change importexit

app.py

@@ -1,19 +1,17 @@
 from fastapi import FastAPI
 from fastapi.responses import FileResponse
-from transformers import AutoProcessor, VitsModel
-import torch
-import soundfile as sf
 import aiofiles
-import uvicorn
 import os
+from TTS.api import TTS  # install via: pip install TTS
+import uvicorn
 
 app = FastAPI()
 
-MODEL_NAME = "wibowo-id/vits-id"
-device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+# Load the Tacotron2 + HiFi-GAN Indonesian model (from Coqui TTS hub)
+# Model ID on HF: "tts_models/id/id_tts_tacotron2"
+tts = TTS(model_name="tts_models/id/id_tts_tacotron2")
 
-processor = AutoProcessor.from_pretrained(MODEL_NAME)
-model = VitsModel.from_pretrained(MODEL_NAME).to(device)
+OUTPUT_FILE = "output.wav"
 
 @app.post("/tts")
 async def tts_api(payload: dict):
@@ -21,21 +19,16 @@ async def tts_api(payload: dict):
     if not text:
         return {"error": "Text is required"}
 
-    inputs = processor(text, return_tensors="pt").to(device)
-    with torch.no_grad():
-        outputs = model(**inputs)
-        audio = outputs.waveform.squeeze().cpu().numpy()
-
-    file_path = "output.wav"
-    sf.write(file_path, audio, samplerate=model.config.sampling_rate)
+    # Generate speech and save to OUTPUT_FILE
+    tts.tts_to_file(text=text, file_path=OUTPUT_FILE)
 
-    async with aiofiles.open(file_path, "rb") as f:
+    async with aiofiles.open(OUTPUT_FILE, "rb") as f:
         audio_data = await f.read()
 
     return {
         "text": text,
-        "file_url": f"/download/{os.path.basename(file_path)}",
-        "size": len(audio_data)
+        "file_url": f"/download/{os.path.basename(OUTPUT_FILE)}",
+        "size": len(audio_data),
     }
 
 @app.get("/download/{filename}")