Add Koltin and Java API for Kokoro TTS models (#1728)

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

@@ -39,6 +39,7 @@ model_dir={{ tts_model.model_dir }}
 model_name={{ tts_model.model_name }}
 acoustic_model_name={{ tts_model.acoustic_model_name }}
 vocoder={{ tts_model.vocoder }}
+voices={{ tts_model.voices }}
 lang={{ tts_model.lang }}
 lang_iso_639_3={{ tts_model.lang_iso_639_3 }}
 
@@ -70,6 +71,10 @@ sed -i.bak s/"lang = null"/"lang = \"$lang_iso_639_3\""/ ./TtsEngine.kt
   sed -i.bak s/"vocoder = null"/"vocoder = \"$vocoder\""/ ./TtsEngine.kt
 {% endif %}
 
+{% if tts_model.voices %}
+  sed -i.bak s/"voices = null"/"voices = \"$voices\""/ ./TtsEngine.kt
+{% endif %}
+
 {% if tts_model.rule_fsts %}
   rule_fsts={{ tts_model.rule_fsts }}
   sed -i.bak s%"ruleFsts = null"%"ruleFsts = \"$rule_fsts\""% ./TtsEngine.kt

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

@@ -39,6 +39,7 @@ model_dir={{ tts_model.model_dir }}
 model_name={{ tts_model.model_name }}
 acoustic_model_name={{ tts_model.acoustic_model_name }}
 vocoder={{ tts_model.vocoder }}
+voices={{ tts_model.voices }}
 lang={{ tts_model.lang }}
 
 wget -qq https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/$model_dir.tar.bz2
@@ -69,6 +70,9 @@ sed -i.bak s/"modelDir = null"/"modelDir = \"$model_dir\""/ ./MainActivity.kt
   sed -i.bak s/"vocoder = null"/"vocoder = \"$vocoder\""/ ./MainActivity.kt
 {% endif %}
 
+{% if tts_model.voices %}
+  sed -i.bak s/"voices = null"/"voices = \"$voices\""/ ./MainActivity.kt
+{% endif %}
 
 {% if tts_model.rule_fsts %}
   rule_fsts={{ tts_model.rule_fsts }}

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

@@ -33,6 +33,7 @@ class TtsModel:
     model_name: str = ""  # for vits
     acoustic_model_name: str = ""  # for matcha
     vocoder: str = ""  # for matcha
+    voices: str = ""  # for kokoro
     lang: str = ""  # en, zh, fr, de, etc.
     rule_fsts: Optional[List[str]] = None
     rule_fars: Optional[List[str]] = None
@@ -409,6 +410,21 @@ def get_matcha_models() -> List[TtsModel]:
     return chinese_models + english_models
 
 
+def get_kokoro_models() -> List[TtsModel]:
+    english_models = [
+        TtsModel(
+            model_dir="kokoro-en-v0_19",
+            model_name="model.onnx",
+            lang="en",
+        )
+    ]
+    for m in english_models:
+        m.data_dir = f"{m.model_dir}/espeak-ng-data"
+        m.voices = "voices.bin"
+
+    return english_models
+
+
 def main():
     args = get_args()
     index = args.index
@@ -421,6 +437,7 @@ def main():
     all_model_list += get_mimic3_models()
     all_model_list += get_coqui_models()
     all_model_list += get_matcha_models()
+    all_model_list += get_kokoro_models()
 
     convert_lang_to_iso_639_3(all_model_list)
     print(all_model_list)