Update app.py

app.py

@@ -4,9 +4,11 @@ import time
 
 import gradio as gr
 import numpy as np
+import torch
 import yt_dlp as youtube_dl
 from gradio_client import Client
 from pyannote.audio import Pipeline
+from transformers.pipelines.audio_utils import ffmpeg_read
 
 
 YT_LENGTH_LIMIT_S = 36000  # limit to 1 hour YouTube files
@@ -222,8 +224,16 @@ def transcribe_yt(yt_url, task="transcribe", group_by_speaker=True, progress=gr.
     with tempfile.TemporaryDirectory() as tmpdirname:
         filepath = os.path.join(tmpdirname, "video.mp4")
         download_yt_audio(yt_url, filepath)
-        diarization = diarization_pipeline(filepath)
+        with open(filepath, "rb") as f:
+            inputs = f.read()
 
+    inputs = ffmpeg_read(inputs, SAMPLING_RATE)
+    inputs = torch.from_numpy(inputs).float()
+    inputs = inputs.unsqueeze(0)
+
+    diarization = diarization_pipeline(
+        {"waveform": inputs, "sample_rate": SAMPLING_RATE},
+    )
     segments = diarization.for_json()["content"]
 
     # only fetch the transcription result after performing diarization