Update app.py

app.py

@@ -17,17 +17,14 @@ def load_data(input_file):
   
   """ Function for resampling to ensure that the speech input is sampled at 16KHz.
   """
+  sampling_rate = 16_000
   #read the file
-  speech, sample_rate = librosa.load(input_file)
-  #make it 1-D
-  if len(speech.shape) > 1: 
-      speech = speech[:,0] + speech[:,1]
-  #Resampling at 16KHz since wav2vec2-base-960h is pretrained and fine-tuned on speech audio sampled at 16 KHz.
-  if sample_rate !=16000:
-    speech = librosa.resample(speech, sample_rate,16000)
-  return speech
+  speech, sample_rate = librosa.load(input_file, sample_rate=sampling_rate, mono=True)
+  speech = librosa.effects.trim(speech, top_db= 10)
+  return speech[0]
     
 def asr_pipe(input_file):
+  load_data(input_file)
   transcription = p(input_file, chunk_length_s=3, stride_length_s=(1, 1))["text"]
   return transcription