diff --git a/README.md b/README.md index b82d92a6ec..c7b73f4fd6 100644 --- a/README.md +++ b/README.md @@ -1854,7 +1854,7 @@ The following extractors use this feature: * `tab`: Which tab to download - one of `new`, `top`, `videos`, `podcasts`, `streams`, `stacks` #### twitter -* `legacy_api`: Force usage of the legacy Twitter API instead of the GraphQL API for tweet extraction. Has no effect if login cookies are passed +* `api`: Select one of `graphql` (default), `legacy` or `syndication` as the API for tweet extraction. Has no effect if logged in #### stacommu, wrestleuniverse * `device_id`: UUID value assigned by the website and used to enforce device limits for paid livestream content. Can be found in browser local storage diff --git a/yt_dlp/extractor/twitter.py b/yt_dlp/extractor/twitter.py index f86216f8ff..4065acbaaa 100644 --- a/yt_dlp/extractor/twitter.py +++ b/yt_dlp/extractor/twitter.py @@ -1,9 +1,10 @@ -import functools import json +import random import re from .common import InfoExtractor from .periscope import PeriscopeBaseIE, PeriscopeIE +from ..compat import functools # isort: split from ..compat import ( compat_parse_qs, compat_urllib_parse_unquote, @@ -147,10 +148,14 @@ class TwitterBaseIE(InfoExtractor): def is_logged_in(self): return bool(self._get_cookies(self._API_BASE).get('auth_token')) + @functools.cached_property + def _selected_api(self): + return self._configuration_arg('api', ['graphql'], ie_key='Twitter')[0] + def _fetch_guest_token(self, display_id): guest_token = traverse_obj(self._download_json( f'{self._API_BASE}guest/activate.json', display_id, 'Downloading guest token', data=b'', - headers=self._set_base_headers(legacy=display_id and self._configuration_arg('legacy_api'))), + headers=self._set_base_headers(legacy=display_id and self._selected_api == 'legacy')), ('guest_token', {str})) if not guest_token: raise ExtractorError('Could not retrieve guest token') @@ -295,7 +300,7 @@ class TwitterBaseIE(InfoExtractor): self.report_login() def _call_api(self, path, video_id, query={}, graphql=False): - headers = self._set_base_headers(legacy=not graphql and self._configuration_arg('legacy_api')) + headers = self._set_base_headers(legacy=not graphql and self._selected_api == 'legacy') headers.update({ 'x-twitter-auth-type': 'OAuth2Session', 'x-twitter-client-language': 'en', @@ -707,6 +712,7 @@ class TwitterIE(TwitterBaseIE): 'tags': [], 'age_limit': 0, }, + 'skip': 'This Tweet is unavailable', }, { # not available in Periscope 'url': 'https://twitter.com/ViviEducation/status/1136534865145286656', @@ -721,6 +727,7 @@ class TwitterIE(TwitterBaseIE): 'view_count': int, }, 'add_ie': ['TwitterBroadcast'], + 'skip': 'Broadcast no longer exists', }, { # unified card 'url': 'https://twitter.com/BrooklynNets/status/1349794411333394432?s=20', @@ -773,9 +780,9 @@ class TwitterIE(TwitterBaseIE): 'url': 'https://twitter.com/UltimaShadowX/status/1577719286659006464', 'info_dict': { 'id': '1577719286659006464', - 'title': 'Ultima📛 | #вʟм - Test', + 'title': 'Ultima📛| New Era - Test', 'description': 'Test https://t.co/Y3KEZD7Dad', - 'uploader': 'Ultima📛 | #вʟм', + 'uploader': 'Ultima📛| New Era', 'uploader_id': 'UltimaShadowX', 'uploader_url': 'https://twitter.com/UltimaShadowX', 'upload_date': '20221005', @@ -811,7 +818,7 @@ class TwitterIE(TwitterBaseIE): 'age_limit': 0, }, }, { - # Adult content, fails if not logged in (GraphQL) + # Adult content, fails if not logged in 'url': 'https://twitter.com/Rizdraws/status/1575199173472927762', 'info_dict': { 'id': '1575199163847000068', @@ -831,9 +838,10 @@ class TwitterIE(TwitterBaseIE): 'age_limit': 18, 'tags': [] }, + 'params': {'skip_download': 'The media could not be played'}, 'skip': 'Requires authentication', }, { - # Playlist result only with auth + # Playlist result only with graphql API 'url': 'https://twitter.com/Srirachachau/status/1395079556562706435', 'playlist_mincount': 2, 'info_dict': { @@ -898,7 +906,7 @@ class TwitterIE(TwitterBaseIE): 'uploader_id': 'MoniqueCamarra', 'live_status': 'was_live', 'release_timestamp': 1658417414, - 'description': 'md5:4dc8e972f1d8b3c6580376fabb02a3ad', + 'description': 'md5:acce559345fd49f129c20dbcda3f1201', 'timestamp': 1658407771, 'release_date': '20220721', 'upload_date': '20220721', @@ -1007,10 +1015,10 @@ class TwitterIE(TwitterBaseIE): 'view_count': int, 'thumbnail': 'https://pbs.twimg.com/ext_tw_video_thumb/1600009362759733248/pu/img/XVhFQivj75H_YxxV.jpg?name=orig', 'age_limit': 0, - 'uploader': 'Mün The Friend Of YWAP', + 'uploader': 'Mün', 'repost_count': int, 'upload_date': '20221206', - 'title': 'Mün The Friend Of YWAP - This is a genius ad by Apple. \U0001f525\U0001f525\U0001f525\U0001f525\U0001f525', + 'title': 'Mün - This is a genius ad by Apple. \U0001f525\U0001f525\U0001f525\U0001f525\U0001f525', 'comment_count': int, 'like_count': int, 'tags': [], @@ -1019,7 +1027,7 @@ class TwitterIE(TwitterBaseIE): 'timestamp': 1670306984.0, }, }, { - # url to retweet id w/ legacy api + # retweeted_status (private) 'url': 'https://twitter.com/liberdalau/status/1623739803874349067', 'info_dict': { 'id': '1623274794488659969', @@ -1039,32 +1047,84 @@ class TwitterIE(TwitterBaseIE): 'like_count': int, 'repost_count': int, }, - 'params': {'extractor_args': {'twitter': {'legacy_api': ['']}}}, 'skip': 'Protected tweet', }, { - # orig tweet w/ graphql - 'url': 'https://twitter.com/liberdalau/status/1623739803874349067', + # retweeted_status + 'url': 'https://twitter.com/playstrumpcard/status/1695424220702888009', 'info_dict': { - 'id': '1623274794488659969', - 'display_id': '1623739803874349067', + 'id': '1694928337846538240', 'ext': 'mp4', - 'title': '@selfisekai@hackerspace.pl 🐀 - RT @Johnnybull3ts: Me after going viral to over 30million people: Whoopsie-daisy', - 'description': 'md5:9258bdbb54793bdc124fe1cd47e96c6a', - 'uploader': '@selfisekai@hackerspace.pl 🐀', - 'uploader_id': 'liberdalau', - 'uploader_url': 'https://twitter.com/liberdalau', + 'display_id': '1695424220702888009', + 'title': 'md5:e8daa9527bc2b947121395494f786d9d', + 'description': 'md5:004f2d37fd58737724ec75bc7e679938', + 'uploader': 'Benny Johnson', + 'uploader_id': 'bennyjohnson', + 'uploader_url': 'https://twitter.com/bennyjohnson', 'age_limit': 0, 'tags': [], - 'duration': 8.033, - 'timestamp': 1675964711.0, - 'upload_date': '20230209', - 'thumbnail': r're:https://pbs\.twimg\.com/ext_tw_video_thumb/.+', + 'duration': 45.001, + 'timestamp': 1692962814.0, + 'upload_date': '20230825', + 'thumbnail': r're:https://pbs\.twimg\.com/amplify_video_thumb/.+', 'like_count': int, - 'view_count': int, 'repost_count': int, + 'view_count': int, 'comment_count': int, }, - 'skip': 'Protected tweet', + }, { + # retweeted_status w/ legacy API + 'url': 'https://twitter.com/playstrumpcard/status/1695424220702888009', + 'info_dict': { + 'id': '1694928337846538240', + 'ext': 'mp4', + 'display_id': '1695424220702888009', + 'title': 'md5:e8daa9527bc2b947121395494f786d9d', + 'description': 'md5:004f2d37fd58737724ec75bc7e679938', + 'uploader': 'Benny Johnson', + 'uploader_id': 'bennyjohnson', + 'uploader_url': 'https://twitter.com/bennyjohnson', + 'age_limit': 0, + 'tags': [], + 'duration': 45.001, + 'timestamp': 1692962814.0, + 'upload_date': '20230825', + 'thumbnail': r're:https://pbs\.twimg\.com/amplify_video_thumb/.+', + 'like_count': int, + 'repost_count': int, + }, + 'params': {'extractor_args': {'twitter': {'api': ['legacy']}}}, + }, { + # Broadcast embedded in tweet + 'url': 'https://twitter.com/JessicaDobsonWX/status/1693057346933600402', + 'info_dict': { + 'id': '1yNGaNLjEblJj', + 'ext': 'mp4', + 'title': 'Jessica Dobson - WAVE Weather Now - Saturday 8/19/23 Update', + 'uploader': 'Jessica Dobson', + 'uploader_id': '1DZEoDwDovRQa', + 'thumbnail': r're:^https?://.*\.jpg', + 'view_count': int, + }, + 'add_ie': ['TwitterBroadcast'], + }, { + # Animated gif and quote tweet video, with syndication API + 'url': 'https://twitter.com/BAKKOOONN/status/1696256659889565950', + 'playlist_mincount': 2, + 'info_dict': { + 'id': '1696256659889565950', + 'title': 'BAKOON - https://t.co/zom968d0a0', + 'description': 'https://t.co/zom968d0a0', + 'tags': [], + 'uploader': 'BAKOON', + 'uploader_id': 'BAKKOOONN', + 'uploader_url': 'https://twitter.com/BAKKOOONN', + 'age_limit': 18, + 'timestamp': 1693254077.0, + 'upload_date': '20230828', + 'like_count': int, + }, + 'params': {'extractor_args': {'twitter': {'api': ['syndication']}}}, + 'expected_warnings': ['Not all metadata'], }, { # onion route 'url': 'https://twitter3e4tixl4xyajtrzo62zg5vztmjuricljdp2c5kshju4avyoid.onion/TwitterBlue/status/1484226494708662273', @@ -1103,6 +1163,14 @@ class TwitterIE(TwitterBaseIE): 'only_matching': True, }] + _MEDIA_ID_RE = re.compile(r'_video/(\d+)/') + + @property + def _GRAPHQL_ENDPOINT(self): + if self.is_logged_in: + return 'zZXycP0V6H7m-2r0mOnFcA/TweetDetail' + return '2ICDjqPd81tulZcYrtpTuQ/TweetResultByRestId' + def _graphql_to_legacy(self, data, twid): result = traverse_obj(data, ( 'threaded_conversation_with_injections_v2', 'instructions', 0, 'entries', @@ -1130,9 +1198,14 @@ class TwitterIE(TwitterBaseIE): 'user': ('core', 'user_results', 'result', 'legacy'), 'card': ('card', 'legacy'), 'quoted_status': ('quoted_status_result', 'result', 'legacy'), + 'retweeted_status': ('legacy', 'retweeted_status_result', 'result', 'legacy'), }, expected_type=dict, default={})) - # extra transformation is needed since result does not match legacy format + # extra transformations needed since result does not match legacy format + if status.get('retweeted_status'): + status['retweeted_status']['user'] = traverse_obj(status, ( + 'retweeted_status_result', 'result', 'core', 'user_results', 'result', 'legacy', {dict})) or {} + binding_values = { binding_value.get('key'): binding_value.get('value') for binding_value in traverse_obj(status, ('card', 'binding_values', ..., {dict})) @@ -1208,33 +1281,42 @@ class TwitterIE(TwitterBaseIE): } def _extract_status(self, twid): - if self.is_logged_in: - return self._graphql_to_legacy( - self._call_graphql_api('zZXycP0V6H7m-2r0mOnFcA/TweetDetail', twid), twid) + if self.is_logged_in or self._selected_api == 'graphql': + status = self._graphql_to_legacy(self._call_graphql_api(self._GRAPHQL_ENDPOINT, twid), twid) - try: - if not self._configuration_arg('legacy_api'): - return self._graphql_to_legacy( - self._call_graphql_api('2ICDjqPd81tulZcYrtpTuQ/TweetResultByRestId', twid), twid) - return traverse_obj(self._call_api(f'statuses/show/{twid}.json', twid, { + elif self._selected_api == 'legacy': + status = self._call_api(f'statuses/show/{twid}.json', twid, { 'cards_platform': 'Web-12', 'include_cards': 1, 'include_reply_count': 1, 'include_user_entities': 0, 'tweet_mode': 'extended', - }), 'retweeted_status', None) + }) - except ExtractorError as e: - if e.expected: - raise + elif self._selected_api == 'syndication': self.report_warning( - f'{e.orig_msg}. Falling back to syndication endpoint; some metadata may be missing', twid) + 'Not all metadata or media is available via syndication endpoint', twid, only_once=True) + status = self._download_json( + 'https://cdn.syndication.twimg.com/tweet-result', twid, 'Downloading syndication JSON', + headers={'User-Agent': 'Googlebot'}, query={ + 'id': twid, + # TODO: token = ((Number(twid) / 1e15) * Math.PI).toString(36).replace(/(0+|\.)/g, '') + 'token': ''.join(random.choices('123456789abcdefghijklmnopqrstuvwxyz', k=10)), + }) + if not status: + raise ExtractorError('Syndication endpoint returned empty JSON response') + # Transform the result so its structure matches that of legacy/graphql + media = [] + for detail in traverse_obj(status, ((None, 'quoted_tweet'), 'mediaDetails', ..., {dict})): + detail['id_str'] = traverse_obj(detail, ( + 'video_info', 'variants', ..., 'url', {self._MEDIA_ID_RE.search}, 1), get_all=False) or twid + media.append(detail) + status['extended_entities'] = {'media': media} - status = self._download_json( - 'https://cdn.syndication.twimg.com/tweet-result', twid, 'Downloading syndication JSON', - headers={'User-Agent': 'Googlebot'}, query={'id': twid}) - status['extended_entities'] = {'media': status.get('mediaDetails')} - return status + else: + raise ExtractorError(f'"{self._selected_api}" is not a valid API selection', expected=True) + + return traverse_obj(status, 'retweeted_status', None, expected_type=dict) or {} def _real_extract(self, url): twid, selected_index = self._match_valid_url(url).group('id', 'index') @@ -1266,10 +1348,7 @@ class TwitterIE(TwitterBaseIE): } def extract_from_video_info(media): - media_id = traverse_obj(media, 'id_str', 'id', ( - 'video_info', 'variants', ..., 'url', - {functools.partial(re.search, r'_video/(\d+)/')}, 1 - ), get_all=False, expected_type=str_or_none) or twid + media_id = traverse_obj(media, 'id_str', 'id', expected_type=str_or_none) self.write_debug(f'Extracting from video info: {media_id}') formats = [] @@ -1503,6 +1582,8 @@ class TwitterBroadcastIE(TwitterBaseIE, PeriscopeBaseIE): broadcast = self._call_api( 'broadcasts/show.json', broadcast_id, {'ids': broadcast_id})['broadcasts'][broadcast_id] + if not broadcast: + raise ExtractorError('Broadcast no longer exists', expected=True) info = self._parse_broadcast_data(broadcast, broadcast_id) media_key = broadcast['media_key'] source = self._call_api(