Update TTS Engine APK to support multi-lang (#2294)

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

@@ -23,6 +23,9 @@ object TtsEngine {
     // cmn for Mandarin
     var lang: String? = null
 
+    // if a model supports two languages, set also lang2
+    var lang2: String? = null
+
 
     val speedState: MutableState<Float> = mutableFloatStateOf(1.0F)
     val speakerIdState: MutableState<Int> = mutableIntStateOf(0)
@@ -76,6 +79,7 @@ object TtsEngine {
         dataDir = null
         dictDir = null
         lang = null
+        lang2 = null
 
         // Please enable one and only one of the examples below
 
@@ -125,6 +129,7 @@ object TtsEngine {
         // lexicon = "lexicon.txt"
         // dictDir = "vits-melo-tts-zh_en/dict"
         // lang = "zho"
+        // lang2 = "eng"
 
         // Example 7
         // matcha-icefall-zh-baker
@@ -162,6 +167,7 @@ object TtsEngine {
         // dictDir = "kokoro-multi-lang-v1_0/dict"
         // lexicon = "kokoro-multi-lang-v1_0/lexicon-us-en.txt,kokoro-multi-lang-v1_0/lexicon-zh.txt"
         // lang = "eng"
+        // lang2 = "zho"
         // ruleFsts = "$modelDir/phone-zh.fst,$modelDir/date-zh.fst,$modelDir/number-zh.fst"
         //
         // This model supports many languages, e.g., English, Chinese, etc.

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

@@ -60,6 +60,9 @@ class TtsService : TextToSpeechService() {
 
         // see https://github.com/Miserlou/Android-SDK-Samples/blob/master/TtsEngine/src/com/example/android/ttsengine/RobotSpeakTtsService.java#L68
         onLoadLanguage(TtsEngine.lang, "", "")
+        if (TtsEngine.lang2 != null) {
+            onLoadLanguage(TtsEngine.lang2, "", "")
+        }
     }
 
     override fun onDestroy() {
@@ -71,7 +74,7 @@ class TtsService : TextToSpeechService() {
     override fun onIsLanguageAvailable(_lang: String?, _country: String?, _variant: String?): Int {
         val lang = _lang ?: ""
 
-        if (lang == TtsEngine.lang) {
+        if (lang == TtsEngine.lang || lang == TtsEngine.lang2) {
             return TextToSpeech.LANG_AVAILABLE
         }
 
@@ -87,12 +90,12 @@ class TtsService : TextToSpeechService() {
         Log.i(TAG, "onLoadLanguage: $_lang, $_country")
         val lang = _lang ?: ""
 
-        return if (lang == TtsEngine.lang) {
+        return if (lang == TtsEngine.lang || lang == TtsEngine.lang2) {
             Log.i(TAG, "creating tts, lang :$lang")
             TtsEngine.createTts(application)
             TextToSpeech.LANG_AVAILABLE
         } else {
-            Log.i(TAG, "lang $lang not supported, tts engine lang: ${TtsEngine.lang}")
+            Log.i(TAG, "lang $lang not supported, tts engine lang: ${TtsEngine.lang}, ${TtsEngine.lang2}")
             TextToSpeech.LANG_NOT_SUPPORTED
         }
     }

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

@@ -42,6 +42,7 @@ vocoder={{ tts_model.vocoder }}
 voices={{ tts_model.voices }}
 lang={{ tts_model.lang }}
 lang_iso_639_3={{ tts_model.lang_iso_639_3 }}
+lang_iso_639_3_2={{ tts_model.lang_iso_639_3_2 }}
 
 wget -qq https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/$model_dir.tar.bz2
 tar xf $model_dir.tar.bz2
@@ -59,6 +60,10 @@ pushd android/SherpaOnnxTtsEngine/app/src/main/java/com/k2fsa/sherpa/onnx/tts/en
 sed -i.bak s/"modelDir = null"/"modelDir = \"$model_dir\""/ ./TtsEngine.kt
 sed -i.bak s/"lang = null"/"lang = \"$lang_iso_639_3\""/ ./TtsEngine.kt
 
+{% if tts_model.lang2 %}
+  sed -i.bak s/"lang2 = null"/"lang2 = \"$lang_iso_639_3_2\""/ ./TtsEngine.kt
+{% endif %}
+
 {% if tts_model.model_name %}
   sed -i.bak s/"modelName = null"/"modelName = \"$model_name\""/ ./TtsEngine.kt
 {% endif %}
@@ -109,7 +114,7 @@ if [[ $model_dir == vits-melo-tts-zh_en ]]; then
   lang=zh_en
 fi
 
-if [[ $model_dir == kokoro-multi-lang-v1_0 ]]; then
+if [[ $model_dir == kokoro-multi-lang-v1_0 || $model_dir == kokoro-multi-lang-v1_1 || $model_dir == kokoro-int8-multi-lang-v1_1 ]]; then
   lang=zh_en
 fi
 

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

@@ -108,7 +108,7 @@ if [[ $model_dir == vits-melo-tts-zh_en ]]; then
   lang=zh_en
 fi
 
-if [[ $model_dir == kokoro-multi-lang-v1_0 ]]; then
+if [[ $model_dir == kokoro-multi-lang-v1_0 || $model_dir == kokoro-multi-lang-v1_1 || $model_dir == kokoro-int8-multi-lang-v1_1 ]]; then
   lang=zh_en
 fi
 

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

@@ -35,12 +35,14 @@ class TtsModel:
     vocoder: str = ""  # for matcha
     voices: str = ""  # for kokoro
     lang: str = ""  # en, zh, fr, de, etc.
+    lang2: str = ""  # en, zh, fr, de, etc.
     rule_fsts: Optional[List[str]] = None
     rule_fars: Optional[List[str]] = None
     data_dir: Optional[str] = None
     dict_dir: Optional[str] = None
     is_char: bool = False
     lang_iso_639_3: str = ""
+    lang_iso_639_3_2: str = ""
     lexicon: str = ""
 
 
@@ -48,6 +50,8 @@ def convert_lang_to_iso_639_3(models: List[TtsModel]):
     for m in models:
         if m.lang_iso_639_3 == "":
             m.lang_iso_639_3 = Lang(m.lang).pt3
+        if m.lang2 != "":
+            m.lang_iso_639_3_2 = Lang(m.lang2).pt3
 
 
 def get_coqui_models() -> List[TtsModel]:
@@ -322,6 +326,7 @@ def get_vits_models() -> List[TtsModel]:
             model_dir="vits-melo-tts-zh_en",
             model_name="model.onnx",
             lang="zh",
+            lang2="en",
         ),
         TtsModel(
             model_dir="vits-zh-hf-fanchen-C",
@@ -438,16 +443,19 @@ def get_kokoro_models() -> List[TtsModel]:
             model_dir="kokoro-multi-lang-v1_0",
             model_name="model.onnx",
             lang="en",
+            lang2="zh",
         ),
         TtsModel(
             model_dir="kokoro-multi-lang-v1_1",
             model_name="model.onnx",
             lang="en",
+            lang2="zh",
         ),
         TtsModel(
             model_dir="kokoro-int8-multi-lang-v1_1",
             model_name="model.int8.onnx",
             lang="en",
+            lang2="zh",
         ),
     ]
     for m in multi_lingual_models: