Update app.py

app.py

@@ -11,19 +11,27 @@ assert torch.cuda.is_available(), "Demo requires a GPU."
 DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
 print(DEVICE)
 
-# Load model once
-model = SopranoTTS(
-    backend="auto",
-    device=DEVICE,
-    cache_size_mb=100,
-    decoder_batch_size=1,
-)
+model = None
+
+def load_model():
+    global model
+    if model is None:
+        # Load model once
+        model = SopranoTTS(
+            backend="auto",
+            device=DEVICE,
+            cache_size_mb=100,
+            decoder_batch_size=1,
+        )
+    return model
 
-SAMPLE_RATE = 32000
 
+SAMPLE_RATE = 32000
 
 @spaces.GPU
 def tts_stream(text, temperature, top_p, repetition_penalty, state):
+    model = load_model()
+    
     if not text.strip():
         yield None, state
         return