Fix

app.py

@@ -135,34 +135,9 @@ def extract_resampled_audio(video_path, target_sr=16000):
     torchaudio.save(resampled_audio_path, waveform, sample_rate=target_sr)
     return resampled_audio_path
 
-import ffmpeg
-import os
-
-def maybe_downsample(video_path):
-    probe = ffmpeg.probe(video_path)
-    video_streams = [s for s in probe['streams'] if s['codec_type'] == 'video']
-    width = int(video_streams[0]['width'])
-    height = int(video_streams[0]['height'])
-
-    if height > 720:
-        downsampled_path = tempfile.mktemp(suffix="_720p.mp4")
-        (
-            ffmpeg
-            .input(video_path)
-            .filter('scale', -1, 720)
-            .output(downsampled_path, vcodec='libx264', acodec='aac')
-            .overwrite_output()
-            .run(quiet=True)
-        )
-        return downsampled_path
-    else:
-        return video_path
 
 @spaces.GPU
 def extract_faces(video_file):
-    #cap = cv2.VideoCapture(video_file)
-    #fps = cap.get(cv2.CAP_PROP_FPS)
-    video_file = maybe_downsample(video_file)
     cap = cv2.VideoCapture(video_file)
     fps = cap.get(cv2.CAP_PROP_FPS)
     frames = []
@@ -242,8 +217,13 @@ def extract_faces(video_file):
 
     enhanced_audio_path = run_avse_inference(output_path, audio_path)
 
+    from moviepy.video.fx import flip_vertical
+    flipped_output_path = os.path.join(tmpdir, "face_only_video_flipped.mp4")
+    flipped_clip = flip_vertical(clip)
+    flipped_clip.write_videofile(flipped_output_path, codec="libx264", audio=False, fps=25)
 
-    return output_path, enhanced_audio_path
+    return flipped_output_path, enhanced_audio_path
+    #return output_path, enhanced_audio_path
     #return output_path, audio_path
 
 iface = gr.Interface(