Update app.py

app.py

@@ -13,14 +13,24 @@ model = Wav2Vec2ForCTC.from_pretrained(model_id)
 def transcribe(audio_data):
     if audio_data is None:
         return "⚠️ No audio received."
-    
-    audio, sample_rate = audio_data  # ✅ Unpack the tuple
 
-    # Resample if needed
+    audio, sample_rate = audio_data
+
+    # Convert stereo to mono if needed
+    if len(audio.shape) == 2:
+        audio = np.mean(audio, axis=1)
+
+    # Ensure sample_rate is an integer
+    sample_rate = int(sample_rate)
+
+    # Resample to 16000 Hz if needed
     if sample_rate != 16000:
         number_of_samples = round(len(audio) * 16000 / sample_rate)
         audio = resample(audio, number_of_samples)
 
+    # Normalize audio
+    audio = audio.astype(np.float32)
+
     # Process and predict
     input_values = processor(audio, sampling_rate=16000, return_tensors="pt").input_values
     with torch.no_grad():