Update app.py

app.py

@@ -62,13 +62,15 @@ def gen_tts(text, description):
             do_sample=True,
             temperature=1.0,
         )
-        
-        # Extract the generated audio
-        audio_arr = generation.cpu().numpy().squeeze()
 
-        # Check for NaN or Inf values and handle them
-        if np.any(np.isnan(audio_arr)) or np.any(np.isinf(audio_arr)):
-            raise ValueError("Generated audio contains invalid values (NaN or Inf).")
+        # Inspect the raw audio generation output
+        print(f"Generated audio shape: {generation.shape}")
+        print(f"Generated audio values: {generation.cpu().numpy().squeeze()}")
+
+        # Check if there are any meaningful values in the audio output
+        audio_arr = generation.cpu().numpy().squeeze()
+        if np.all(audio_arr == 0):
+            raise ValueError("Generated audio is empty or silent.")
 
         # Normalize the audio array to the range [-1, 1]
         audio_arr = audio_arr / np.max(np.abs(audio_arr))