diff --git a/README.md b/README.md index 6ff6d93d6..838131e8c 100644 --- a/README.md +++ b/README.md @@ -1337,12 +1337,12 @@ # EXTRACTOR ARGUMENTS The following extractors use this feature: * **youtube** * `skip`: `hls` or `dash` (or both) to skip download of the respective manifests - * `player_client`: `web` (default) or `android` (force use the android client fallbacks for video extraction) - * `player_skip`: `configs` - skip requests if applicable for client configs and use defaults + * `player_client`: Clients to extract video data from - one or more of `web`, `android`, `ios`, `web_music`, `android_music`, `ios_music`. By default, `android,web` is used. If the URL is from `music.youtube.com`, `android,web,android_music,web_music` is used + * `player_skip`: `configs` - skip any requests for client configs and use defaults * `comment_sort`: `top` or `new` (default) - choose comment sorting mode (on YouTube's side). * `max_comments`: maximum amount of comments to download (default all). * `max_comment_depth`: maximum depth for nested comments. YouTube supports depths 1 or 2 (default). - + * **funimation** * `language`: Languages to extract. Eg: `funimation:language=english,japanese` * `version`: The video version to extract - `uncut` or `simulcast` diff --git a/yt_dlp/downloader/youtube_live_chat.py b/yt_dlp/downloader/youtube_live_chat.py index 35e88e367..5e05426e6 100644 --- a/yt_dlp/downloader/youtube_live_chat.py +++ b/yt_dlp/downloader/youtube_live_chat.py @@ -116,7 +116,7 @@ def download_and_parse_fragment(url, frag_index, request_data=None, headers=None if not success: return False, None, None, None try: - data = ie._extract_yt_initial_data(video_id, raw_fragment.decode('utf-8', 'replace')) + data = ie.extract_yt_initial_data(video_id, raw_fragment.decode('utf-8', 'replace')) except RegexNotFoundError: data = None if not data: @@ -146,7 +146,7 @@ def download_and_parse_fragment(url, frag_index, request_data=None, headers=None if not success: return False try: - data = ie._extract_yt_initial_data(video_id, raw_fragment.decode('utf-8', 'replace')) + data = ie.extract_yt_initial_data(video_id, raw_fragment.decode('utf-8', 'replace')) except RegexNotFoundError: return False continuation_id = try_get( @@ -155,7 +155,7 @@ def download_and_parse_fragment(url, frag_index, request_data=None, headers=None # no data yet but required to call _append_fragment self._append_fragment(ctx, b'') - ytcfg = ie._extract_ytcfg(video_id, raw_fragment.decode('utf-8', 'replace')) + ytcfg = ie.extract_ytcfg(video_id, raw_fragment.decode('utf-8', 'replace')) if not ytcfg: return False @@ -183,7 +183,7 @@ def download_and_parse_fragment(url, frag_index, request_data=None, headers=None request_data['currentPlayerState'] = {'playerOffsetMs': str(max(offset - 5000, 0))} if click_tracking_params: request_data['context']['clickTracking'] = {'clickTrackingParams': click_tracking_params} - headers = ie._generate_api_headers(ytcfg, visitor_data=visitor_data) + headers = ie.generate_api_headers(ytcfg, visitor_data=visitor_data) headers.update({'content-type': 'application/json'}) fragment_request_data = json.dumps(request_data, ensure_ascii=False).encode('utf-8') + b'\n' success, continuation_id, offset, click_tracking_params = download_and_parse_fragment( diff --git a/yt_dlp/extractor/youtube.py b/yt_dlp/extractor/youtube.py index b3b43188a..aa0421a72 100644 --- a/yt_dlp/extractor/youtube.py +++ b/yt_dlp/extractor/youtube.py @@ -28,18 +28,18 @@ ) from ..jsinterp import JSInterpreter from ..utils import ( - bool_or_none, bytes_to_intlist, clean_html, - dict_get, datetime_from_str, + dict_get, error_to_compat_str, ExtractorError, - format_field, float_or_none, + format_field, int_or_none, intlist_to_bytes, mimetype2ext, + orderedSet, parse_codecs, parse_count, parse_duration, @@ -392,6 +392,49 @@ def _real_initialize(self): } }, 'INNERTUBE_CONTEXT_CLIENT_NAME': 21 + }, + 'IOS': { + 'INNERTUBE_API_VERSION': 'v1', + 'INNERTUBE_CLIENT_NAME': 'IOS', + 'INNERTUBE_CLIENT_VERSION': '16.20', + 'INNERTUBE_API_KEY': 'AIzaSyAO_FJ2SlqU8Q4STEHLGCilw_Y9_11qcW8', + 'INNERTUBE_CONTEXT': { + 'client': { + 'clientName': 'IOS', + 'clientVersion': '16.20', + 'hl': 'en', + } + }, + 'INNERTUBE_CONTEXT_CLIENT_NAME': 5 + + }, + 'IOS_MUSIC': { + 'INNERTUBE_API_VERSION': 'v1', + 'INNERTUBE_CLIENT_NAME': 'IOS_MUSIC', + 'INNERTUBE_CLIENT_VERSION': '4.32', + 'INNERTUBE_API_KEY': 'AIzaSyDK3iBpDP9nHVTk2qL73FLJICfOC3c51Og', + 'INNERTUBE_CONTEXT': { + 'client': { + 'clientName': 'IOS_MUSIC', + 'clientVersion': '4.32', + 'hl': 'en', + } + }, + 'INNERTUBE_CONTEXT_CLIENT_NAME': 26 + }, + 'IOS_MESSAGES_EXTENSION': { + 'INNERTUBE_API_VERSION': 'v1', + 'INNERTUBE_CLIENT_NAME': 'IOS_MESSAGES_EXTENSION', + 'INNERTUBE_CLIENT_VERSION': '16.20', + 'INNERTUBE_API_KEY': 'AIzaSyDCU8hByM-4DrUqRUYnGn-3llEO78bcxq8', + 'INNERTUBE_CONTEXT': { + 'client': { + 'clientName': 'IOS_MESSAGES_EXTENSION', + 'clientVersion': '16.20', + 'hl': 'en', + } + }, + 'INNERTUBE_CONTEXT_CLIENT_NAME': 66 } } @@ -402,6 +445,22 @@ def _real_initialize(self): 'ANDROID_MUSIC': 'music.youtube.com' } + # clients starting with _ cannot be explicity requested by the user + _YT_CLIENTS = { + 'web': 'WEB', + 'web_music': 'WEB_REMIX', + '_web_embedded': 'WEB_EMBEDDED_PLAYER', + '_web_agegate': 'TVHTML5', + 'android': 'ANDROID', + 'android_music': 'ANDROID_MUSIC', + '_android_embedded': 'ANDROID_EMBEDDED_PLAYER', + '_android_agegate': 'ANDROID', + 'ios': 'IOS', + 'ios_music': 'IOS_MUSIC', + '_ios_embedded': 'IOS_MESSAGES_EXTENSION', + '_ios_agegate': 'IOS' + } + def _get_default_ytcfg(self, client='WEB'): if client in self._YT_DEFAULT_YTCFGS: return copy.deepcopy(self._YT_DEFAULT_YTCFGS[client]) @@ -420,8 +479,11 @@ def _extract_client_name(self, ytcfg, default_client='WEB'): return self._ytcfg_get_safe(ytcfg, lambda x: x['INNERTUBE_CLIENT_NAME'], compat_str, default_client) @staticmethod - def _extract_session_index(ytcfg): - return int_or_none(try_get(ytcfg, lambda x: x['SESSION_INDEX'])) + def _extract_session_index(*data): + for ytcfg in data: + session_index = int_or_none(try_get(ytcfg, lambda x: x['SESSION_INDEX'])) + if session_index is not None: + return session_index def _extract_client_version(self, ytcfg, default_client='WEB'): return self._ytcfg_get_safe(ytcfg, lambda x: x['INNERTUBE_CLIENT_VERSION'], compat_str, default_client) @@ -473,7 +535,7 @@ def _call_api(self, ep, query, video_id, fatal=True, headers=None, data = {'context': context} if context else {'context': self._extract_context(default_client=default_client)} data.update(query) - real_headers = self._generate_api_headers(client=default_client) + real_headers = self.generate_api_headers(default_client=default_client) real_headers.update({'content-type': 'application/json'}) if headers: real_headers.update(headers) @@ -483,7 +545,7 @@ def _call_api(self, ep, query, video_id, fatal=True, headers=None, data=json.dumps(data).encode('utf8'), headers=real_headers, query={'key': api_key or self._extract_api_key()}) - def _extract_yt_initial_data(self, video_id, webpage): + def extract_yt_initial_data(self, video_id, webpage): return self._parse_json( self._search_regex( (r'%s\s*%s' % (self._YT_INITIAL_DATA_RE, self._YT_INITIAL_BOUNDARY_RE), @@ -491,7 +553,9 @@ def _extract_yt_initial_data(self, video_id, webpage): video_id) def _extract_identity_token(self, webpage, item_id): - ytcfg = self._extract_ytcfg(item_id, webpage) + if not webpage: + return None + ytcfg = self.extract_ytcfg(item_id, webpage) if ytcfg: token = try_get(ytcfg, lambda x: x['ID_TOKEN'], compat_str) if token: @@ -519,7 +583,7 @@ def _extract_account_syncid(*args): # and just "user_syncid||" for primary channel. We only want the channel_syncid return sync_ids[0] - def _extract_ytcfg(self, video_id, webpage): + def extract_ytcfg(self, video_id, webpage): if not webpage: return {} return self._parse_json( @@ -527,18 +591,19 @@ def _extract_ytcfg(self, video_id, webpage): r'ytcfg\.set\s*\(\s*({.+?})\s*\)\s*;', webpage, 'ytcfg', default='{}'), video_id, fatal=False) or {} - def _generate_api_headers(self, ytcfg=None, identity_token=None, account_syncid=None, - visitor_data=None, api_hostname=None, client='WEB', session_index=None): - origin = 'https://' + (api_hostname if api_hostname else self._get_innertube_host(client)) + def generate_api_headers( + self, ytcfg=None, identity_token=None, account_syncid=None, + visitor_data=None, api_hostname=None, default_client='WEB', session_index=None): + origin = 'https://' + (api_hostname if api_hostname else self._get_innertube_host(default_client)) headers = { 'X-YouTube-Client-Name': compat_str( - self._ytcfg_get_safe(ytcfg, lambda x: x['INNERTUBE_CONTEXT_CLIENT_NAME'], default_client=client)), - 'X-YouTube-Client-Version': self._extract_client_version(ytcfg, client), + self._ytcfg_get_safe(ytcfg, lambda x: x['INNERTUBE_CONTEXT_CLIENT_NAME'], default_client=default_client)), + 'X-YouTube-Client-Version': self._extract_client_version(ytcfg, default_client), 'Origin': origin } if not visitor_data and ytcfg: visitor_data = try_get( - self._extract_context(ytcfg, client), lambda x: x['client']['visitorData'], compat_str) + self._extract_context(ytcfg, default_client), lambda x: x['client']['visitorData'], compat_str) if identity_token: headers['X-Youtube-Identity-Token'] = identity_token if account_syncid: @@ -1122,7 +1187,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor): 'uploader_id': 'olympic', 'uploader_url': r're:https?://(?:www\.)?youtube\.com/user/olympic', 'description': 'HO09 - Women - GER-AUS - Hockey - 31 July 2012 - London 2012 Olympic Games', - 'uploader': 'Olympic', + 'uploader': 'Olympics', 'title': 'Hockey - Women - GER-AUS - London 2012 Olympic Games', }, 'params': { @@ -1287,16 +1352,16 @@ class YoutubeIE(YoutubeBaseInfoExtractor): 'id': 'lsguqyKfVQg', 'ext': 'mp4', 'title': '{dark walk}; Loki/AC/Dishonored; collab w/Elflover21', - 'alt_title': 'Dark Walk - Position Music', + 'alt_title': 'Dark Walk', 'description': 'md5:8085699c11dc3f597ce0410b0dcbb34a', 'duration': 133, 'upload_date': '20151119', 'uploader_id': 'IronSoulElf', 'uploader_url': r're:https?://(?:www\.)?youtube\.com/user/IronSoulElf', 'uploader': 'IronSoulElf', - 'creator': 'Todd Haberman, Daniel Law Heath and Aaron Kaplan', - 'track': 'Dark Walk - Position Music', - 'artist': 'Todd Haberman, Daniel Law Heath and Aaron Kaplan', + 'creator': 'Todd Haberman;\nDaniel Law Heath and Aaron Kaplan', + 'track': 'Dark Walk', + 'artist': 'Todd Haberman;\nDaniel Law Heath and Aaron Kaplan', 'album': 'Position Music - Production Music Vol. 143 - Dark Walk', }, 'params': { @@ -1615,6 +1680,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor): 'title': 'San Diego teen commits suicide after bullying over embarrassing video', 'channel_id': 'UC-SJ6nODDmufqBzPBwCvYvQ', 'uploader': 'CBS This Morning', + 'uploader_id': 'CBSThisMorning', 'upload_date': '20140716', 'description': 'md5:acde3a73d3f133fc97e837a9f76b53b7' } @@ -1652,13 +1718,14 @@ class YoutubeIE(YoutubeBaseInfoExtractor): 'url': 'https://www.youtube.com/watch?v=YOelRv7fMxY', 'info_dict': { 'id': 'YOelRv7fMxY', - 'title': 'Digging a Secret Tunnel from my Workshop', + 'title': 'DIGGING A SECRET TUNNEL Part 1', 'ext': '3gp', 'upload_date': '20210624', 'channel_id': 'UCp68_FLety0O-n9QU6phsgw', 'uploader': 'colinfurze', + 'uploader_id': 'colinfurze', 'channel_url': r're:https?://(?:www\.)?youtube\.com/channel/UCp68_FLety0O-n9QU6phsgw', - 'description': 'md5:ecb672623246d98c6c562eed6ae798c3' + 'description': 'md5:b5096f56af7ccd7a555c84db81738b22' }, 'params': { 'format': '17', # 3gp format available on android @@ -1692,10 +1759,12 @@ def __init__(self, *args, **kwargs): def _extract_player_url(self, ytcfg=None, webpage=None): player_url = try_get(ytcfg, (lambda x: x['PLAYER_JS_URL']), str) - if not player_url: + if not player_url and webpage: player_url = self._search_regex( r'"(?:PLAYER_JS_URL|jsUrl)"\s*:\s*"([^"]+)"', webpage, 'player URL', fatal=False) + if not player_url: + return None if player_url.startswith('//'): player_url = 'https:' + player_url elif not re.match(r'https?://', player_url): @@ -1859,10 +1928,10 @@ def _extract_signature_timestamp(self, video_id, player_url, ytcfg=None, fatal=F 'JS player signature timestamp', group='sts', fatal=fatal)) return sts - def _mark_watched(self, video_id, player_response): - playback_url = url_or_none(try_get( - player_response, - lambda x: x['playbackTracking']['videostatsPlaybackUrl']['baseUrl'])) + def _mark_watched(self, video_id, player_responses): + playback_url = url_or_none((traverse_obj( + player_responses, ('playbackTracking', 'videostatsPlaybackUrl', 'baseUrl'), + expected_type=str) or [None])[0]) if not playback_url: return parsed_playback_url = compat_urlparse.urlparse(playback_url) @@ -2122,7 +2191,7 @@ def extract_thread(contents): for page_num in itertools.count(0): if not continuation: break - headers = self._generate_api_headers(ytcfg, identity_token, account_syncid, visitor_data) + headers = self.generate_api_headers(ytcfg, identity_token, account_syncid, visitor_data) comment_prog_str = '(%d/%d)' % (comment_counts[0], comment_counts[1]) if page_num == 0: if is_first_continuation: @@ -2277,6 +2346,10 @@ def _get_video_info_params(video_id, client='TVHTML5'): 'TVHTML5': { 'c': 'TVHTML5', 'cver': '6.20180913', + }, + 'IOS': { + 'c': 'IOS', + 'cver': '16.20' } } query = { @@ -2287,215 +2360,111 @@ def _get_video_info_params(video_id, client='TVHTML5'): query.update(GVI_CLIENTS.get(client)) return query - def _real_extract(self, url): - url, smuggled_data = unsmuggle_url(url, {}) - video_id = self._match_id(url) + def _extract_player_response(self, client, video_id, master_ytcfg, player_ytcfg, identity_token, player_url, initial_pr): - is_music_url = smuggled_data.get('is_music_url') or self.is_music_url(url) + session_index = self._extract_session_index(player_ytcfg, master_ytcfg) + syncid = self._extract_account_syncid(player_ytcfg, master_ytcfg, initial_pr) + sts = self._extract_signature_timestamp(video_id, player_url, master_ytcfg, fatal=False) + headers = self.generate_api_headers( + player_ytcfg, identity_token, syncid, + default_client=self._YT_CLIENTS[client], session_index=session_index) - base_url = self.http_scheme() + '//www.youtube.com/' - webpage_url = base_url + 'watch?v=' + video_id - webpage = self._download_webpage( - webpage_url + '&bpctr=9999999999&has_verified=1', video_id, fatal=False) + yt_query = {'videoId': video_id} + yt_query.update(self._generate_player_context(sts)) + return self._extract_response( + item_id=video_id, ep='player', query=yt_query, + ytcfg=player_ytcfg, headers=headers, fatal=False, + default_client=self._YT_CLIENTS[client], + note='Downloading %s player API JSON' % client.replace('_', ' ').strip() + ) or None - ytcfg = self._extract_ytcfg(video_id, webpage) or self._get_default_ytcfg() - identity_token = self._extract_identity_token(webpage, video_id) - session_index = self._extract_session_index(ytcfg) - player_url = self._extract_player_url(ytcfg, webpage) + def _extract_age_gated_player_response(self, client, video_id, ytcfg, identity_token, player_url, initial_pr): + gvi_client = self._YT_CLIENTS.get(f'_{client}_agegate') + if not gvi_client: + return - player_client = self._configuration_arg('player_client', [''])[0] - if player_client not in ('web', 'android', ''): - self.report_warning(f'Invalid player_client {player_client} given. Falling back to android client.') - force_mobile_client = player_client != 'web' - player_skip = self._configuration_arg('player_skip') - player_response = None + pr = self._parse_json(traverse_obj( + compat_parse_qs(self._download_webpage( + self.http_scheme() + '//www.youtube.com/get_video_info', video_id, + 'Refetching age-gated %s info webpage' % gvi_client.lower(), + 'unable to download video info webpage', fatal=False, + query=self._get_video_info_params(video_id, client=gvi_client))), + ('player_response', 0), expected_type=str) or '{}', video_id) + if pr: + return pr + + self.report_warning('Falling back to embedded-only age-gate workaround') + embed_webpage = None + if client == 'web' and 'configs' not in self._configuration_arg('player_skip'): + embed_webpage = self._download_webpage( + 'https://www.youtube.com/embed/%s?html5=1' % video_id, + video_id=video_id, note=f'Downloading age-gated {client} embed config') + + ytcfg_age = self.extract_ytcfg(video_id, embed_webpage) or {} + # If we extracted the embed webpage, it'll tell us if we can view the video + embedded_pr = self._parse_json( + traverse_obj(ytcfg_age, ('PLAYER_VARS', 'embedded_player_response'), expected_type=str) or '{}', + video_id=video_id) + embedded_ps_reason = traverse_obj(embedded_pr, ('playabilityStatus', 'reason'), expected_type=str) or '' + if embedded_ps_reason in self._AGE_GATE_REASONS: + return + return self._extract_player_response( + f'_{client}_embedded', video_id, + ytcfg_age or ytcfg, ytcfg_age if client == 'web' else {}, + identity_token, player_url, initial_pr) + + def _get_requested_clients(self, url, smuggled_data): + requested_clients = [client for client in self._configuration_arg('player_client') + if client[:0] != '_' and client in self._YT_CLIENTS] + if not requested_clients: + requested_clients = ['android', 'web'] + + if smuggled_data.get('is_music_url') or self.is_music_url(url): + requested_clients.extend( + f'{client}_music' for client in requested_clients if not client.endswith('_music')) + + return orderedSet(requested_clients) + + def _extract_player_responses(self, clients, video_id, webpage, master_ytcfg, player_url, identity_token): + initial_pr = None if webpage: - player_response = self._extract_yt_initial_variable( + initial_pr = self._extract_yt_initial_variable( webpage, self._YT_INITIAL_PLAYER_RESPONSE_RE, video_id, 'initial player response') - syncid = self._extract_account_syncid(ytcfg, player_response) - headers = self._generate_api_headers(ytcfg, identity_token, syncid, session_index=session_index) - - ytm_streaming_data = {} - if is_music_url: - ytm_webpage = None - sts = self._extract_signature_timestamp(video_id, player_url, ytcfg, fatal=False) - if sts and not force_mobile_client and 'configs' not in player_skip: - ytm_webpage = self._download_webpage( - 'https://music.youtube.com', - video_id, fatal=False, note='Downloading remix client config') - - ytm_cfg = self._extract_ytcfg(video_id, ytm_webpage) or {} - ytm_client = 'WEB_REMIX' - if not sts or force_mobile_client: - # Android client already has signature descrambled - # See: https://github.com/TeamNewPipe/NewPipeExtractor/issues/562 - if not sts: - self.report_warning('Falling back to android remix client for player API.') - ytm_client = 'ANDROID_MUSIC' - ytm_cfg = {} - - ytm_headers = self._generate_api_headers( - ytm_cfg, identity_token, syncid, - client=ytm_client, session_index=session_index) - ytm_query = {'videoId': video_id} - ytm_query.update(self._generate_player_context(sts)) - - ytm_player_response = self._extract_response( - item_id=video_id, ep='player', query=ytm_query, - ytcfg=ytm_cfg, headers=ytm_headers, fatal=False, - default_client=ytm_client, - note='Downloading %sremix player API JSON' % ('android ' if force_mobile_client else '')) - ytm_streaming_data = try_get(ytm_player_response, lambda x: x['streamingData'], dict) or {} - - if not player_response or force_mobile_client: - sts = self._extract_signature_timestamp(video_id, player_url, ytcfg, fatal=False) - yt_client = 'WEB' - ytpcfg = ytcfg - ytp_headers = headers - if not sts or force_mobile_client: - # Android client already has signature descrambled - # See: https://github.com/TeamNewPipe/NewPipeExtractor/issues/562 - if not sts: - self.report_warning('Falling back to android client for player API.') - yt_client = 'ANDROID' - ytpcfg = {} - ytp_headers = self._generate_api_headers(ytpcfg, identity_token, syncid, - client=yt_client, session_index=session_index) - - yt_query = {'videoId': video_id} - yt_query.update(self._generate_player_context(sts)) - player_response = self._extract_response( - item_id=video_id, ep='player', query=yt_query, - ytcfg=ytpcfg, headers=ytp_headers, fatal=False, - default_client=yt_client, - note='Downloading %splayer API JSON' % ('android ' if force_mobile_client else '') - ) or player_response - - # Age-gate workarounds - playability_status = player_response.get('playabilityStatus') or {} - if playability_status.get('reason') in self._AGE_GATE_REASONS: - gvi_clients = ('ANDROID', 'TVHTML5') if force_mobile_client else ('TVHTML5', 'ANDROID') - for gvi_client in gvi_clients: - pr = self._parse_json(try_get(compat_parse_qs( - self._download_webpage( - base_url + 'get_video_info', video_id, - 'Refetching age-gated %s info webpage' % gvi_client.lower(), - 'unable to download video info webpage', fatal=False, - query=self._get_video_info_params(video_id, client=gvi_client))), - lambda x: x['player_response'][0], - compat_str) or '{}', video_id) - if pr: - break - if not pr: - self.report_warning('Falling back to embedded-only age-gate workaround.') - embed_webpage = None - sts = self._extract_signature_timestamp(video_id, player_url, ytcfg, fatal=False) - if sts and not force_mobile_client and 'configs' not in player_skip: - embed_webpage = self._download_webpage( - 'https://www.youtube.com/embed/%s?html5=1' % video_id, - video_id=video_id, note='Downloading age-gated embed config') - - ytcfg_age = self._extract_ytcfg(video_id, embed_webpage) or {} - # If we extracted the embed webpage, it'll tell us if we can view the video - embedded_pr = self._parse_json( - try_get(ytcfg_age, lambda x: x['PLAYER_VARS']['embedded_player_response'], str) or '{}', - video_id=video_id) - embedded_ps_reason = try_get(embedded_pr, lambda x: x['playabilityStatus']['reason'], str) or '' - if embedded_ps_reason not in self._AGE_GATE_REASONS: - yt_client = 'WEB_EMBEDDED_PLAYER' - if not sts or force_mobile_client: - # Android client already has signature descrambled - # See: https://github.com/TeamNewPipe/NewPipeExtractor/issues/562 - if not sts: - self.report_warning( - 'Falling back to android embedded client for player API (note: some formats may be missing).') - yt_client = 'ANDROID_EMBEDDED_PLAYER' - ytcfg_age = {} - - ytage_headers = self._generate_api_headers( - ytcfg_age, identity_token, syncid, - client=yt_client, session_index=session_index) - yt_age_query = {'videoId': video_id} - yt_age_query.update(self._generate_player_context(sts)) - pr = self._extract_response( - item_id=video_id, ep='player', query=yt_age_query, - ytcfg=ytcfg_age, headers=ytage_headers, fatal=False, - default_client=yt_client, - note='Downloading %sage-gated player API JSON' % ('android ' if force_mobile_client else '') - ) or {} - - if pr: - player_response = pr - - trailer_video_id = try_get( - playability_status, - lambda x: x['errorScreen']['playerLegacyDesktopYpcTrailerRenderer']['trailerVideoId'], - compat_str) - if trailer_video_id: - return self.url_result( - trailer_video_id, self.ie_key(), trailer_video_id) - - search_meta = ( - lambda x: self._html_search_meta(x, webpage, default=None)) \ - if webpage else lambda x: None - - video_details = player_response.get('videoDetails') or {} - microformat = try_get( - player_response, - lambda x: x['microformat']['playerMicroformatRenderer'], - dict) or {} - video_title = video_details.get('title') \ - or self._get_text(microformat.get('title')) \ - or search_meta(['og:title', 'twitter:title', 'title']) - video_description = video_details.get('shortDescription') - - if not smuggled_data.get('force_singlefeed', False): - if not self.get_param('noplaylist'): - multifeed_metadata_list = try_get( - player_response, - lambda x: x['multicamera']['playerLegacyMulticameraRenderer']['metadataList'], - compat_str) - if multifeed_metadata_list: - entries = [] - feed_ids = [] - for feed in multifeed_metadata_list.split(','): - # Unquote should take place before split on comma (,) since textual - # fields may contain comma as well (see - # https://github.com/ytdl-org/youtube-dl/issues/8536) - feed_data = compat_parse_qs( - compat_urllib_parse_unquote_plus(feed)) - - def feed_entry(name): - return try_get( - feed_data, lambda x: x[name][0], compat_str) - - feed_id = feed_entry('id') - if not feed_id: - continue - feed_title = feed_entry('title') - title = video_title - if feed_title: - title += ' (%s)' % feed_title - entries.append({ - '_type': 'url_transparent', - 'ie_key': 'Youtube', - 'url': smuggle_url( - base_url + 'watch?v=' + feed_data['id'][0], - {'force_singlefeed': True}), - 'title': title, - }) - feed_ids.append(feed_id) - self.to_screen( - 'Downloading multifeed video (%s) - add --no-playlist to just download video %s' - % (', '.join(feed_ids), video_id)) - return self.playlist_result( - entries, video_id, video_title, video_description) + age_gated = False + for client in clients: + player_ytcfg = master_ytcfg if client == 'web' else {} + if age_gated: + pr = None + elif client == 'web' and initial_pr: + pr = initial_pr else: - self.to_screen('Downloading just video %s because of --no-playlist' % video_id) + if client == 'web_music' and 'configs' not in self._configuration_arg('player_skip'): + ytm_webpage = self._download_webpage( + 'https://music.youtube.com', + video_id, fatal=False, note='Downloading remix client config') + player_ytcfg = self.extract_ytcfg(video_id, ytm_webpage) or {} + pr = self._extract_player_response( + client, video_id, player_ytcfg or master_ytcfg, player_ytcfg, identity_token, player_url, initial_pr) + if pr: + yield pr + if age_gated or traverse_obj(pr, ('playabilityStatus', 'reason')) in self._AGE_GATE_REASONS: + age_gated = True + pr = self._extract_age_gated_player_response( + client, video_id, player_ytcfg or master_ytcfg, identity_token, player_url, initial_pr) + if pr: + yield pr + # Android player_response does not have microFormats which are needed for + # extraction of some data. So we return the initial_pr with formats + # stripped out even if not requested by the user + # See: https://github.com/yt-dlp/yt-dlp/issues/501 + if initial_pr and 'web' not in clients: + initial_pr['streamingData'] = None + yield initial_pr - formats, itags, stream_ids = [], [], [] + def _extract_formats(self, streaming_data, video_id, player_url, is_live): + itags, stream_ids = [], [] itag_qualities = {} q = qualities([ # "tiny" is the smallest video-only format. But some audio-only formats @@ -2503,12 +2472,7 @@ def feed_entry(name): 'tiny', 'audio_quality_low', 'audio_quality_medium', 'audio_quality_high', # Audio only formats 'small', 'medium', 'large', 'hd720', 'hd1080', 'hd1440', 'hd2160', 'hd2880', 'highres' ]) - - streaming_data = player_response.get('streamingData') or {} - streaming_formats = streaming_data.get('formats') or [] - streaming_formats.extend(streaming_data.get('adaptiveFormats') or []) - streaming_formats.extend(ytm_streaming_data.get('formats') or []) - streaming_formats.extend(ytm_streaming_data.get('adaptiveFormats') or []) + streaming_formats = traverse_obj(streaming_data, (..., ('formats', 'adaptiveFormats'), ...), default=[]) for fmt in streaming_formats: if fmt.get('targetDurationSec') or fmt.get('drmFamilies'): @@ -2554,7 +2518,8 @@ def feed_entry(name): 'asr': int_or_none(fmt.get('audioSampleRate')), 'filesize': int_or_none(fmt.get('contentLength')), 'format_id': itag, - 'format_note': audio_track.get('displayName') or fmt.get('qualityLabel') or quality, + 'format_note': ', '.join(filter(None, ( + audio_track.get('displayName'), fmt.get('qualityLabel') or quality))), 'fps': int_or_none(fmt.get('fps')), 'height': int_or_none(fmt.get('height')), 'quality': q(quality), @@ -2572,6 +2537,7 @@ def feed_entry(name): # but is actually worse than all other formats if dct['ext'] == '3gp': dct['quality'] = q('tiny') + dct['preference'] = -10 no_audio = dct.get('acodec') == 'none' no_video = dct.get('vcodec') == 'none' if no_audio: @@ -2585,22 +2551,25 @@ def feed_entry(name): } if dct.get('ext'): dct['container'] = dct['ext'] + '_dash' - formats.append(dct) + yield dct skip_manifests = self._configuration_arg('skip') - get_dash = 'dash' not in skip_manifests and self.get_param('youtube_include_dash_manifest', True) + get_dash = not is_live and 'dash' not in skip_manifests and self.get_param('youtube_include_dash_manifest', True) get_hls = 'hls' not in skip_manifests and self.get_param('youtube_include_hls_manifest', True) - for sd in (streaming_data, ytm_streaming_data): + for sd in streaming_data: hls_manifest_url = get_hls and sd.get('hlsManifestUrl') if hls_manifest_url: for f in self._extract_m3u8_formats( hls_manifest_url, video_id, 'mp4', fatal=False): itag = self._search_regex( r'/itag/(\d+)', f['url'], 'itag', default=None) + if itag in itags: + continue if itag: f['format_id'] = itag - formats.append(f) + itags.append(itag) + yield f dash_manifest_url = get_dash and sd.get('dashManifestUrl') if dash_manifest_url: @@ -2609,6 +2578,8 @@ def feed_entry(name): itag = f['format_id'] if itag in itags: continue + if itag: + itags.append(itag) if itag in itag_qualities: f['quality'] = q(itag_qualities[itag]) filesize = int_or_none(self._search_regex( @@ -2616,33 +2587,142 @@ def feed_entry(name): or f['url'], 'file size', default=None)) if filesize: f['filesize'] = filesize - formats.append(f) + yield f + + def _real_extract(self, url): + url, smuggled_data = unsmuggle_url(url, {}) + video_id = self._match_id(url) + + base_url = self.http_scheme() + '//www.youtube.com/' + webpage_url = base_url + 'watch?v=' + video_id + webpage = self._download_webpage( + webpage_url + '&bpctr=9999999999&has_verified=1', video_id, fatal=False) + + master_ytcfg = self.extract_ytcfg(video_id, webpage) or self._get_default_ytcfg() + player_url = self._extract_player_url(master_ytcfg, webpage) + identity_token = self._extract_identity_token(webpage, video_id) + + player_responses = list(self._extract_player_responses( + self._get_requested_clients(url, smuggled_data), + video_id, webpage, master_ytcfg, player_url, identity_token)) + + get_first = lambda obj, keys, **kwargs: ( + traverse_obj(obj, (..., *variadic(keys)), **kwargs) or [None])[0] + + playability_statuses = traverse_obj( + player_responses, (..., 'playabilityStatus'), expected_type=dict, default=[]) + + trailer_video_id = get_first( + playability_statuses, + ('errorScreen', 'playerLegacyDesktopYpcTrailerRenderer', 'trailerVideoId'), + expected_type=str) + if trailer_video_id: + return self.url_result( + trailer_video_id, self.ie_key(), trailer_video_id) + + search_meta = ((lambda x: self._html_search_meta(x, webpage, default=None)) + if webpage else (lambda x: None)) + + video_details = traverse_obj( + player_responses, (..., 'videoDetails'), expected_type=dict, default=[]) + microformats = traverse_obj( + player_responses, (..., 'microformat', 'playerMicroformatRenderer'), + expected_type=dict, default=[]) + video_title = ( + get_first(video_details, 'title') + or self._get_text(microformats, (..., 'title')) + or search_meta(['og:title', 'twitter:title', 'title'])) + video_description = get_first(video_details, 'shortDescription') + + if not smuggled_data.get('force_singlefeed', False): + if not self.get_param('noplaylist'): + multifeed_metadata_list = get_first( + player_responses, + ('multicamera', 'playerLegacyMulticameraRenderer', 'metadataList'), + expected_type=str) + if multifeed_metadata_list: + entries = [] + feed_ids = [] + for feed in multifeed_metadata_list.split(','): + # Unquote should take place before split on comma (,) since textual + # fields may contain comma as well (see + # https://github.com/ytdl-org/youtube-dl/issues/8536) + feed_data = compat_parse_qs( + compat_urllib_parse_unquote_plus(feed)) + + def feed_entry(name): + return try_get( + feed_data, lambda x: x[name][0], compat_str) + + feed_id = feed_entry('id') + if not feed_id: + continue + feed_title = feed_entry('title') + title = video_title + if feed_title: + title += ' (%s)' % feed_title + entries.append({ + '_type': 'url_transparent', + 'ie_key': 'Youtube', + 'url': smuggle_url( + '%swatch?v=%s' % (base_url, feed_data['id'][0]), + {'force_singlefeed': True}), + 'title': title, + }) + feed_ids.append(feed_id) + self.to_screen( + 'Downloading multifeed video (%s) - add --no-playlist to just download video %s' + % (', '.join(feed_ids), video_id)) + return self.playlist_result( + entries, video_id, video_title, video_description) + else: + self.to_screen('Downloading just video %s because of --no-playlist' % video_id) + + category = get_first(microformats, 'category') or search_meta('genre') + channel_id = get_first(video_details, 'channelId') \ + or get_first(microformats, 'externalChannelId') \ + or search_meta('channelId') + duration = int_or_none( + get_first(video_details, 'lengthSeconds') + or get_first(microformats, 'lengthSeconds')) \ + or parse_duration(search_meta('duration')) + is_live = get_first(video_details, 'isLive') + is_upcoming = get_first(video_details, 'isUpcoming') + owner_profile_url = get_first(microformats, 'ownerProfileUrl') + + streaming_data = traverse_obj(player_responses, (..., 'streamingData'), default=[]) + formats = list(self._extract_formats(streaming_data, video_id, player_url, is_live)) if not formats: - if not self.get_param('allow_unplayable_formats') and streaming_data.get('licenseInfos'): + if not self.get_param('allow_unplayable_formats') and traverse_obj(streaming_data, (..., 'licenseInfos')): self.raise_no_formats( 'This video is DRM protected.', expected=True) - pemr = try_get( - playability_status, - lambda x: x['errorScreen']['playerErrorMessageRenderer'], - dict) or {} - reason = self._get_text(pemr.get('reason')) or playability_status.get('reason') - subreason = pemr.get('subreason') + pemr = get_first( + playability_statuses, + ('errorScreen', 'playerErrorMessageRenderer'), expected_type=dict) or {} + reason = self._get_text(pemr, 'reason') or get_first(playability_statuses, 'reason') + subreason = clean_html(self._get_text(pemr, 'subreason') or '') if subreason: - subreason = clean_html(self._get_text(subreason)) if subreason == 'The uploader has not made this video available in your country.': - countries = microformat.get('availableCountries') + countries = get_first(microformats, 'availableCountries') if not countries: regions_allowed = search_meta('regionsAllowed') countries = regions_allowed.split(',') if regions_allowed else None self.raise_geo_restricted(subreason, countries, metadata_available=True) - reason += '\n' + subreason + reason += f'. {subreason}' if reason: self.raise_no_formats(reason, expected=True) + for f in formats: + # TODO: detect if throttled + if '&n=' in f['url']: # possibly throttled + f['source_preference'] = -10 + # note = f.get('format_note') + # f['format_note'] = f'{note} (throttled)' if note else '(throttled)' + self._sort_formats(formats) - keywords = video_details.get('keywords') or [] + keywords = get_first(video_details, 'keywords', expected_type=list) or [] if not keywords and webpage: keywords = [ unescapeHTML(m.group('content')) @@ -2660,36 +2740,24 @@ def feed_entry(name): f['stretched_ratio'] = ratio break - category = microformat.get('category') or search_meta('genre') - channel_id = video_details.get('channelId') \ - or microformat.get('externalChannelId') \ - or search_meta('channelId') - duration = int_or_none( - video_details.get('lengthSeconds') - or microformat.get('lengthSeconds')) \ - or parse_duration(search_meta('duration')) - is_live = video_details.get('isLive') - is_upcoming = video_details.get('isUpcoming') - owner_profile_url = microformat.get('ownerProfileUrl') - thumbnails = [] - for container in (video_details, microformat): - for thumbnail in (try_get( - container, - lambda x: x['thumbnail']['thumbnails'], list) or []): - thumbnail_url = thumbnail.get('url') - if not thumbnail_url: - continue - # Sometimes youtube gives a wrong thumbnail URL. See: - # https://github.com/yt-dlp/yt-dlp/issues/233 - # https://github.com/ytdl-org/youtube-dl/issues/28023 - if 'maxresdefault' in thumbnail_url: - thumbnail_url = thumbnail_url.split('?')[0] - thumbnails.append({ - 'url': thumbnail_url, - 'height': int_or_none(thumbnail.get('height')), - 'width': int_or_none(thumbnail.get('width')), - }) + thumbnail_dicts = traverse_obj( + (video_details, microformats), (..., ..., 'thumbnail', 'thumbnails', ...), + expected_type=dict, default=[]) + for thumbnail in thumbnail_dicts: + thumbnail_url = thumbnail.get('url') + if not thumbnail_url: + continue + # Sometimes youtube gives a wrong thumbnail URL. See: + # https://github.com/yt-dlp/yt-dlp/issues/233 + # https://github.com/ytdl-org/youtube-dl/issues/28023 + if 'maxresdefault' in thumbnail_url: + thumbnail_url = thumbnail_url.split('?')[0] + thumbnails.append({ + 'url': thumbnail_url, + 'height': int_or_none(thumbnail.get('height')), + 'width': int_or_none(thumbnail.get('width')), + }) thumbnail_url = search_meta(['og:image', 'twitter:image']) if thumbnail_url: thumbnails.append({ @@ -2725,34 +2793,31 @@ def feed_entry(name): 'thumbnails': thumbnails, 'description': video_description, 'upload_date': unified_strdate( - microformat.get('uploadDate') + get_first(microformats, 'uploadDate') or search_meta('uploadDate')), - 'uploader': video_details['author'], + 'uploader': get_first(video_details, 'author'), 'uploader_id': self._search_regex(r'/(?:channel|user)/([^/?&#]+)', owner_profile_url, 'uploader id') if owner_profile_url else None, 'uploader_url': owner_profile_url, 'channel_id': channel_id, - 'channel_url': 'https://www.youtube.com/channel/' + channel_id if channel_id else None, + 'channel_url': f'https://www.youtube.com/channel/{channel_id}' if channel_id else None, 'duration': duration, 'view_count': int_or_none( - video_details.get('viewCount') - or microformat.get('viewCount') + get_first((video_details, microformats), (..., 'viewCount')) or search_meta('interactionCount')), - 'average_rating': float_or_none(video_details.get('averageRating')), + 'average_rating': float_or_none(get_first(video_details, 'averageRating')), 'age_limit': 18 if ( - microformat.get('isFamilySafe') is False + get_first(microformats, 'isFamilySafe') is False or search_meta('isFamilyFriendly') == 'false' or search_meta('og:restrictions:age') == '18+') else 0, 'webpage_url': webpage_url, 'categories': [category] if category else None, 'tags': keywords, 'is_live': is_live, - 'playable_in_embed': playability_status.get('playableInEmbed'), - 'was_live': video_details.get('isLiveContent'), + 'playable_in_embed': get_first(playability_statuses, 'playableInEmbed'), + 'was_live': get_first(video_details, 'isLiveContent'), } - pctr = try_get( - player_response, - lambda x: x['captions']['playerCaptionsTracklistRenderer'], dict) + pctr = get_first(player_responses, ('captions', 'playerCaptionsTracklistRenderer'), expected_type=dict) subtitles = {} if pctr: def process_language(container, base_url, lang_code, sub_name, query): @@ -2827,9 +2892,13 @@ def process_language(container, base_url, lang_code, sub_name, query): webpage, self._YT_INITIAL_DATA_RE, video_id, 'yt initial data') if not initial_data: + headers = self.generate_api_headers( + master_ytcfg, identity_token, self._extract_account_syncid(master_ytcfg), + session_index=self._extract_session_index(master_ytcfg)) + initial_data = self._extract_response( item_id=video_id, ep='next', fatal=False, - ytcfg=ytcfg, headers=headers, query={'videoId': video_id}, + ytcfg=master_ytcfg, headers=headers, query={'videoId': video_id}, note='Downloading initial data API JSON') try: @@ -2943,8 +3012,8 @@ def process_language(container, base_url, lang_code, sub_name, query): if v: info[d_k] = v - is_private = bool_or_none(video_details.get('isPrivate')) - is_unlisted = bool_or_none(microformat.get('isUnlisted')) + is_private = get_first(video_details, 'isPrivate', expected_type=bool) + is_unlisted = get_first(microformats, 'isUnlisted', expected_type=bool) is_membersonly = None is_premium = None if initial_data and is_private is not None: @@ -2976,9 +3045,8 @@ def process_language(container, base_url, lang_code, sub_name, query): get_comments = self.get_param('getcomments', False) if get_annotations or get_comments: xsrf_token = None - ytcfg = self._extract_ytcfg(video_id, webpage) - if ytcfg: - xsrf_token = try_get(ytcfg, lambda x: x['XSRF_TOKEN'], compat_str) + if master_ytcfg: + xsrf_token = try_get(master_ytcfg, lambda x: x['XSRF_TOKEN'], compat_str) if not xsrf_token: xsrf_token = self._search_regex( r'([\'"])XSRF_TOKEN\1\s*:\s*([\'"])(?P(?:(?!\2).)+)\2', @@ -2986,12 +3054,14 @@ def process_language(container, base_url, lang_code, sub_name, query): # annotations if get_annotations: - invideo_url = try_get( - player_response, lambda x: x['annotations'][0]['playerAnnotationsUrlsRenderer']['invideoUrl'], compat_str) + invideo_url = get_first( + player_responses, + ('annotations', 0, 'playerAnnotationsUrlsRenderer', 'invideoUrl'), + expected_type=str) if xsrf_token and invideo_url: xsrf_field_name = None - if ytcfg: - xsrf_field_name = try_get(ytcfg, lambda x: x['XSRF_FIELD_NAME'], compat_str) + if master_ytcfg: + xsrf_field_name = try_get(master_ytcfg, lambda x: x['XSRF_FIELD_NAME'], compat_str) if not xsrf_field_name: xsrf_field_name = self._search_regex( r'([\'"])XSRF_FIELD_NAME\1\s*:\s*([\'"])(?P\w+)\2', @@ -3004,9 +3074,9 @@ def process_language(container, base_url, lang_code, sub_name, query): data=urlencode_postdata({xsrf_field_name: xsrf_token})) if get_comments: - info['__post_extractor'] = lambda: self._extract_comments(ytcfg, video_id, contents, webpage) + info['__post_extractor'] = lambda: self._extract_comments(master_ytcfg, video_id, contents, webpage) - self.mark_watched(video_id, player_response) + self.mark_watched(video_id, player_responses) return info @@ -3287,7 +3357,7 @@ class YoutubeTabIE(YoutubeBaseInfoExtractor): }, { 'url': 'https://www.youtube.com/channel/UCoMdktPbSTixAyNGwb-UYkQ/live', 'info_dict': { - 'id': 'X1whbWASnNQ', # This will keep changing + 'id': 'FMtPN8yp5LU', # This will keep changing 'ext': 'mp4', 'title': compat_str, 'uploader': 'Sky News', @@ -3693,7 +3763,7 @@ def extract_entries(parent_renderer): # this needs to called again for continua for page_num in itertools.count(1): if not continuation: break - headers = self._generate_api_headers(ytcfg, identity_token, account_syncid, visitor_data) + headers = self.generate_api_headers(ytcfg, identity_token, account_syncid, visitor_data) response = self._extract_response( item_id='%s page %s' % (item_id, page_num), query=continuation, headers=headers, ytcfg=ytcfg, @@ -3843,7 +3913,7 @@ def _extract_from_tabs(self, item_id, webpage, data, tabs): 'channel': metadata['uploader'], 'channel_id': metadata['uploader_id'], 'channel_url': metadata['uploader_url']}) - ytcfg = self._extract_ytcfg(item_id, webpage) + ytcfg = self.extract_ytcfg(item_id, webpage) return self.playlist_result( self._entries( selected_tab, playlist_id, @@ -3853,8 +3923,8 @@ def _extract_from_tabs(self, item_id, webpage, data, tabs): def _extract_mix_playlist(self, playlist, playlist_id, data, webpage): first_id = last_id = None - ytcfg = self._extract_ytcfg(playlist_id, webpage) - headers = self._generate_api_headers( + ytcfg = self.extract_ytcfg(playlist_id, webpage) + headers = self.generate_api_headers( ytcfg, account_syncid=self._extract_account_syncid(ytcfg, data), identity_token=self._extract_identity_token(webpage, item_id=playlist_id)) for page_num in itertools.count(1): @@ -3971,8 +4041,8 @@ def _reload_with_unavailable_videos(self, item_id, data, webpage): params = browse_endpoint.get('params') break - ytcfg = self._extract_ytcfg(item_id, webpage) - headers = self._generate_api_headers( + ytcfg = self.extract_ytcfg(item_id, webpage) + headers = self.generate_api_headers( ytcfg, account_syncid=self._extract_account_syncid(ytcfg, data), identity_token=self._extract_identity_token(webpage, item_id=item_id), visitor_data=try_get( @@ -3999,7 +4069,7 @@ def _extract_webpage(self, url, item_id): webpage = self._download_webpage( url, item_id, 'Downloading webpage%s' % (' (retry #%d)' % count if count else '')) - data = self._extract_yt_initial_data(item_id, webpage) + data = self.extract_yt_initial_data(item_id, webpage) if data.get('contents') or data.get('currentVideoEndpoint'): break # Extract alerts here only when there is error @@ -4165,6 +4235,7 @@ class YoutubePlaylistIE(InfoExtractor): 'id': 'PLBB231211A4F62143', 'uploader': 'Wickydoo', 'uploader_id': 'UCKSpbfbl5kRQpTdL7kMc-1Q', + 'description': 'md5:8fa6f52abb47a9552002fa3ddfc57fc2', }, 'playlist_mincount': 29, }, { @@ -4187,12 +4258,13 @@ class YoutubePlaylistIE(InfoExtractor): } }, { 'url': 'http://www.youtube.com/embed/_xDOZElKyNU?list=PLsyOSbh5bs16vubvKePAQ1x3PhKavfBIl', - 'playlist_mincount': 982, + 'playlist_mincount': 654, 'info_dict': { 'title': '2018 Chinese New Singles (11/6 updated)', 'id': 'PLsyOSbh5bs16vubvKePAQ1x3PhKavfBIl', 'uploader': 'LBK', 'uploader_id': 'UC21nz3_MesPLqtDqwdvnoxA', + 'description': 'md5:da521864744d60a198e3a88af4db0d9d', } }, { 'url': 'TLGGrESM50VT6acwMjAyMjAxNw', @@ -4366,7 +4438,7 @@ def _entries(self, query, n): def _get_n_results(self, query, n): """Get a specified number of results for a query""" - return self.playlist_result(self._entries(query, n), query) + return self.playlist_result(self._entries(query, n), query, query) class YoutubeSearchDateIE(YoutubeSearchIE): @@ -4385,6 +4457,7 @@ class YoutubeSearchURLIE(YoutubeSearchIE): 'url': 'https://www.youtube.com/results?baz=bar&search_query=youtube-dl+test+video&filters=video&lclk=video', 'playlist_mincount': 5, 'info_dict': { + 'id': 'youtube-dl test video', 'title': 'youtube-dl test video', } }, {