Enhance error handling in audio generation

Updated the seed input validation in the generate_audio_from_prompt function to include OverflowError, ensuring that cases like infinity are handled gracefully. This change improves the robustness of the audio generation process by providing a fallback to system entropy when the seed is invalid.

app.py

@@ -23,8 +23,8 @@ def generate_audio_from_prompt(prompt, duration, seed):
         try:
             seed_int = int(seed)
             np.random.seed(seed_int)
-        except (ValueError, TypeError):
-            # If seed can't be converted to int, use system entropy
+        except (ValueError, TypeError, OverflowError):
+            # If seed can't be converted to int (including overflow cases like infinity), use system entropy
             pass
 
     # Extract features from prompt to influence audio