feat: Add Typecast TTS (JaeYi voice) - https://typecast.ai/

models.py

@@ -565,6 +565,7 @@ def insert_initial_models():
     has_supertone = bool(os.getenv("SUPERTONE_API_KEY"))
     has_clova = bool(os.getenv("CLOVA_CLIENT_ID") and os.getenv("CLOVA_API_KEY"))
     has_humelo = bool(os.getenv("HUMELO_API_KEY"))
+    has_typecast = bool(os.getenv("TYPECAST_API_KEY"))
     
     tts_models = [
         # 채널톡 TTS (한국어 특화) - 항상 활성화
@@ -638,6 +639,15 @@ def insert_initial_models():
             is_active=has_humelo,
             model_url="https://humelo.com/",
         ),
+        # Typecast TTS (한국어 특화) - API 키 있을 때만 활성화
+        Model(
+            id="typecast-jaeyi",
+            name="Typecast (JaeYi)",
+            model_type=ModelType.TTS,
+            is_open=False,
+            is_active=has_typecast,
+            model_url="https://typecast.ai/",
+        ),
     ]
 
     for model in tts_models:

requirements.txt

@@ -13,4 +13,5 @@ waitress
 huggingface-hub
 scipy
 numpy
-pydub
+pydub
+typecast-python

tts.py

@@ -52,6 +52,9 @@ CLOVA_API_KEY = os.getenv("CLOVA_API_KEY")
 HUMELO_API_KEY = os.getenv("HUMELO_API_KEY")
 HUMELO_API_URL = "https://agitvxptajouhvoatxio.supabase.co/functions/v1/dive-synthesize-v1"
 
+# Typecast TTS
+TYPECAST_API_KEY = os.getenv("TYPECAST_API_KEY")
+
 def resample_wav_to_16khz(input_path: str) -> str:
     """
     Resample a WAV file to 16kHz for fair comparison.
@@ -204,6 +207,12 @@ model_mapping = {
         "voice": "리아",
         "emotion": "neutral",
     },
+    # Typecast TTS (한국어 특화)
+    "typecast-jaeyi": {
+        "provider": "typecast",
+        "voice_id": "tc_612ed01c7eb720fddd3ddedf",
+        "model": "ssfm-v21",
+    },
 }
 
 
@@ -402,6 +411,33 @@ def predict_humelo_tts(text: str, voice: str = "리아", emotion: str = "neutral
         return f.name
 
 
+def predict_typecast_tts(text: str, voice_id: str = "tc_612ed01c7eb720fddd3ddedf", model: str = "ssfm-v21") -> str:
+    """Typecast TTS API 호출"""
+    api_key = TYPECAST_API_KEY
+    if not api_key:
+        raise ValueError("TYPECAST_API_KEY 환경 변수가 설정되지 않았습니다.")
+    
+    try:
+        from typecast.client import Typecast
+        from typecast.models import TTSRequest
+        
+        cli = Typecast(api_key=api_key)
+        response = cli.text_to_speech(TTSRequest(
+            text=text,
+            model=model,
+            voice_id=voice_id,
+        ))
+        
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as f:
+            f.write(response.audio_data)
+            return f.name
+            
+    except ImportError:
+        raise ValueError("typecast-python 패키지가 설치되지 않았습니다. pip install typecast-python")
+    except Exception as e:
+        raise ValueError(f"Typecast TTS API 오류: {str(e)}")
+
+
 def predict_google_tts(text: str, voice: str = "ko-KR-Wavenet-A") -> str:
     """Google Cloud TTS API 호출"""
     api_key = os.getenv("GOOGLE_API_KEY")
@@ -494,6 +530,14 @@ def predict_tts(text: str, model: str) -> str:
         # Humelo might return MP3 or WAV, check extension
         is_mp3 = audio_path.endswith(".mp3")
     
+    elif provider == "typecast":
+        audio_path = predict_typecast_tts(
+            text,
+            config.get("voice_id", "tc_612ed01c7eb720fddd3ddedf"),
+            config.get("model", "ssfm-v21"),
+        )
+        # Typecast returns WAV
+    
     else:
         raise ValueError(f"알 수 없는 provider: {provider}")