ffmpeg

app.py

@@ -56,9 +56,9 @@ def encode_video(video_path):
     return video
 
 def extract_audio(video_path):
-    video = mp.VideoFileClip(video_path)
+    import subprocess
     audio_path = "temp_audio.wav"
-    video.audio.write_audiofile(audio_path)
+    subprocess.call(['ffmpeg', '-i', video_path, '-ab', '160k', '-ac', '2', '-ar', '44100', '-vn', audio_path])
     return audio_path
 
 def transcribe_audio(audio_file):
@@ -68,7 +68,7 @@ def transcribe_audio(audio_file):
     inputs = ffmpeg_read(inputs, asr_pipeline.feature_extractor.sampling_rate)
     inputs = {"array": inputs, "sampling_rate": asr_pipeline.feature_extractor.sampling_rate}
 
-    transcription = asr_pipeline(inputs, batch_size=8, generate_kwargs={"task": "transcribe"}, return_timestamps=False)["text"]
+    transcription = asr_pipeline(inputs, batch_size=8, generate_kwargs={"task": "transcribe"}, return_timestamps=True)["text"]
     return transcription
 
 @spaces.GPU

packages.txt

@@ -0,0 +1 @@
+ffmpeg