Add JavaScript (node-addon) API for speech enhancement GTCRN models (#1996)

.github/scripts/test-nodejs-addon-npm.sh

@@ -10,6 +10,15 @@ arch=$(node -p "require('os').arch()")
 platform=$(node -p "require('os').platform()")
 node_version=$(node -p "process.versions.node.split('.')[0]")
 
+echo "----------non-streaming speech denoiser----------"
+
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/speech-enhancement-models/gtcrn_simple.onnx
+curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/speech-enhancement-models/inp_16k.wav
+
+node ./test_offline_speech_enhancement_gtcrn.js
+rm gtcrn_simple.onnx
+ls -lh *.wav
+
 echo "----------non-streaming asr FireRedAsr----------"
 curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16.tar.bz2
 tar xvf sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16.tar.bz2

flutter/sherpa_onnx/example/example.md

@@ -22,5 +22,5 @@ Hint: All of the following functions can be used in Flutter, even if some of the
 |Speaker identification and verification| [Address](https://github.com/k2-fsa/sherpa-onnx/tree/master/dart-api-examples/speaker-identification)| macOS, Windows, Linux|
 |Audio tagging| [Address](https://github.com/k2-fsa/sherpa-onnx/tree/master/dart-api-examples/audio-tagging)| macOS, Windows, Linux|
 |Keyword spotter| [Address](https://github.com/k2-fsa/sherpa-onnx/tree/master/dart-api-examples/keyword-spotter)| macOS, Windows, Linux|
-|Add punctuions| [Address](https://github.com/k2-fsa/sherpa-onnx/tree/master/dart-api-examples/add-punctuations)| macOS, Windows, Linux|
+|Add punctuations| [Address](https://github.com/k2-fsa/sherpa-onnx/tree/master/dart-api-examples/add-punctuations)| macOS, Windows, Linux|
 |Speech enhancement/denoising| [Address](https://github.com/k2-fsa/sherpa-onnx/tree/master/dart-api-examples/speech-enhancement-gtcrn)| macOS, Windows, Linux|

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/CMakeLists.txt

@@ -44,6 +44,7 @@ add_library(sherpa_onnx SHARED
   keyword-spotting.cc
   non-streaming-asr.cc
   non-streaming-speaker-diarization.cc
+  non-streaming-speech-denoiser.cc
   non-streaming-tts.cc
   punctuation.cc
   sherpa-onnx-node-addon-api.cc

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/audio-tagging.cc

@@ -121,7 +121,7 @@ AudioTaggingCreateOfflineStreamWrapper(const Napi::CallbackInfo &info) {
     return {};
   }
 
-  SherpaOnnxAudioTagging *at =
+  const SherpaOnnxAudioTagging *at =
       info[0].As<Napi::External<SherpaOnnxAudioTagging>>().Data();
 
   const SherpaOnnxOfflineStream *stream =
@@ -169,10 +169,10 @@ static Napi::Object AudioTaggingComputeWrapper(const Napi::CallbackInfo &info) {
     return {};
   }
 
-  SherpaOnnxAudioTagging *at =
+  const SherpaOnnxAudioTagging *at =
       info[0].As<Napi::External<SherpaOnnxAudioTagging>>().Data();
 
-  SherpaOnnxOfflineStream *stream =
+  const SherpaOnnxOfflineStream *stream =
       info[1].As<Napi::External<SherpaOnnxOfflineStream>>().Data();
 
   int32_t top_k = info[2].As<Napi::Number>().Int32Value();

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/non-streaming-asr.cc

@@ -341,7 +341,7 @@ static Napi::External<SherpaOnnxOfflineStream> CreateOfflineStreamWrapper(
     return {};
   }
 
-  SherpaOnnxOfflineRecognizer *recognizer =
+  const SherpaOnnxOfflineRecognizer *recognizer =
       info[0].As<Napi::External<SherpaOnnxOfflineRecognizer>>().Data();
 
   const SherpaOnnxOfflineStream *stream =
@@ -373,7 +373,7 @@ static void AcceptWaveformOfflineWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxOfflineStream *stream =
+  const SherpaOnnxOfflineStream *stream =
       info[0].As<Napi::External<SherpaOnnxOfflineStream>>().Data();
 
   if (!info[1].IsObject()) {
@@ -454,10 +454,10 @@ static void DecodeOfflineStreamWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxOfflineRecognizer *recognizer =
+  const SherpaOnnxOfflineRecognizer *recognizer =
       info[0].As<Napi::External<SherpaOnnxOfflineRecognizer>>().Data();
 
-  SherpaOnnxOfflineStream *stream =
+  const SherpaOnnxOfflineStream *stream =
       info[1].As<Napi::External<SherpaOnnxOfflineStream>>().Data();
 
   SherpaOnnxDecodeOfflineStream(recognizer, stream);
@@ -482,7 +482,7 @@ static Napi::String GetOfflineStreamResultAsJsonWrapper(
     return {};
   }
 
-  SherpaOnnxOfflineStream *stream =
+  const SherpaOnnxOfflineStream *stream =
       info[0].As<Napi::External<SherpaOnnxOfflineStream>>().Data();
 
   const char *json = SherpaOnnxGetOfflineStreamResultAsJson(stream);

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/non-streaming-speech-denoiser.cc

@@ -0,0 +1,278 @@
+// scripts/node-addon-api/src/non-streaming-speech-denoiser.cc
+//
+// Copyright (c)  2025  Xiaomi Corporation
+#include <sstream>
+
+#include "macros.h"  // NOLINT
+#include "napi.h"    // NOLINT
+#include "sherpa-onnx/c-api/c-api.h"
+
+static SherpaOnnxOfflineSpeechDenoiserGtcrnModelConfig
+GetOfflineSpeechDenoiserGtcrnModelConfig(Napi::Object obj) {
+  SherpaOnnxOfflineSpeechDenoiserGtcrnModelConfig c;
+  memset(&c, 0, sizeof(c));
+
+  if (!obj.Has("gtcrn") || !obj.Get("gtcrn").IsObject()) {
+    return c;
+  }
+
+  Napi::Object o = obj.Get("gtcrn").As<Napi::Object>();
+
+  SHERPA_ONNX_ASSIGN_ATTR_STR(model, model);
+
+  return c;
+}
+
+static SherpaOnnxOfflineSpeechDenoiserModelConfig
+GetOfflineSpeechDenoiserModelConfig(Napi::Object obj) {
+  SherpaOnnxOfflineSpeechDenoiserModelConfig c;
+  memset(&c, 0, sizeof(c));
+
+  if (!obj.Has("model") || !obj.Get("model").IsObject()) {
+    return c;
+  }
+
+  Napi::Object o = obj.Get("model").As<Napi::Object>();
+
+  c.gtcrn = GetOfflineSpeechDenoiserGtcrnModelConfig(o);
+
+  SHERPA_ONNX_ASSIGN_ATTR_INT32(num_threads, numThreads);
+
+  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();
+    }
+  }
+
+  SHERPA_ONNX_ASSIGN_ATTR_STR(provider, provider);
+
+  return c;
+}
+
+static Napi::External<SherpaOnnxOfflineSpeechDenoiser>
+CreateOfflineSpeechDenoiserWrapper(const Napi::CallbackInfo &info) {
+  Napi::Env env = info.Env();
+#if __OHOS__
+  // the last argument is the NativeResourceManager
+  if (info.Length() != 2) {
+    std::ostringstream os;
+    os << "Expect only 2 arguments. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return {};
+  }
+#else
+  if (info.Length() != 1) {
+    std::ostringstream os;
+    os << "Expect only 1 argument. Given: " << info.Length();
+
+    Napi::TypeError::New(env, os.str()).ThrowAsJavaScriptException();
+
+    return {};
+  }
+#endif
+
+  if (!info[0].IsObject()) {
+    Napi::TypeError::New(env, "Expect an object as the argument")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  Napi::Object o = info[0].As<Napi::Object>();
+
+  SherpaOnnxOfflineSpeechDenoiserConfig c;
+  memset(&c, 0, sizeof(c));
+  c.model = GetOfflineSpeechDenoiserModelConfig(o);
+
+#if __OHOS__
+  std::unique_ptr<NativeResourceManager,
+                  decltype(&OH_ResourceManager_ReleaseNativeResourceManager)>
+      mgr(OH_ResourceManager_InitNativeResourceManager(env, info[1]),
+          &OH_ResourceManager_ReleaseNativeResourceManager);
+
+  const SherpaOnnxOfflineSpeechDenoiser *sd =
+      SherpaOnnxCreateOfflineSpeechDenoiserOHOS(&c, mgr.get());
+#else
+  const SherpaOnnxOfflineSpeechDenoiser *sd =
+      SherpaOnnxCreateOfflineSpeechDenoiser(&c);
+#endif
+
+  SHERPA_ONNX_DELETE_C_STR(c.model.gtcrn.model);
+  SHERPA_ONNX_DELETE_C_STR(c.model.provider);
+
+  if (!sd) {
+    Napi::TypeError::New(env, "Please check your config!")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  return Napi::External<SherpaOnnxOfflineSpeechDenoiser>::New(
+      env, const_cast<SherpaOnnxOfflineSpeechDenoiser *>(sd),
+      [](Napi::Env env, SherpaOnnxOfflineSpeechDenoiser *sd) {
+        SherpaOnnxDestroyOfflineSpeechDenoiser(sd);
+      });
+}
+
+static Napi::Object OfflineSpeechDenoiserRunWrapper(
+    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 offline speech denoiser pointer.")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  const SherpaOnnxOfflineSpeechDenoiser *sd =
+      info[0].As<Napi::External<SherpaOnnxOfflineSpeechDenoiser>>().Data();
+
+  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();
+
+  const SherpaOnnxDenoisedAudio *audio;
+
+#if __OHOS__
+  // Note(fangjun): For unknown reasons on HarmonyOS, we need to divide it by
+  // sizeof(float) here
+  audio = SherpaOnnxOfflineSpeechDenoiserRun(
+      sd, samples.Data(), samples.ElementLength() / sizeof(float), sample_rate);
+#else
+  audio = SherpaOnnxOfflineSpeechDenoiserRun(
+      sd, samples.Data(), samples.ElementLength(), sample_rate);
+#endif
+
+  bool enable_external_buffer = true;
+  if (obj.Has("enableExternalBuffer") &&
+      obj.Get("enableExternalBuffer").IsBoolean()) {
+    enable_external_buffer =
+        obj.Get("enableExternalBuffer").As<Napi::Boolean>().Value();
+  }
+
+  if (enable_external_buffer) {
+    Napi::ArrayBuffer arrayBuffer = Napi::ArrayBuffer::New(
+        env, const_cast<float *>(audio->samples), sizeof(float) * audio->n,
+        [](Napi::Env /*env*/, void * /*data*/,
+           const SherpaOnnxDenoisedAudio *hint) {
+          SherpaOnnxDestroyDenoisedAudio(hint);
+        },
+        audio);
+    Napi::Float32Array float32Array =
+        Napi::Float32Array::New(env, audio->n, arrayBuffer, 0);
+
+    Napi::Object ans = Napi::Object::New(env);
+    ans.Set(Napi::String::New(env, "samples"), float32Array);
+    ans.Set(Napi::String::New(env, "sampleRate"), audio->sample_rate);
+    return ans;
+  } else {
+    // don't use external buffer
+    Napi::ArrayBuffer arrayBuffer =
+        Napi::ArrayBuffer::New(env, sizeof(float) * audio->n);
+
+    Napi::Float32Array float32Array =
+        Napi::Float32Array::New(env, audio->n, arrayBuffer, 0);
+
+    std::copy(audio->samples, audio->samples + audio->n, float32Array.Data());
+
+    Napi::Object ans = Napi::Object::New(env);
+    ans.Set(Napi::String::New(env, "samples"), float32Array);
+    ans.Set(Napi::String::New(env, "sampleRate"), audio->sample_rate);
+    SherpaOnnxDestroyDenoisedAudio(audio);
+    return ans;
+  }
+}
+
+static Napi::Number OfflineSpeechDenoiserGetSampleRateWrapper(
+    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 offline speech denoiser pointer.")
+        .ThrowAsJavaScriptException();
+
+    return {};
+  }
+
+  const SherpaOnnxOfflineSpeechDenoiser *sd =
+      info[0].As<Napi::External<SherpaOnnxOfflineSpeechDenoiser>>().Data();
+
+  int32_t sample_rate = SherpaOnnxOfflineSpeechDenoiserGetSampleRate(sd);
+
+  return Napi::Number::New(env, sample_rate);
+}
+
+void InitNonStreamingSpeechDenoiser(Napi::Env env, Napi::Object exports) {
+  exports.Set(Napi::String::New(env, "createOfflineSpeechDenoiser"),
+              Napi::Function::New(env, CreateOfflineSpeechDenoiserWrapper));
+
+  exports.Set(Napi::String::New(env, "offlineSpeechDenoiserRunWrapper"),
+              Napi::Function::New(env, OfflineSpeechDenoiserRunWrapper));
+
+  exports.Set(
+      Napi::String::New(env, "offlineSpeechDenoiserGetSampleRateWrapper"),
+      Napi::Function::New(env, OfflineSpeechDenoiserGetSampleRateWrapper));
+}

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/non-streaming-tts.cc

@@ -217,7 +217,7 @@ static Napi::Number OfflineTtsSampleRateWrapper(
     return {};
   }
 
-  SherpaOnnxOfflineTts *tts =
+  const SherpaOnnxOfflineTts *tts =
       info[0].As<Napi::External<SherpaOnnxOfflineTts>>().Data();
 
   int32_t sample_rate = SherpaOnnxOfflineTtsSampleRate(tts);
@@ -245,7 +245,7 @@ static Napi::Number OfflineTtsNumSpeakersWrapper(
     return {};
   }
 
-  SherpaOnnxOfflineTts *tts =
+  const SherpaOnnxOfflineTts *tts =
       info[0].As<Napi::External<SherpaOnnxOfflineTts>>().Data();
 
   int32_t num_speakers = SherpaOnnxOfflineTtsNumSpeakers(tts);
@@ -273,7 +273,7 @@ static Napi::Object OfflineTtsGenerateWrapper(const Napi::CallbackInfo &info) {
     return {};
   }
 
-  SherpaOnnxOfflineTts *tts =
+  const SherpaOnnxOfflineTts *tts =
       info[0].As<Napi::External<SherpaOnnxOfflineTts>>().Data();
 
   if (!info[1].IsObject()) {
@@ -418,9 +418,9 @@ using TSFN = Napi::TypedThreadSafeFunction<Napi::Reference<Napi::Value>,
 
 class TtsGenerateWorker : public Napi::AsyncWorker {
  public:
-  TtsGenerateWorker(const Napi::Env &env, TSFN tsfn, SherpaOnnxOfflineTts *tts,
+  TtsGenerateWorker(const Napi::Env &env, TSFN tsfn,
-                    const std::string &text, float speed, int32_t sid,
+                    const SherpaOnnxOfflineTts *tts, const std::string &text,
-                    bool use_external_buffer)
+                    float speed, int32_t sid, bool use_external_buffer)
       : tsfn_(tsfn),
         Napi::AsyncWorker{env, "TtsGenerateWorker"},
         deferred_(env),
@@ -506,7 +506,7 @@ class TtsGenerateWorker : public Napi::AsyncWorker {
  private:
   TSFN tsfn_;
   Napi::Promise::Deferred deferred_;
-  SherpaOnnxOfflineTts *tts_;
+  const SherpaOnnxOfflineTts *tts_;
   std::string text_;
   float speed_;
   int32_t sid_;
@@ -537,7 +537,7 @@ static Napi::Object OfflineTtsGenerateAsyncWrapper(
     return {};
   }
 
-  SherpaOnnxOfflineTts *tts =
+  const SherpaOnnxOfflineTts *tts =
       info[0].As<Napi::External<SherpaOnnxOfflineTts>>().Data();
 
   if (!info[1].IsObject()) {

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/punctuation.cc

@@ -108,7 +108,7 @@ static Napi::String OfflinePunctuationAddPunctWraper(
     return {};
   }
 
-  SherpaOnnxOfflinePunctuation *punct =
+  const SherpaOnnxOfflinePunctuation *punct =
       info[0].As<Napi::External<SherpaOnnxOfflinePunctuation>>().Data();
   Napi::String js_text = info[1].As<Napi::String>();
   std::string text = js_text.Utf8Value();

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/sherpa-onnx-node-addon-api.cc

@@ -27,6 +27,8 @@ void InitKeywordSpotting(Napi::Env env, Napi::Object exports);
 
 void InitNonStreamingSpeakerDiarization(Napi::Env env, Napi::Object exports);
 
+void InitNonStreamingSpeechDenoiser(Napi::Env env, Napi::Object exports);
+
 #if __OHOS__
 void InitUtils(Napi::Env env, Napi::Object exports);
 #endif
@@ -44,6 +46,7 @@ Napi::Object Init(Napi::Env env, Napi::Object exports) {
   InitPunctuation(env, exports);
   InitKeywordSpotting(env, exports);
   InitNonStreamingSpeakerDiarization(env, exports);
+  InitNonStreamingSpeechDenoiser(env, exports);
 
 #if __OHOS__
   InitUtils(env, exports);

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/speaker-identification.cc

@@ -108,7 +108,7 @@ static Napi::Number SpeakerEmbeddingExtractorDimWrapper(
     return {};
   }
 
-  SherpaOnnxSpeakerEmbeddingExtractor *extractor =
+  const SherpaOnnxSpeakerEmbeddingExtractor *extractor =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingExtractor>>().Data();
 
   int32_t dim = SherpaOnnxSpeakerEmbeddingExtractorDim(extractor);
@@ -137,7 +137,7 @@ SpeakerEmbeddingExtractorCreateStreamWrapper(const Napi::CallbackInfo &info) {
     return {};
   }
 
-  SherpaOnnxSpeakerEmbeddingExtractor *extractor =
+  const SherpaOnnxSpeakerEmbeddingExtractor *extractor =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingExtractor>>().Data();
 
   const SherpaOnnxOnlineStream *stream =
@@ -177,10 +177,10 @@ static Napi::Boolean SpeakerEmbeddingExtractorIsReadyWrapper(
     return {};
   }
 
-  SherpaOnnxSpeakerEmbeddingExtractor *extractor =
+  const SherpaOnnxSpeakerEmbeddingExtractor *extractor =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingExtractor>>().Data();
 
-  SherpaOnnxOnlineStream *stream =
+  const SherpaOnnxOnlineStream *stream =
       info[1].As<Napi::External<SherpaOnnxOnlineStream>>().Data();
 
   int32_t is_ready =
@@ -226,10 +226,10 @@ static Napi::Float32Array SpeakerEmbeddingExtractorComputeEmbeddingWrapper(
     }
   }
 
-  SherpaOnnxSpeakerEmbeddingExtractor *extractor =
+  const SherpaOnnxSpeakerEmbeddingExtractor *extractor =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingExtractor>>().Data();
 
-  SherpaOnnxOnlineStream *stream =
+  const SherpaOnnxOnlineStream *stream =
       info[1].As<Napi::External<SherpaOnnxOnlineStream>>().Data();
 
   const float *v =
@@ -329,7 +329,7 @@ static Napi::Boolean SpeakerEmbeddingManagerAddWrapper(
     return {};
   }
 
-  SherpaOnnxSpeakerEmbeddingManager *manager =
+  const SherpaOnnxSpeakerEmbeddingManager *manager =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingManager>>().Data();
 
   Napi::Object obj = info[1].As<Napi::Object>();
@@ -399,7 +399,7 @@ static Napi::Boolean SpeakerEmbeddingManagerAddListFlattenedWrapper(
     return {};
   }
 
-  SherpaOnnxSpeakerEmbeddingManager *manager =
+  const SherpaOnnxSpeakerEmbeddingManager *manager =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingManager>>().Data();
 
   Napi::Object obj = info[1].As<Napi::Object>();
@@ -486,7 +486,7 @@ static Napi::Boolean SpeakerEmbeddingManagerRemoveWrapper(
     return {};
   }
 
-  SherpaOnnxSpeakerEmbeddingManager *manager =
+  const SherpaOnnxSpeakerEmbeddingManager *manager =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingManager>>().Data();
 
   Napi::String js_name = info[1].As<Napi::String>();
@@ -525,7 +525,7 @@ static Napi::String SpeakerEmbeddingManagerSearchWrapper(
     return {};
   }
 
-  SherpaOnnxSpeakerEmbeddingManager *manager =
+  const SherpaOnnxSpeakerEmbeddingManager *manager =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingManager>>().Data();
 
   Napi::Object obj = info[1].As<Napi::Object>();
@@ -603,7 +603,7 @@ static Napi::Boolean SpeakerEmbeddingManagerVerifyWrapper(
     return {};
   }
 
-  SherpaOnnxSpeakerEmbeddingManager *manager =
+  const SherpaOnnxSpeakerEmbeddingManager *manager =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingManager>>().Data();
 
   Napi::Object obj = info[1].As<Napi::Object>();
@@ -691,7 +691,7 @@ static Napi::Boolean SpeakerEmbeddingManagerContainsWrapper(
     return {};
   }
 
-  SherpaOnnxSpeakerEmbeddingManager *manager =
+  const SherpaOnnxSpeakerEmbeddingManager *manager =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingManager>>().Data();
 
   Napi::String js_name = info[1].As<Napi::String>();
@@ -724,7 +724,7 @@ static Napi::Number SpeakerEmbeddingManagerNumSpeakersWrapper(
     return {};
   }
 
-  SherpaOnnxSpeakerEmbeddingManager *manager =
+  const SherpaOnnxSpeakerEmbeddingManager *manager =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingManager>>().Data();
 
   int32_t num_speakers = SherpaOnnxSpeakerEmbeddingManagerNumSpeakers(manager);
@@ -753,7 +753,7 @@ static Napi::Array SpeakerEmbeddingManagerGetAllSpeakersWrapper(
     return {};
   }
 
-  SherpaOnnxSpeakerEmbeddingManager *manager =
+  const SherpaOnnxSpeakerEmbeddingManager *manager =
       info[0].As<Napi::External<SherpaOnnxSpeakerEmbeddingManager>>().Data();
 
   int32_t num_speakers = SherpaOnnxSpeakerEmbeddingManagerNumSpeakers(manager);

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/spoken-language-identification.cc

@@ -106,16 +106,17 @@ SpokenLanguageIdentificationCreateOfflineStreamWrapper(
     return {};
   }
 
-  SherpaOnnxSpokenLanguageIdentification *slid =
+  const SherpaOnnxSpokenLanguageIdentification *slid =
       info[0]
           .As<Napi::External<SherpaOnnxSpokenLanguageIdentification>>()
           .Data();
 
-  SherpaOnnxOfflineStream *stream =
+  const SherpaOnnxOfflineStream *stream =
       SherpaOnnxSpokenLanguageIdentificationCreateOfflineStream(slid);
 
   return Napi::External<SherpaOnnxOfflineStream>::New(
-      env, stream, [](Napi::Env env, SherpaOnnxOfflineStream *stream) {
+      env, const_cast<SherpaOnnxOfflineStream *>(stream),
+      [](Napi::Env env, SherpaOnnxOfflineStream *stream) {
         SherpaOnnxDestroyOfflineStream(stream);
       });
 }
@@ -147,12 +148,12 @@ static Napi::String SpokenLanguageIdentificationComputeWrapper(
     return {};
   }
 
-  SherpaOnnxSpokenLanguageIdentification *slid =
+  const SherpaOnnxSpokenLanguageIdentification *slid =
       info[0]
           .As<Napi::External<SherpaOnnxSpokenLanguageIdentification>>()
           .Data();
 
-  SherpaOnnxOfflineStream *stream =
+  const SherpaOnnxOfflineStream *stream =
       info[1].As<Napi::External<SherpaOnnxOfflineStream>>().Data();
 
   const SherpaOnnxSpokenLanguageIdentificationResult *r =

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/streaming-asr.cc

@@ -278,7 +278,7 @@ static Napi::External<SherpaOnnxOnlineStream> CreateOnlineStreamWrapper(
     return {};
   }
 
-  SherpaOnnxOnlineRecognizer *recognizer =
+  const SherpaOnnxOnlineRecognizer *recognizer =
       info[0].As<Napi::External<SherpaOnnxOnlineRecognizer>>().Data();
 
   const SherpaOnnxOnlineStream *stream =
@@ -310,7 +310,7 @@ static void AcceptWaveformWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxOnlineStream *stream =
+  const SherpaOnnxOnlineStream *stream =
       info[0].As<Napi::External<SherpaOnnxOnlineStream>>().Data();
 
   if (!info[1].IsObject()) {
@@ -390,10 +390,10 @@ static Napi::Boolean IsOnlineStreamReadyWrapper(
     return {};
   }
 
-  SherpaOnnxOnlineRecognizer *recognizer =
+  const SherpaOnnxOnlineRecognizer *recognizer =
       info[0].As<Napi::External<SherpaOnnxOnlineRecognizer>>().Data();
 
-  SherpaOnnxOnlineStream *stream =
+  const SherpaOnnxOnlineStream *stream =
       info[1].As<Napi::External<SherpaOnnxOnlineStream>>().Data();
 
   int32_t is_ready = SherpaOnnxIsOnlineStreamReady(recognizer, stream);
@@ -427,10 +427,10 @@ static void DecodeOnlineStreamWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxOnlineRecognizer *recognizer =
+  const SherpaOnnxOnlineRecognizer *recognizer =
       info[0].As<Napi::External<SherpaOnnxOnlineRecognizer>>().Data();
 
-  SherpaOnnxOnlineStream *stream =
+  const SherpaOnnxOnlineStream *stream =
       info[1].As<Napi::External<SherpaOnnxOnlineStream>>().Data();
 
   SherpaOnnxDecodeOnlineStream(recognizer, stream);
@@ -463,10 +463,10 @@ static Napi::String GetOnlineStreamResultAsJsonWrapper(
     return {};
   }
 
-  SherpaOnnxOnlineRecognizer *recognizer =
+  const SherpaOnnxOnlineRecognizer *recognizer =
       info[0].As<Napi::External<SherpaOnnxOnlineRecognizer>>().Data();
 
-  SherpaOnnxOnlineStream *stream =
+  const SherpaOnnxOnlineStream *stream =
       info[1].As<Napi::External<SherpaOnnxOnlineStream>>().Data();
 
   const char *json = SherpaOnnxGetOnlineStreamResultAsJson(recognizer, stream);
@@ -496,7 +496,7 @@ static void InputFinishedWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxOnlineStream *stream =
+  const SherpaOnnxOnlineStream *stream =
       info[0].As<Napi::External<SherpaOnnxOnlineStream>>().Data();
 
   SherpaOnnxOnlineStreamInputFinished(stream);
@@ -528,10 +528,10 @@ static void ResetOnlineStreamWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxOnlineRecognizer *recognizer =
+  const SherpaOnnxOnlineRecognizer *recognizer =
       info[0].As<Napi::External<SherpaOnnxOnlineRecognizer>>().Data();
 
-  SherpaOnnxOnlineStream *stream =
+  const SherpaOnnxOnlineStream *stream =
       info[1].As<Napi::External<SherpaOnnxOnlineStream>>().Data();
 
   SherpaOnnxOnlineStreamReset(recognizer, stream);
@@ -563,10 +563,10 @@ static Napi::Boolean IsEndpointWrapper(const Napi::CallbackInfo &info) {
     return {};
   }
 
-  SherpaOnnxOnlineRecognizer *recognizer =
+  const SherpaOnnxOnlineRecognizer *recognizer =
       info[0].As<Napi::External<SherpaOnnxOnlineRecognizer>>().Data();
 
-  SherpaOnnxOnlineStream *stream =
+  const SherpaOnnxOnlineStream *stream =
       info[1].As<Napi::External<SherpaOnnxOnlineStream>>().Data();
 
   int32_t is_endpoint = SherpaOnnxOnlineStreamIsEndpoint(recognizer, stream);
@@ -636,7 +636,7 @@ static void PrintWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxDisplay *display =
+  const SherpaOnnxDisplay *display =
       info[0].As<Napi::External<SherpaOnnxDisplay>>().Data();
 
   int32_t idx = info[1].As<Napi::Number>().Int32Value();

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/vad.cc

@@ -28,11 +28,12 @@ static Napi::External<SherpaOnnxCircularBuffer> CreateCircularBufferWrapper(
     return {};
   }
 
-  SherpaOnnxCircularBuffer *buf =
+  const SherpaOnnxCircularBuffer *buf =
       SherpaOnnxCreateCircularBuffer(info[0].As<Napi::Number>().Int32Value());
 
   return Napi::External<SherpaOnnxCircularBuffer>::New(
-      env, buf, [](Napi::Env env, SherpaOnnxCircularBuffer *p) {
+      env, const_cast<SherpaOnnxCircularBuffer *>(buf),
+      [](Napi::Env env, SherpaOnnxCircularBuffer *p) {
         SherpaOnnxDestroyCircularBuffer(p);
       });
 }
@@ -56,7 +57,7 @@ static void CircularBufferPushWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxCircularBuffer *buf =
+  const SherpaOnnxCircularBuffer *buf =
       info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
 
   if (!info[1].IsTypedArray()) {
@@ -101,7 +102,7 @@ static Napi::Float32Array CircularBufferGetWrapper(
     return {};
   }
 
-  SherpaOnnxCircularBuffer *buf =
+  const SherpaOnnxCircularBuffer *buf =
       info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
 
   if (!info[1].IsNumber()) {
@@ -179,7 +180,7 @@ static void CircularBufferPopWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxCircularBuffer *buf =
+  const SherpaOnnxCircularBuffer *buf =
       info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
 
   if (!info[1].IsNumber()) {
@@ -213,7 +214,7 @@ static Napi::Number CircularBufferSizeWrapper(const Napi::CallbackInfo &info) {
     return {};
   }
 
-  SherpaOnnxCircularBuffer *buf =
+  const SherpaOnnxCircularBuffer *buf =
       info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
 
   int32_t size = SherpaOnnxCircularBufferSize(buf);
@@ -240,7 +241,7 @@ static Napi::Number CircularBufferHeadWrapper(const Napi::CallbackInfo &info) {
     return {};
   }
 
-  SherpaOnnxCircularBuffer *buf =
+  const SherpaOnnxCircularBuffer *buf =
       info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
 
   int32_t size = SherpaOnnxCircularBufferHead(buf);
@@ -267,7 +268,7 @@ static void CircularBufferResetWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxCircularBuffer *buf =
+  const SherpaOnnxCircularBuffer *buf =
       info[0].As<Napi::External<SherpaOnnxCircularBuffer>>().Data();
 
   SherpaOnnxCircularBufferReset(buf);
@@ -360,18 +361,19 @@ CreateVoiceActivityDetectorWrapper(const Napi::CallbackInfo &info) {
       mgr(OH_ResourceManager_InitNativeResourceManager(env, info[2]),
           &OH_ResourceManager_ReleaseNativeResourceManager);
 
-  SherpaOnnxVoiceActivityDetector *vad =
+  const SherpaOnnxVoiceActivityDetector *vad =
       SherpaOnnxCreateVoiceActivityDetectorOHOS(&c, buffer_size_in_seconds,
                                                 mgr.get());
 #else
-  SherpaOnnxVoiceActivityDetector *vad =
+  const SherpaOnnxVoiceActivityDetector *vad =
       SherpaOnnxCreateVoiceActivityDetector(&c, buffer_size_in_seconds);
 #endif
   SHERPA_ONNX_DELETE_C_STR(c.silero_vad.model);
   SHERPA_ONNX_DELETE_C_STR(c.provider);
 
   return Napi::External<SherpaOnnxVoiceActivityDetector>::New(
-      env, vad, [](Napi::Env env, SherpaOnnxVoiceActivityDetector *p) {
+      env, const_cast<SherpaOnnxVoiceActivityDetector *>(vad),
+      [](Napi::Env env, SherpaOnnxVoiceActivityDetector *p) {
         SherpaOnnxDestroyVoiceActivityDetector(p);
       });
 }
@@ -396,7 +398,7 @@ static void VoiceActivityDetectorAcceptWaveformWrapper(
     return;
   }
 
-  SherpaOnnxVoiceActivityDetector *vad =
+  const SherpaOnnxVoiceActivityDetector *vad =
       info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
 
   if (!info[1].IsTypedArray()) {
@@ -440,7 +442,7 @@ static Napi::Boolean VoiceActivityDetectorEmptyWrapper(
     return {};
   }
 
-  SherpaOnnxVoiceActivityDetector *vad =
+  const SherpaOnnxVoiceActivityDetector *vad =
       info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
 
   int32_t is_empty = SherpaOnnxVoiceActivityDetectorEmpty(vad);
@@ -468,7 +470,7 @@ static Napi::Boolean VoiceActivityDetectorDetectedWrapper(
     return {};
   }
 
-  SherpaOnnxVoiceActivityDetector *vad =
+  const SherpaOnnxVoiceActivityDetector *vad =
       info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
 
   int32_t is_detected = SherpaOnnxVoiceActivityDetectorDetected(vad);
@@ -495,7 +497,7 @@ static void VoiceActivityDetectorPopWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxVoiceActivityDetector *vad =
+  const SherpaOnnxVoiceActivityDetector *vad =
       info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
 
   SherpaOnnxVoiceActivityDetectorPop(vad);
@@ -520,7 +522,7 @@ static void VoiceActivityDetectorClearWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxVoiceActivityDetector *vad =
+  const SherpaOnnxVoiceActivityDetector *vad =
       info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
 
   SherpaOnnxVoiceActivityDetectorClear(vad);
@@ -556,7 +558,7 @@ static Napi::Object VoiceActivityDetectorFrontWrapper(
     }
   }
 
-  SherpaOnnxVoiceActivityDetector *vad =
+  const SherpaOnnxVoiceActivityDetector *vad =
       info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
 
   const SherpaOnnxSpeechSegment *segment =
@@ -618,7 +620,7 @@ static void VoiceActivityDetectorResetWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxVoiceActivityDetector *vad =
+  const SherpaOnnxVoiceActivityDetector *vad =
       info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
 
   SherpaOnnxVoiceActivityDetectorReset(vad);
@@ -643,7 +645,7 @@ static void VoiceActivityDetectorFlushWrapper(const Napi::CallbackInfo &info) {
     return;
   }
 
-  SherpaOnnxVoiceActivityDetector *vad =
+  const SherpaOnnxVoiceActivityDetector *vad =
       info[0].As<Napi::External<SherpaOnnxVoiceActivityDetector>>().Data();
 
   SherpaOnnxVoiceActivityDetectorFlush(vad);

nodejs-addon-examples/README.md

@@ -43,6 +43,12 @@ export LD_LIBRARY_PATH=$PWD/node_modules/.pnpm/sherpa-onnx-node@<REPLACE-THIS-WI
 
 The following tables list the examples in this folder.
 
+## Speech enhancement/denoising
+
+|File| Description|
+|---|---|
+|[./test_offline_speech_enhancement_gtcrn.js](./test_offline_speech_enhancement_gtcrn.js)| It demonstrates how to use sherpa-onnx JavaScript API for speech enhancement.|
+
 ## Speaker diarization
 
 |File| Description|
@@ -159,6 +165,15 @@ curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/speaker-segm
 node ./test_offline_speaker_diarization.js
 ```
 
+### Speech enhancement/denoising
+
+```bash
+wget https://github.com/k2-fsa/sherpa-onnx/releases/download/speech-enhancement-models/gtcrn_simple.onnx
+wget https://github.com/k2-fsa/sherpa-onnx/releases/download/speech-enhancement-models/inp_16k.wav
+
+node ./test_offline_speech_enhancement_gtcrn.js
+```
+
 ### Voice Activity detection (VAD)
 
 ```bash

nodejs-addon-examples/test_offline_speech_enhancement_gtcrn.js

@@ -0,0 +1,32 @@
+// Copyright (c)  2025  Xiaomi Corporation (authors: Fangjun Kuang)
+
+const sherpa_onnx = require('sherpa-onnx-node');
+
+function createOfflineSpeechDenoiser() {
+  // please download models from
+  // https://github.com/k2-fsa/sherpa-onnx/releases/tag/speech-enhancement-models
+  const config = {
+    model: {
+      gtcrn: {model: './gtcrn_simple.onnx'},
+      debug: true,
+      numThreads: 1,
+    },
+  };
+
+  return new sherpa_onnx.OfflineSpeechDenoiser(config);
+}
+
+const sd = createOfflineSpeechDenoiser();
+
+const waveFilename = './inp_16k.wav';
+const wave = sherpa_onnx.readWave(waveFilename);
+const denoised = sd.run({
+  samples: wave.samples,
+  sampleRate: wave.sampleRate,
+  enableExternalBuffer: true
+});
+sherpa_onnx.writeWave(
+    './enhanced-16k.wav',
+    {samples: denoised.samples, sampleRate: denoised.sampleRate});
+
+console.log(`Saved to ./enhanced-16k.wav`);

scripts/node-addon-api/CMakeLists.txt

@@ -20,6 +20,7 @@ set(srcs
   src/keyword-spotting.cc
   src/non-streaming-asr.cc
   src/non-streaming-speaker-diarization.cc
+  src/non-streaming-speech-denoiser.cc
   src/non-streaming-tts.cc
   src/punctuation.cc
   src/sherpa-onnx-node-addon-api.cc

scripts/node-addon-api/lib/non-streaming-speech-denoiser.js

@@ -0,0 +1,28 @@
+const addon = require('./addon.js');
+
+class OfflineSpeechDenoiser {
+  constructor(config) {
+    this.handle = addon.createOfflineSpeechDenoiser(config);
+    this.config = config;
+
+    this.sampleRate =
+        addon.offlineSpeechDenoiserGetSampleRateWrapper(this.handle);
+  }
+
+  /*
+    obj is
+    {samples: samples, sampleRate: sampleRate, enableExternalBuffer: true}
+
+    samples is a float32 array containing samples in the range [-1, 1]
+    sampleRate is a number
+
+   return an object {samples: Float32Array, sampleRate: <a number>}
+   */
+  run(obj) {
+    return addon.offlineSpeechDenoiserRunWrapper(this.handle, obj);
+  }
+}
+
+module.exports = {
+  OfflineSpeechDenoiser,
+}

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

@@ -9,6 +9,7 @@ const at = require('./audio-tagg.js');
 const punct = require('./punctuation.js');
 const kws = require('./keyword-spotter.js');
 const sd = require('./non-streaming-speaker-diarization.js');
+const speech_denoiser = require('./non-streaming-speech-denoiser.js');
 
 module.exports = {
   OnlineRecognizer: streaming_asr.OnlineRecognizer,
@@ -26,4 +27,5 @@ module.exports = {
   Punctuation: punct.Punctuation,
   KeywordSpotter: kws.KeywordSpotter,
   OfflineSpeakerDiarization: sd.OfflineSpeakerDiarization,
+  OfflineSpeechDenoiser: speech_denoiser.OfflineSpeechDenoiser,
 }

scripts/node-addon-api/src/non-streaming-speech-denoiser.cc

@@ -0,0 +1 @@
+../../../harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/non-streaming-speech-denoiser.cc

sherpa-onnx/c-api/c-api.cc

@@ -922,22 +922,23 @@ struct SherpaOnnxCircularBuffer {
   std::unique_ptr<sherpa_onnx::CircularBuffer> impl;
 };
 
-SherpaOnnxCircularBuffer *SherpaOnnxCreateCircularBuffer(int32_t capacity) {
+const SherpaOnnxCircularBuffer *SherpaOnnxCreateCircularBuffer(
+    int32_t capacity) {
   SherpaOnnxCircularBuffer *buffer = new SherpaOnnxCircularBuffer;
   buffer->impl = std::make_unique<sherpa_onnx::CircularBuffer>(capacity);
   return buffer;
 }
 
-void SherpaOnnxDestroyCircularBuffer(SherpaOnnxCircularBuffer *buffer) {
+void SherpaOnnxDestroyCircularBuffer(const SherpaOnnxCircularBuffer *buffer) {
   delete buffer;
 }
 
-void SherpaOnnxCircularBufferPush(SherpaOnnxCircularBuffer *buffer,
+void SherpaOnnxCircularBufferPush(const SherpaOnnxCircularBuffer *buffer,
                                   const float *p, int32_t n) {
   buffer->impl->Push(p, n);
 }
 
-const float *SherpaOnnxCircularBufferGet(SherpaOnnxCircularBuffer *buffer,
+const float *SherpaOnnxCircularBufferGet(const SherpaOnnxCircularBuffer *buffer,
                                          int32_t start_index, int32_t n) {
   std::vector<float> v = buffer->impl->Get(start_index, n);
 
@@ -948,19 +949,20 @@ const float *SherpaOnnxCircularBufferGet(SherpaOnnxCircularBuffer *buffer,
 
 void SherpaOnnxCircularBufferFree(const float *p) { delete[] p; }
 
-void SherpaOnnxCircularBufferPop(SherpaOnnxCircularBuffer *buffer, int32_t n) {
+void SherpaOnnxCircularBufferPop(const SherpaOnnxCircularBuffer *buffer,
+                                 int32_t n) {
   buffer->impl->Pop(n);
 }
 
-int32_t SherpaOnnxCircularBufferSize(SherpaOnnxCircularBuffer *buffer) {
+int32_t SherpaOnnxCircularBufferSize(const SherpaOnnxCircularBuffer *buffer) {
   return buffer->impl->Size();
 }
 
-int32_t SherpaOnnxCircularBufferHead(SherpaOnnxCircularBuffer *buffer) {
+int32_t SherpaOnnxCircularBufferHead(const SherpaOnnxCircularBuffer *buffer) {
   return buffer->impl->Head();
 }
 
-void SherpaOnnxCircularBufferReset(SherpaOnnxCircularBuffer *buffer) {
+void SherpaOnnxCircularBufferReset(const SherpaOnnxCircularBuffer *buffer) {
   buffer->impl->Reset();
 }
 
@@ -1008,7 +1010,7 @@ sherpa_onnx::VadModelConfig GetVadModelConfig(
   return vad_config;
 }
 
-SherpaOnnxVoiceActivityDetector *SherpaOnnxCreateVoiceActivityDetector(
+const SherpaOnnxVoiceActivityDetector *SherpaOnnxCreateVoiceActivityDetector(
     const SherpaOnnxVadModelConfig *config, float buffer_size_in_seconds) {
   auto vad_config = GetVadModelConfig(config);
 
@@ -1025,35 +1027,37 @@ SherpaOnnxVoiceActivityDetector *SherpaOnnxCreateVoiceActivityDetector(
 }
 
 void SherpaOnnxDestroyVoiceActivityDetector(
-    SherpaOnnxVoiceActivityDetector *p) {
+    const SherpaOnnxVoiceActivityDetector *p) {
   delete p;
 }
 
 void SherpaOnnxVoiceActivityDetectorAcceptWaveform(
-    SherpaOnnxVoiceActivityDetector *p, const float *samples, int32_t n) {
+    const SherpaOnnxVoiceActivityDetector *p, const float *samples, int32_t n) {
   p->impl->AcceptWaveform(samples, n);
 }
 
 int32_t SherpaOnnxVoiceActivityDetectorEmpty(
-    SherpaOnnxVoiceActivityDetector *p) {
+    const SherpaOnnxVoiceActivityDetector *p) {
   return p->impl->Empty();
 }
 
 int32_t SherpaOnnxVoiceActivityDetectorDetected(
-    SherpaOnnxVoiceActivityDetector *p) {
+    const SherpaOnnxVoiceActivityDetector *p) {
   return p->impl->IsSpeechDetected();
 }
 
-void SherpaOnnxVoiceActivityDetectorPop(SherpaOnnxVoiceActivityDetector *p) {
+void SherpaOnnxVoiceActivityDetectorPop(
+    const SherpaOnnxVoiceActivityDetector *p) {
   p->impl->Pop();
 }
 
-void SherpaOnnxVoiceActivityDetectorClear(SherpaOnnxVoiceActivityDetector *p) {
+void SherpaOnnxVoiceActivityDetectorClear(
+    const SherpaOnnxVoiceActivityDetector *p) {
   p->impl->Clear();
 }
 
 const SherpaOnnxSpeechSegment *SherpaOnnxVoiceActivityDetectorFront(
-    SherpaOnnxVoiceActivityDetector *p) {
+    const SherpaOnnxVoiceActivityDetector *p) {
   const sherpa_onnx::SpeechSegment &segment = p->impl->Front();
 
   SherpaOnnxSpeechSegment *ans = new SherpaOnnxSpeechSegment;
@@ -1072,11 +1076,13 @@ void SherpaOnnxDestroySpeechSegment(const SherpaOnnxSpeechSegment *p) {
   }
 }
 
-void SherpaOnnxVoiceActivityDetectorReset(SherpaOnnxVoiceActivityDetector *p) {
+void SherpaOnnxVoiceActivityDetectorReset(
+    const SherpaOnnxVoiceActivityDetector *p) {
   p->impl->Reset();
 }
 
-void SherpaOnnxVoiceActivityDetectorFlush(SherpaOnnxVoiceActivityDetector *p) {
+void SherpaOnnxVoiceActivityDetectorFlush(
+    const SherpaOnnxVoiceActivityDetector *p) {
   p->impl->Flush();
 }
 
@@ -1915,7 +1921,7 @@ struct SherpaOnnxLinearResampler {
   std::unique_ptr<sherpa_onnx::LinearResample> impl;
 };
 
-SherpaOnnxLinearResampler *SherpaOnnxCreateLinearResampler(
+const SherpaOnnxLinearResampler *SherpaOnnxCreateLinearResampler(
     int32_t samp_rate_in_hz, int32_t samp_rate_out_hz, float filter_cutoff_hz,
     int32_t num_zeros) {
   SherpaOnnxLinearResampler *p = new SherpaOnnxLinearResampler;
@@ -1925,12 +1931,12 @@ SherpaOnnxLinearResampler *SherpaOnnxCreateLinearResampler(
   return p;
 }
 
-void SherpaOnnxDestroyLinearResampler(SherpaOnnxLinearResampler *p) {
+void SherpaOnnxDestroyLinearResampler(const SherpaOnnxLinearResampler *p) {
   delete p;
 }
 
 const SherpaOnnxResampleOut *SherpaOnnxLinearResamplerResample(
-    SherpaOnnxLinearResampler *p, const float *input, int32_t input_dim,
+    const SherpaOnnxLinearResampler *p, const float *input, int32_t input_dim,
     int32_t flush) {
   std::vector<float> o;
   p->impl->Resample(input, input_dim, flush, &o);
@@ -2320,7 +2326,7 @@ const SherpaOnnxOfflineSpeechDenoiser *
 SherpaOnnxCreateOfflineSpeechDenoiserOHOS(
     const SherpaOnnxOfflineSpeechDenoiserConfig *config,
     NativeResourceManager *mgr) {
-  auto sd_config = GetOfflineSpeechDenoiserConfia(config);
+  auto sd_config = GetOfflineSpeechDenoiserConfig(config);
 
   SherpaOnnxOfflineSpeechDenoiser *sd = new SherpaOnnxOfflineSpeechDenoiser;
 
@@ -2361,7 +2367,8 @@ const SherpaOnnxOfflineRecognizer *SherpaOnnxCreateOfflineRecognizerOHOS(
   return recognizer;
 }
 
-SherpaOnnxVoiceActivityDetector *SherpaOnnxCreateVoiceActivityDetectorOHOS(
+const SherpaOnnxVoiceActivityDetector *
+SherpaOnnxCreateVoiceActivityDetectorOHOS(
     const SherpaOnnxVadModelConfig *config, float buffer_size_in_seconds,
     NativeResourceManager *mgr) {
   if (mgr == nullptr) {

sherpa-onnx/c-api/c-api.h

@@ -808,15 +808,15 @@ SHERPA_ONNX_API typedef struct SherpaOnnxCircularBuffer
 // Return an instance of circular buffer. The user has to use
 // SherpaOnnxDestroyCircularBuffer() to free the returned pointer to avoid
 // memory leak.
-SHERPA_ONNX_API SherpaOnnxCircularBuffer *SherpaOnnxCreateCircularBuffer(
+SHERPA_ONNX_API const SherpaOnnxCircularBuffer *SherpaOnnxCreateCircularBuffer(
     int32_t capacity);
 
 // Free the pointer returned by SherpaOnnxCreateCircularBuffer()
 SHERPA_ONNX_API void SherpaOnnxDestroyCircularBuffer(
-    SherpaOnnxCircularBuffer *buffer);
+    const SherpaOnnxCircularBuffer *buffer);
 
 SHERPA_ONNX_API void SherpaOnnxCircularBufferPush(
-    SherpaOnnxCircularBuffer *buffer, const float *p, int32_t n);
+    const SherpaOnnxCircularBuffer *buffer, const float *p, int32_t n);
 
 // Return n samples starting at the given index.
 //
@@ -824,27 +824,27 @@ SHERPA_ONNX_API void SherpaOnnxCircularBufferPush(
 // The user has to use SherpaOnnxCircularBufferFree() to free the returned
 // pointer to avoid memory leak.
 SHERPA_ONNX_API const float *SherpaOnnxCircularBufferGet(
-    SherpaOnnxCircularBuffer *buffer, int32_t start_index, int32_t n);
+    const SherpaOnnxCircularBuffer *buffer, int32_t start_index, int32_t n);
 
 // Free the pointer returned by SherpaOnnxCircularBufferGet().
 SHERPA_ONNX_API void SherpaOnnxCircularBufferFree(const float *p);
 
 // Remove n elements from the buffer
 SHERPA_ONNX_API void SherpaOnnxCircularBufferPop(
-    SherpaOnnxCircularBuffer *buffer, int32_t n);
+    const SherpaOnnxCircularBuffer *buffer, int32_t n);
 
 // Return number of elements in the buffer.
 SHERPA_ONNX_API int32_t
-SherpaOnnxCircularBufferSize(SherpaOnnxCircularBuffer *buffer);
+SherpaOnnxCircularBufferSize(const SherpaOnnxCircularBuffer *buffer);
 
 // Return the head of the buffer. It's always non-decreasing until you
 // invoke SherpaOnnxCircularBufferReset() which resets head to 0.
 SHERPA_ONNX_API int32_t
-SherpaOnnxCircularBufferHead(SherpaOnnxCircularBuffer *buffer);
+SherpaOnnxCircularBufferHead(const SherpaOnnxCircularBuffer *buffer);
 
 // Clear all elements in the buffer
 SHERPA_ONNX_API void SherpaOnnxCircularBufferReset(
-    SherpaOnnxCircularBuffer *buffer);
+    const SherpaOnnxCircularBuffer *buffer);
 
 SHERPA_ONNX_API typedef struct SherpaOnnxSpeechSegment {
   // The start index in samples of this segment
@@ -862,40 +862,40 @@ typedef struct SherpaOnnxVoiceActivityDetector SherpaOnnxVoiceActivityDetector;
 // Return an instance of VoiceActivityDetector.
 // The user has to use SherpaOnnxDestroyVoiceActivityDetector() to free
 // the returned pointer to avoid memory leak.
-SHERPA_ONNX_API SherpaOnnxVoiceActivityDetector *
+SHERPA_ONNX_API const SherpaOnnxVoiceActivityDetector *
 SherpaOnnxCreateVoiceActivityDetector(const SherpaOnnxVadModelConfig *config,
                                       float buffer_size_in_seconds);
 
 SHERPA_ONNX_API void SherpaOnnxDestroyVoiceActivityDetector(
-    SherpaOnnxVoiceActivityDetector *p);
+    const SherpaOnnxVoiceActivityDetector *p);
 
 SHERPA_ONNX_API void SherpaOnnxVoiceActivityDetectorAcceptWaveform(
-    SherpaOnnxVoiceActivityDetector *p, const float *samples, int32_t n);
+    const SherpaOnnxVoiceActivityDetector *p, const float *samples, int32_t n);
 
 // Return 1 if there are no speech segments available.
 // Return 0 if there are speech segments.
 SHERPA_ONNX_API int32_t
-SherpaOnnxVoiceActivityDetectorEmpty(SherpaOnnxVoiceActivityDetector *p);
+SherpaOnnxVoiceActivityDetectorEmpty(const SherpaOnnxVoiceActivityDetector *p);
 
 // Return 1 if there is voice detected.
 // Return 0 if voice is silent.
-SHERPA_ONNX_API int32_t
+SHERPA_ONNX_API int32_t SherpaOnnxVoiceActivityDetectorDetected(
-SherpaOnnxVoiceActivityDetectorDetected(SherpaOnnxVoiceActivityDetector *p);
+    const SherpaOnnxVoiceActivityDetector *p);
 
 // Return the first speech segment.
 // It throws if SherpaOnnxVoiceActivityDetectorEmpty() returns 1.
 SHERPA_ONNX_API void SherpaOnnxVoiceActivityDetectorPop(
-    SherpaOnnxVoiceActivityDetector *p);
+    const SherpaOnnxVoiceActivityDetector *p);
 
 // Clear current speech segments.
 SHERPA_ONNX_API void SherpaOnnxVoiceActivityDetectorClear(
-    SherpaOnnxVoiceActivityDetector *p);
+    const SherpaOnnxVoiceActivityDetector *p);
 
 // Return the first speech segment.
 // The user has to use SherpaOnnxDestroySpeechSegment() to free the returned
 // pointer to avoid memory leak.
 SHERPA_ONNX_API const SherpaOnnxSpeechSegment *
-SherpaOnnxVoiceActivityDetectorFront(SherpaOnnxVoiceActivityDetector *p);
+SherpaOnnxVoiceActivityDetectorFront(const SherpaOnnxVoiceActivityDetector *p);
 
 // Free the pointer returned SherpaOnnxVoiceActivityDetectorFront().
 SHERPA_ONNX_API void SherpaOnnxDestroySpeechSegment(
@@ -903,10 +903,10 @@ SHERPA_ONNX_API void SherpaOnnxDestroySpeechSegment(
 
 // Re-initialize the voice activity detector.
 SHERPA_ONNX_API void SherpaOnnxVoiceActivityDetectorReset(
-    SherpaOnnxVoiceActivityDetector *p);
+    const SherpaOnnxVoiceActivityDetector *p);
 
 SHERPA_ONNX_API void SherpaOnnxVoiceActivityDetectorFlush(
-    SherpaOnnxVoiceActivityDetector *p);
+    const SherpaOnnxVoiceActivityDetector *p);
 
 // ============================================================
 // For offline Text-to-Speech (i.e., non-streaming TTS)
@@ -1481,15 +1481,16 @@ SHERPA_ONNX_API typedef struct SherpaOnnxLinearResampler
 */
 // The user has to invoke SherpaOnnxDestroyLinearResampler()
 // to free the returned pointer to avoid memory leak
-SHERPA_ONNX_API SherpaOnnxLinearResampler *SherpaOnnxCreateLinearResampler(
+SHERPA_ONNX_API const SherpaOnnxLinearResampler *
-    int32_t samp_rate_in_hz, int32_t samp_rate_out_hz, float filter_cutoff_hz,
+SherpaOnnxCreateLinearResampler(int32_t samp_rate_in_hz,
-    int32_t num_zeros);
+                                int32_t samp_rate_out_hz,
+                                float filter_cutoff_hz, int32_t num_zeros);
 
 SHERPA_ONNX_API void SherpaOnnxDestroyLinearResampler(
-    SherpaOnnxLinearResampler *p);
+    const SherpaOnnxLinearResampler *p);
 
 SHERPA_ONNX_API void SherpaOnnxLinearResamplerReset(
-    SherpaOnnxLinearResampler *p);
+    const SherpaOnnxLinearResampler *p);
 
 typedef struct SherpaOnnxResampleOut {
   const float *samples;
@@ -1501,7 +1502,7 @@ typedef struct SherpaOnnxResampleOut {
 // If this is the last segment, you can set flush to 1; otherwise, please
 // set flush to 0
 SHERPA_ONNX_API const SherpaOnnxResampleOut *SherpaOnnxLinearResamplerResample(
-    SherpaOnnxLinearResampler *p, const float *input, int32_t input_dim,
+    const SherpaOnnxLinearResampler *p, const float *input, int32_t input_dim,
     int32_t flush);
 
 SHERPA_ONNX_API void SherpaOnnxLinearResamplerResampleFree(
@@ -1724,7 +1725,7 @@ SherpaOnnxCreateOfflineRecognizerOHOS(
 // Return an instance of VoiceActivityDetector.
 // The user has to use SherpaOnnxDestroyVoiceActivityDetector() to free
 // the returned pointer to avoid memory leak.
-SHERPA_ONNX_API SherpaOnnxVoiceActivityDetector *
+SHERPA_ONNX_API const SherpaOnnxVoiceActivityDetector *
 SherpaOnnxCreateVoiceActivityDetectorOHOS(
     const SherpaOnnxVadModelConfig *config, float buffer_size_in_seconds,
     NativeResourceManager *mgr);