Add Android TTS Engine APK for KittenTTS

android/SherpaOnnxTts/app/src/main/java/com/k2fsa/sherpa/onnx/MainActivity.kt

@@ -192,6 +192,7 @@ class MainActivity : AppCompatActivity() {
         var dataDir: String?
         var dictDir: String?
         var assets: AssetManager? = application.assets
+        var isKitten = false
 
         // The purpose of such a design is to make the CI test easier
         // Please see
@@ -291,6 +292,14 @@ class MainActivity : AppCompatActivity() {
         // lexicon = "kokoro-multi-lang-v1_0/lexicon-us-en.txt,kokoro-multi-lang-v1_0/lexicon-zh.txt"
         // ruleFsts = "$modelDir/phone-zh.fst,$modelDir/date-zh.fst,$modelDir/number-zh.fst"
 
+        // Example 11
+        // kitten-nano-en-v0_1-fp16
+        // modelDir = "kitten-nano-en-v0_1-fp16"
+        // modelName = "model.fp16.onnx"
+        // voices = "voices.bin"
+        // dataDir = "kokoro-multi-lang-v1_0/espeak-ng-data"
+        // isKitten = true
+
         if (dataDir != null) {
             val newDir = copyDataDir(dataDir!!)
             dataDir = "$newDir/$dataDir"
@@ -315,6 +324,7 @@ class MainActivity : AppCompatActivity() {
             dictDir = dictDir ?: "",
             ruleFsts = ruleFsts ?: "",
             ruleFars = ruleFars ?: "",
+            isKitten = isKitten,
         )!!
 
         tts = OfflineTts(assetManager = assets, config = config)

android/SherpaOnnxTtsEngine/app/src/main/java/com/k2fsa/sherpa/onnx/tts/engine/TtsEngine.kt

@@ -53,6 +53,7 @@ object TtsEngine {
     private var dataDir: String? = null
     private var dictDir: String? = null
     private var assets: AssetManager? = null
+    private var isKitten = false
 
     init {
         // The purpose of such a design is to make the CI test easier
@@ -172,6 +173,15 @@ object TtsEngine {
         //
         // This model supports many languages, e.g., English, Chinese, etc.
         // We set lang to eng here.
+
+        // Example 11
+        // kitten-nano-en-v0_1-fp16
+        // modelDir = "kitten-nano-en-v0_1-fp16"
+        // modelName = "model.fp16.onnx"
+        // voices = "voices.bin"
+        // dataDir = "kitten-nano-en-v0_1-fp16/espeak-ng-data"
+        // lang = "eng"
+        // isKitten = true
     }
 
     fun createTts(context: Context) {
@@ -207,7 +217,8 @@ object TtsEngine {
             dataDir = dataDir ?: "",
             dictDir = dictDir ?: "",
             ruleFsts = ruleFsts ?: "",
-            ruleFars = ruleFars ?: ""
+            ruleFars = ruleFars ?: "",
+            isKitten = isKitten,
         )
 
         speed = PreferenceHelper(context).getSpeed()

scripts/apk/build-apk-tts-engine.sh.in

@@ -64,6 +64,11 @@ sed -i.bak s/"lang = null"/"lang = \"$lang_iso_639_3\""/ ./TtsEngine.kt
   sed -i.bak s/"lang2 = null"/"lang2 = \"$lang_iso_639_3_2\""/ ./TtsEngine.kt
 {% endif %}
 
+{% if tts_model.is_kitten %}
+  sed -i.bak s/"isKitten = false"/"isKitten = true"/ ./TtsEngine.kt
+{% endif %}
+
+
 {% if tts_model.model_name %}
   sed -i.bak s/"modelName = null"/"modelName = \"$model_name\""/ ./TtsEngine.kt
 {% endif %}

scripts/apk/build-apk-tts.sh.in

@@ -101,6 +101,10 @@ sed -i.bak s/"modelDir = null"/"modelDir = \"$model_dir\""/ ./MainActivity.kt
   sed -i.bak s%"lexicon = null"%"lexicon = \"$lexicon\""% ./MainActivity.kt
 {% endif %}
 
+{% if tts_model.is_kitten %}
+  sed -i.bak s/"isKitten = false"/"isKitten = true"/ ./MainActivity.kt
+{% endif %}
+
 git diff
 popd
 

scripts/apk/generate-tts-apk-script.py

@@ -44,6 +44,7 @@ class TtsModel:
     lang_iso_639_3: str = ""
     lang_iso_639_3_2: str = ""
     lexicon: str = ""
+    is_kitten: bool = False
 
 
 def convert_lang_to_iso_639_3(models: List[TtsModel]):
@@ -494,6 +495,22 @@ def get_kokoro_models() -> List[TtsModel]:
     return english_models + multi_lingual_models
 
 
+def get_kitten_models() -> List[TtsModel]:
+    english_models = [
+        TtsModel(
+            model_dir="kitten-nano-en-v0_1-fp16",
+            model_name="model.fp16.onnx",
+            lang="en",
+        )
+    ]
+    for m in english_models:
+        m.data_dir = f"{m.model_dir}/espeak-ng-data"
+        m.voices = "voices.bin"
+        m.is_kitten = True
+
+    return english_models
+
+
 def main():
     args = get_args()
     index = args.index
@@ -507,6 +524,7 @@ def main():
     all_model_list += get_coqui_models()
     all_model_list += get_matcha_models()
     all_model_list += get_kokoro_models()
+    all_model_list += get_kitten_models()
 
     convert_lang_to_iso_639_3(all_model_list)
     print(all_model_list)

sherpa-onnx/csrc/piper-phonemize-lexicon.cc

@@ -535,6 +535,10 @@ template PiperPhonemizeLexicon::PiperPhonemizeLexicon(
 template PiperPhonemizeLexicon::PiperPhonemizeLexicon(
     AAssetManager *mgr, const std::string &tokens, const std::string &data_dir,
     const OfflineTtsKokoroModelMetaData &kokoro_meta_data);
+
+template PiperPhonemizeLexicon::PiperPhonemizeLexicon(
+    AAssetManager *mgr, const std::string &tokens, const std::string &data_dir,
+    const OfflineTtsKittenModelMetaData &kokoro_meta_data);
 #endif
 
 #if __OHOS__
@@ -552,6 +556,11 @@ template PiperPhonemizeLexicon::PiperPhonemizeLexicon(
     NativeResourceManager *mgr, const std::string &tokens,
     const std::string &data_dir,
     const OfflineTtsKokoroModelMetaData &kokoro_meta_data);
+
+template PiperPhonemizeLexicon::PiperPhonemizeLexicon(
+    NativeResourceManager *mgr, const std::string &tokens,
+    const std::string &data_dir,
+    const OfflineTtsKittenModelMetaData &kokoro_meta_data);
 #endif
 
 }  // namespace sherpa_onnx