Merge 8cd107a6e0 into 36b29bb353

yt_dlp/extractor/_extractors.py

@@ -1912,6 +1912,7 @@ from .sonyliv import (
     SonyLIVIE,
     SonyLIVSeriesIE,
 )
+from .sora import SoraExtractorIE
 from .soundcloud import (
     SoundcloudEmbedIE,
     SoundcloudIE,

yt_dlp/extractor/sora.py

@@ -0,0 +1,74 @@
+import re
+import urllib.parse
+
+from .common import InfoExtractor
+from ..utils import (
+    lowercase_escape,
+    unescapeHTML,
+    url_or_none,
+)
+
+
+class SoraExtractorIE(InfoExtractor):
+    _VALID_URL = r'https://sora\.chatgpt\.com/p/(?P<id>[^/]+)'
+    _TESTS = [{
+        'url': 'https://sora.chatgpt.com/p/s_68ddd8c38ae8819180aa11a37b1d1e88',
+        'info_dict': {
+            'id': 's_68ddd8c38ae8819180aa11a37b1d1e88',
+            'ext': 'mp4',
+            'title': 'unitof on Sora',
+            'description': 'md5:d7a614c169fa6da1dbff019e736bd2b0',
+        },
+    }]
+
+    def _real_extract(self, url):
+        video_id = self._match_id(url)
+        webpage = self._download_webpage(url, video_id, impersonate=True)
+
+        title = self._og_search_title(webpage)
+        description = self._og_search_description(webpage)
+
+        # - https://videos.openai.com/vg-assets/.../videos/00000/src.mp4?... (unwatermarked)
+        # - https://videos.openai.com/vg-assets/.../videos/00000_wm/src.mp4?... (watermarked)
+        url_pattern = r'https?://videos\.openai\.com/[^\s"\'<>]*src\.mp4[^\s"\'<>]*'
+
+        found_urls = [
+            unescapeHTML(lowercase_escape(unescapeHTML(m)))
+            for m in re.findall(url_pattern, webpage)
+        ]
+
+        formats = []
+        for vurl in found_urls:
+            vurl = url_or_none(vurl)
+            if not vurl:
+                continue
+            parsed = urllib.parse.urlparse(vurl)
+            decoded_path = urllib.parse.unquote(parsed.path or '')
+
+            # We have not yet figured out signed unwatermarked URLs are generated in iOS app, due to SSL pinning
+            is_wm = ('/00000_wm/' in decoded_path)
+            formats.append({
+                'url': vurl,
+                'ext': 'mp4',
+                'format_id': 'watermarked' if is_wm else 'unwatermarked',
+                'preference': -10 if is_wm else 10,
+            })
+
+        if formats:
+            return {
+                'id': video_id,
+                'title': title,
+                'formats': formats,
+                'description': description,
+            }
+
+        # Fallback to og:video if direct URLs not found
+        og_video = self._og_search_video_url(webpage, default=None)
+        if og_video:
+            return {
+                'id': video_id,
+                'title': title,
+                'url': og_video,
+            }
+
+        self.raise_no_formats('No playable video URLs found on the page', expected=True)