feat: 后端封装tts，可以通过配置文件切换openvoice和vits

utils/tts/vits_utils.py

@@ -2,6 +2,7 @@ import os
 import numpy as np
 import torch
 from torch import LongTensor
+from typing import Optional
 import soundfile as sf
 # vits
 from .vits import utils, commons
@@ -79,19 +80,19 @@ class TextToSpeech:
             print(f"Synthesis time: {time.time() - start_time} s")
         return audio
     
-    def synthesize(self, text, language, speaker_id, noise_scale, noise_scale_w, length_scale, save_audio=False, return_bytes=False):
+    def synthesize(self, text, tts_info,target_se: Optional[np.ndarray]=None, save_audio=False, return_bytes=True):
         if not len(text):
             return "输入文本不能为空！", None
         text = text.replace('\n', ' ').replace('\r', '').replace(" ", "")
         if len(text) > 100 and self.limitation:
             return f"输入文字过长！{len(text)}>100", None
-        text = self._preprocess_text(text, language)
-        audio = self._generate_audio(text, speaker_id, noise_scale, noise_scale_w, length_scale)
+        text = self._preprocess_text(text, tts_info['language'])
+        audio = self._generate_audio(text, tts_info['speaker_id'], tts_info['noise_scale'], tts_info['noise_scale_w'], tts_info['length_scale'])
         if self.debug or save_audio:
             self.save_audio(audio, self.RATE, 'output_file.wav')
         if return_bytes:
             audio = self.convert_numpy_to_bytes(audio)
-        return self.RATE, audio
+        return audio, self.RATE
 
     def convert_numpy_to_bytes(self, audio_data):
         if isinstance(audio_data, np.ndarray):