Update app.py

app.py

@@ -30,21 +30,28 @@ speech_granite = AutoModelForSpeechSeq2Seq.from_pretrained(
     trust_remote_code=True
 )
 print("Model loaded successfully")
-
 def transcribe_audio(audio_input):
     """Process audio input and return transcription"""
     start_time = time.time()
     
+    print(f"Audio input received: {type(audio_input)}, value: {audio_input}")
+    
     if audio_input is None:
         return "Error: No audio provided.", 0.0
     
     try:
         # Load audio file
         if isinstance(audio_input, tuple):  # From microphone
+            print("Processing microphone input")
             sr, wav_np = audio_input
             wav = torch.from_numpy(wav_np).float().unsqueeze(0)
-        else:  # From file
+        else:  # From file upload
+            print(f"Processing file input: {audio_input}")
+            if not os.path.exists(audio_input):
+                return f"Error: File does not exist: {audio_input}", 0.0
+            
             wav, sr = torchaudio.load(audio_input, normalize=True)
+            print(f"Loaded audio file: {audio_input}")
         
         print(f"Original audio: sample rate {sr}Hz, shape {wav.shape}")
         
@@ -129,13 +136,14 @@ Upload an audio file or use your microphone to record speech.
 
 iface = gr.Interface(
     fn=transcribe_audio,
-    inputs=gr.Audio(sources=["upload", "microphone"], type="filepath"),
+    inputs=gr.Audio(sources=["upload", "microphone"], type="filepath", label="Upload or record audio"),
     outputs=[
         gr.Textbox(label="Transcription", lines=5),
         gr.Number(label="Processing Time (seconds)")
     ],
     title=title,
     description=description,
+    examples=[["example.wav"]] if os.path.exists("example.wav") else None,
 )
 
 if __name__ == "__main__":