[InfoExtractor] Support DASH subtitle extraction (yt-dlp back-port)

This commit is contained in:
dirkf 2024-01-27 15:29:25 +00:00
parent dc512e3a8a
commit c58b655a9e
2 changed files with 300 additions and 125 deletions

View File

@ -993,7 +993,8 @@ jwplayer("mediaplayer").setup({"abouttext":"Visit Indie DB","aboutlink":"http:\/
'tbr': 5997.485, 'tbr': 5997.485,
'width': 1920, 'width': 1920,
'height': 1080, 'height': 1080,
}] }],
{},
), ( ), (
# https://github.com/ytdl-org/youtube-dl/pull/14844 # https://github.com/ytdl-org/youtube-dl/pull/14844
'urls_only', 'urls_only',
@ -1076,7 +1077,8 @@ jwplayer("mediaplayer").setup({"abouttext":"Visit Indie DB","aboutlink":"http:\/
'tbr': 4400, 'tbr': 4400,
'width': 1920, 'width': 1920,
'height': 1080, 'height': 1080,
}] }],
{},
), ( ), (
# https://github.com/ytdl-org/youtube-dl/issues/20346 # https://github.com/ytdl-org/youtube-dl/issues/20346
# Media considered unfragmented even though it contains # Media considered unfragmented even though it contains
@ -1122,18 +1124,119 @@ jwplayer("mediaplayer").setup({"abouttext":"Visit Indie DB","aboutlink":"http:\/
'width': 360, 'width': 360,
'height': 360, 'height': 360,
'fps': 30, 'fps': 30,
}] }],
{},
), (
'subtitles',
'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/',
[{
'format_id': 'audio=128001',
'manifest_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'ext': 'm4a',
'tbr': 128.001,
'asr': 48000,
'format_note': 'DASH audio',
'container': 'm4a_dash',
'vcodec': 'none',
'acodec': 'mp4a.40.2',
'url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'fragment_base_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/dash/',
'protocol': 'http_dash_segments',
}, {
'format_id': 'video=100000',
'manifest_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'ext': 'mp4',
'width': 336,
'height': 144,
'tbr': 100,
'format_note': 'DASH video',
'container': 'mp4_dash',
'vcodec': 'avc1.4D401F',
'acodec': 'none',
'url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'fragment_base_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/dash/',
'protocol': 'http_dash_segments',
}, {
'format_id': 'video=326000',
'manifest_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'ext': 'mp4',
'width': 562,
'height': 240,
'tbr': 326,
'format_note': 'DASH video',
'container': 'mp4_dash',
'vcodec': 'avc1.4D401F',
'acodec': 'none',
'url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'fragment_base_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/dash/',
'protocol': 'http_dash_segments',
}, {
'format_id': 'video=698000',
'manifest_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'ext': 'mp4',
'width': 844,
'height': 360,
'tbr': 698,
'format_note': 'DASH video',
'container': 'mp4_dash',
'vcodec': 'avc1.4D401F',
'acodec': 'none',
'url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'fragment_base_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/dash/',
'protocol': 'http_dash_segments',
}, {
'format_id': 'video=1493000',
'manifest_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'ext': 'mp4',
'width': 1126,
'height': 480,
'tbr': 1493,
'format_note': 'DASH video',
'container': 'mp4_dash',
'vcodec': 'avc1.4D401F',
'acodec': 'none',
'url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'fragment_base_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/dash/',
'protocol': 'http_dash_segments',
}, {
'format_id': 'video=4482000',
'manifest_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'ext': 'mp4',
'width': 1688,
'height': 720,
'tbr': 4482,
'format_note': 'DASH video',
'container': 'mp4_dash',
'vcodec': 'avc1.4D401F',
'acodec': 'none',
'url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'fragment_base_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/dash/',
'protocol': 'http_dash_segments',
}],
{
'en': [
{
'ext': 'mp4',
'manifest_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/manifest.mpd',
'fragment_base_url': 'https://sdn-global-streaming-cache-3qsdn.akamaized.net/stream/3144/files/17/07/672975/3144-kZT4LWMQw6Rh7Kpd.ism/dash/',
'protocol': 'http_dash_segments',
}
]
},
) )
] ]
for mpd_file, mpd_url, mpd_base_url, expected_formats in _TEST_CASES: for mpd_file, mpd_url, mpd_base_url, expected_formats, expected_subtitles in _TEST_CASES:
with open('./test/testdata/mpd/%s.mpd' % mpd_file, with open('./test/testdata/mpd/%s.mpd' % mpd_file,
mode='r', encoding='utf-8') as f: mode='r', encoding='utf-8') as f:
formats = self.ie._parse_mpd_formats( formats, subtitles = self.ie._parse_mpd_formats_and_subtitles(
compat_etree_fromstring(f.read().encode('utf-8')), compat_etree_fromstring(f.read().encode('utf-8')),
mpd_base_url=mpd_base_url, mpd_url=mpd_url) mpd_base_url=mpd_base_url, mpd_url=mpd_url)
self.ie._sort_formats(formats) self.ie._sort_formats(formats)
expect_value(self, formats, expected_formats, None) expect_value(self, formats, expected_formats, None)
expect_value(self, subtitles, expected_subtitles, None)
def test_parse_f4m_formats(self): def test_parse_f4m_formats(self):
_TEST_CASES = [ _TEST_CASES = [

View File

@ -2,6 +2,7 @@
from __future__ import unicode_literals from __future__ import unicode_literals
import base64 import base64
import collections
import datetime import datetime
import functools import functools
import hashlib import hashlib
@ -58,6 +59,7 @@ from ..utils import (
GeoRestrictedError, GeoRestrictedError,
GeoUtils, GeoUtils,
int_or_none, int_or_none,
join_nonempty,
js_to_json, js_to_json,
JSON_LD_RE, JSON_LD_RE,
mimetype2ext, mimetype2ext,
@ -74,6 +76,7 @@ from ..utils import (
str_or_none, str_or_none,
str_to_int, str_to_int,
strip_or_none, strip_or_none,
T,
traverse_obj, traverse_obj,
try_get, try_get,
unescapeHTML, unescapeHTML,
@ -1751,6 +1754,12 @@ class InfoExtractor(object):
'format_note': 'Quality selection URL', 'format_note': 'Quality selection URL',
} }
def _report_ignoring_subs(self, name):
self.report_warning(bug_reports_message(
'Ignoring subtitle tracks found in the {0} manifest; '
'if any subtitle tracks are missing,'.format(name)
), only_once=True)
def _extract_m3u8_formats(self, m3u8_url, video_id, ext=None, def _extract_m3u8_formats(self, m3u8_url, video_id, ext=None,
entry_protocol='m3u8', preference=None, entry_protocol='m3u8', preference=None,
m3u8_id=None, note=None, errnote=None, m3u8_id=None, note=None, errnote=None,
@ -2191,23 +2200,46 @@ class InfoExtractor(object):
}) })
return entries return entries
def _extract_mpd_formats(self, mpd_url, video_id, mpd_id=None, note=None, errnote=None, fatal=True, data=None, headers={}, query={}): def _extract_mpd_formats(self, *args, **kwargs):
fmts, subs = self._extract_mpd_formats_and_subtitles(*args, **kwargs)
if subs:
self._report_ignoring_subs('DASH')
return fmts
def _extract_mpd_formats_and_subtitles(
self, mpd_url, video_id, mpd_id=None, note=None, errnote=None,
fatal=True, data=None, headers=None, query=None):
# TODO: or not? param not yet implemented
if self.get_param('ignore_no_formats_error'):
fatal = False
res = self._download_xml_handle( res = self._download_xml_handle(
mpd_url, video_id, mpd_url, video_id,
note=note or 'Downloading MPD manifest', note='Downloading MPD manifest' if note is None else note,
errnote=errnote or 'Failed to download MPD manifest', errnote='Failed to download MPD manifest' if errnote is None else errnote,
fatal=fatal, data=data, headers=headers, query=query) fatal=fatal, data=data, headers=headers or {}, query=query or {})
if res is False: if res is False:
return [] return [], {}
mpd_doc, urlh = res mpd_doc, urlh = res
if mpd_doc is None: if mpd_doc is None:
return [] return [], {}
mpd_base_url = base_url(urlh.geturl())
return self._parse_mpd_formats( # We could have been redirected to a new url when we retrieved our mpd file.
mpd_url = urlh.geturl()
mpd_base_url = base_url(mpd_url)
return self._parse_mpd_formats_and_subtitles(
mpd_doc, mpd_id, mpd_base_url, mpd_url) mpd_doc, mpd_id, mpd_base_url, mpd_url)
def _parse_mpd_formats(self, mpd_doc, mpd_id=None, mpd_base_url='', mpd_url=None): def _parse_mpd_formats(self, *args, **kwargs):
fmts, subs = self._parse_mpd_formats_and_subtitles(*args, **kwargs)
if subs:
self._report_ignoring_subs('DASH')
return fmts
def _parse_mpd_formats_and_subtitles(
self, mpd_doc, mpd_id=None, mpd_base_url='', mpd_url=None):
""" """
Parse formats from MPD manifest. Parse formats from MPD manifest.
References: References:
@ -2215,8 +2247,10 @@ class InfoExtractor(object):
http://standards.iso.org/ittf/PubliclyAvailableStandards/c065274_ISO_IEC_23009-1_2014.zip http://standards.iso.org/ittf/PubliclyAvailableStandards/c065274_ISO_IEC_23009-1_2014.zip
2. https://en.wikipedia.org/wiki/Dynamic_Adaptive_Streaming_over_HTTP 2. https://en.wikipedia.org/wiki/Dynamic_Adaptive_Streaming_over_HTTP
""" """
# TODO: param not yet implemented: default like previous yt-dl logic
if not self.get_param('dynamic_mpd', False):
if mpd_doc.get('type') == 'dynamic': if mpd_doc.get('type') == 'dynamic':
return [] return [], {}
namespace = self._search_regex(r'(?i)^{([^}]+)?}MPD$', mpd_doc.tag, 'namespace', default=None) namespace = self._search_regex(r'(?i)^{([^}]+)?}MPD$', mpd_doc.tag, 'namespace', default=None)
@ -2228,6 +2262,7 @@ class InfoExtractor(object):
def extract_multisegment_info(element, ms_parent_info): def extract_multisegment_info(element, ms_parent_info):
ms_info = ms_parent_info.copy() ms_info = ms_parent_info.copy()
base_url = ms_info.get('base_url')
# As per [1, 5.3.9.2.2] SegmentList and SegmentTemplate share some # As per [1, 5.3.9.2.2] SegmentList and SegmentTemplate share some
# common attributes and elements. We will only extract relevant # common attributes and elements. We will only extract relevant
@ -2285,7 +2320,8 @@ class InfoExtractor(object):
return ms_info return ms_info
mpd_duration = parse_duration(mpd_doc.get('mediaPresentationDuration')) mpd_duration = parse_duration(mpd_doc.get('mediaPresentationDuration'))
formats = [] formats, subtitles = [], {}
stream_numbers = collections.defaultdict(int)
for period in mpd_doc.findall(_add_ns('Period')): for period in mpd_doc.findall(_add_ns('Period')):
period_duration = parse_duration(period.get('duration')) or mpd_duration period_duration = parse_duration(period.get('duration')) or mpd_duration
period_ms_info = extract_multisegment_info(period, { period_ms_info = extract_multisegment_info(period, {
@ -2295,7 +2331,7 @@ class InfoExtractor(object):
for adaptation_set in period.findall(_add_ns('AdaptationSet')): for adaptation_set in period.findall(_add_ns('AdaptationSet')):
if is_drm_protected(adaptation_set): if is_drm_protected(adaptation_set):
continue continue
adaption_set_ms_info = extract_multisegment_info(adaptation_set, period_ms_info) adaptation_set_ms_info = extract_multisegment_info(adaptation_set, period_ms_info)
for representation in adaptation_set.findall(_add_ns('Representation')): for representation in adaptation_set.findall(_add_ns('Representation')):
if is_drm_protected(representation): if is_drm_protected(representation):
continue continue
@ -2303,27 +2339,35 @@ class InfoExtractor(object):
representation_attrib.update(representation.attrib) representation_attrib.update(representation.attrib)
# According to [1, 5.3.7.2, Table 9, page 41], @mimeType is mandatory # According to [1, 5.3.7.2, Table 9, page 41], @mimeType is mandatory
mime_type = representation_attrib['mimeType'] mime_type = representation_attrib['mimeType']
content_type = mime_type.split('/')[0] content_type = representation_attrib.get('contentType') or mime_type.split('/')[0]
if content_type == 'text': codec_str = representation_attrib.get('codecs', '')
# TODO implement WebVTT downloading # Some kind of binary subtitle found in some youtube livestreams
pass if mime_type == 'application/x-rawcc':
elif content_type in ('video', 'audio'): codecs = {'scodec': codec_str}
base_url = '' else:
for element in (representation, adaptation_set, period, mpd_doc): codecs = parse_codecs(codec_str)
base_url_e = element.find(_add_ns('BaseURL')) if content_type not in ('video', 'audio', 'text'):
if base_url_e is not None: if mime_type == 'image/jpeg':
base_url = base_url_e.text + base_url content_type = mime_type
if re.match(r'^https?://', base_url): elif codecs.get('vcodec', 'none') != 'none':
break content_type = 'video'
if mpd_base_url and not re.match(r'^https?://', base_url): elif codecs.get('acodec', 'none') != 'none':
if not mpd_base_url.endswith('/') and not base_url.startswith('/'): content_type = 'audio'
mpd_base_url += '/' elif codecs.get('scodec', 'none') != 'none':
base_url = mpd_base_url + base_url content_type = 'text'
elif mimetype2ext(mime_type) in ('tt', 'dfxp', 'ttml', 'xml', 'json'):
content_type = 'text'
else:
self.report_warning('Unknown MIME type %s in DASH manifest' % mime_type)
continue
representation_id = representation_attrib.get('id') representation_id = representation_attrib.get('id')
lang = representation_attrib.get('lang') lang = representation_attrib.get('lang')
url_el = representation.find(_add_ns('BaseURL')) url_el = representation.find(_add_ns('BaseURL'))
filesize = int_or_none(url_el.attrib.get('{http://youtube.com/yt/2012/10/10}contentLength') if url_el is not None else None) filesize = int_or_none(url_el.get('{http://youtube.com/yt/2012/10/10}contentLength') if url_el is not None else None)
bandwidth = int_or_none(representation_attrib.get('bandwidth')) bandwidth = int_or_none(representation_attrib.get('bandwidth'))
format_id = join_nonempty(representation_id or content_type, mpd_id)
if content_type in ('video', 'audio'):
f = { f = {
'format_id': '%s-%s' % (mpd_id, representation_id) if mpd_id else representation_id, 'format_id': '%s-%s' % (mpd_id, representation_id) if mpd_id else representation_id,
'manifest_url': mpd_url, 'manifest_url': mpd_url,
@ -2338,8 +2382,27 @@ class InfoExtractor(object):
'filesize': filesize, 'filesize': filesize,
'container': mimetype2ext(mime_type) + '_dash', 'container': mimetype2ext(mime_type) + '_dash',
} }
f.update(parse_codecs(representation_attrib.get('codecs'))) f.update(codecs)
representation_ms_info = extract_multisegment_info(representation, adaption_set_ms_info) elif content_type == 'text':
f = {
'ext': mimetype2ext(mime_type),
'manifest_url': mpd_url,
'filesize': filesize,
}
elif content_type == 'image/jpeg':
# See test case in VikiIE
# https://www.viki.com/videos/1175236v-choosing-spouse-by-lottery-episode-1
f = {
'format_id': format_id,
'ext': 'mhtml',
'manifest_url': mpd_url,
'format_note': 'DASH storyboards (jpeg)',
'acodec': 'none',
'vcodec': 'none',
}
if is_drm_protected(adaptation_set) or is_drm_protected(representation):
f['has_drm'] = True
representation_ms_info = extract_multisegment_info(representation, adaptation_set_ms_info)
def prepare_template(template_name, identifiers): def prepare_template(template_name, identifiers):
tmpl = representation_ms_info[template_name] tmpl = representation_ms_info[template_name]
@ -2391,7 +2454,8 @@ class InfoExtractor(object):
segment_duration = None segment_duration = None
if 'total_number' not in representation_ms_info and 'segment_duration' in representation_ms_info: if 'total_number' not in representation_ms_info and 'segment_duration' in representation_ms_info:
segment_duration = float_or_none(representation_ms_info['segment_duration'], representation_ms_info['timescale']) segment_duration = float_or_none(representation_ms_info['segment_duration'], representation_ms_info['timescale'])
representation_ms_info['total_number'] = int(math.ceil(float(period_duration) / segment_duration)) representation_ms_info['total_number'] = int(math.ceil(
float_or_none(period_duration, segment_duration, default=0)))
representation_ms_info['fragments'] = [{ representation_ms_info['fragments'] = [{
media_location_key: media_template % { media_location_key: media_template % {
'Number': segment_number, 'Number': segment_number,
@ -2431,11 +2495,12 @@ class InfoExtractor(object):
add_segment_url() add_segment_url()
segment_number += 1 segment_number += 1
segment_time += segment_d segment_time += segment_d
elif 'segment_urls' in representation_ms_info and 's' in representation_ms_info: elif 'segment_urls' in representation_ms_info:
fragments = []
if 's' in representation_ms_info:
# No media template # No media template
# Example: https://www.youtube.com/watch?v=iXZV5uAYMJI # Example: https://www.youtube.com/watch?v=iXZV5uAYMJI
# or any YouTube dashsegments video # or any YouTube dashsegments video
fragments = []
segment_index = 0 segment_index = 0
timescale = representation_ms_info['timescale'] timescale = representation_ms_info['timescale']
for s in representation_ms_info['s']: for s in representation_ms_info['s']:
@ -2487,8 +2552,15 @@ class InfoExtractor(object):
f['url'] = base_url f['url'] = base_url
formats.append(f) formats.append(f)
else: else:
self.report_warning('Unknown MIME type %s in DASH manifest' % mime_type) # Assuming direct URL to unfragmented media.
return formats f['url'] = representation_ms_info['base_url']
if content_type in ('video', 'audio', 'image/jpeg'):
f['manifest_stream_number'] = stream_numbers[f['url']]
stream_numbers[f['url']] += 1
formats.append(f)
elif content_type == 'text':
subtitles.setdefault(lang or 'und', []).append(f)
return formats, subtitles
def _extract_ism_formats(self, ism_url, video_id, ism_id=None, note=None, errnote=None, fatal=True, data=None, headers={}, query={}): def _extract_ism_formats(self, ism_url, video_id, ism_id=None, note=None, errnote=None, fatal=True, data=None, headers={}, query={}):
res = self._download_xml_handle( res = self._download_xml_handle(