Update app.py

app.py

@@ -17,10 +17,10 @@ torch.set_num_threads(os.cpu_count())
 
 # Pre-load models
 models = {
-    "tiny": whisperx.load_model("tiny", device, compute_type=compute_type),
-    "base": whisperx.load_model("base", device, compute_type=compute_type),
-    "small": whisperx.load_model("small", device, compute_type=compute_type),
-    "large": whisperx.load_model("large", device, compute_type=compute_type),
+    "tiny": whisperx.load_model("tiny", device, compute_type=compute_type, vad_method='silero'),
+    "base": whisperx.load_model("base", device, compute_type=compute_type, vad_method='silero'),
+    "small": whisperx.load_model("small", device, compute_type=compute_type, vad_method='silero'),
+    "large": whisperx.load_model("large", device, compute_type=compute_type, vad_method='silero'),
 }
 
 def transcribe(audio_file, model_size="base", debug=False):
@@ -35,7 +35,7 @@ def transcribe(audio_file, model_size="base", debug=False):
         # Run inference
         model = models[model_size]
         batch_size = 8 if model_size == "tiny" else 4
-        transcript = model.transcribe(audio, batch_size=batch_size, vad_method='silero')
+        transcript = model.transcribe(audio, batch_size=batch_size)
         
         # Align whisper output
         model_a, metadata = whisperx.load_align_model(