Add node-addon-api for VAD (#864)

nodejs-addon-examples/.gitignore

@@ -0,0 +1 @@
+crash.log

nodejs-addon-examples/README.md

@@ -38,3 +38,12 @@ node ./test_asr_streaming_transducer.js
 
 node ./test_asr_streaming_transducer_microphone.js
 ```
+
+# VAD
+
+```bash
+wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
+
+node ./test_vad_microphone.js
+```
+

nodejs-addon-examples/test_asr_streaming_transducer.js

@@ -36,10 +36,10 @@ console.log('Started')
 let start = performance.now();
 const stream = recognizer.createStream();
 const wave = sherpa_onnx.readWave(waveFilename);
-stream.acceptWaveform(wave.samples, wave.sampleRate);
+stream.acceptWaveform({sampleRate: wave.sampleRate, samples: wave.samples});
 
 const tailPadding = new Float32Array(wave.sampleRate * 0.4);
-stream.acceptWaveform(tailPadding, wave.sampleRate);
+stream.acceptWaveform({samples: tailPadding, sampleRate: wave.sampleRate});
 
 while (recognizer.isReady(stream)) {
   recognizer.decode(stream);

nodejs-addon-examples/test_asr_streaming_transducer_microphone.js

@@ -60,7 +60,8 @@ const display = new sherpa_onnx.Display(50);
 ai.on('data', data => {
   const samples = new Float32Array(data.buffer);
 
-  stream.acceptWaveform(samples, recognizer.config.featConfig.sampleRate);
+  stream.acceptWaveform(
+      {sampleRate: recognizer.config.featConfig.sampleRate, samples: samples});
 
   while (recognizer.isReady(stream)) {
     recognizer.decode(stream);

nodejs-addon-examples/test_vad_microphone.js

@@ -0,0 +1,88 @@
+// Copyright (c)  2023-2024  Xiaomi Corporation (authors: Fangjun Kuang)
+
+const portAudio = require('naudiodon2');
+// console.log(portAudio.getDevices());
+
+const sherpa_onnx = require('sherpa-onnx-node');
+
+function createVad() {
+  // please download silero_vad.onnx from
+  // https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
+  const config = {
+    sileroVad: {
+      model: './silero_vad.onnx',
+      threshold: 0.5,
+      minSpeechDuration: 0.25,
+      minSilenceDuration: 0.5,
+      windowSize: 512,
+    },
+    sampleRate: 16000,
+    debug: true,
+    numThreads: 1,
+  };
+
+  const bufferSizeInSeconds = 60;
+
+  return new sherpa_onnx.Vad(config, bufferSizeInSeconds);
+}
+
+vad = createVad();
+
+const bufferSizeInSeconds = 30;
+const buffer =
+    new sherpa_onnx.CircularBuffer(bufferSizeInSeconds * vad.config.sampleRate);
+
+
+const ai = new portAudio.AudioIO({
+  inOptions: {
+    channelCount: 1,
+    closeOnError: true,  // Close the stream if an audio error is detected, if
+                         // set false then just log the error
+    deviceId: -1,  // Use -1 or omit the deviceId to select the default device
+    sampleFormat: portAudio.SampleFormatFloat32,
+    sampleRate: vad.config.sampleRate,
+  }
+});
+
+let printed = false;
+let index = 0;
+ai.on('data', data => {
+  const windowSize = vad.config.sileroVad.windowSize;
+  buffer.push(new Float32Array(data.buffer));
+  while (buffer.size() > windowSize) {
+    const samples = buffer.get(buffer.head(), windowSize);
+    buffer.pop(windowSize);
+    vad.acceptWaveform(samples)
+    if (vad.isDetected() && !printed) {
+      console.log(`${index}: Detected speech`)
+      printed = true;
+    }
+
+    if (!vad.isDetected()) {
+      printed = false;
+    }
+
+    while (!vad.isEmpty()) {
+      const segment = vad.front();
+      vad.pop();
+      const filename = `${index}-${
+          new Date()
+              .toLocaleTimeString('en-US', {hour12: false})
+              .split(' ')[0]}.wav`;
+      sherpa_onnx.writeWave(
+          filename,
+          {samples: segment.samples, sampleRate: vad.config.sampleRate})
+      const duration = segment.samples.length / vad.config.sampleRate;
+      console.log(`${index} End of speech. Duration: ${duration} seconds`);
+      console.log(`Saved to ${filename}`);
+      index += 1;
+    }
+  }
+});
+
+ai.on('close', () => {
+  console.log('Free resources');
+});
+
+ai.start();
+console.log('Started! Please speak')

python-api-examples/offline-tts-play.py

@@ -47,7 +47,7 @@ python3 ./python-api-examples/offline-tts-play.py \
  --vits-model=./sherpa-onnx-vits-zh-ll/model.onnx \
  --vits-lexicon=./sherpa-onnx-vits-zh-ll/lexicon.txt \
  --vits-tokens=./sherpa-onnx-vits-zh-ll/tokens.txt \
- --tts-rule-fsts='./sherpa-onnx-vits-zh-ll/phone.fst,./sherpa-onnx-vits-zh-ll/date.fst,./sherpa-onnx-vits-zh-ll/number.fst' \
+ --tts-rule-fsts=./sherpa-onnx-vits-zh-ll/phone.fst,./sherpa-onnx-vits-zh-ll/date.fst,./sherpa-onnx-vits-zh-ll/number.fst \
  --vits-dict-dir=./sherpa-onnx-vits-zh-ll/dict \
  --sid=2 \
  --output-filename=./test-2.wav \

python-api-examples/offline-tts.py

@@ -48,7 +48,7 @@ python3 ./python-api-examples/offline-tts.py \
  --vits-model=./sherpa-onnx-vits-zh-ll/model.onnx \
  --vits-lexicon=./sherpa-onnx-vits-zh-ll/lexicon.txt \
  --vits-tokens=./sherpa-onnx-vits-zh-ll/tokens.txt \
- --tts-rule-fsts='./sherpa-onnx-vits-zh-ll/phone.fst,./sherpa-onnx-vits-zh-ll/date.fst,./sherpa-onnx-vits-zh-ll/number.fst' \
+ --tts-rule-fsts=./sherpa-onnx-vits-zh-ll/phone.fst,./sherpa-onnx-vits-zh-ll/date.fst,./sherpa-onnx-vits-zh-ll/number.fst \
  --vits-dict-dir=./sherpa-onnx-vits-zh-ll/dict \
  --sid=2 \
  --output-filename=./test-2.wav \

scripts/node-addon-api/CMakeLists.txt

@@ -20,7 +20,9 @@ include_directories(${CMAKE_JS_INC})
 set(srcs
   src/sherpa-onnx-node-addon-api.cc
   src/streaming-asr.cc
+  src/vad.cc
   src/wave-reader.cc
+  src/wave-writer.cc
 )
 
 if(NOT DEFINED ENV{SHERPA_ONNX_INSTALL_DIR})

scripts/node-addon-api/lib/sherpa-onnx.js

@@ -1,8 +1,12 @@
 const addon = require('./addon.js')
 const streaming_asr = require('./streaming-asr.js');
+const vad = require('./vad.js');
 
 module.exports = {
   OnlineRecognizer: streaming_asr.OnlineRecognizer,
   readWave: addon.readWave,
+  writeWave: addon.writeWave,
   Display: streaming_asr.Display,
+  Vad: vad.Vad,
+  CircularBuffer: vad.CircularBuffer,
 }

scripts/node-addon-api/lib/streaming-asr.js

@@ -15,10 +15,11 @@ class OnlineStream {
     this.handle = handle;
   }
 
+  // obj is {samples: samples, sampleRate: sampleRate}
   // samples is a float32 array containing samples in the range [-1, 1]
-  acceptWaveform(samples, sampleRate) {
+  // sampleRate is a number
-    addon.acceptWaveformOnline(
+  acceptWaveform(obj) {
-        this.handle, {samples: samples, sampleRate: sampleRate})
+    addon.acceptWaveformOnline(this.handle, obj)
   }
 
   inputFinished() {

scripts/node-addon-api/lib/vad.js

@@ -0,0 +1,88 @@
+const addon = require('./addon.js');
+
+class CircularBuffer {
+  constructor(capacity) {
+    this.handle = addon.createCircularBuffer(capacity);
+  }
+
+  // samples is a float32 array
+  push(samples) {
+    addon.circularBufferPush(this.handle, samples);
+  }
+
+  // return a float32 array
+  get(startIndex, n) {
+    return addon.circularBufferGet(this.handle, startIndex, n);
+  }
+
+  pop(n) {
+    return addon.circularBufferPop(this.handle, n);
+  }
+
+  size() {
+    return addon.circularBufferSize(this.handle);
+  }
+
+  head() {
+    return addon.circularBufferHead(this.handle);
+  }
+
+  reset() {
+    return addon.circularBufferReset(this.handle);
+  }
+}
+
+class Vad {
+  /*
+config = {
+  sileroVad: {
+    model: "./silero_vad.onnx",
+    threshold: 0.5,
+  }
+}
+   */
+  constructor(config, bufferSizeInSeconds) {
+    this.handle =
+        addon.createVoiceActivityDetector(config, bufferSizeInSeconds);
+    this.config = config;
+  }
+
+  acceptWaveform(samples) {
+    addon.voiceActivityDetectorAcceptWaveform(this.handle, samples)
+  }
+
+  isEmpty() {
+    return addon.voiceActivityDetectorIsEmpty(this.handle)
+  }
+
+  isDetected() {
+    return addon.voiceActivityDetectorIsDetected(this.handle)
+  }
+
+  pop() {
+    addon.voiceActivityDetectorPop(this.handle)
+  }
+
+  clear() {
+    addon.VoiceActivityDetectorClearWrapper(this.handle)
+  }
+
+  /*
+{
+  samples: a 1-d float32 array,
+  start: a int32
+}
+   */
+  front() {
+    return addon.voiceActivityDetectorFront(this.handle)
+  }
+
+  reset() {
+    return addon.VoiceActivityDetectorResetWrapper(this.handle)
+  }
+}
+
+module.exports = {
+  Vad,
+  CircularBuffer,
+}

scripts/node-addon-api/src/sherpa-onnx-node-addon-api.cc

@@ -5,10 +5,14 @@
 
 void InitStreamingAsr(Napi::Env env, Napi::Object exports);
 void InitWaveReader(Napi::Env env, Napi::Object exports);
+void InitWaveWriter(Napi::Env env, Napi::Object exports);
+void InitVad(Napi::Env env, Napi::Object exports);
 
 Napi::Object Init(Napi::Env env, Napi::Object exports) {
   InitStreamingAsr(env, exports);
   InitWaveReader(env, exports);
+  InitWaveWriter(env, exports);
+  InitVad(env, exports);
 
   return exports;
 }

scripts/node-addon-api/src/streaming-asr.cc

@@ -125,8 +125,13 @@ static SherpaOnnxOnlineModelConfig GetOnlineModelConfig(Napi::Object obj) {
     config.provider = p;
   }
 
-  if (o.Has("debug") && o.Get("debug").IsNumber()) {
+  if (o.Has("debug") &&
-    config.debug = o.Get("debug").As<Napi::Number>().Int32Value();
+      (o.Get("debug").IsNumber() || o.Get("debug").IsBoolean())) {
+    if (o.Get("debug").IsBoolean()) {
+      config.debug = o.Get("debug").As<Napi::Boolean>().Value();
+    } else {
+      config.debug = o.Get("debug").As<Napi::Number>().Int32Value();
+    }
   }
 
   if (o.Has("modelType") && o.Get("modelType").IsString()) {

scripts/node-addon-api/src/vad.cc

@@ -0,0 +1,620 @@
+// scripts/node-addon-api/src/vad.cc
+//
+// Copyright (c)  2024  Xiaomi Corporation
+
+#include <sstream>
+
+#include "napi.h"  // NOLINT
+#include "sherpa-onnx/c-api/c-api.h"
+
+static Napi::External<SherpaOnnxCircularBuffer> CreateCircularBufferWrapper(
+    const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+  if (info.Length() != 1) {
+    std::ostringstream os;
+    os << "Expect only 1 argument. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[0].IsNumber()) {
+    Napi::TypeError::New(env, "You should pass an integer as the argument.")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  SherpaOnnxCircularBuffer *buf =
+      SherpaOnnxCreateCircularBuffer(info[0].As<Napi::Number>().Int32Value());
+
+  return Napi::External<SherpaOnnxCircularBuffer>::New(
+      env, buf, [](Napi::Env env, SherpaOnnxCircularBuffer *p) {
+        SherpaOnnxDestroyCircularBuffer(p);
+      });
+}
+
+static void CircularBufferPushWrapper(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 2) {
+    std::ostringstream os;
+    os << "Expect only 2 arguments. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be an CircularBuffer pointer.")
+        .ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  SherpaOnnxCircularBuffer *buf =
+      info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
+
+  if (!info[1].IsTypedArray()) {
+    Napi::TypeError::New(env, "Argument 1 should be a Float32Array.")
+        .ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  Napi::Float32Array data = info[1].As<Napi::Float32Array>();
+  SherpaOnnxCircularBufferPush(buf, data.Data(), data.ElementLength());
+}
+
+// see https://github.com/nodejs/node-addon-api/blob/main/doc/typed_array.md
+// https://github.com/nodejs/node-addon-examples/blob/main/src/2-js-to-native-conversion/typed_array_to_native/node-addon-api/typed_array_to_native.cc
+static Napi::Float32Array CircularBufferGetWrapper(
+    const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 3) {
+    std::ostringstream os;
+    os << "Expect only 3 arguments. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be an CircularBuffer pointer.")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  SherpaOnnxCircularBuffer *buf =
+      info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
+
+  if (!info[1].IsNumber()) {
+    Napi::TypeError::New(env, "Argument 1 should be an integer (startIndex).")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[2].IsNumber()) {
+    Napi::TypeError::New(env, "Argument 2 should be an integer (n).")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  int32_t start_index = info[1].As<Napi::Number>().Int32Value();
+  int32_t n = info[2].As<Napi::Number>().Int32Value();
+
+  const float *data = SherpaOnnxCircularBufferGet(buf, start_index, n);
+
+  Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New(
+      env, const_cast<float *>(data), sizeof(float) * n,
+      [](Napi::Env /*env*/, void *p) {
+        SherpaOnnxCircularBufferFree(reinterpret_cast<const float *>(p));
+      });
+
+  Napi::Float32Array float32Array =
+      Napi::Float32Array::New(env, n, arrayBuffer, 0);
+
+  return float32Array;
+}
+
+static void CircularBufferPopWrapper(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 2) {
+    std::ostringstream os;
+    os << "Expect only 2 arguments. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be an CircularBuffer pointer.")
+        .ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  SherpaOnnxCircularBuffer *buf =
+      info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
+
+  if (!info[1].IsNumber()) {
+    Napi::TypeError::New(env, "Argument 1 should be an integer (n).")
+        .ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  int32_t n = info[1].As<Napi::Number>().Int32Value();
+
+  SherpaOnnxCircularBufferPop(buf, n);
+}
+
+static Napi::Number CircularBufferSizeWrapper(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 1) {
+    std::ostringstream os;
+    os << "Expect only 1 argument. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be an CircularBuffer pointer.")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  SherpaOnnxCircularBuffer *buf =
+      info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
+
+  int32_t size = SherpaOnnxCircularBufferSize(buf);
+
+  return Napi::Number::New(env, size);
+}
+
+static Napi::Number CircularBufferHeadWrapper(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 1) {
+    std::ostringstream os;
+    os << "Expect only 1 argument. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be an CircularBuffer pointer.")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  SherpaOnnxCircularBuffer *buf =
+      info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
+
+  int32_t size = SherpaOnnxCircularBufferHead(buf);
+
+  return Napi::Number::New(env, size);
+}
+
+static void CircularBufferResetWrapper(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 1) {
+    std::ostringstream os;
+    os << "Expect only 1 argument. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be an CircularBuffer pointer.")
+        .ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  SherpaOnnxCircularBuffer *buf =
+      info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
+
+  SherpaOnnxCircularBufferReset(buf);
+}
+
+static SherpaOnnxSileroVadModelConfig GetSileroVadConfig(
+    const Napi::Object &obj) {
+  SherpaOnnxSileroVadModelConfig c;
+  memset(&c, 0, sizeof(c));
+
+  if (!obj.Has("sileroVad") || !obj.Get("sileroVad").IsObject()) {
+    return c;
+  }
+
+  Napi::Object o = obj.Get("sileroVad").As<Napi::Object>();
+
+  if (o.Has("model") && o.Get("model").IsString()) {
+    Napi::String model = o.Get("model").As<Napi::String>();
+    std::string s = model.Utf8Value();
+    char *p = new char[s.size() + 1];
+    std::copy(s.begin(), s.end(), p);
+    p[s.size()] = 0;
+
+    c.model = p;
+  }
+
+  if (o.Has("threshold") && o.Get("threshold").IsNumber()) {
+    c.threshold = o.Get("threshold").As<Napi::Number>().FloatValue();
+  }
+
+  if (o.Has("minSilenceDuration") && o.Get("minSilenceDuration").IsNumber()) {
+    c.min_silence_duration =
+        o.Get("minSilenceDuration").As<Napi::Number>().FloatValue();
+  }
+
+  if (o.Has("minSpeechDuration") && o.Get("minSpeechDuration").IsNumber()) {
+    c.min_speech_duration =
+        o.Get("minSpeechDuration").As<Napi::Number>().FloatValue();
+  }
+
+  if (o.Has("windowSize") && o.Get("windowSize").IsNumber()) {
+    c.window_size = o.Get("windowSize").As<Napi::Number>().Int32Value();
+  }
+
+  return c;
+}
+
+static Napi::External<SherpaOnnxVoiceActivityDetector>
+CreateVoiceActivityDetectorWrapper(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+  if (info.Length() != 2) {
+    std::ostringstream os;
+    os << "Expect only 2 arguments. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[0].IsObject()) {
+    Napi::TypeError::New(env,
+                         "You should pass an object as the first argument.")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[1].IsNumber()) {
+    Napi::TypeError::New(env,
+                         "You should pass an integer as the second argument.")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  Napi::Object o = info[0].As<Napi::Object>();
+
+  SherpaOnnxVadModelConfig c;
+  memset(&c, 0, sizeof(c));
+  c.silero_vad = GetSileroVadConfig(o);
+
+  if (o.Has("sampleRate") && o.Get("sampleRate").IsNumber()) {
+    c.sample_rate = o.Get("sampleRate").As<Napi::Number>().Int32Value();
+  }
+
+  if (o.Has("numThreads") && o.Get("numThreads").IsNumber()) {
+    c.num_threads = o.Get("numThreads").As<Napi::Number>().Int32Value();
+  }
+
+  if (o.Has("provider") && o.Get("provider").IsString()) {
+    Napi::String provider = o.Get("provider").As<Napi::String>();
+    std::string s = provider.Utf8Value();
+    char *p = new char[s.size() + 1];
+    std::copy(s.begin(), s.end(), p);
+    p[s.size()] = 0;
+
+    c.provider = p;
+  }
+
+  if (o.Has("debug") &&
+      (o.Get("debug").IsNumber() || o.Get("debug").IsBoolean())) {
+    if (o.Get("debug").IsBoolean()) {
+      c.debug = o.Get("debug").As<Napi::Boolean>().Value();
+    } else {
+      c.debug = o.Get("debug").As<Napi::Number>().Int32Value();
+    }
+  }
+
+  float buffer_size_in_seconds = info[1].As<Napi::Number>().FloatValue();
+
+  SherpaOnnxVoiceActivityDetector *vad =
+      SherpaOnnxCreateVoiceActivityDetector(&c, buffer_size_in_seconds);
+
+  if (c.silero_vad.model) {
+    delete[] c.silero_vad.model;
+  }
+
+  if (c.provider) {
+    delete[] c.provider;
+  }
+
+  return Napi::External<SherpaOnnxVoiceActivityDetector>::New(
+      env, vad, [](Napi::Env env, SherpaOnnxVoiceActivityDetector *p) {
+        SherpaOnnxDestroyVoiceActivityDetector(p);
+      });
+}
+
+static void VoiceActivityDetectorAcceptWaveformWrapper(
+    const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 2) {
+    std::ostringstream os;
+    os << "Expect only 2 arguments. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be a VAD pointer.")
+        .ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  SherpaOnnxVoiceActivityDetector *vad =
+      info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
+
+  if (!info[1].IsTypedArray()) {
+    Napi::TypeError::New(
+        env, "Argument 1 should be a Float32Array containing samples")
+        .ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  Napi::Float32Array samples = info[1].As<Napi::Float32Array>();
+
+  SherpaOnnxVoiceActivityDetectorAcceptWaveform(vad, samples.Data(),
+                                                samples.ElementLength());
+}
+
+static Napi::Boolean VoiceActivityDetectorEmptyWrapper(
+    const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 1) {
+    std::ostringstream os;
+    os << "Expect only 1 argument. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be a VAD pointer.")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  SherpaOnnxVoiceActivityDetector *vad =
+      info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
+
+  int32_t is_empty = SherpaOnnxVoiceActivityDetectorEmpty(vad);
+
+  return Napi::Boolean::New(env, is_empty);
+}
+
+static Napi::Boolean VoiceActivityDetectorDetectedWrapper(
+    const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 1) {
+    std::ostringstream os;
+    os << "Expect only 1 argument. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be a VAD pointer.")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  SherpaOnnxVoiceActivityDetector *vad =
+      info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
+
+  int32_t is_detected = SherpaOnnxVoiceActivityDetectorDetected(vad);
+
+  return Napi::Boolean::New(env, is_detected);
+}
+
+static void VoiceActivityDetectorPopWrapper(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 1) {
+    std::ostringstream os;
+    os << "Expect only 1 argument. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be a VAD pointer.")
+        .ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  SherpaOnnxVoiceActivityDetector *vad =
+      info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
+
+  SherpaOnnxVoiceActivityDetectorPop(vad);
+}
+
+static void VoiceActivityDetectorClearWrapper(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 1) {
+    std::ostringstream os;
+    os << "Expect only 1 argument. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be a VAD pointer.")
+        .ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  SherpaOnnxVoiceActivityDetector *vad =
+      info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
+
+  SherpaOnnxVoiceActivityDetectorClear(vad);
+}
+
+static Napi::Object VoiceActivityDetectorFrontWrapper(
+    const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 1) {
+    std::ostringstream os;
+    os << "Expect only 1 argument. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be a VAD pointer.")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  SherpaOnnxVoiceActivityDetector *vad =
+      info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
+
+  const SherpaOnnxSpeechSegment *segment =
+      SherpaOnnxVoiceActivityDetectorFront(vad);
+
+  Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New(
+      env, const_cast<float *>(segment->samples), sizeof(float) * segment->n,
+      [](Napi::Env /*env*/, void * /*data*/,
+         const SherpaOnnxSpeechSegment *hint) {
+        SherpaOnnxDestroySpeechSegment(hint);
+      },
+      segment);
+
+  Napi::Float32Array float32Array =
+      Napi::Float32Array::New(env, segment->n, arrayBuffer, 0);
+
+  Napi::Object obj = Napi::Object::New(env);
+  obj.Set(Napi::String::New(env, "start"), segment->start);
+  obj.Set(Napi::String::New(env, "samples"), float32Array);
+
+  return obj;
+}
+
+static void VoiceActivityDetectorResetWrapper(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 1) {
+    std::ostringstream os;
+    os << "Expect only 1 argument. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  if (!info[0].IsExternal()) {
+    Napi::TypeError::New(env, "Argument 0 should be a VAD pointer.")
+        .ThrowAsJavaScriptException();
+
+    return;
+  }
+
+  SherpaOnnxVoiceActivityDetector *vad =
+      info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
+
+  SherpaOnnxVoiceActivityDetectorReset(vad);
+}
+
+void InitVad(Napi::Env env, Napi::Object exports) {
+  exports.Set(Napi::String::New(env, "createCircularBuffer"),
+              Napi::Function::New(env, CreateCircularBufferWrapper));
+
+  exports.Set(Napi::String::New(env, "circularBufferPush"),
+              Napi::Function::New(env, CircularBufferPushWrapper));
+
+  exports.Set(Napi::String::New(env, "circularBufferGet"),
+              Napi::Function::New(env, CircularBufferGetWrapper));
+
+  exports.Set(Napi::String::New(env, "circularBufferPop"),
+              Napi::Function::New(env, CircularBufferPopWrapper));
+
+  exports.Set(Napi::String::New(env, "circularBufferSize"),
+              Napi::Function::New(env, CircularBufferSizeWrapper));
+
+  exports.Set(Napi::String::New(env, "circularBufferHead"),
+              Napi::Function::New(env, CircularBufferHeadWrapper));
+
+  exports.Set(Napi::String::New(env, "circularBufferReset"),
+              Napi::Function::New(env, CircularBufferResetWrapper));
+
+  exports.Set(Napi::String::New(env, "createVoiceActivityDetector"),
+              Napi::Function::New(env, CreateVoiceActivityDetectorWrapper));
+
+  exports.Set(
+      Napi::String::New(env, "voiceActivityDetectorAcceptWaveform"),
+      Napi::Function::New(env, VoiceActivityDetectorAcceptWaveformWrapper));
+
+  exports.Set(Napi::String::New(env, "voiceActivityDetectorIsEmpty"),
+              Napi::Function::New(env, VoiceActivityDetectorEmptyWrapper));
+
+  exports.Set(Napi::String::New(env, "voiceActivityDetectorIsDetected"),
+              Napi::Function::New(env, VoiceActivityDetectorDetectedWrapper));
+
+  exports.Set(Napi::String::New(env, "voiceActivityDetectorPop"),
+              Napi::Function::New(env, VoiceActivityDetectorPopWrapper));
+
+  exports.Set(Napi::String::New(env, "voiceActivityDetectorClear"),
+              Napi::Function::New(env, VoiceActivityDetectorClearWrapper));
+
+  exports.Set(Napi::String::New(env, "voiceActivityDetectorFront"),
+              Napi::Function::New(env, VoiceActivityDetectorFrontWrapper));
+
+  exports.Set(Napi::String::New(env, "voiceActivityDetectorReset"),
+              Napi::Function::New(env, VoiceActivityDetectorResetWrapper));
+}

scripts/node-addon-api/src/wave-writer.cc

@@ -0,0 +1,81 @@
+// scripts/node-addon-api/src/wave-writer.cc
+//
+// Copyright (c)  2024  Xiaomi Corporation
+
+#include <sstream>
+
+#include "napi.h"  // NOLINT
+#include "sherpa-onnx/c-api/c-api.h"
+
+// (filename, {samples: samples, sampleRate: sampleRate}
+static Napi::Boolean WriteWaveWrapper(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+
+  if (info.Length() != 2) {
+    std::ostringstream os;
+    os << "Expect only 2 argument2. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[0].IsString()) {
+    Napi::TypeError::New(env, "Argument 0 should be a string")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!info[1].IsObject()) {
+    Napi::TypeError::New(env, "Argument 1 should be an object")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  Napi::Object obj = info[1].As<Napi::Object>();
+
+  if (!obj.Has("samples")) {
+    Napi::TypeError::New(env, "The argument object should have a field samples")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!obj.Get("samples").IsTypedArray()) {
+    Napi::TypeError::New(env, "The object['samples'] should be a typed array")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!obj.Has("sampleRate")) {
+    Napi::TypeError::New(env,
+                         "The argument object should have a field sampleRate")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  if (!obj.Get("sampleRate").IsNumber()) {
+    Napi::TypeError::New(env, "The object['samples'] should be a number")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  Napi::Float32Array samples = obj.Get("samples").As<Napi::Float32Array>();
+  int32_t sample_rate = obj.Get("sampleRate").As<Napi::Number>().Int32Value();
+
+  int32_t ok =
+      SherpaOnnxWriteWave(samples.Data(), samples.ElementLength(), sample_rate,
+                          info[0].As<Napi::String>().Utf8Value().c_str());
+
+  return Napi::Boolean::New(env, ok);
+}
+
+void InitWaveWriter(Napi::Env env, Napi::Object exports) {
+  exports.Set(Napi::String::New(env, "writeWave"),
+              Napi::Function::New(env, WriteWaveWrapper));
+}