fix first speaker bug

app.py

@@ -21,7 +21,8 @@ import numpy as np
 model = whisper.load_model("large-v2")
 embedding_model = PretrainedSpeakerEmbedding( 
     "speechbrain/spkrec-ecapa-voxceleb",
-    device=torch.device("cuda"))
+    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+)
 
 def transcribe(audio, num_speakers):
   path = convert_to_wav(audio)
@@ -84,7 +85,7 @@ def get_output(segments):
   for (i, segment) in enumerate(segments):
     if i > 0:
       output += '\n\n'
-    if segments[i - 1]["speaker"] != segment["speaker"]:
+    if i == 0 or segments[i - 1]["speaker"] != segment["speaker"]:
       output += segment["speaker"] + ' ' + str(time(segment["start"])) + '\n\n'
     output += segment["text"][1:] + ' '
   return output