Update app.py

app.py

@@ -4,7 +4,7 @@ import shutil
 import secrets
 from fastapi import FastAPI, Depends, HTTPException, status, File, UploadFile, Form
 from fastapi.security import HTTPBasic, HTTPBasicCredentials
-from fastapi.responses import FileResponse, JSONResponse
+from fastapi.responses import FileResponse
 from TTS.api import TTS
 
 # --- KONFIGURASI FOLDER TEMP ---
@@ -13,82 +13,79 @@ AUDIO_DIR = "/tmp/audio_output"
 os.makedirs(TEMP_DIR, exist_ok=True)
 os.makedirs(AUDIO_DIR, exist_ok=True)
 
-# Paksa environment variable untuk simpan modul bahasa ke /tmp
+# Paksa semua unduhan model masuk ke /tmp
 os.environ["TTS_HOME"] = TEMP_DIR
 
 app = FastAPI(
-    title="Coqui TTS API Interaktif",
-    description="API lengkap untuk Coqui TTS dengan otentikasi. Semua cache dan unduhan diarahkan ke /tmp."
+    title="PasBlast XTTS-v2 API",
+    description="API Interaktif menggunakan model terhebat Coqui XTTS-v2 yang di-host di Hugging Face."
 )
 security = HTTPBasic()
 
+# Global variable untuk lazy loading agar startup tidak timeout
+tts_model = None
+
+def get_xtts_instance() -> TTS:
+    global tts_model
+    if tts_model === None:
+        try:
+            # Memuat XTTS-v2 secara lokal di CPU
+            tts_model = TTS(model_name="tts_models/multilingual/multi-dataset/xtts_v2", gpu=False)
+        except Exception as e:
+            raise HTTPException(status_code=500, detail=f"Gagal memuat model XTTS-v2: {str(e)}")
+    return tts_model
+
 # --- OTENTIKASI ---
 def verify_auth(credentials: HTTPBasicCredentials = Depends(security)):
-    # Username default: admin
     correct_username = secrets.compare_digest(credentials.username, "admin")
     correct_password = secrets.compare_digest(credentials.password, "Rahasia1234")
     if not (correct_username and correct_password):
         raise HTTPException(
             status_code=status.HTTP_401_UNAUTHORIZED,
-            detail="Autentikasi gagal. Cek username dan password.",
+            detail="Akses ditolak. Kredensial salah.",
             headers={"WWW-Authenticate": "Basic"},
         )
     return credentials.username
 
-# --- HELPER FUNCTION ---
-def get_tts_instance(model_name: str) -> TTS:
-    """Load model secara dinamis (akan download ke /tmp jika belum ada)"""
-    try:
-        # Gunakan CPU jika Space gratis (tidak ada GPU)
-        return TTS(model_name=model_name, progress_bar=False, gpu=False)
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Gagal memuat model: {str(e)}")
-
 # --- ENDPOINTS ---
 
 @app.get("/")
 def root():
-    return {"message": "Coqui TTS API Aktif. Akses /docs untuk interaksi UI API."}
+    return {"message": "XTTS-v2 API Aktif. Akses /docs untuk mencoba secara interaktif."}
 
-@app.get("/models", tags=["Info"])
-def list_models(username: str = Depends(verify_auth)):
-    """Mengambil daftar semua model TTS dan Voice Conversion yang tersedia."""
-    try:
-        models = TTS().list_models()
-        return {"models": models}
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=str(e))
+@app.get("/speakers", tags=["Info"])
+def list_speakers(username: str = Depends(verify_auth)):
+    """Melihat daftar nama karakter suara (speakers) bawaan XTTS-v2 yang bisa digunakan."""
+    tts = get_xtts_instance()
+    return {"speakers": tts.speakers}
 
 @app.post("/tts", tags=["Generation"])
 def generate_tts(
-    text: str = Form(..., description="Teks yang akan diubah menjadi suara"),
-    model_name: str = Form("tts_models/en/ljspeech/vits", description="Nama model dari endpoint /models"),
-    speaker: str = Form(None, description="Nama speaker (jika model multi-speaker)"),
-    language: str = Form(None, description="Kode bahasa (jika model multi-lingual)"),
+    text: str = Form(..., description="Teks Syarat & Ketentuan PasBlast"),
+    speaker: str = Form("Ana Mendes", description="Pilih nama speaker dari endpoint /speakers"),
+    language: str = Form("id", description="Gunakan 'id' untuk Bahasa Indonesia"),
     username: str = Depends(verify_auth)
 ):
-    """Sintesis teks ke suara dasar."""
-    tts = get_tts_instance(model_name)
+    """Sintesis teks menggunakan suara karakter bawaan XTTS-v2."""
+    tts = get_xtts_instance()
     output_path = os.path.join(AUDIO_DIR, f"tts_{uuid.uuid4().hex}.wav")
     
     try:
         tts.tts_to_file(text=text, speaker=speaker, language=language, file_path=output_path)
-        return FileResponse(output_path, media_type="audio/wav", filename="output.wav")
+        return FileResponse(output_path, media_type="audio/wav", filename="pasblast_normal.wav")
     except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error sintesis TTS: {str(e)}")
+        raise HTTPException(status_code=500, detail=f"Gagal memproses TTS: {str(e)}")
 
 @app.post("/tts_voice_clone", tags=["Generation"])
 def generate_tts_voice_clone(
     text: str = Form(...),
-    model_name: str = Form("tts_models/multilingual/multi-dataset/xtts_v2"),
-    language: str = Form("id", description="Kode bahasa output (misal: 'id', 'en')"),
-    reference_audio: UploadFile = File(..., description="File audio WAV singkat untuk di-clone suaranya"),
+    language: str = Form("id"),
+    reference_audio: UploadFile = File(..., description="File WAV suara Anda (durasi 5-10 detik) untuk ditiru"),
     username: str = Depends(verify_auth)
 ):
-    """Sintesis teks dengan meniru suara dari file audio referensi (Zero-shot Voice Cloning)."""
-    tts = get_tts_instance(model_name)
+    """Sintesis teks dengan meniru (*cloning*) suara dari file audio yang Anda unggah."""
+    tts = get_xtts_instance()
     
-    # Simpan file audio referensi ke /tmp
     ref_path = os.path.join(AUDIO_DIR, f"ref_{uuid.uuid4().hex}.wav")
     with open(ref_path, "wb") as buffer:
         shutil.copyfileobj(reference_audio.file, buffer)
@@ -96,41 +93,7 @@ def generate_tts_voice_clone(
     output_path = os.path.join(AUDIO_DIR, f"clone_{uuid.uuid4().hex}.wav")
     
     try:
-        tts.tts_to_file(
-            text=text, 
-            language=language, 
-            speaker_wav=ref_path, 
-            file_path=output_path
-        )
-        return FileResponse(output_path, media_type="audio/wav", filename="cloned_output.wav")
-    except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error Voice Cloning: {str(e)}")
-
-@app.post("/voice_conversion", tags=["Generation"])
-def voice_conversion(
-    source_audio: UploadFile = File(..., description="Audio sumber yang ingin diubah suaranya"),
-    reference_audio: UploadFile = File(..., description="Audio target/referensi (suara yang akan ditiru)"),
-    model_name: str = Form("voice_conversion_models/multilingual/vctk/freevc24"),
-    username: str = Depends(verify_auth)
-):
-    """Mengubah suara dari satu file audio menjadi suara di file audio referensi (Audio-to-Audio)."""
-    tts = get_tts_instance(model_name)
-    
-    source_path = os.path.join(AUDIO_DIR, f"src_{uuid.uuid4().hex}.wav")
-    ref_path = os.path.join(AUDIO_DIR, f"ref_vc_{uuid.uuid4().hex}.wav")
-    output_path = os.path.join(AUDIO_DIR, f"vc_{uuid.uuid4().hex}.wav")
-    
-    with open(source_path, "wb") as buffer:
-        shutil.copyfileobj(source_audio.file, buffer)
-    with open(ref_path, "wb") as buffer:
-        shutil.copyfileobj(reference_audio.file, buffer)
-        
-    try:
-        tts.voice_conversion_to_file(
-            source_wav=source_path, 
-            target_wav=ref_path, 
-            file_path=output_path
-        )
-        return FileResponse(output_path, media_type="audio/wav", filename="converted_output.wav")
+        tts.tts_to_file(text=text, language=language, speaker_wav=ref_path, file_path=output_path)
+        return FileResponse(output_path, media_type="audio/wav", filename="pasblast_cloned.wav")
     except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Error Voice Conversion: {str(e)}")
+        raise HTTPException(status_code=500, detail=f"Gagal melakukan Voice Cloning: {str(e)}")