Add HarmonyOS support for text-to-speech.

k2-fsa · Dec 1, 2024 · 2c99013 · 2c99013
1 parent a3d6e1a
commit 2c99013
Show file tree

Hide file tree

Showing 15 changed files with 189 additions and 102 deletions.
diff --git a/sherpa-onnx/c-api/c-api.cc b/sherpa-onnx/c-api/c-api.cc
@@ -485,9 +485,9 @@ static sherpa_onnx::OfflineRecognizerConfig GetOfflineRecognizerConfig(
 
   if (config->model_config.debug) {
 #if __OHOS__
-    SHERPA_ONNX_LOGE("%{public}s", recognizer_config.ToString().c_str());
+    SHERPA_ONNX_LOGE("%{public}s\n", recognizer_config.ToString().c_str());
 #else
-    SHERPA_ONNX_LOGE("%s", recognizer_config.ToString().c_str());
+    SHERPA_ONNX_LOGE("%s\n", recognizer_config.ToString().c_str());
 #endif
   }
 
@@ -967,9 +967,9 @@ sherpa_onnx::VadModelConfig GetVadModelConfig(
 
   if (vad_config.debug) {
 #if __OHOS__
-    SHERPA_ONNX_LOGE("%{public}s", vad_config.ToString().c_str());
+    SHERPA_ONNX_LOGE("%{public}s\n", vad_config.ToString().c_str());
 #else
-    SHERPA_ONNX_LOGE("%s", vad_config.ToString().c_str());
+    SHERPA_ONNX_LOGE("%s\n", vad_config.ToString().c_str());
 #endif
   }
 
@@ -1053,7 +1053,7 @@ struct SherpaOnnxOfflineTts {
   std::unique_ptr<sherpa_onnx::OfflineTts> impl;
 };
 
-SherpaOnnxOfflineTts *SherpaOnnxCreateOfflineTts(
+static sherpa_onnx::OfflineTtsConfig GetOfflineTtsConfig(
     const SherpaOnnxOfflineTtsConfig *config) {
   sherpa_onnx::OfflineTtsConfig tts_config;
 
@@ -1084,9 +1084,20 @@ SherpaOnnxOfflineTts *SherpaOnnxCreateOfflineTts(
   tts_config.max_num_sentences = SHERPA_ONNX_OR(config->max_num_sentences, 2);
 
   if (tts_config.model.debug) {
+#if __OHOS__
+    SHERPA_ONNX_LOGE("%{public}s\n", tts_config.ToString().c_str());
+#else
     SHERPA_ONNX_LOGE("%s\n", tts_config.ToString().c_str());
+#endif
   }
 
+  return tts_config;
+}
+
+SherpaOnnxOfflineTts *SherpaOnnxCreateOfflineTts(
+    const SherpaOnnxOfflineTtsConfig *config) {
+  auto tts_config = GetOfflineTtsConfig(config);
+
   if (!tts_config.Validate()) {
     SHERPA_ONNX_LOGE("Errors in config");
     return nullptr;
@@ -1908,6 +1919,7 @@ SherpaOnnxOfflineSpeakerDiarizationProcessWithCallbackNoArg(
 
   return ans;
 }
+#endif
 
 #ifdef __OHOS__
 
@@ -1959,6 +1971,23 @@ SherpaOnnxVoiceActivityDetector *SherpaOnnxCreateVoiceActivityDetectorOHOS(
 
   return p;
 }
-#endif
 
-#endif
+#if SHERPA_ONNX_ENABLE_TTS == 1
+SherpaOnnxOfflineTts *SherpaOnnxCreateOfflineTtsOHOS(
+    const SherpaOnnxOfflineTtsConfig *config, NativeResourceManager *mgr) {
+  if (!mgr) {
+    return SherpaOnnxCreateOfflineTts(config);
+  }
+
+  auto tts_config = GetOfflineTtsConfig(config);
+
+  SherpaOnnxOfflineTts *tts = new SherpaOnnxOfflineTts;
+
+  tts->impl = std::make_unique<sherpa_onnx::OfflineTts>(mgr, tts_config);
+
+  return tts;
+}
+
+#endif  // #if SHERPA_ONNX_ENABLE_TTS == 1
+
+#endif  // #ifdef __OHOS__
diff --git a/sherpa-onnx/c-api/c-api.h b/sherpa-onnx/c-api/c-api.h
@@ -1558,6 +1558,9 @@ SHERPA_ONNX_API SherpaOnnxVoiceActivityDetector *
 SherpaOnnxCreateVoiceActivityDetectorOHOS(
     const SherpaOnnxVadModelConfig *config, float buffer_size_in_seconds,
     NativeResourceManager *mgr);
+
+SHERPA_ONNX_API SherpaOnnxOfflineTts *SherpaOnnxCreateOfflineTtsOHOS(
+    const SherpaOnnxOfflineTtsConfig *config, NativeResourceManager *mgr);
 #endif
 
 #if defined(__GNUC__)

diff --git a/sherpa-onnx/csrc/lexicon.cc b/sherpa-onnx/csrc/lexicon.cc
@@ -7,17 +7,19 @@
 #include <algorithm>
 #include <cctype>
 #include <fstream>
+#include <memory>
 #include <sstream>
+#include <strstream>
 #include <utility>
 
 #if __ANDROID_API__ >= 9
-#include <strstream>
-
 #include "android/asset_manager.h"
 #include "android/asset_manager_jni.h"
 #endif
 
-#include <memory>
+#if __OHOS__
+#include "rawfile/raw_file_manager.h"
+#endif
 
 #include "sherpa-onnx/csrc/macros.h"
 #include "sherpa-onnx/csrc/onnx-utils.h"
@@ -110,8 +112,8 @@ Lexicon::Lexicon(const std::string &lexicon, const std::string &tokens,
   InitPunctuations(punctuations);
 }
 
-#if __ANDROID_API__ >= 9
-Lexicon::Lexicon(AAssetManager *mgr, const std::string &lexicon,
+template <typename Manager>
+Lexicon::Lexicon(Manager *mgr, const std::string &lexicon,
                  const std::string &tokens, const std::string &punctuations,
                  const std::string &language, bool debug /*= false*/
                  )
@@ -132,7 +134,6 @@ Lexicon::Lexicon(AAssetManager *mgr, const std::string &lexicon,
 
   InitPunctuations(punctuations);
 }
-#endif
 
 std::vector<TokenIDs> Lexicon::ConvertTextToTokenIds(
     const std::string &text, const std::string & /*voice*/ /*= ""*/) const {
@@ -371,4 +372,18 @@ void Lexicon::InitPunctuations(const std::string &punctuations) {
   }
 }
 
+#if __ANDROID_API__ >= 9
+template Lexicon::Lexicon(AAssetManager *mgr, const std::string &lexicon,
+                          const std::string &tokens,
+                          const std::string &punctuations,
+                          const std::string &language, bool debug = false);
+#endif
+
+#if __OHOS__
+template Lexicon::Lexicon(NativeResourceManager *mgr,
+                          const std::string &lexicon, const std::string &tokens,
+                          const std::string &punctuations,
+                          const std::string &language, bool debug = false);
+#endif
+
 }  // namespace sherpa_onnx
diff --git a/sherpa-onnx/csrc/lexicon.h b/sherpa-onnx/csrc/lexicon.h
@@ -13,11 +13,6 @@
 #include <unordered_set>
 #include <vector>
 
-#if __ANDROID_API__ >= 9
-#include "android/asset_manager.h"
-#include "android/asset_manager_jni.h"
-#endif
-
 #include "sherpa-onnx/csrc/offline-tts-frontend.h"
 
 namespace sherpa_onnx {
@@ -31,11 +26,10 @@ class Lexicon : public OfflineTtsFrontend {
           const std::string &punctuations, const std::string &language,
           bool debug = false);
 
-#if __ANDROID_API__ >= 9
-  Lexicon(AAssetManager *mgr, const std::string &lexicon,
-          const std::string &tokens, const std::string &punctuations,
-          const std::string &language, bool debug = false);
-#endif
+  template <typename Manager>
+  Lexicon(Manager *mgr, const std::string &lexicon, const std::string &tokens,
+          const std::string &punctuations, const std::string &language,
+          bool debug = false);
 
   std::vector<TokenIDs> ConvertTextToTokenIds(
       const std::string &text, const std::string &voice = "") const override;

diff --git a/sherpa-onnx/csrc/offline-tts-character-frontend.cc b/sherpa-onnx/csrc/offline-tts-character-frontend.cc
@@ -2,20 +2,24 @@
 //
 // Copyright (c)  2023  Xiaomi Corporation
 
-#if __ANDROID_API__ >= 9
-#include <strstream>
-
-#include "android/asset_manager.h"
-#include "android/asset_manager_jni.h"
-#endif
 #include <algorithm>
 #include <cctype>
 #include <codecvt>
 #include <fstream>
 #include <locale>
 #include <sstream>
+#include <strstream>
 #include <utility>
 
+#if __ANDROID_API__ >= 9
+#include "android/asset_manager.h"
+#include "android/asset_manager_jni.h"
+#endif
+
+#if __OHOS__
+#include "rawfile/raw_file_manager.h"
+#endif
+
 #include "sherpa-onnx/csrc/macros.h"
 #include "sherpa-onnx/csrc/offline-tts-character-frontend.h"
 #include "sherpa-onnx/csrc/onnx-utils.h"
@@ -82,18 +86,16 @@ OfflineTtsCharacterFrontend::OfflineTtsCharacterFrontend(
   token2id_ = ReadTokens(is);
 }
 
-#if __ANDROID_API__ >= 9
+template <typename Manager>
 OfflineTtsCharacterFrontend::OfflineTtsCharacterFrontend(
-    AAssetManager *mgr, const std::string &tokens,
+    Manager *mgr, const std::string &tokens,
     const OfflineTtsVitsModelMetaData &meta_data)
     : meta_data_(meta_data) {
   auto buf = ReadFile(mgr, tokens);
   std::istrstream is(buf.data(), buf.size());
   token2id_ = ReadTokens(is);
 }
 
-#endif
-
 std::vector<TokenIDs> OfflineTtsCharacterFrontend::ConvertTextToTokenIds(
     const std::string &_text, const std::string & /*voice = ""*/) const {
   // see
@@ -189,4 +191,18 @@ std::vector<TokenIDs> OfflineTtsCharacterFrontend::ConvertTextToTokenIds(
   return ans;
 }
 
+#if __ANDROID_API__ >= 9
+template OfflineTtsCharacterFrontend::OfflineTtsCharacterFrontend(
+    AAssetManager *mgr, const std::string &tokens,
+    const OfflineTtsVitsModelMetaData &meta_data);
+
+#endif
+
+#if __OHOS__
+template OfflineTtsCharacterFrontend::OfflineTtsCharacterFrontend(
+    NativeResourceManager *mgr, const std::string &tokens,
+    const OfflineTtsVitsModelMetaData &meta_data);
+
+#endif
+
 }  // namespace sherpa_onnx
diff --git a/sherpa-onnx/csrc/offline-tts-character-frontend.h b/sherpa-onnx/csrc/offline-tts-character-frontend.h
@@ -9,11 +9,6 @@
 #include <unordered_map>
 #include <vector>
 
-#if __ANDROID_API__ >= 9
-#include "android/asset_manager.h"
-#include "android/asset_manager_jni.h"
-#endif
-
 #include "sherpa-onnx/csrc/offline-tts-frontend.h"
 #include "sherpa-onnx/csrc/offline-tts-vits-model-metadata.h"
 
@@ -24,11 +19,10 @@ class OfflineTtsCharacterFrontend : public OfflineTtsFrontend {
   OfflineTtsCharacterFrontend(const std::string &tokens,
                               const OfflineTtsVitsModelMetaData &meta_data);
 
-#if __ANDROID_API__ >= 9
-  OfflineTtsCharacterFrontend(AAssetManager *mgr, const std::string &tokens,
+  template <typename Manager>
+  OfflineTtsCharacterFrontend(Manager *mgr, const std::string &tokens,
                               const OfflineTtsVitsModelMetaData &meta_data);
 
-#endif
   /** Convert a string to token IDs.
    *
    * @param text The input text.

diff --git a/sherpa-onnx/csrc/offline-tts-impl.cc b/sherpa-onnx/csrc/offline-tts-impl.cc
@@ -6,6 +6,15 @@
 
 #include <memory>
 
+#if __ANDROID_API__ >= 9
+#include "android/asset_manager.h"
+#include "android/asset_manager_jni.h"
+#endif
+
+#if __OHOS__
+#include "rawfile/raw_file_manager.h"
+#endif
+
 #include "sherpa-onnx/csrc/offline-tts-vits-impl.h"
 
 namespace sherpa_onnx {
@@ -16,12 +25,21 @@ std::unique_ptr<OfflineTtsImpl> OfflineTtsImpl::Create(
   return std::make_unique<OfflineTtsVitsImpl>(config);
 }
 
-#if __ANDROID_API__ >= 9
+template <typename Manager>
 std::unique_ptr<OfflineTtsImpl> OfflineTtsImpl::Create(
-    AAssetManager *mgr, const OfflineTtsConfig &config) {
+    Manager *mgr, const OfflineTtsConfig &config) {
   // TODO(fangjun): Support other types
   return std::make_unique<OfflineTtsVitsImpl>(mgr, config);
 }
+
+#if __ANDROID_API__ >= 9
+template std::unique_ptr<OfflineTtsImpl> OfflineTtsImpl::Create(
+    AAssetManager *mgr, const OfflineTtsConfig &config);
+#endif
+
+#if __OHOS__
+template std::unique_ptr<OfflineTtsImpl> OfflineTtsImpl::Create(
+    NativeResourceManager *mgr, const OfflineTtsConfig &config);
 #endif
 
 }  // namespace sherpa_onnx
diff --git a/sherpa-onnx/csrc/offline-tts-impl.h b/sherpa-onnx/csrc/offline-tts-impl.h
@@ -8,11 +8,6 @@
 #include <memory>
 #include <string>
 
-#if __ANDROID_API__ >= 9
-#include "android/asset_manager.h"
-#include "android/asset_manager_jni.h"
-#endif
-
 #include "sherpa-onnx/csrc/offline-tts.h"
 
 namespace sherpa_onnx {
@@ -23,10 +18,9 @@ class OfflineTtsImpl {
 
   static std::unique_ptr<OfflineTtsImpl> Create(const OfflineTtsConfig &config);
 
-#if __ANDROID_API__ >= 9
-  static std::unique_ptr<OfflineTtsImpl> Create(AAssetManager *mgr,
+  template <typename Manager>
+  static std::unique_ptr<OfflineTtsImpl> Create(Manager *mgr,
                                                 const OfflineTtsConfig &config);
-#endif
 
   virtual GeneratedAudio Generate(
       const std::string &text, int64_t sid = 0, float speed = 1.0,