fix: Improve SouthParkDeIE to handle various page layouts

yt_dlp/extractor/southpark.py

@@ -110,10 +110,17 @@ class SouthParkDeIE(SouthParkIE):  # XXX: Do not subclass from concrete IE
         data = self._parse_json(self._search_regex(
             r'window\.__DATA__\s*=\s*({.+?});', webpage, 'data'), display_id)
 
-        video_detail = traverse_obj(data, (
-            'children', lambda _, v: v.get('type') == 'MainContainer',
-            'children', 0, 'children', 0, 'props', 'videoDetail'
-        ), default=traverse_obj(data, ('children', 0, 'videoDetail')))
+        # Try multiple paths to find the video data, handling both regular and special episodes
+        video_detail = traverse_obj(data, [
+            # Path for regular episodes (more complex)
+            ('children', lambda _, v: v.get('type') == 'MainContainer',
+             'children', 0, 'children', 0, 'props', 'videoDetail'),
+            # Fallback path for special episodes (simpler)
+            ('children', 0, 'videoDetail'),
+        ])
+
+        if not video_detail:
+            raise ExtractorError('Could not find video data in page')
 
         api_url = video_detail['videoServiceUrl']