Update app.py

app.py

@@ -15,19 +15,24 @@ model = Wav2Vec2ForCTC.from_pretrained("jonatasgrosman/wav2vec2-large-xlsr-53-it
 # Function to perform ASR on audio data
 def transcribe_audio(audio_data):
     print("Received audio data:", audio_data)  # Debug print
-    if audio_data is None or len(audio_data) != 2:
+
+    # Check if audio_data is None or not a tuple of length 2
+    if audio_data is None or not isinstance(audio_data, tuple) or len(audio_data) != 2:
         return "Invalid audio data format."
 
-    try:
-        # Extract sample rate and audio waveform from the tuple
-        sample_rate, waveform = audio_data
+    sample_rate, waveform = audio_data
 
+    # Check if waveform is None or not a NumPy array
+    if waveform is None or not isinstance(waveform, torch.Tensor):
+        return "Invalid audio data format."
+
+    try:
         # Convert audio data to mono and normalize
         audio_data = torchaudio.transforms.Resample(sample_rate, 16000)(waveform)
         audio_data = torchaudio.functional.gain(audio_data, gain_db=5.0)
 
         # Apply custom preprocessing to the audio data if needed
-        input_values = processor(torch.tensor(audio_data[0]), return_tensors="pt").input_values
+        input_values = processor(audio_data[0], return_tensors="pt").input_values
 
         # Perform ASR
         with torch.no_grad():