feat: tts

src/asr/asr.py

@@ -6,11 +6,10 @@ from fastapi.responses import JSONResponse
 
 from .rapid_paraformer.utils import read_yaml
 from .rapid_paraformer import RapidParaformer
-from .asr_service import ASRService
 from ..blackbox.blackbox import Blackbox
 
 class ASR(Blackbox):
-    
+
     def __init__(self, config: any) -> None:
         config = read_yaml(config)
         self.paraformer = RapidParaformer(config)

src/blackbox/blackbox.py

@@ -8,7 +8,7 @@ class Blackbox(ABC):
     the methods processing, valid and fast_api_handler.
     If implemented correctly, the blackbox class can be used in the main.py file
     """
-    def __init__(self, config: any) -> None:
+    def __init__(self, *args, **kwargs) -> None:
         pass
 
     """

src/blackbox/blackbox_factory.py

@@ -1,3 +1,4 @@
+from .tts import TTS
 from ..asr.asr import ASR
 from .audio_to_text import AudioToText
 from .blackbox import Blackbox
@@ -8,8 +9,8 @@ from .text_to_audio import TextToAudio
 class BlackboxFactory:
 
     def __init__(self) -> None:
+        self.tts = TTS()
         self.asr = ASR("./.env.yaml")
-        pass
 
     def create_blackbox(self, blackbox_name: str, blackbox_config: dict) -> Blackbox:
         if blackbox_name == "audio_to_text":
@@ -20,4 +21,6 @@ class BlackboxFactory:
             return Calculator(blackbox_config)
         if blackbox_name == "asr":
             return self.asr
+        if blackbox_name == "tts":
+            return self.tts
         raise ValueError("Invalid blockbox type")

src/blackbox/tts.py

@@ -0,0 +1,36 @@
+import io
+from typing import Any, Coroutine
+
+from fastapi import Request, Response, status
+from fastapi.responses import JSONResponse
+from .blackbox import Blackbox
+from tts.tts_service import TTService
+
+class TTS(Blackbox):
+
+    def __init__(self) -> None:
+        config = {
+            'paimon': ['resources/tts/models/paimon6k.json', 'resources/tts/models/paimon6k_390k.pth', 'character_paimon', 1],
+            'yunfei': ['resources/tts/models/yunfeimix2.json', 'resources/tts/models/yunfeimix2_53k.pth', 'character_yunfei', 1.1],
+            'catmaid': ['resources/tts/models/catmix.json', 'resources/tts/models/catmix_107k.pth', 'character_catmaid', 1.2]
+        }
+        self.tts_service = TTService(*config['catmaid'])
+        super().__init__(config)
+    
+    def processing(self, text: str) -> io.BytesIO:
+        audio = self.tts_service.read(text)
+        return audio
+
+    def valid(self, txt: any) -> bool:
+        return isinstance(txt, str)
+
+    async def fast_api_handler(self, request: Request) -> Response:
+        try:
+            data = await request.json()
+        except:
+            return JSONResponse(content={"error": "json parse error"}, status_code=status.HTTP_400_BAD_REQUEST)
+        text = data.get("text")
+        if text is None:
+            return JSONResponse(content={"error": "text is required"}, status_code=status.HTTP_400_BAD_REQUEST)
+        by = self.processing(text)
+        return Response(content=by.read(), media_type="audio/wav", headers={"Content-Disposition": "attachment; filename=audio.wav"})