Update app.py

app.py

@@ -19,7 +19,11 @@ import numpy as np
 
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 
-
+model = whisper.load_model("large-v2")
+embedding_model = PretrainedSpeakerEmbedding( 
+    "speechbrain/spkrec-ecapa-voxceleb",
+    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+    )
 ######################################################
 def audio_to_text(audio, num_speakers):
   path, error = convert_to_wav(audio)
@@ -31,12 +35,6 @@ def audio_to_text(audio, num_speakers):
     return "Audio duration too long"
 
 
-  model = whisper.load_model("large-v2")
-  embedding_model = PretrainedSpeakerEmbedding( 
-    "speechbrain/spkrec-ecapa-voxceleb",
-    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
-    )
-
   result = model.transcribe(path)
   segments = result["segments"]