mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-10-31 23:02:40 +00:00
[ie/niconico] Support DMS formats (#9282)
Closes #8389, Closes #8758, Closes #9254 Authored by: pzhlkj6612, xpadev-net
This commit is contained in:
parent
8f423cf805
commit
aa13a8e3dd
1 changed files with 113 additions and 45 deletions
|
@ -13,13 +13,11 @@
|
||||||
from ..utils import (
|
from ..utils import (
|
||||||
ExtractorError,
|
ExtractorError,
|
||||||
OnDemandPagedList,
|
OnDemandPagedList,
|
||||||
bug_reports_message,
|
|
||||||
clean_html,
|
clean_html,
|
||||||
float_or_none,
|
float_or_none,
|
||||||
int_or_none,
|
int_or_none,
|
||||||
join_nonempty,
|
join_nonempty,
|
||||||
parse_duration,
|
parse_duration,
|
||||||
parse_filesize,
|
|
||||||
parse_iso8601,
|
parse_iso8601,
|
||||||
parse_resolution,
|
parse_resolution,
|
||||||
qualities,
|
qualities,
|
||||||
|
@ -55,25 +53,31 @@ class NiconicoIE(InfoExtractor):
|
||||||
'duration': 33,
|
'duration': 33,
|
||||||
'view_count': int,
|
'view_count': int,
|
||||||
'comment_count': int,
|
'comment_count': int,
|
||||||
|
'genres': ['未設定'],
|
||||||
|
'tags': [],
|
||||||
|
'expected_protocol': str,
|
||||||
},
|
},
|
||||||
'skip': 'Requires an account',
|
|
||||||
}, {
|
}, {
|
||||||
# File downloaded with and without credentials are different, so omit
|
# File downloaded with and without credentials are different, so omit
|
||||||
# the md5 field
|
# the md5 field
|
||||||
'url': 'http://www.nicovideo.jp/watch/nm14296458',
|
'url': 'http://www.nicovideo.jp/watch/nm14296458',
|
||||||
'info_dict': {
|
'info_dict': {
|
||||||
'id': 'nm14296458',
|
'id': 'nm14296458',
|
||||||
'ext': 'swf',
|
'ext': 'mp4',
|
||||||
'title': '【鏡音リン】Dance on media【オリジナル】take2!',
|
'title': '【Kagamine Rin】Dance on media【Original】take2!',
|
||||||
'description': 'md5:689f066d74610b3b22e0f1739add0f58',
|
'description': 'md5:9368f2b1f4178de64f2602c2f3d6cbf5',
|
||||||
'thumbnail': r're:https?://.*',
|
'thumbnail': r're:https?://.*',
|
||||||
'uploader': 'りょうた',
|
'uploader': 'りょうた',
|
||||||
'uploader_id': '18822557',
|
'uploader_id': '18822557',
|
||||||
'upload_date': '20110429',
|
'upload_date': '20110429',
|
||||||
'timestamp': 1304065916,
|
'timestamp': 1304065916,
|
||||||
'duration': 209,
|
'duration': 208.0,
|
||||||
|
'comment_count': int,
|
||||||
|
'view_count': int,
|
||||||
|
'genres': ['音楽・サウンド'],
|
||||||
|
'tags': ['Translation_Request', 'Kagamine_Rin', 'Rin_Original'],
|
||||||
|
'expected_protocol': str,
|
||||||
},
|
},
|
||||||
'skip': 'Requires an account',
|
|
||||||
}, {
|
}, {
|
||||||
# 'video exists but is marked as "deleted"
|
# 'video exists but is marked as "deleted"
|
||||||
# md5 is unstable
|
# md5 is unstable
|
||||||
|
@ -107,22 +111,24 @@ class NiconicoIE(InfoExtractor):
|
||||||
}, {
|
}, {
|
||||||
# video not available via `getflv`; "old" HTML5 video
|
# video not available via `getflv`; "old" HTML5 video
|
||||||
'url': 'http://www.nicovideo.jp/watch/sm1151009',
|
'url': 'http://www.nicovideo.jp/watch/sm1151009',
|
||||||
'md5': '8fa81c364eb619d4085354eab075598a',
|
'md5': 'f95a3d259172667b293530cc2e41ebda',
|
||||||
'info_dict': {
|
'info_dict': {
|
||||||
'id': 'sm1151009',
|
'id': 'sm1151009',
|
||||||
'ext': 'mp4',
|
'ext': 'mp4',
|
||||||
'title': 'マスターシステム本体内蔵のスペハリのメインテーマ(PSG版)',
|
'title': 'マスターシステム本体内蔵のスペハリのメインテーマ(PSG版)',
|
||||||
'description': 'md5:6ee077e0581ff5019773e2e714cdd0b7',
|
'description': 'md5:f95a3d259172667b293530cc2e41ebda',
|
||||||
'thumbnail': r're:https?://.*',
|
'thumbnail': r're:https?://.*',
|
||||||
'duration': 184,
|
'duration': 184,
|
||||||
'timestamp': 1190868283,
|
'timestamp': 1190835883,
|
||||||
'upload_date': '20070927',
|
'upload_date': '20070926',
|
||||||
'uploader': 'denden2',
|
'uploader': 'denden2',
|
||||||
'uploader_id': '1392194',
|
'uploader_id': '1392194',
|
||||||
'view_count': int,
|
'view_count': int,
|
||||||
'comment_count': int,
|
'comment_count': int,
|
||||||
|
'genres': ['ゲーム'],
|
||||||
|
'tags': [],
|
||||||
|
'expected_protocol': str,
|
||||||
},
|
},
|
||||||
'skip': 'Requires an account',
|
|
||||||
}, {
|
}, {
|
||||||
# "New" HTML5 video
|
# "New" HTML5 video
|
||||||
# md5 is unstable
|
# md5 is unstable
|
||||||
|
@ -132,16 +138,18 @@ class NiconicoIE(InfoExtractor):
|
||||||
'ext': 'mp4',
|
'ext': 'mp4',
|
||||||
'title': '新作TVアニメ「戦姫絶唱シンフォギアAXZ」PV 最高画質',
|
'title': '新作TVアニメ「戦姫絶唱シンフォギアAXZ」PV 最高画質',
|
||||||
'description': 'md5:e52974af9a96e739196b2c1ca72b5feb',
|
'description': 'md5:e52974af9a96e739196b2c1ca72b5feb',
|
||||||
'timestamp': 1498514060,
|
'timestamp': 1498481660,
|
||||||
'upload_date': '20170626',
|
'upload_date': '20170626',
|
||||||
'uploader': 'ゲスト',
|
'uploader': 'no-namamae',
|
||||||
'uploader_id': '40826363',
|
'uploader_id': '40826363',
|
||||||
'thumbnail': r're:https?://.*',
|
'thumbnail': r're:https?://.*',
|
||||||
'duration': 198,
|
'duration': 198,
|
||||||
'view_count': int,
|
'view_count': int,
|
||||||
'comment_count': int,
|
'comment_count': int,
|
||||||
|
'genres': ['アニメ'],
|
||||||
|
'tags': [],
|
||||||
|
'expected_protocol': str,
|
||||||
},
|
},
|
||||||
'skip': 'Requires an account',
|
|
||||||
}, {
|
}, {
|
||||||
# Video without owner
|
# Video without owner
|
||||||
'url': 'http://www.nicovideo.jp/watch/sm18238488',
|
'url': 'http://www.nicovideo.jp/watch/sm18238488',
|
||||||
|
@ -151,7 +159,7 @@ class NiconicoIE(InfoExtractor):
|
||||||
'ext': 'mp4',
|
'ext': 'mp4',
|
||||||
'title': '【実写版】ミュータントタートルズ',
|
'title': '【実写版】ミュータントタートルズ',
|
||||||
'description': 'md5:15df8988e47a86f9e978af2064bf6d8e',
|
'description': 'md5:15df8988e47a86f9e978af2064bf6d8e',
|
||||||
'timestamp': 1341160408,
|
'timestamp': 1341128008,
|
||||||
'upload_date': '20120701',
|
'upload_date': '20120701',
|
||||||
'uploader': None,
|
'uploader': None,
|
||||||
'uploader_id': None,
|
'uploader_id': None,
|
||||||
|
@ -159,8 +167,10 @@ class NiconicoIE(InfoExtractor):
|
||||||
'duration': 5271,
|
'duration': 5271,
|
||||||
'view_count': int,
|
'view_count': int,
|
||||||
'comment_count': int,
|
'comment_count': int,
|
||||||
|
'genres': ['エンターテイメント'],
|
||||||
|
'tags': [],
|
||||||
|
'expected_protocol': str,
|
||||||
},
|
},
|
||||||
'skip': 'Requires an account',
|
|
||||||
}, {
|
}, {
|
||||||
'url': 'http://sp.nicovideo.jp/watch/sm28964488?ss_pos=1&cp_in=wt_tg',
|
'url': 'http://sp.nicovideo.jp/watch/sm28964488?ss_pos=1&cp_in=wt_tg',
|
||||||
'only_matching': True,
|
'only_matching': True,
|
||||||
|
@ -353,15 +363,10 @@ def _extract_format_for_quality(self, video_id, audio_quality, video_quality, dm
|
||||||
if not audio_quality.get('isAvailable') or not video_quality.get('isAvailable'):
|
if not audio_quality.get('isAvailable') or not video_quality.get('isAvailable'):
|
||||||
return None
|
return None
|
||||||
|
|
||||||
def extract_video_quality(video_quality):
|
|
||||||
return parse_filesize('%sB' % self._search_regex(
|
|
||||||
r'\| ([0-9]*\.?[0-9]*[MK])', video_quality, 'vbr', default=''))
|
|
||||||
|
|
||||||
format_id = '-'.join(
|
format_id = '-'.join(
|
||||||
[remove_start(s['id'], 'archive_') for s in (video_quality, audio_quality)] + [dmc_protocol])
|
[remove_start(s['id'], 'archive_') for s in (video_quality, audio_quality)] + [dmc_protocol])
|
||||||
|
|
||||||
vid_qual_label = traverse_obj(video_quality, ('metadata', 'label'))
|
vid_qual_label = traverse_obj(video_quality, ('metadata', 'label'))
|
||||||
vid_quality = traverse_obj(video_quality, ('metadata', 'bitrate'))
|
|
||||||
|
|
||||||
return {
|
return {
|
||||||
'url': 'niconico_dmc:%s/%s/%s' % (video_id, video_quality['id'], audio_quality['id']),
|
'url': 'niconico_dmc:%s/%s/%s' % (video_id, video_quality['id'], audio_quality['id']),
|
||||||
|
@ -370,10 +375,15 @@ def extract_video_quality(video_quality):
|
||||||
'ext': 'mp4', # Session API are used in HTML5, which always serves mp4
|
'ext': 'mp4', # Session API are used in HTML5, which always serves mp4
|
||||||
'acodec': 'aac',
|
'acodec': 'aac',
|
||||||
'vcodec': 'h264',
|
'vcodec': 'h264',
|
||||||
'abr': float_or_none(traverse_obj(audio_quality, ('metadata', 'bitrate')), 1000),
|
**traverse_obj(audio_quality, ('metadata', {
|
||||||
'vbr': float_or_none(vid_quality if vid_quality > 0 else extract_video_quality(vid_qual_label), 1000),
|
'abr': ('bitrate', {functools.partial(float_or_none, scale=1000)}),
|
||||||
'height': traverse_obj(video_quality, ('metadata', 'resolution', 'height')),
|
'asr': ('samplingRate', {int_or_none}),
|
||||||
'width': traverse_obj(video_quality, ('metadata', 'resolution', 'width')),
|
})),
|
||||||
|
**traverse_obj(video_quality, ('metadata', {
|
||||||
|
'vbr': ('bitrate', {functools.partial(float_or_none, scale=1000)}),
|
||||||
|
'height': ('resolution', 'height', {int_or_none}),
|
||||||
|
'width': ('resolution', 'width', {int_or_none}),
|
||||||
|
})),
|
||||||
'quality': -2 if 'low' in video_quality['id'] else None,
|
'quality': -2 if 'low' in video_quality['id'] else None,
|
||||||
'protocol': 'niconico_dmc',
|
'protocol': 'niconico_dmc',
|
||||||
'expected_protocol': dmc_protocol, # XXX: This is not a documented field
|
'expected_protocol': dmc_protocol, # XXX: This is not a documented field
|
||||||
|
@ -383,6 +393,63 @@ def extract_video_quality(video_quality):
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
def _yield_dmc_formats(self, api_data, video_id):
|
||||||
|
dmc_data = traverse_obj(api_data, ('media', 'delivery', 'movie'))
|
||||||
|
audios = traverse_obj(dmc_data, ('audios', ..., {dict}))
|
||||||
|
videos = traverse_obj(dmc_data, ('videos', ..., {dict}))
|
||||||
|
protocols = traverse_obj(dmc_data, ('session', 'protocols', ..., {str}))
|
||||||
|
if not all((audios, videos, protocols)):
|
||||||
|
return
|
||||||
|
|
||||||
|
for audio_quality, video_quality, protocol in itertools.product(audios, videos, protocols):
|
||||||
|
if fmt := self._extract_format_for_quality(video_id, audio_quality, video_quality, protocol):
|
||||||
|
yield fmt
|
||||||
|
|
||||||
|
def _yield_dms_formats(self, api_data, video_id):
|
||||||
|
fmt_filter = lambda _, v: v['isAvailable'] and v['id']
|
||||||
|
videos = traverse_obj(api_data, ('media', 'domand', 'videos', fmt_filter))
|
||||||
|
audios = traverse_obj(api_data, ('media', 'domand', 'audios', fmt_filter))
|
||||||
|
access_key = traverse_obj(api_data, ('media', 'domand', 'accessRightKey', {str}))
|
||||||
|
track_id = traverse_obj(api_data, ('client', 'watchTrackId', {str}))
|
||||||
|
if not all((videos, audios, access_key, track_id)):
|
||||||
|
return
|
||||||
|
|
||||||
|
dms_m3u8_url = self._download_json(
|
||||||
|
f'https://nvapi.nicovideo.jp/v1/watch/{video_id}/access-rights/hls', video_id,
|
||||||
|
data=json.dumps({
|
||||||
|
'outputs': list(itertools.product((v['id'] for v in videos), (a['id'] for a in audios)))
|
||||||
|
}).encode(), query={'actionTrackId': track_id}, headers={
|
||||||
|
'x-access-right-key': access_key,
|
||||||
|
'x-frontend-id': 6,
|
||||||
|
'x-frontend-version': 0,
|
||||||
|
'x-request-with': 'https://www.nicovideo.jp',
|
||||||
|
})['data']['contentUrl']
|
||||||
|
# Getting all audio formats results in duplicate video formats which we filter out later
|
||||||
|
dms_fmts = self._extract_m3u8_formats(dms_m3u8_url, video_id)
|
||||||
|
|
||||||
|
# m3u8 extraction does not provide audio bitrates, so extract from the API data and fix
|
||||||
|
for audio_fmt in traverse_obj(dms_fmts, lambda _, v: v['vcodec'] == 'none'):
|
||||||
|
yield {
|
||||||
|
**audio_fmt,
|
||||||
|
**traverse_obj(audios, (lambda _, v: audio_fmt['format_id'].startswith(v['id']), {
|
||||||
|
'format_id': ('id', {str}),
|
||||||
|
'abr': ('bitRate', {functools.partial(float_or_none, scale=1000)}),
|
||||||
|
'asr': ('samplingRate', {int_or_none}),
|
||||||
|
}), get_all=False),
|
||||||
|
'acodec': 'aac',
|
||||||
|
'ext': 'm4a',
|
||||||
|
}
|
||||||
|
|
||||||
|
# Sort before removing dupes to keep the format dicts with the lowest tbr
|
||||||
|
video_fmts = sorted((fmt for fmt in dms_fmts if fmt['vcodec'] != 'none'), key=lambda f: f['tbr'])
|
||||||
|
self._remove_duplicate_formats(video_fmts)
|
||||||
|
# Calculate the true vbr/tbr by subtracting the lowest abr
|
||||||
|
min_abr = min(traverse_obj(audios, (..., 'bitRate', {float_or_none})), default=0) / 1000
|
||||||
|
for video_fmt in video_fmts:
|
||||||
|
video_fmt['tbr'] -= min_abr
|
||||||
|
video_fmt['format_id'] = f'video-{video_fmt["tbr"]:.0f}'
|
||||||
|
yield video_fmt
|
||||||
|
|
||||||
def _real_extract(self, url):
|
def _real_extract(self, url):
|
||||||
video_id = self._match_id(url)
|
video_id = self._match_id(url)
|
||||||
|
|
||||||
|
@ -409,19 +476,17 @@ def _real_extract(self, url):
|
||||||
webpage, 'error reason', default=None)
|
webpage, 'error reason', default=None)
|
||||||
if not error_msg:
|
if not error_msg:
|
||||||
raise
|
raise
|
||||||
raise ExtractorError(re.sub(r'\s+', ' ', error_msg), expected=True)
|
raise ExtractorError(clean_html(error_msg), expected=True)
|
||||||
|
|
||||||
formats = []
|
club_joined = traverse_obj(api_data, ('channel', 'viewer', 'follow', 'isFollowed', {bool}))
|
||||||
|
if club_joined is None:
|
||||||
def get_video_info(*items, get_first=True, **kwargs):
|
fail_msg = self._html_search_regex(
|
||||||
return traverse_obj(api_data, ('video', *items), get_all=not get_first, **kwargs)
|
r'<p[^>]+\bclass="fail-message"[^>]*>(?P<msg>.+?)</p>',
|
||||||
|
webpage, 'fail message', default=None, group='msg')
|
||||||
quality_info = api_data['media']['delivery']['movie']
|
if fail_msg:
|
||||||
session_api_data = quality_info['session']
|
self.raise_login_required(clean_html(fail_msg), metadata_available=True)
|
||||||
for (audio_quality, video_quality, protocol) in itertools.product(quality_info['audios'], quality_info['videos'], session_api_data['protocols']):
|
elif not club_joined:
|
||||||
fmt = self._extract_format_for_quality(video_id, audio_quality, video_quality, protocol)
|
self.raise_login_required('This video is for members only', metadata_available=True)
|
||||||
if fmt:
|
|
||||||
formats.append(fmt)
|
|
||||||
|
|
||||||
# Start extracting information
|
# Start extracting information
|
||||||
tags = None
|
tags = None
|
||||||
|
@ -440,11 +505,15 @@ def get_video_info(*items, get_first=True, **kwargs):
|
||||||
|
|
||||||
thumb_prefs = qualities(['url', 'middleUrl', 'largeUrl', 'player', 'ogp'])
|
thumb_prefs = qualities(['url', 'middleUrl', 'largeUrl', 'player', 'ogp'])
|
||||||
|
|
||||||
|
def get_video_info(*items, get_first=True, **kwargs):
|
||||||
|
return traverse_obj(api_data, ('video', *items), get_all=not get_first, **kwargs)
|
||||||
|
|
||||||
return {
|
return {
|
||||||
'id': video_id,
|
'id': video_id,
|
||||||
'_api_data': api_data,
|
'_api_data': api_data,
|
||||||
'title': get_video_info(('originalTitle', 'title')) or self._og_search_title(webpage, default=None),
|
'title': get_video_info(('originalTitle', 'title')) or self._og_search_title(webpage, default=None),
|
||||||
'formats': formats,
|
'formats': [*self._yield_dmc_formats(api_data, video_id),
|
||||||
|
*self._yield_dms_formats(api_data, video_id)],
|
||||||
'thumbnails': [{
|
'thumbnails': [{
|
||||||
'id': key,
|
'id': key,
|
||||||
'url': url,
|
'url': url,
|
||||||
|
@ -472,8 +541,11 @@ def get_video_info(*items, get_first=True, **kwargs):
|
||||||
|
|
||||||
def _get_subtitles(self, video_id, api_data):
|
def _get_subtitles(self, video_id, api_data):
|
||||||
comments_info = traverse_obj(api_data, ('comment', 'nvComment', {dict})) or {}
|
comments_info = traverse_obj(api_data, ('comment', 'nvComment', {dict})) or {}
|
||||||
|
if not comments_info.get('server'):
|
||||||
|
return
|
||||||
|
|
||||||
danmaku = traverse_obj(self._download_json(
|
danmaku = traverse_obj(self._download_json(
|
||||||
f'{comments_info.get("server")}/v1/threads', video_id, data=json.dumps({
|
f'{comments_info["server"]}/v1/threads', video_id, data=json.dumps({
|
||||||
'additionals': {},
|
'additionals': {},
|
||||||
'params': comments_info.get('params'),
|
'params': comments_info.get('params'),
|
||||||
'threadKey': comments_info.get('threadKey'),
|
'threadKey': comments_info.get('threadKey'),
|
||||||
|
@ -489,10 +561,6 @@ def _get_subtitles(self, video_id, api_data):
|
||||||
note='Downloading comments', errnote='Failed to download comments'),
|
note='Downloading comments', errnote='Failed to download comments'),
|
||||||
('data', 'threads', ..., 'comments', ...))
|
('data', 'threads', ..., 'comments', ...))
|
||||||
|
|
||||||
if not danmaku:
|
|
||||||
self.report_warning(f'Failed to get comments. {bug_reports_message()}')
|
|
||||||
return
|
|
||||||
|
|
||||||
return {
|
return {
|
||||||
'comments': [{
|
'comments': [{
|
||||||
'ext': 'json',
|
'ext': 'json',
|
||||||
|
|
Loading…
Reference in a new issue