refactor: do not use the model method to convert in WAV

tts/tts.py

@@ -1,6 +1,7 @@
 from typing import TYPE_CHECKING
 from pathlib import Path
 from io import BytesIO
+import wave
 
 import torch
 from torch.package import PackageImporter
@@ -42,7 +43,8 @@ def generate(self, text: str, speaker: str, sample_rate: int) -> bytes:
             raise InvalidSampleRateException(sample_rate)
 
         text = self._delete_dashes(text)
-        return self._generate_audio(model, text, speaker, sample_rate)
+        tensor = self._generate_audio(model, text, speaker, sample_rate)
+        return self._convert_to_wav(tensor, sample_rate)
 
     def _load_model(self, model_path: Path):
         package = PackageImporter(model_path)
@@ -70,30 +72,29 @@ def _delete_dashes(self, text: str) -> str:
 
     def _generate_audio(
         self, model: "TTSModelMultiAcc_v3", text: str, speaker: str, sample_rate: int
-    ) -> bytes:
+    ) -> torch.Tensor:
         try:
-            audio: torch.Tensor = model.apply_tts(text=text, speaker=speaker, sample_rate=sample_rate)
+            return model.apply_tts(text=text, speaker=speaker, sample_rate=sample_rate)
         except ValueError:
             raise NotCorrectTextException(text)
         except Exception as error:
             if str(error) == "Model couldn't generate your text, probably it's too long":
                 raise TextTooLongException(text)
             raise
-        else:
-            return self._convert_to_wav(model, audio, sample_rate)
-
-    def _convert_to_wav(self, model: "TTSModelMultiAcc_v3", audio: torch.Tensor, sample_rate: int) -> bytes:
-        with BytesIO() as buffer:
-            model.write_wave(
-                buffer,
-                audio=self._normalize_audio(audio),
-                sample_rate=sample_rate,
-            )
+
+    def _convert_to_wav(self, tensor: torch.Tensor, sample_rate: int) -> bytes:
+        audio = self._normalize_audio(tensor)
+        with BytesIO() as buffer, wave.open(buffer, 'wb') as wav:
+            wav.setnchannels(1)  # mono
+            wav.setsampwidth(2)  # quality is 16 bit. Do not change
+            wav.setframerate(sample_rate)
+            wav.writeframes(audio)
+
             buffer.seek(0)
             return buffer.read()
 
-    def _normalize_audio(self, audio: torch.Tensor):
-        audio: np.ndarray = audio.numpy() * MAX_INT16
+    def _normalize_audio(self, tensor: torch.Tensor):
+        audio: np.ndarray = tensor.numpy() * MAX_INT16
         return audio.astype(np.int16)
 
 tts = TTS()