Add C++ runtime for vocos (#2014)

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

@@ -395,7 +395,7 @@ def get_matcha_models() -> List[TtsModel]:
         s = [f"{m.model_dir}/{r}" for r in rule_fsts]
         m.rule_fsts = ",".join(s)
         m.dict_dir = m.model_dir + "/dict"
-        m.vocoder = "hifigan_v2.onnx"
+        m.vocoder = "vocos-22khz-univ.onnx"
 
     english_persian_models = [
         TtsModel(
@@ -416,7 +416,7 @@ def get_matcha_models() -> List[TtsModel]:
     ]
     for m in english_persian_models:
         m.data_dir = f"{m.model_dir}/espeak-ng-data"
-        m.vocoder = "hifigan_v2.onnx"
+        m.vocoder = "vocos-22khz-univ.onnx"
 
     return chinese_models + english_persian_models
 

scripts/matcha-tts/fa-en/run.sh

@@ -20,8 +20,8 @@ if [ ! -f male/tokens.txt ]; then
   curl -SL --output male/tokens.txt https://huggingface.co/mah92/Musa-FA_EN-Matcha-TTS-Model/resolve/main/tokens_sherpa_with_fa.txt
 fi
 
-if [ ! -f hifigan_v2.onnx ]; then
-  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/vocoder-models/hifigan_v2.onnx
+if [ ! -f vocos-22khz-univ.onnx ]; then
+  curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/vocoder-models/vocos-22khz-univ.onnx
 fi
 
 if [ ! -f .add-meta-data.done ]; then
@@ -31,14 +31,14 @@ fi
 
 python3 ./test.py \
   --am ./female/model.onnx \
-  --vocoder ./hifigan_v2.onnx \
+  --vocoder ./vocos-22khz-univ.onnx \
   --tokens ./female/tokens.txt \
   --text "This is a test. این یک نمونه ی تست فارسی است." \
   --out-wav "./female-en-fa.wav"
 
 python3 ./test.py \
   --am ./male/model.onnx \
-  --vocoder ./hifigan_v2.onnx \
+  --vocoder ./vocos-22khz-univ.onnx \
   --tokens ./male/tokens.txt \
   --text "This is a test. این یک نمونه ی تست فارسی است." \
   --out-wav "./male-en-fa.wav"