fix: enhance audio input handling in transcribe function

app.py

@@ -24,6 +24,16 @@ def transcribe(audio, state=""):
     if audio is None:
         return state, state
 
+    if isinstance(audio, tuple):
+        # If audio is a tuple, assume the first element is the file path
+        print("Received tuple input, extracting first element as file path")
+        audio = audio[0] if len(audio) > 0 else None
+    elif not isinstance(audio, str):
+        raise ValueError(f"Expected audio as a file path (str), got {type(audio)}")
+
+    if not audio:
+        raise ValueError("No valid audio input provided")
+    
     global model
     # Move model to GPU if available
     if torch.cuda.is_available():