(
       JSI_EXPORT_FUNCTION(AudioBufferSourceNodeHostObject, start),
-      JSI_EXPORT_FUNCTION(AudioBufferSourceNodeHostObject, setBuffer));
+      JSI_EXPORT_FUNCTION(AudioBufferSourceNodeHostObject, setBuffer),
+      JSI_EXPORT_FUNCTION(AudioBufferSourceNodeHostObject, getInputLatency),
+      JSI_EXPORT_FUNCTION(AudioBufferSourceNodeHostObject, getOutputLatency));
 }
 
 AudioBufferSourceNodeHostObject::~AudioBufferSourceNodeHostObject() {
@@ -149,4 +151,18 @@ JSI_HOST_FUNCTION_IMPL(AudioBufferSourceNodeHostObject, setBuffer) {
   return jsi::Value::undefined();
 }
 
+JSI_HOST_FUNCTION_IMPL(AudioBufferSourceNodeHostObject, getInputLatency) {
+  auto audioBufferSourceNode =
+      std::static_pointer_cast<AudioBufferSourceNode>(node_);
+
+  return audioBufferSourceNode->getInputLatency();
+}
+
+JSI_HOST_FUNCTION_IMPL(AudioBufferSourceNodeHostObject, getOutputLatency) {
+  auto audioBufferSourceNode =
+      std::static_pointer_cast<AudioBufferSourceNode>(node_);
+
+  return audioBufferSourceNode->getOutputLatency();
+}
+
 } // namespace audioapi

packages/react-native-audio-api/common/cpp/audioapi/HostObjects/sources/AudioBufferSourceNodeHostObject.h

@@ -32,6 +32,8 @@ class AudioBufferSourceNodeHostObject
 
   JSI_HOST_FUNCTION_DECL(start);
   JSI_HOST_FUNCTION_DECL(setBuffer);
+  JSI_HOST_FUNCTION_DECL(getInputLatency);
+  JSI_HOST_FUNCTION_DECL(getOutputLatency);
 };
 
 } // namespace audioapi

packages/react-native-audio-api/common/cpp/audioapi/core/sources/AudioBufferQueueSourceNode.cpp

@@ -17,6 +17,12 @@ AudioBufferQueueSourceNode::AudioBufferQueueSourceNode(
   buffers_ = {};
   stretch_->presetDefault(channelCount_, context_->getSampleRate());
 
+  if (pitchCorrection) {
+    // If pitch correction is enabled, add extra frames at the end
+    // to compensate for processing latency.
+    addExtraTailFrames_ = true;
+  }
+
   isInitialized_ = true;
 }
 
@@ -161,7 +167,19 @@ void AudioBufferQueueSourceNode::processWithoutInterpolation(
       context_->audioEventHandlerRegistry_->invokeHandlerWithEventBody(
           "ended", onEndedCallbackId_, body);
 
-      if (buffers_.empty()) {
+      if (buffers_.empty() && addExtraTailFrames_) {
+        int extraTailFrames = static_cast<int>(
+            stretch_->inputLatency() + stretch_->outputLatency());
+        auto tailBuffer = std::make_shared<AudioBuffer>(
+            channelCount_, extraTailFrames, context_->getSampleRate());
+
+        tailBuffer->bus_->zero();
+
+        buffers_.emplace(bufferId_, tailBuffer);
+        bufferId_++;
+
+        addExtraTailFrames_ = false;
+      } else if (buffers_.empty()) {
         processingBus->zero(writeIndex, framesLeft);
         readIndex = 0;
 

packages/react-native-audio-api/common/cpp/audioapi/core/sources/AudioBufferQueueSourceNode.h

@@ -38,6 +38,7 @@ class AudioBufferQueueSourceNode : public AudioBufferBaseSourceNode {
     size_t bufferId_ = 0;
 
     bool isPaused_ = false;
+    bool addExtraTailFrames_ = false;
 
     double playedBuffersDuration_ = 0;
 

packages/react-native-audio-api/common/cpp/audioapi/core/sources/AudioBufferSourceNode.cpp

@@ -51,6 +51,22 @@ std::shared_ptr<AudioBuffer> AudioBufferSourceNode::getBuffer() const {
   return buffer_;
 }
 
+double AudioBufferSourceNode::getInputLatency() const {
+  if (pitchCorrection_) {
+    return static_cast<double>(stretch_->inputLatency()) /
+        context_->getSampleRate();
+  }
+  return 0;
+}
+
+double AudioBufferSourceNode::getOutputLatency() const {
+  if (pitchCorrection_) {
+    return static_cast<double>(stretch_->outputLatency()) /
+        context_->getSampleRate();
+  }
+  return 0;
+}
+
 void AudioBufferSourceNode::setLoop(bool loop) {
   loop_ = loop;
 }
@@ -82,17 +98,30 @@ void AudioBufferSourceNode::setBuffer(
   }
 
   buffer_ = buffer;
-  alignedBus_ = std::make_shared<AudioBus>(*buffer_->bus_);
   channelCount_ = buffer_->getNumberOfChannels();
 
+  stretch_->presetDefault(channelCount_, buffer_->getSampleRate());
+
+  if (pitchCorrection_) {
+    int extraTailFrames = static_cast<int>(
+        (getInputLatency() + getOutputLatency()) * context_->getSampleRate());
+    size_t totalSize = buffer_->getLength() + extraTailFrames;
+
+    alignedBus_ = std::make_shared<AudioBus>(
+        totalSize, channelCount_, buffer_->getSampleRate());
+    alignedBus_->copy(buffer_->bus_.get(), 0, 0, buffer_->getLength());
+
+    alignedBus_->zero(buffer_->getLength(), extraTailFrames);
+  } else {
+    alignedBus_ = std::make_shared<AudioBus>(*buffer_->bus_);
+  }
+
   audioBus_ = std::make_shared<AudioBus>(
       RENDER_QUANTUM_SIZE, channelCount_, context_->getSampleRate());
   playbackRateBus_ = std::make_shared<AudioBus>(
       RENDER_QUANTUM_SIZE * 3, channelCount_, context_->getSampleRate());
 
   loopEnd_ = buffer_->getDuration();
-
-  stretch_->presetDefault(channelCount_, buffer_->getSampleRate());
 }
 
 void AudioBufferSourceNode::start(double when, double offset, double duration) {

packages/react-native-audio-api/common/cpp/audioapi/core/sources/AudioBufferSourceNode.h

@@ -4,9 +4,9 @@
 #include <audioapi/core/sources/AudioBufferBaseSourceNode.h>
 #include <audioapi/libs/signalsmith-stretch/signalsmith-stretch.h>
 
-#include <memory>
-#include <cstddef>
 #include <algorithm>
+#include <cstddef>
+#include <memory>
 #include <string>
 
 namespace audioapi {
@@ -24,6 +24,8 @@ class AudioBufferSourceNode : public AudioBufferBaseSourceNode {
   [[nodiscard]] double getLoopStart() const;
   [[nodiscard]] double getLoopEnd() const;
   [[nodiscard]] std::shared_ptr<AudioBuffer> getBuffer() const;
+  [[nodiscard]] double getInputLatency() const;
+  [[nodiscard]] double getOutputLatency() const;
 
   void setLoop(bool loop);
   void setLoopSkip(bool loopSkip);
@@ -38,7 +40,7 @@ class AudioBufferSourceNode : public AudioBufferBaseSourceNode {
   void setOnLoopEndedCallbackId(uint64_t callbackId);
 
  protected:
-  std::shared_ptr<AudioBus> processNode(const std::shared_ptr<AudioBus>& processingBus, int framesToProcess) override;
+  std::shared_ptr<AudioBus> processNode(const std::shared_ptr<AudioBus> &processingBus, int framesToProcess) override;
   double getCurrentPosition() const override;
 
  private:
@@ -56,13 +58,13 @@ class AudioBufferSourceNode : public AudioBufferBaseSourceNode {
   void sendOnLoopEndedEvent();
 
   void processWithoutInterpolation(
-      const std::shared_ptr<AudioBus>& processingBus,
+      const std::shared_ptr<AudioBus> &processingBus,
       size_t startOffset,
       size_t offsetLength,
       float playbackRate) override;
 
   void processWithInterpolation(
-      const std::shared_ptr<AudioBus>& processingBus,
+      const std::shared_ptr<AudioBus> &processingBus,
       size_t startOffset,
       size_t offsetLength,
       float playbackRate) override;

packages/react-native-audio-api/src/core/AudioBufferSourceNode.ts

@@ -118,4 +118,12 @@ export default class AudioBufferSourceNode extends AudioBufferBaseSourceNode {
     (this.node as IAudioBufferSourceNode).onLoopEnded =
       this.onLoopEndedSubscription.subscriptionId;
   }
+
+  public getLatency(): number {
+    return (
+      (this.node as IAudioBufferSourceNode).getOutputLatency() +
+      (this.node as IAudioBufferSourceNode).getInputLatency() *
+        (this.node as IAudioBufferSourceNode).playbackRate.value
+    );
+  }
 }

packages/react-native-audio-api/src/interfaces.ts

@@ -175,6 +175,8 @@ export interface IAudioBufferSourceNode extends IAudioBufferBaseSourceNode {
 
   start: (when?: number, offset?: number, duration?: number) => void;
   setBuffer: (audioBuffer: IAudioBuffer | null) => void;
+  getInputLatency: () => number;
+  getOutputLatency: () => number;
 
   // passing subscriptionId(uint_64 in cpp, string in js) to the cpp
   onLoopEnded: string;