[extractor/tiktok] Improve `TikTokLive` extractor (#6520)

Closes #6459
Authored by: bashonly

yt_dlp/extractor/genius.py

@@ -10,7 +10,7 @@ from ..utils import (
 
 
 class GeniusIE(InfoExtractor):
-    _VALID_URL = r'https?://(?:www\.)?genius\.com/videos/(?P<id>[^?/#]+)'
+    _VALID_URL = r'https?://(?:www\.)?genius\.com/(?:videos|(?P<article>a))/(?P<id>[^?/#]+)'
     _TESTS = [{
         'url': 'https://genius.com/videos/Vince-staples-breaks-down-the-meaning-of-when-sparks-fly',
         'md5': '64c2ad98cfafcfda23bfa0ad0c512f4c',
@@ -41,19 +41,37 @@ class GeniusIE(InfoExtractor):
             'timestamp': 1631209167,
             'thumbnail': r're:^https?://.*\.jpg$',
         },
+    }, {
+        'url': 'https://genius.com/a/cordae-anderson-paak-break-down-the-meaning-of-two-tens',
+        'md5': 'f98a4e03b16b0a2821bd6e52fb3cc9d7',
+        'info_dict': {
+            'id': '6321509903112',
+            'ext': 'mp4',
+            'title': 'Cordae & Anderson .Paak Breaks Down The Meaning Of “Two Tens”',
+            'description': 'md5:1255f0e1161d07342ce56a8464ac339d',
+            'tags': ['song id: 5457554'],
+            'uploader_id': '4863540648001',
+            'duration': 361.813,
+            'upload_date': '20230301',
+            'timestamp': 1677703908,
+            'thumbnail': r're:^https?://.*\.jpg$',
+        },
     }]
 
     def _real_extract(self, url):
-        display_id = self._match_id(url)
+        display_id, is_article = self._match_valid_url(url).group('id', 'article')
         webpage = self._download_webpage(url, display_id)
 
         metadata = self._search_json(
-            r'<meta content="', webpage, 'metadata', display_id, transform_source=unescapeHTML)
-        video_id = traverse_obj(
-            metadata, ('video', 'provider_id'),
-            ('dfp_kv', lambda _, x: x['name'] == 'brightcove_video_id', 'values', 0), get_all=False)
+            r'<meta content="', webpage, 'metadata', display_id,
+            end_pattern=r'"\s+itemprop="page_data"', transform_source=unescapeHTML)
+        video_id = traverse_obj(metadata, (
+            (('article', 'media', ...), ('video', None)),
+            ('provider_id', ('dfp_kv', lambda _, v: v['name'] == 'brightcove_video_id', 'values', ...))),
+            get_all=False)
         if not video_id:
-            raise ExtractorError('Brightcove video id not found in webpage')
+            # Not all article pages have videos, expect the error
+            raise ExtractorError('Brightcove video ID not found in webpage', expected=bool(is_article))
 
         config = self._search_json(r'var\s*APP_CONFIG\s*=', webpage, 'config', video_id, default={})
         account_id = config.get('brightcove_account_id', '4863540648001')
@@ -68,7 +86,7 @@ class GeniusIE(InfoExtractor):
 
 
 class GeniusLyricsIE(InfoExtractor):
-    _VALID_URL = r'https?://(?:www\.)?genius\.com/(?P<id>[^?/#]+)-lyrics[?/#]?'
+    _VALID_URL = r'https?://(?:www\.)?genius\.com/(?P<id>[^?/#]+)-lyrics(?:[?/#]|$)'
     _TESTS = [{
         'url': 'https://genius.com/Lil-baby-heyy-lyrics',
         'playlist_mincount': 2,

yt_dlp/extractor/mediastream.py

@@ -2,16 +2,44 @@ import re
 
 from .common import InfoExtractor
 from ..utils import (
+    clean_html,
     remove_end,
-    str_or_none,
-    strip_or_none,
     traverse_obj,
     urljoin,
 )
 
 
-class MediaStreamIE(InfoExtractor):
-    _VALID_URL = r'https?://mdstrm.com/(?:embed|live-stream)/(?P<id>\w+)'
+class MediaStreamBaseIE(InfoExtractor):
+    _EMBED_BASE_URL = 'https://mdstrm.com/embed'
+    _BASE_URL_RE = r'https?://mdstrm\.com/(?:embed|live-stream)'
+
+    def _extract_mediastream_urls(self, webpage):
+        yield from traverse_obj(list(self._yield_json_ld(webpage, None)), (
+            lambda _, v: v['@type'] == 'VideoObject', ('embedUrl', 'contentUrl'),
+            {lambda x: x if re.match(rf'{self._BASE_URL_RE}/\w+', x) else None}))
+
+        for mobj in re.finditer(r'<script[^>]+>[^>]*playerMdStream\.mdstreamVideo\(\s*[\'"](?P<video_id>\w+)', webpage):
+            yield f'{self._EMBED_BASE_URL}/{mobj.group("video_id")}'
+
+        yield from re.findall(
+            rf'<iframe[^>]+\bsrc="({self._BASE_URL_RE}/\w+)', webpage)
+
+        for mobj in re.finditer(
+            r'''(?x)
+                <(?:div|ps-mediastream)[^>]+
+                (class="[^"]*MediaStreamVideoPlayer)[^"]*"[^>]+
+                data-video-id="(?P<video_id>\w+)"
+                (?:\s*data-video-type="(?P<video_type>[^"]+))?
+                (?:[^>]*>\s*<div[^>]+\1[^"]*"[^>]+data-mediastream=["\'][^>]+
+                    https://mdstrm\.com/(?P<live>live-stream))?
+                ''', webpage):
+
+            video_type = 'live-stream' if mobj.group('video_type') == 'live' or mobj.group('live') else 'embed'
+            yield f'https://mdstrm.com/{video_type}/{mobj.group("video_id")}'
+
+
+class MediaStreamIE(MediaStreamBaseIE):
+    _VALID_URL = MediaStreamBaseIE._BASE_URL_RE + r'/(?P<id>\w+)'
 
     _TESTS = [{
         'url': 'https://mdstrm.com/embed/6318e3f1d1d316083ae48831',
@@ -23,6 +51,7 @@ class MediaStreamIE(InfoExtractor):
             'thumbnail': r're:^https?://[^?#]+6318e3f1d1d316083ae48831',
             'ext': 'mp4',
         },
+        'params': {'skip_download': 'm3u8'},
     }]
 
     _WEBPAGE_TESTS = [{
@@ -35,9 +64,7 @@ class MediaStreamIE(InfoExtractor):
             'ext': 'mp4',
             'live_status': 'is_live',
         },
-        'params': {
-            'skip_download': 'Livestream'
-        },
+        'params': {'skip_download': 'Livestream'},
     }, {
         'url': 'https://www.multimedios.com/television/clases-de-llaves-y-castigos-quien-sabe-mas',
         'md5': 'de31f0b1ecc321fb35bf22d58734ea40',
@@ -48,6 +75,7 @@ class MediaStreamIE(InfoExtractor):
             'thumbnail': 're:^https?://[^?#]+63731bab8ec9b308a2c9ed28',
             'ext': 'mp4',
         },
+        'params': {'skip_download': 'm3u8'},
     }, {
         'url': 'https://www.americatv.com.pe/videos/esto-es-guerra/facundo-gonzalez-sufrio-fuerte-golpe-durante-competencia-frente-hugo-garcia-eeg-noticia-139120',
         'info_dict': {
@@ -57,6 +85,7 @@ class MediaStreamIE(InfoExtractor):
             'thumbnail': 're:^https?://[^?#]+63756df1c638b008a5659dec',
             'ext': 'mp4',
         },
+        'params': {'skip_download': 'm3u8'},
     }, {
         'url': 'https://www.americatv.com.pe/videos/al-fondo-hay-sitio/nuevas-lomas-town-bernardo-mata-se-enfrento-sujeto-luchar-amor-macarena-noticia-139083',
         'info_dict': {
@@ -66,26 +95,12 @@ class MediaStreamIE(InfoExtractor):
             'thumbnail': 're:^https?://[^?#]+637307669609130f74cd3a6e',
             'ext': 'mp4',
         },
+        'params': {'skip_download': 'm3u8'},
     }]
 
-    @classmethod
-    def _extract_embed_urls(cls, url, webpage):
-        for mobj in re.finditer(r'<script[^>]+>[^>]*playerMdStream.mdstreamVideo\(\s*[\'"](?P<video_id>\w+)', webpage):
-            yield f'https://mdstrm.com/embed/{mobj.group("video_id")}'
-
-        yield from re.findall(
-            r'<iframe[^>]src\s*=\s*"(https://mdstrm.com/[\w-]+/\w+)', webpage)
-
-        for mobj in re.finditer(
-            r'''(?x)
-                <(?:div|ps-mediastream)[^>]+
-                class\s*=\s*"[^"]*MediaStreamVideoPlayer[^"]*"[^>]+
-                data-video-id\s*=\s*"(?P<video_id>\w+)\s*"
-                (?:\s*data-video-type\s*=\s*"(?P<video_type>[^"]+))?
-                ''', webpage):
-
-            video_type = 'live-stream' if mobj.group('video_type') == 'live' else 'embed'
-            yield f'https://mdstrm.com/{video_type}/{mobj.group("video_id")}'
+    def _extract_from_webpage(self, url, webpage):
+        for embed_url in self._extract_mediastream_urls(webpage):
+            yield self.url_result(embed_url, MediaStreamIE, None)
 
     def _real_extract(self, url):
         video_id = self._match_id(url)
@@ -94,7 +109,7 @@ class MediaStreamIE(InfoExtractor):
         if 'Debido a tu ubicación no puedes ver el contenido' in webpage:
             self.raise_geo_restricted()
 
-        player_config = self._search_json(r'window.MDSTRM.OPTIONS\s*=', webpage, 'metadata', video_id)
+        player_config = self._search_json(r'window\.MDSTRM\.OPTIONS\s*=', webpage, 'metadata', video_id)
 
         formats, subtitles = [], {}
         for video_format in player_config['src']:
@@ -122,7 +137,7 @@ class MediaStreamIE(InfoExtractor):
         }
 
 
-class WinSportsVideoIE(InfoExtractor):
+class WinSportsVideoIE(MediaStreamBaseIE):
     _VALID_URL = r'https?://www\.winsports\.co/videos/(?P<id>[\w-]+)'
 
     _TESTS = [{
@@ -158,21 +173,36 @@ class WinSportsVideoIE(InfoExtractor):
             'ext': 'mp4',
         },
         'params': {'skip_download': 'm3u8'},
+    }, {
+        'url': 'https://www.winsports.co/videos/bucaramanga-se-quedo-con-el-grito-de-gol-en-la-garganta',
+        'info_dict': {
+            'id': '6402adb62bbf3b18d454e1b0',
+            'display_id': 'bucaramanga-se-quedo-con-el-grito-de-gol-en-la-garganta',
+            'title': '⚽Bucaramanga se quedó con el grito de gol en la garganta',
+            'description': 'Gol anulado Bucaramanga',
+            'thumbnail': r're:^https?://[^?#]+6402adb62bbf3b18d454e1b0',
+            'ext': 'mp4',
+        },
+        'params': {'skip_download': 'm3u8'},
     }]
 
     def _real_extract(self, url):
         display_id = self._match_id(url)
         webpage = self._download_webpage(url, display_id)
-        json_ld = self._search_json_ld(webpage, display_id, expected_type='VideoObject', default={})
-        media_setting_json = self._search_json(
-            r'<script\s*[^>]+data-drupal-selector="drupal-settings-json">', webpage, 'drupal-setting-json', display_id)
+        data = self._search_json(
+            r'<script\s*[^>]+data-drupal-selector="drupal-settings-json">', webpage, 'data', display_id)
 
-        mediastream_id = traverse_obj(
-            media_setting_json, ('settings', 'mediastream_formatter', ..., 'mediastream_id', {str_or_none}),
-            get_all=False) or json_ld.get('url')
-        if not mediastream_id:
+        mediastream_url = urljoin(f'{self._EMBED_BASE_URL}/', (
+            traverse_obj(data, (
+                (('settings', 'mediastream_formatter', ..., 'mediastream_id'), 'url'), {str}), get_all=False)
+            or next(self._extract_mediastream_urls(webpage), None)))
+
+        if not mediastream_url:
             self.raise_no_formats('No MediaStream embed found in webpage')
 
+        title = clean_html(remove_end(
+            self._search_json_ld(webpage, display_id, expected_type='VideoObject', default={}).get('title')
+            or self._og_search_title(webpage), '| Win Sports'))
+
         return self.url_result(
-            urljoin('https://mdstrm.com/embed/', mediastream_id), MediaStreamIE, display_id, url_transparent=True,
-            display_id=display_id, video_title=strip_or_none(remove_end(json_ld.get('title'), '| Win Sports')))
+            mediastream_url, MediaStreamIE, display_id, url_transparent=True, display_id=display_id, video_title=title)

yt_dlp/extractor/tiktok.py

@@ -1,6 +1,7 @@
 import itertools
 import json
 import random
+import re
 import string
 import time
 
@@ -12,15 +13,18 @@ from ..utils import (
     LazyList,
     UnsupportedError,
     UserNotLive,
+    format_field,
     get_element_by_id,
     get_first,
     int_or_none,
     join_nonempty,
+    merge_dicts,
     qualities,
     remove_start,
     srt_subtitles_timecode,
     str_or_none,
     traverse_obj,
+    try_call,
     try_get,
     url_or_none,
 )
@@ -563,7 +567,7 @@ class TikTokIE(TikTokBaseIE):
             self.report_warning(f'{e}; trying with webpage')
 
         url = self._create_url(user_id, video_id)
-        webpage = self._download_webpage(url, video_id, headers={'User-Agent': 'User-Agent:Mozilla/5.0'})
+        webpage = self._download_webpage(url, video_id, headers={'User-Agent': 'Mozilla/5.0'})
         next_data = self._search_nextjs_data(webpage, video_id, default='{}')
         if next_data:
             status = traverse_obj(next_data, ('props', 'pageProps', 'statusCode'), expected_type=int) or 0
@@ -983,40 +987,173 @@ class TikTokVMIE(InfoExtractor):
         return self.url_result(new_url)
 
 
-class TikTokLiveIE(InfoExtractor):
-    _VALID_URL = r'https?://(?:www\.)?tiktok\.com/@(?P<id>[\w\.-]+)/live'
+class TikTokLiveIE(TikTokBaseIE):
+    _VALID_URL = r'''(?x)https?://(?:
+        (?:www\.)?tiktok\.com/@(?P<uploader>[\w.-]+)/live|
+        m\.tiktok\.com/share/live/(?P<id>\d+)
+    )'''
     IE_NAME = 'tiktok:live'
 
     _TESTS = [{
+        'url': 'https://www.tiktok.com/@weathernewslive/live',
+        'info_dict': {
+            'id': '7210809319192726273',
+            'ext': 'mp4',
+            'title': r're:ウェザーニュースLiVE[\d\s:-]*',
+            'creator': 'ウェザーニュースLiVE',
+            'uploader': 'weathernewslive',
+            'uploader_id': '6621496731283095554',
+            'uploader_url': 'https://www.tiktok.com/@weathernewslive',
+            'live_status': 'is_live',
+            'concurrent_view_count': int,
+        },
+        'params': {'skip_download': 'm3u8'},
+    }, {
+        'url': 'https://www.tiktok.com/@pilarmagenta/live',
+        'info_dict': {
+            'id': '7209423610325322522',
+            'ext': 'mp4',
+            'title': str,
+            'creator': 'Pilarmagenta',
+            'uploader': 'pilarmagenta',
+            'uploader_id': '6624846890674683909',
+            'uploader_url': 'https://www.tiktok.com/@pilarmagenta',
+            'live_status': 'is_live',
+            'concurrent_view_count': int,
+        },
+        'skip': 'Livestream',
+    }, {
+        'url': 'https://m.tiktok.com/share/live/7209423610325322522/?language=en',
+        'only_matching': True,
+    }, {
         'url': 'https://www.tiktok.com/@iris04201/live',
         'only_matching': True,
     }]
 
+    def _call_api(self, url, param, room_id, uploader, key=None):
+        response = traverse_obj(self._download_json(
+            url, room_id, fatal=False, query={
+                'aid': '1988',
+                param: room_id,
+            }), (key, {dict}), default={})
+
+        # status == 2 if live else 4
+        if int_or_none(response.get('status')) == 2:
+            return response
+        # If room_id is obtained via mobile share URL and cannot be refreshed, do not wait for live
+        elif not uploader:
+            raise ExtractorError('This livestream has ended', expected=True)
+        raise UserNotLive(video_id=uploader)
+
     def _real_extract(self, url):
-        uploader = self._match_id(url)
-        webpage = self._download_webpage(url, uploader, headers={'User-Agent': 'User-Agent:Mozilla/5.0'})
-        room_id = self._html_search_regex(r'snssdk\d*://live\?room_id=(\d+)', webpage, 'room ID', default=None)
+        uploader, room_id = self._match_valid_url(url).group('uploader', 'id')
+        webpage = self._download_webpage(
+            url, uploader or room_id, headers={'User-Agent': 'Mozilla/5.0'}, fatal=not room_id)
+
+        if webpage:
+            data = try_call(lambda: self._get_sigi_state(webpage, uploader or room_id))
+            room_id = (traverse_obj(data, ('UserModule', 'users', ..., 'roomId', {str_or_none}), get_all=False)
+                       or self._search_regex(r'snssdk\d*://live\?room_id=(\d+)', webpage, 'room ID', default=None)
+                       or room_id)
+            uploader = uploader or traverse_obj(
+                data, ('LiveRoom', 'liveRoomUserInfo', 'user', 'uniqueId'),
+                ('UserModule', 'users', ..., 'uniqueId'), get_all=False, expected_type=str)
+
         if not room_id:
             raise UserNotLive(video_id=uploader)
-        live_info = traverse_obj(self._download_json(
-            'https://www.tiktok.com/api/live/detail/', room_id, query={
-                'aid': '1988',
-                'roomID': room_id,
-            }), 'LiveRoomInfo', expected_type=dict, default={})
 
-        if 'status' not in live_info:
-            raise ExtractorError('Unexpected response from TikTok API')
-        # status = 2 if live else 4
-        if not int_or_none(live_info['status']) == 2:
-            raise UserNotLive(video_id=uploader)
+        formats = []
+        live_info = self._call_api(
+            'https://webcast.tiktok.com/webcast/room/info', 'room_id', room_id, uploader, key='data')
+
+        get_quality = qualities(('SD1', 'ld', 'SD2', 'sd', 'HD1', 'hd', 'FULL_HD1', 'uhd', 'ORIGION', 'origin'))
+        parse_inner = lambda x: self._parse_json(x, None)
+
+        for quality, stream in traverse_obj(live_info, (
+                'stream_url', 'live_core_sdk_data', 'pull_data', 'stream_data',
+                {parse_inner}, 'data', {dict}), default={}).items():
+
+            sdk_params = traverse_obj(stream, ('main', 'sdk_params', {parse_inner}, {
+                'vcodec': ('VCodec', {str}),
+                'tbr': ('vbitrate', {lambda x: int_or_none(x, 1000)}),
+                'resolution': ('resolution', {lambda x: re.match(r'(?i)\d+x\d+|\d+p', x).group().lower()}),
+            }))
+
+            flv_url = traverse_obj(stream, ('main', 'flv', {url_or_none}))
+            if flv_url:
+                formats.append({
+                    'url': flv_url,
+                    'ext': 'flv',
+                    'format_id': f'flv-{quality}',
+                    'quality': get_quality(quality),
+                    **sdk_params,
+                })
+
+            hls_url = traverse_obj(stream, ('main', 'hls', {url_or_none}))
+            if hls_url:
+                formats.append({
+                    'url': hls_url,
+                    'ext': 'mp4',
+                    'protocol': 'm3u8_native',
+                    'format_id': f'hls-{quality}',
+                    'quality': get_quality(quality),
+                    **sdk_params,
+                })
+
+        def get_vcodec(*keys):
+            return traverse_obj(live_info, (
+                'stream_url', *keys, {parse_inner}, 'VCodec', {str}))
+
+        for stream in ('hls', 'rtmp'):
+            stream_url = traverse_obj(live_info, ('stream_url', f'{stream}_pull_url', {url_or_none}))
+            if stream_url:
+                formats.append({
+                    'url': stream_url,
+                    'ext': 'mp4' if stream == 'hls' else 'flv',
+                    'protocol': 'm3u8_native' if stream == 'hls' else 'https',
+                    'format_id': f'{stream}-pull',
+                    'vcodec': get_vcodec(f'{stream}_pull_url_params'),
+                    'quality': get_quality('ORIGION'),
+                })
+
+        for f_id, f_url in traverse_obj(live_info, ('stream_url', 'flv_pull_url', {dict}), default={}).items():
+            if not url_or_none(f_url):
+                continue
+            formats.append({
+                'url': f_url,
+                'ext': 'flv',
+                'format_id': f'flv-{f_id}'.lower(),
+                'vcodec': get_vcodec('flv_pull_url_params', f_id),
+                'quality': get_quality(f_id),
+            })
+
+        # If uploader is a guest on another's livestream, primary endpoint will not have m3u8 URLs
+        if not traverse_obj(formats, lambda _, v: v['ext'] == 'mp4'):
+            live_info = merge_dicts(live_info, self._call_api(
+                'https://www.tiktok.com/api/live/detail/', 'roomID', room_id, uploader, key='LiveRoomInfo'))
+            if url_or_none(live_info.get('liveUrl')):
+                formats.append({
+                    'url': live_info['liveUrl'],
+                    'ext': 'mp4',
+                    'protocol': 'm3u8_native',
+                    'format_id': 'hls-fallback',
+                    'vcodec': 'h264',
+                    'quality': get_quality('origin'),
+                })
+
+        uploader = uploader or traverse_obj(live_info, ('ownerInfo', 'uniqueId'), ('owner', 'display_id'))
 
         return {
             'id': room_id,
-            'title': live_info.get('title') or self._html_search_meta(['og:title', 'twitter:title'], webpage, default=''),
             'uploader': uploader,
-            'uploader_id': traverse_obj(live_info, ('ownerInfo', 'id')),
-            'creator': traverse_obj(live_info, ('ownerInfo', 'nickname')),
-            'concurrent_view_count': traverse_obj(live_info, ('liveRoomStats', 'userCount'), expected_type=int),
-            'formats': self._extract_m3u8_formats(live_info['liveUrl'], room_id, 'mp4', live=True),
+            'uploader_url': format_field(uploader, None, self._UPLOADER_URL_FORMAT) or None,
             'is_live': True,
+            'formats': formats,
+            '_format_sort_fields': ('quality', 'ext'),
+            **traverse_obj(live_info, {
+                'title': 'title',
+                'uploader_id': (('ownerInfo', 'owner'), 'id', {str_or_none}),
+                'creator': (('ownerInfo', 'owner'), 'nickname'),
+                'concurrent_view_count': (('user_count', ('liveRoomStats', 'userCount')), {int_or_none}),
+            }, get_all=False),
         }