Add HarmonyOS support for text-to-speech. (#1584)

Author: csukuangfj

.github/workflows/lazarus.yaml

@@ -56,7 +56,7 @@ jobs:
           key: ${{ matrix.os }}
 
       # See https://github.com/gcarreno/setup-lazarus
-      - uses: gcarreno/setup-lazarus@v3
+      - uses: gcarreno/setup-lazarus@v3.3.1
         with:
           lazarus-version: "stable"
           with-cache: true

.github/workflows/test-nodejs-npm.yaml

@@ -26,7 +26,7 @@ jobs:
       fail-fast: false
       matrix:
         os: [ubuntu-latest, macos-latest, windows-2019]
-        python-version: ["3.8"]
+        python-version: ["3.10"]
 
     steps:
       - uses: actions/checkout@v4

harmony-os/SherpaOnnxHar/sherpa_onnx/BuildProfile.ets

@@ -2,8 +2,8 @@
  * Use these variables when you tailor your ArkTS code. They must be of the const type.
  */
 export const HAR_VERSION = '1.10.32';
-export const BUILD_MODE_NAME = 'release';
-export const DEBUG = false;
+export const BUILD_MODE_NAME = 'debug';
+export const DEBUG = true;
 export const TARGET_NAME = 'default';
 
 /**

harmony-os/SherpaOnnxHar/sherpa_onnx/Index.ets

@@ -38,3 +38,12 @@ export {
   OnlineRecognizerResult,
   OnlineRecognizer,
 } from './src/main/ets/components/StreamingAsr';
+
+export {
+  OfflineTtsVitsModelConfig,
+  OfflineTtsModelConfig,
+  OfflineTtsConfig,
+  OfflineTts,
+  TtsOutput,
+  TtsInput,
+} from './src/main/ets/components/NonStreamingTts';

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/macros.h

@@ -8,8 +8,8 @@
 #include <string>
 
 #if __OHOS__
-#include "rawfile/raw_file_manager.h"
 #include "hilog/log.h"
+#include "rawfile/raw_file_manager.h"
 
 #undef LOG_DOMAIN
 #undef LOG_TAG

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

@@ -236,7 +236,10 @@ CreateOfflineRecognizerWrapper(const Napi::CallbackInfo &info) {
   SHERPA_ONNX_ASSIGN_ATTR_FLOAT(blank_penalty, blankPenalty);
 
 #if __OHOS__
-  std::unique_ptr<NativeResourceManager, decltype(&OH_ResourceManager_ReleaseNativeResourceManager)> mgr (OH_ResourceManager_InitNativeResourceManager(env, info[1]), &OH_ResourceManager_ReleaseNativeResourceManager);
+  std::unique_ptr<NativeResourceManager,
+                  decltype(&OH_ResourceManager_ReleaseNativeResourceManager)>
+      mgr(OH_ResourceManager_InitNativeResourceManager(env, info[1]),
+          &OH_ResourceManager_ReleaseNativeResourceManager);
 
   const SherpaOnnxOfflineRecognizer *recognizer =
       SherpaOnnxCreateOfflineRecognizerOHOS(&c, mgr.get());

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

@@ -63,6 +63,17 @@ static SherpaOnnxOfflineTtsModelConfig GetOfflineTtsModelConfig(
 static Napi::External<SherpaOnnxOfflineTts> CreateOfflineTtsWrapper(
     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();
@@ -71,6 +82,7 @@ static Napi::External<SherpaOnnxOfflineTts> CreateOfflineTtsWrapper(
 
     return {};
   }
+#endif
 
   if (!info[0].IsObject()) {
     Napi::TypeError::New(env, "Expect an object as the argument")
@@ -90,7 +102,15 @@ static Napi::External<SherpaOnnxOfflineTts> CreateOfflineTtsWrapper(
   SHERPA_ONNX_ASSIGN_ATTR_INT32(max_num_sentences, maxNumSentences);
   SHERPA_ONNX_ASSIGN_ATTR_STR(rule_fars, ruleFars);
 
+#if __OHOS__
+  std::unique_ptr<NativeResourceManager,
+                  decltype(&OH_ResourceManager_ReleaseNativeResourceManager)>
+      mgr(OH_ResourceManager_InitNativeResourceManager(env, info[1]),
+          &OH_ResourceManager_ReleaseNativeResourceManager);
+  SherpaOnnxOfflineTts *tts = SherpaOnnxCreateOfflineTtsOHOS(&c, mgr.get());
+#else
   SherpaOnnxOfflineTts *tts = SherpaOnnxCreateOfflineTts(&c);
+#endif
 
   if (c.model.vits.model) {
     delete[] c.model.vits.model;

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

@@ -211,7 +211,10 @@ static Napi::External<SherpaOnnxOnlineRecognizer> CreateOnlineRecognizerWrapper(
   c.ctc_fst_decoder_config = GetCtcFstDecoderConfig(o);
 
 #if __OHOS__
-  std::unique_ptr<NativeResourceManager, decltype(&OH_ResourceManager_ReleaseNativeResourceManager)> mgr (OH_ResourceManager_InitNativeResourceManager(env, info[1]), &OH_ResourceManager_ReleaseNativeResourceManager);
+  std::unique_ptr<NativeResourceManager,
+                  decltype(&OH_ResourceManager_ReleaseNativeResourceManager)>
+      mgr(OH_ResourceManager_InitNativeResourceManager(env, info[1]),
+          &OH_ResourceManager_ReleaseNativeResourceManager);
 
   const SherpaOnnxOnlineRecognizer *recognizer =
       SherpaOnnxCreateOnlineRecognizerOHOS(&c, mgr.get());

harmony-os/SherpaOnnxHar/sherpa_onnx/src/main/cpp/types/libsherpa_onnx/Index.d.ts

@@ -33,3 +33,8 @@ export const decodeOnlineStream: (handle: object, streamHandle: object) => void;
 export const isEndpoint: (handle: object, streamHandle: object) => boolean;
 export const reset: (handle: object, streamHandle: object) => void;
 export const getOnlineStreamResultAsJson: (handle: object, streamHandle: object) => string;
+
+export const createOfflineTts: (config: object, mgr?: object) => object;
+export const getOfflineTtsNumSpeakers: (handle: object) => number;
+export const getOfflineTtsSampleRate: (handle: object) => number;
+export const offlineTtsGenerate: (handle: object, input: object) => object;

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

@@ -70,8 +70,10 @@ static void CircularBufferPushWrapper(const Napi::CallbackInfo &info) {
 
 #if __OHOS__
   // Note(fangjun): Normally, we don't need to divied it by sizeof(float).
-  // However, data.ElementLength() here returns number of bytes, not number of elements.
-  SherpaOnnxCircularBufferPush(buf, data.Data(), data.ElementLength() / sizeof(float));
+  // However, data.ElementLength() here returns number of bytes, not number of
+  // elements.
+  SherpaOnnxCircularBufferPush(buf, data.Data(),
+                               data.ElementLength() / sizeof(float));
 #else
   SherpaOnnxCircularBufferPush(buf, data.Data(), data.ElementLength());
 #endif
@@ -353,10 +355,14 @@ CreateVoiceActivityDetectorWrapper(const Napi::CallbackInfo &info) {
   float buffer_size_in_seconds = info[1].As<Napi::Number>().FloatValue();
 
 #if __OHOS__
-  std::unique_ptr<NativeResourceManager, decltype(&OH_ResourceManager_ReleaseNativeResourceManager)> mgr(OH_ResourceManager_InitNativeResourceManager(env, info[2]), &OH_ResourceManager_ReleaseNativeResourceManager);
+  std::unique_ptr<NativeResourceManager,
+                  decltype(&OH_ResourceManager_ReleaseNativeResourceManager)>
+      mgr(OH_ResourceManager_InitNativeResourceManager(env, info[2]),
+          &OH_ResourceManager_ReleaseNativeResourceManager);
 
   SherpaOnnxVoiceActivityDetector *vad =
-      SherpaOnnxCreateVoiceActivityDetectorOHOS(&c, buffer_size_in_seconds, mgr.get());
+      SherpaOnnxCreateVoiceActivityDetectorOHOS(&c, buffer_size_in_seconds,
+                                                mgr.get());
 #else
   SherpaOnnxVoiceActivityDetector *vad =
       SherpaOnnxCreateVoiceActivityDetector(&c, buffer_size_in_seconds);
@@ -410,9 +416,10 @@ static void VoiceActivityDetectorAcceptWaveformWrapper(
   Napi::Float32Array samples = info[1].As<Napi::Float32Array>();
 
 #if __OHOS__
-  // Note(fangjun): For unknown reasons, we need to use `/sizeof(float)` here for Huawei
-  SherpaOnnxVoiceActivityDetectorAcceptWaveform(vad, samples.Data(),
-                                                samples.ElementLength() / sizeof(float));
+  // Note(fangjun): For unknown reasons, we need to use `/sizeof(float)` here
+  // for Huawei
+  SherpaOnnxVoiceActivityDetectorAcceptWaveform(
+      vad, samples.Data(), samples.ElementLength() / sizeof(float));
 #else
   SherpaOnnxVoiceActivityDetectorAcceptWaveform(vad, samples.Data(),
                                                 samples.ElementLength());