Add limitations

app.py

@@ -136,26 +136,28 @@ def extract_resampled_audio(video_path, target_sr=16000):
     torchaudio.save(resampled_audio_path, waveform, sample_rate=target_sr)
     return resampled_audio_path
 
-@spaces.GPU
-def extract_faces(video_file):
-    # Step 0: Check resolution
-    cap = cv2.VideoCapture(video_file)
+import cv2
+import ffmpeg
+
+def downsample_if_needed(video_path):
+    cap = cv2.VideoCapture(video_path)
     width  = cap.get(cv2.CAP_PROP_FRAME_WIDTH)
     height = cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
     cap.release()
 
-    # Step 1: Downsample if needed
-    if width > 1280 or height > 720:
-        resized_path = tempfile.mktemp(suffix=".mp4")
-        subprocess.run([
-            "ffmpeg", "-y", "-i", video_file,
-            "-vf", "scale='min(1280,iw)':-2",
-            "-c:v", "libx264", "-crf", "28",
-            "-preset", "fast", "-an", resized_path
-        ])
-        video_file = resized_path
-
-    cap = cv2.VideoCapture(video_file)
+    if max(width, height) > 720:
+        downsampled_path = video_path.replace(".mp4", "_720p.mp4")
+        ffmpeg.input(video_path).output(
+            downsampled_path, vf="scale='min(720,iw)':-2", **{"c:v": "libx264"}
+        ).overwrite_output().run()
+        return downsampled_path
+    return video_path
+
+@spaces.GPU
+def extract_faces(video_file):
+    video_path = downsample_if_needed(video_file)
+
+    cap = cv2.VideoCapture(video_path)
     fps = cap.get(cv2.CAP_PROP_FPS)
     frames = []