Update app.py

app.py

@@ -1,7 +1,7 @@
 import os
 import tempfile
 import whisperx
-from whisperx import diarize
+from pyannote.audio import Pipeline
 import pandas as pd
 import librosa
 import soundfile as sf
@@ -25,9 +25,19 @@ Segment: Any = None
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
 token = os.environ.get("HF_TOKEN")
-if not token:
-    print("Warning: HF_TOKEN not set. Diarization will be skipped.")
-perform_diarization = True if token else False
+try:
+    if token:
+        diarization_pipeline = Pipeline.from_pretrained(
+            "pyannote/speaker-diarization-3.1", 
+            use_auth_token=token
+        )
+        diarization_pipeline.to(torch.device(device))
+    else:
+        diarization_pipeline = None
+except Exception as e:
+    print(f"Error loading pyannote pipeline: {type(e).__name__}: {e}. Diarization will be skipped.")
+    diarization_pipeline = None
+global_diarizer = diarization_pipeline
 model_name = "medium"
 class TimelineItem(BaseModel):
     start: float
@@ -192,7 +202,7 @@ def analyze_audio(audio_file: str,
         if perform_diarization:
             print("Performing speaker diarization (Requires HF_TOKEN)...")
             try:
-                diarize_output = diarize(audio_for_model, token=token)
+                diarize_output = global_diarizer(audio_for_model)
                 for segment, _, label in diarize_output.itertracks(yield_label=True):
                     print(f"start={segment.start:.1f}s stop={segment.end:.1f}s {label}")
             except Exception as e: