0
0
Fork 0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-26 02:55:17 +00:00

[extractor/twitch] Support storyboards for VODs (#4342)

Authored by: ftk
This commit is contained in:
ftk 2022-07-12 20:27:50 +00:00 committed by GitHub
parent 45e8a04e48
commit 418bbfd722
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23

View file

@ -12,6 +12,7 @@
compat_urllib_parse_urlparse,
)
from ..utils import (
base_url,
clean_html,
dict_get,
ExtractorError,
@ -52,6 +53,7 @@ class TwitchBaseIE(InfoExtractor):
'VideoPreviewOverlay': '3006e77e51b128d838fa4e835723ca4dc9a05c5efd4466c1085215c6e437e65c',
'VideoMetadata': '226edb3e692509f727fd56821f5653c05740242c82b0388883e0c0e75dcbf687',
'VideoPlayer_ChapterSelectButtonVideo': '8d2793384aac3773beab5e59bd5d6f585aedb923d292800119e03d40cd0f9b41',
'VideoPlayer_VODSeekbarPreviewVideo': '07e99e4d56c5a7c67117a154777b0baf85a5ffefa393b213f4bc712ccaf85dd6',
}
def _perform_login(self, username, password):
@ -202,6 +204,8 @@ class TwitchVodIE(TwitchBaseIE):
'uploader_id': 'riotgames',
'view_count': int,
'start_time': 310,
'chapters': [],
'live_status': 'was_live',
},
'params': {
# m3u8 download
@ -270,9 +274,52 @@ class TwitchVodIE(TwitchBaseIE):
'title': 'Art'
}
],
'live_status': 'was_live',
'thumbnail': r're:^https?://.*\.jpg$',
'view_count': int,
},
'params': {
'skip_download': True
},
}, {
'note': 'Storyboards',
'url': 'https://www.twitch.tv/videos/635475444',
'info_dict': {
'id': 'v635475444',
'format_id': 'sb0',
'ext': 'mhtml',
'title': 'Riot Games',
'duration': 11643,
'uploader': 'Riot Games',
'uploader_id': 'riotgames',
'timestamp': 1590770569,
'upload_date': '20200529',
'chapters': [
{
'start_time': 0,
'end_time': 573,
'title': 'League of Legends'
},
{
'start_time': 573,
'end_time': 3922,
'title': 'Legends of Runeterra'
},
{
'start_time': 3922,
'end_time': 11643,
'title': 'Art'
}
],
'live_status': 'was_live',
'thumbnail': r're:^https?://.*\.jpg$',
'view_count': int,
'columns': int,
'rows': int,
},
'params': {
'format': 'mhtml',
'skip_download': True
}
}]
@ -290,16 +337,23 @@ def _download_info(self, item_id):
'includePrivate': False,
'videoID': item_id,
},
}, {
'operationName': 'VideoPlayer_VODSeekbarPreviewVideo',
'variables': {
'includePrivate': False,
'videoID': item_id,
},
}],
'Downloading stream metadata GraphQL')
video = traverse_obj(data, (0, 'data', 'video'))
video['moments'] = traverse_obj(data, (1, 'data', 'video', 'moments', 'edges', ..., 'node'))
video['storyboard'] = traverse_obj(data, (2, 'data', 'video', 'seekPreviewsURL'), expected_type=url_or_none)
if video is None:
raise ExtractorError(
'Video %s does not exist' % item_id, expected=True)
return self._extract_info_gql(video, item_id)
return video
def _extract_info(self, info):
status = info.get('status')
@ -383,10 +437,44 @@ def _extract_info_gql(self, info, item_id):
'was_live': True,
}
def _extract_storyboard(self, item_id, storyboard_json_url, duration):
if not duration or not storyboard_json_url:
return
spec = self._download_json(storyboard_json_url, item_id, 'Downloading storyboard metadata JSON', fatal=False) or []
# sort from highest quality to lowest
# This makes sb0 the highest-quality format, sb1 - lower, etc which is consistent with youtube sb ordering
spec.sort(key=lambda x: int_or_none(x.get('width')) or 0, reverse=True)
base = base_url(storyboard_json_url)
for i, s in enumerate(spec):
count = int_or_none(s.get('count'))
images = s.get('images')
if not (images and count):
continue
fragment_duration = duration / len(images)
yield {
'format_id': f'sb{i}',
'format_note': 'storyboard',
'ext': 'mhtml',
'protocol': 'mhtml',
'acodec': 'none',
'vcodec': 'none',
'url': urljoin(base, images[0]),
'width': int_or_none(s.get('width')),
'height': int_or_none(s.get('height')),
'fps': count / duration,
'rows': int_or_none(s.get('rows')),
'columns': int_or_none(s.get('cols')),
'fragments': [{
'url': urljoin(base, path),
'duration': fragment_duration,
} for path in images],
}
def _real_extract(self, url):
vod_id = self._match_id(url)
info = self._download_info(vod_id)
video = self._download_info(vod_id)
info = self._extract_info_gql(video, vod_id)
access_token = self._download_access_token(vod_id, 'video', 'id')
formats = self._extract_m3u8_formats(
@ -403,6 +491,8 @@ def _real_extract(self, url):
})),
vod_id, 'mp4', entry_protocol='m3u8_native')
formats.extend(self._extract_storyboard(vod_id, video.get('storyboard'), info.get('duration')))
self._prefer_source(formats)
info['formats'] = formats