feature: 集成bert_vits

app/concrete.py

@@ -6,7 +6,8 @@ from .abstract import *
 from .public import *
 from .exception import *
 from .dependency import get_logger
-from utils.vits_utils import TextToSpeech
+from utils.vits_utils import TextToSpeech as VITS_TextToSpeech
+from utils.bert_vits2_utils import TextToSpeech as BertVits_TextToSpeech
 from config import Config
 import threading
 import requests
@@ -17,7 +18,11 @@ import time
 import json
 
 # ----------- 初始化vits ----------- #
-vits = TextToSpeech()
+vits = VITS_TextToSpeech()
+# ---------------------------------- #
+
+# -------- 初始化bert-vits --------- #
+bert_vits = BertVits_TextToSpeech()
 # ---------------------------------- #
 
 # ---------- 初始化logger ---------- #
@@ -294,6 +299,14 @@ class VITS_TTS(TTS):
     def synthetize(self, assistant, text):
         tts_info = json.loads(assistant.tts_info)
         return vits.synthesize(text, tts_info)
+
+class BertVits_TTS(TTS):
+    def __init__(self):
+        pass
+    
+    def synthetize(self, assistant, text):
+        tts_info = json.loads(assistant.tts_info)
+        return bert_vits.synthesize(text, tts_info)
 # --------------------------------- #
 
 
@@ -319,6 +332,8 @@ class TTSFactory:
     def create_tts(self,tts_type:str) -> TTS:
         if tts_type == 'VITS':
             return VITS_TTS()
+        if tts_type == 'BertVits':
+            return BertVits_TTS()
 # --------------------------------- #
 
 
@@ -420,7 +435,12 @@ class Agent():
         self.tts_audio_service_chain.add_service(TTSAudioRecordService())
     
     def init_recorder(self,user_id):
-        self.recorder = Recorder(user_id)
+        input_sr = 16000
+        if isinstance(self.tts, BertVits_TTS):
+            output_sr = 44100
+        elif isinstance(self.tts, VITS_TTS):
+            output_sr = 22050
+        self.recorder = Recorder(user_id,input_sr,output_sr)
     
     # 对用户输入的音频进行预处理
     def user_audio_process(self, audio):