[YouTube] Rework n-sig processing, realigning with yt-dlp

* apply n-sig before chunked fragments, fixes #32692
This commit is contained in:
dirkf 2024-01-15 18:34:21 +00:00
parent 640d39f03a
commit f8b0135850

View File

@ -2,6 +2,7 @@
from __future__ import unicode_literals from __future__ import unicode_literals
import collections
import itertools import itertools
import json import json
import os.path import os.path
@ -23,10 +24,10 @@ from ..compat import (
) )
from ..jsinterp import JSInterpreter from ..jsinterp import JSInterpreter
from ..utils import ( from ..utils import (
ExtractorError,
clean_html, clean_html,
dict_get, dict_get,
error_to_compat_str, error_to_compat_str,
ExtractorError,
float_or_none, float_or_none,
extract_attributes, extract_attributes,
get_element_by_attribute, get_element_by_attribute,
@ -36,6 +37,7 @@ from ..utils import (
LazyList, LazyList,
merge_dicts, merge_dicts,
mimetype2ext, mimetype2ext,
NO_DEFAULT,
parse_codecs, parse_codecs,
parse_duration, parse_duration,
parse_qs, parse_qs,
@ -45,6 +47,7 @@ from ..utils import (
str_or_none, str_or_none,
str_to_int, str_to_int,
traverse_obj, traverse_obj,
try_call,
try_get, try_get,
txt_or_none, txt_or_none,
unescapeHTML, unescapeHTML,
@ -1460,6 +1463,30 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
self._code_cache = {} self._code_cache = {}
self._player_cache = {} self._player_cache = {}
# *ytcfgs, webpage=None
def _extract_player_url(self, *ytcfgs, **kw_webpage):
if ytcfgs and not isinstance(ytcfgs[0], dict):
webpage = kw_webpage.get('webpage') or ytcfgs[0]
if webpage:
player_url = self._search_regex(
r'"(?:PLAYER_JS_URL|jsUrl)"\s*:\s*"([^"]+)"',
webpage or '', 'player URL', fatal=False)
if player_url:
ytcfgs = ytcfgs + ({'PLAYER_JS_URL': player_url},)
return traverse_obj(
ytcfgs, (Ellipsis, 'PLAYER_JS_URL'), (Ellipsis, 'WEB_PLAYER_CONTEXT_CONFIGS', Ellipsis, 'jsUrl'),
get_all=False, expected_type=lambda u: urljoin('https://www.youtube.com', u))
def _download_player_url(self, video_id, fatal=False):
res = self._download_webpage(
'https://www.youtube.com/iframe_api',
note='Downloading iframe API JS', video_id=video_id, fatal=fatal)
player_version = self._search_regex(
r'player\\?/([0-9a-fA-F]{8})\\?/', res or '', 'player version', fatal=fatal,
default=NO_DEFAULT if res else None)
if player_version:
return 'https://www.youtube.com/s/player/{0}/player_ias.vflset/en_US/base.js'.format(player_version)
def _signature_cache_id(self, example_sig): def _signature_cache_id(self, example_sig):
""" Return a string representation of a signature """ """ Return a string representation of a signature """
return '.'.join(compat_str(len(part)) for part in example_sig.split('.')) return '.'.join(compat_str(len(part)) for part in example_sig.split('.'))
@ -1474,46 +1501,49 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
raise ExtractorError('Cannot identify player %r' % player_url) raise ExtractorError('Cannot identify player %r' % player_url)
return id_m.group('id') return id_m.group('id')
def _get_player_code(self, video_id, player_url, player_id=None): def _load_player(self, video_id, player_url, fatal=True, player_id=None):
if not player_id: if not player_id:
player_id = self._extract_player_info(player_url) player_id = self._extract_player_info(player_url)
if player_id not in self._code_cache: if player_id not in self._code_cache:
self._code_cache[player_id] = self._download_webpage( code = self._download_webpage(
player_url, video_id, player_url, video_id, fatal=fatal,
note='Downloading player ' + player_id, note='Downloading player ' + player_id,
errnote='Download of %s failed' % player_url) errnote='Download of %s failed' % player_url)
return self._code_cache[player_id] if code:
self._code_cache[player_id] = code
return self._code_cache[player_id] if fatal else self._code_cache.get(player_id)
def _extract_signature_function(self, video_id, player_url, example_sig): def _extract_signature_function(self, video_id, player_url, example_sig):
player_id = self._extract_player_info(player_url) player_id = self._extract_player_info(player_url)
# Read from filesystem cache # Read from filesystem cache
func_id = 'js_%s_%s' % ( func_id = 'js_{0}_{1}'.format(
player_id, self._signature_cache_id(example_sig)) player_id, self._signature_cache_id(example_sig))
assert os.path.basename(func_id) == func_id assert os.path.basename(func_id) == func_id
cache_spec = self._downloader.cache.load('youtube-sigfuncs', func_id) self.write_debug('Extracting signature function {0}'.format(func_id))
if cache_spec is not None: cache_spec, code = self.cache.load('youtube-sigfuncs', func_id), None
if not cache_spec:
code = self._load_player(video_id, player_url, player_id)
if code:
res = self._parse_sig_js(code)
test_string = ''.join(map(compat_chr, range(len(example_sig))))
cache_spec = [ord(c) for c in res(test_string)]
self.cache.store('youtube-sigfuncs', func_id, cache_spec)
return lambda s: ''.join(s[i] for i in cache_spec) return lambda s: ''.join(s[i] for i in cache_spec)
code = self._get_player_code(video_id, player_url, player_id)
res = self._parse_sig_js(code)
test_string = ''.join(map(compat_chr, range(len(example_sig))))
cache_res = res(test_string)
cache_spec = [ord(c) for c in cache_res]
self._downloader.cache.store('youtube-sigfuncs', func_id, cache_spec)
return res
def _print_sig_code(self, func, example_sig): def _print_sig_code(self, func, example_sig):
if not self.get_param('youtube_print_sig_code'):
return
def gen_sig_code(idxs): def gen_sig_code(idxs):
def _genslice(start, end, step): def _genslice(start, end, step):
starts = '' if start == 0 else str(start) starts = '' if start == 0 else str(start)
ends = (':%d' % (end + step)) if end + step >= 0 else ':' ends = (':%d' % (end + step)) if end + step >= 0 else ':'
steps = '' if step == 1 else (':%d' % step) steps = '' if step == 1 else (':%d' % step)
return 's[%s%s%s]' % (starts, ends, steps) return 's[{0}{1}{2}]'.format(starts, ends, steps)
step = None step = None
# Quelch pyflakes warnings - start will be set when step is set # Quelch pyflakes warnings - start will be set when step is set
@ -1564,143 +1594,137 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
jscode, 'Initial JS player signature function name', group='sig') jscode, 'Initial JS player signature function name', group='sig')
jsi = JSInterpreter(jscode) jsi = JSInterpreter(jscode)
initial_function = jsi.extract_function(funcname) initial_function = jsi.extract_function(funcname)
return lambda s: initial_function([s]) return lambda s: initial_function([s])
def _cached(self, func, *cache_id):
def inner(*args, **kwargs):
if cache_id not in self._player_cache:
try:
self._player_cache[cache_id] = func(*args, **kwargs)
except ExtractorError as e:
self._player_cache[cache_id] = e
except Exception as e:
self._player_cache[cache_id] = ExtractorError(traceback.format_exc(), cause=e)
ret = self._player_cache[cache_id]
if isinstance(ret, Exception):
raise ret
return ret
return inner
def _decrypt_signature(self, s, video_id, player_url): def _decrypt_signature(self, s, video_id, player_url):
"""Turn the encrypted s field into a working signature""" """Turn the encrypted s field into a working signature"""
extract_sig = self._cached(
if player_url is None: self._extract_signature_function, 'sig', player_url, self._signature_cache_id(s))
raise ExtractorError('Cannot decrypt signature without player_url') func = extract_sig(video_id, player_url, s)
try:
player_id = (player_url, self._signature_cache_id(s))
if player_id not in self._player_cache:
func = self._extract_signature_function(
video_id, player_url, s
)
self._player_cache[player_id] = func
func = self._player_cache[player_id]
if self._downloader.params.get('youtube_print_sig_code'):
self._print_sig_code(func, s) self._print_sig_code(func, s)
return func(s) return func(s)
except Exception as e:
tb = traceback.format_exc()
raise ExtractorError(
'Signature extraction failed: ' + tb, cause=e)
def _extract_player_url(self, webpage):
player_url = self._search_regex(
r'"(?:PLAYER_JS_URL|jsUrl)"\s*:\s*"([^"]+)"',
webpage or '', 'player URL', fatal=False)
if not player_url:
return
if player_url.startswith('//'):
player_url = 'https:' + player_url
elif not re.match(r'https?://', player_url):
player_url = compat_urllib_parse.urljoin(
'https://www.youtube.com', player_url)
return player_url
# from yt-dlp # from yt-dlp
# See also: # See also:
# 1. https://github.com/ytdl-org/youtube-dl/issues/29326#issuecomment-894619419 # 1. https://github.com/ytdl-org/youtube-dl/issues/29326#issuecomment-894619419
# 2. https://code.videolan.org/videolan/vlc/-/blob/4fb284e5af69aa9ac2100ccbdd3b88debec9987f/share/lua/playlist/youtube.lua#L116 # 2. https://code.videolan.org/videolan/vlc/-/blob/4fb284e5af69aa9ac2100ccbdd3b88debec9987f/share/lua/playlist/youtube.lua#L116
# 3. https://github.com/ytdl-org/youtube-dl/issues/30097#issuecomment-950157377 # 3. https://github.com/ytdl-org/youtube-dl/issues/30097#issuecomment-950157377
def _extract_n_function_name(self, jscode): def _decrypt_nsig(self, n, video_id, player_url):
target = r'(?P<nfunc>[a-zA-Z_$][\w$]*)(?:\[(?P<idx>\d+)\])?' """Turn the encrypted n field into a working signature"""
nfunc_and_idx = self._search_regex( if player_url is None:
r'\.get\("n"\)\)&&\(b=(%s)\([\w$]+\)' % (target, ), raise ExtractorError('Cannot decrypt nsig without player_url')
jscode, 'Initial JS player n function name')
nfunc, idx = re.match(target, nfunc_and_idx).group('nfunc', 'idx')
if not idx:
return nfunc
VAR_RE_TMPL = r'var\s+%s\s*=\s*(?P<name>\[(?P<alias>%s)\])[;,]'
note = 'Initial JS player n function {0} (%s[%s])' % (nfunc, idx)
def search_function_code(needle, group):
return self._search_regex(
VAR_RE_TMPL % (re.escape(nfunc), needle), jscode,
note.format(group), group=group)
if int_or_none(idx) == 0:
real_nfunc = search_function_code(r'[a-zA-Z_$][\w$]*', group='alias')
if real_nfunc:
return real_nfunc
return self._parse_json(
search_function_code('.+?', group='name'),
nfunc, transform_source=js_to_json)[int(idx)]
def _extract_n_function(self, video_id, player_url):
player_id = self._extract_player_info(player_url)
func_code = self._downloader.cache.load('youtube-nsig', player_id)
if func_code:
jsi = JSInterpreter(func_code)
else:
jscode = self._get_player_code(video_id, player_url, player_id)
funcname = self._extract_n_function_name(jscode)
jsi = JSInterpreter(jscode)
func_code = jsi.extract_function_code(funcname)
self._downloader.cache.store('youtube-nsig', player_id, func_code)
if self._downloader.params.get('youtube_print_sig_code'):
self.to_screen('Extracted nsig function from {0}:\n{1}\n'.format(player_id, func_code[1]))
return lambda s: jsi.extract_function_from_code(*func_code)([s])
def _n_descramble(self, n_param, player_url, video_id):
"""Compute the response to YT's "n" parameter challenge,
or None
Args:
n_param -- challenge string that is the value of the
URL's "n" query parameter
player_url -- URL of YT player JS
video_id
"""
sig_id = ('nsig_value', n_param)
if sig_id in self._player_cache:
return self._player_cache[sig_id]
try: try:
player_id = ('nsig', player_url) jsi, player_id, func_code = self._extract_n_function_code(video_id, player_url)
if player_id not in self._player_cache: except ExtractorError as e:
self._player_cache[player_id] = self._extract_n_function(video_id, player_url) raise ExtractorError('Unable to extract nsig jsi, player_id, func_codefunction code', cause=e)
func = self._player_cache[player_id] if self.get_param('youtube_print_sig_code'):
ret = func(n_param) self.to_screen('Extracted nsig function from {0}:\n{1}\n'.format(
if ret.startswith('enhanced_except_'): player_id, func_code[1]))
raise ExtractorError('Unhandled exception in decode')
self._player_cache[sig_id] = ret try:
if self._downloader.params.get('verbose', False): extract_nsig = self._cached(self._extract_n_function_from_code, 'nsig func', player_url)
self._downloader.to_screen('[debug] [%s] %s' % (self.IE_NAME, 'Decrypted nsig {0} => {1}'.format(n_param, self._player_cache[sig_id]))) ret = extract_nsig(jsi, func_code)(n)
return self._player_cache[sig_id] except JSInterpreter.Exception as e:
except Exception as e: self.report_warning(
self._downloader.report_warning( '%s (%s %s)' % (
'[%s] %s (%s %s)' % ( self.__ie_msg(
self.IE_NAME, 'Unable to decode n-parameter: download likely to be throttled'),
'Unable to decode n-parameter: download likely to be throttled',
error_to_compat_str(e), error_to_compat_str(e),
traceback.format_exc())) traceback.format_exc()))
return
self.write_debug('Decrypted nsig {0} => {1}'.format(n, ret))
return ret
def _extract_n_function_name(self, jscode):
func_name, idx = self._search_regex(
r'\.get\("n"\)\)&&\(b=(?P<nfunc>[a-zA-Z_$][\w$]*)(?:\[(?P<idx>\d+)\])?\([\w$]+\)',
jscode, 'Initial JS player n function name', group=('nfunc', 'idx'))
if not idx:
return func_name
return self._parse_json(self._search_regex(
r'var {0}\s*=\s*(\[.+?\])\s*[,;]'.format(re.escape(func_name)), jscode,
'Initial JS player n function list ({0}.{1})'.format(func_name, idx)),
func_name, transform_source=js_to_json)[int(idx)]
def _extract_n_function_code(self, video_id, player_url):
player_id = self._extract_player_info(player_url)
func_code = self.cache.load('youtube-nsig', player_id)
jscode = func_code or self._load_player(video_id, player_url)
jsi = JSInterpreter(jscode)
if func_code:
return jsi, player_id, func_code
func_name = self._extract_n_function_name(jscode)
# For redundancy
func_code = self._search_regex(
r'''(?xs)%s\s*=\s*function\s*\((?P<var>[\w$]+)\)\s*
# NB: The end of the regex is intentionally kept strict
{(?P<code>.+?}\s*return\ [\w$]+.join\(""\))};''' % func_name,
jscode, 'nsig function', group=('var', 'code'), default=None)
if func_code:
func_code = ([func_code[0]], func_code[1])
else:
self.write_debug('Extracting nsig function with jsinterp')
func_code = jsi.extract_function_code(func_name)
self.cache.store('youtube-nsig', player_id, func_code)
return jsi, player_id, func_code
def _extract_n_function_from_code(self, jsi, func_code):
func = jsi.extract_function_from_code(*func_code)
def extract_nsig(s):
try:
ret = func([s])
except JSInterpreter.Exception:
raise
except Exception as e:
raise JSInterpreter.Exception(traceback.format_exc(), cause=e)
if ret.startswith('enhanced_except_'):
raise JSInterpreter.Exception('Signature function returned an exception')
return ret
return extract_nsig
def _unthrottle_format_urls(self, video_id, player_url, *formats):
def decrypt_nsig(n):
return self._cached(self._decrypt_nsig, 'nsig', n, player_url)
def _unthrottle_format_urls(self, video_id, player_url, formats):
for fmt in formats: for fmt in formats:
parsed_fmt_url = compat_urllib_parse.urlparse(fmt['url']) parsed_fmt_url = compat_urllib_parse.urlparse(fmt['url'])
n_param = compat_parse_qs(parsed_fmt_url.query).get('n') n_param = compat_parse_qs(parsed_fmt_url.query).get('n')
if not n_param: if not n_param:
continue continue
n_param = n_param[-1] n_param = n_param[-1]
n_response = self._n_descramble(n_param, player_url, video_id) n_response = decrypt_nsig(n_param)(n_param, video_id, player_url)
if n_response is None: if n_response is None:
# give up if descrambling failed # give up if descrambling failed
break break
for fmt_dct in traverse_obj(fmt, (None, (None, ('fragments', Ellipsis))), expected_type=dict): fmt['url'] = update_url_query(fmt['url'], {'n': n_response})
fmt_dct['url'] = update_url(
fmt_dct['url'], query_update={'n': [n_response]})
# from yt-dlp, with tweaks # from yt-dlp, with tweaks
def _extract_signature_timestamp(self, video_id, player_url, ytcfg=None, fatal=False): def _extract_signature_timestamp(self, video_id, player_url, ytcfg=None, fatal=False):
@ -1708,16 +1732,16 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
Extract signatureTimestamp (sts) Extract signatureTimestamp (sts)
Required to tell API what sig/player version is in use. Required to tell API what sig/player version is in use.
""" """
sts = int_or_none(ytcfg.get('STS')) if isinstance(ytcfg, dict) else None sts = traverse_obj(ytcfg, 'STS', expected_type=int)
if not sts: if not sts:
# Attempt to extract from player # Attempt to extract from player
if player_url is None: if player_url is None:
error_msg = 'Cannot extract signature timestamp without player_url.' error_msg = 'Cannot extract signature timestamp without player_url.'
if fatal: if fatal:
raise ExtractorError(error_msg) raise ExtractorError(error_msg)
self._downloader.report_warning(error_msg) self.report_warning(error_msg)
return return
code = self._get_player_code(video_id, player_url) code = self._load_player(video_id, player_url, fatal=fatal)
sts = int_or_none(self._search_regex( sts = int_or_none(self._search_regex(
r'(?:signatureTimestamp|sts)\s*:\s*(?P<sts>[0-9]{5})', code or '', r'(?:signatureTimestamp|sts)\s*:\s*(?P<sts>[0-9]{5})', code or '',
'JS player signature timestamp', group='sts', fatal=fatal)) 'JS player signature timestamp', group='sts', fatal=fatal))
@ -1733,12 +1757,18 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
# cpn generation algorithm is reverse engineered from base.js. # cpn generation algorithm is reverse engineered from base.js.
# In fact it works even with dummy cpn. # In fact it works even with dummy cpn.
CPN_ALPHABET = 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789-_' CPN_ALPHABET = 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789-_'
cpn = ''.join((CPN_ALPHABET[random.randint(0, 256) & 63] for _ in range(0, 16))) cpn = ''.join(CPN_ALPHABET[random.randint(0, 256) & 63] for _ in range(0, 16))
playback_url = update_url( # more consistent results setting it to right before the end
playback_url, query_update={ qs = parse_qs(playback_url)
'ver': ['2'], video_length = '{0}'.format(float((qs.get('len') or ['1.5'])[0]) - 1)
'cpn': [cpn],
playback_url = update_url_query(
playback_url, {
'ver': '2',
'cpn': cpn,
'cmt': video_length,
'el': 'detailpage', # otherwise defaults to "shorts"
}) })
self._download_webpage( self._download_webpage(
@ -1986,8 +2016,11 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
else: else:
self.to_screen('Downloading just video %s because of --no-playlist' % video_id) self.to_screen('Downloading just video %s because of --no-playlist' % video_id)
if not player_url:
player_url = self._extract_player_url(webpage)
formats = [] formats = []
itags = [] itags = collections.defaultdict(set)
itag_qualities = {} itag_qualities = {}
q = qualities(['tiny', 'small', 'medium', 'large', 'hd720', 'hd1080', 'hd1440', 'hd2160', 'hd2880', 'highres']) q = qualities(['tiny', 'small', 'medium', 'large', 'hd720', 'hd1080', 'hd1440', 'hd2160', 'hd2880', 'highres'])
CHUNK_SIZE = 10 << 20 CHUNK_SIZE = 10 << 20
@ -2003,58 +2036,92 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
}) })
} for range_start in range(0, f['filesize'], CHUNK_SIZE)) } for range_start in range(0, f['filesize'], CHUNK_SIZE))
lower = lambda s: s.lower()
for fmt in streaming_formats: for fmt in streaming_formats:
if fmt.get('targetDurationSec') or fmt.get('drmFamilies'): if fmt.get('targetDurationSec'):
continue continue
itag = str_or_none(fmt.get('itag')) itag = str_or_none(fmt.get('itag'))
quality = fmt.get('quality') audio_track = traverse_obj(fmt, ('audioTrack', T(dict))) or {}
if itag and quality:
quality = traverse_obj(fmt, ((
# The 3gp format (17) in android client has a quality of "small",
# but is actually worse than other formats
T(lambda _: 'tiny' if itag == 17 else None),
('quality', T(lambda q: q if q and q != 'tiny' else None)),
('audioQuality', T(lower)),
'quality'), T(txt_or_none)), get_all=False)
if quality and itag:
itag_qualities[itag] = quality itag_qualities[itag] = quality
# FORMAT_STREAM_TYPE_OTF(otf=1) requires downloading the init fragment # FORMAT_STREAM_TYPE_OTF(otf=1) requires downloading the init fragment
# (adding `&sq=0` to the URL) and parsing emsg box to determine the # (adding `&sq=0` to the URL) and parsing emsg box to determine the
# number of fragment that would subsequently requested with (`&sq=N`) # number of fragments that would subsequently be requested with (`&sq=N`)
if fmt.get('type') == 'FORMAT_STREAM_TYPE_OTF': if fmt.get('type') == 'FORMAT_STREAM_TYPE_OTF':
continue continue
fmt_url = fmt.get('url') fmt_url = fmt.get('url')
if not fmt_url: if not fmt_url:
sc = compat_parse_qs(fmt.get('signatureCipher')) sc = compat_parse_qs(fmt.get('signatureCipher'))
fmt_url = url_or_none(try_get(sc, lambda x: x['url'][0])) fmt_url = traverse_obj(sc, ('url', -1, T(url_or_none)))
encrypted_sig = try_get(sc, lambda x: x['s'][0]) encrypted_sig = traverse_obj(sc, ('s', -1))
if not (sc and fmt_url and encrypted_sig): if not (fmt_url and encrypted_sig):
continue continue
if not player_url: player_url = player_url or self._extract_player_url(webpage)
player_url = self._extract_player_url(webpage)
if not player_url: if not player_url:
continue continue
signature = self._decrypt_signature(sc['s'][0], video_id, player_url) try:
sp = try_get(sc, lambda x: x['sp'][0]) or 'signature' fmt_url = update_url_query(fmt_url, {
fmt_url += '&' + sp + '=' + signature traverse_obj(sc, ('sp', -1)) or 'signature':
[self._decrypt_signature(encrypted_sig, video_id, player_url)],
})
except ExtractorError as e:
self.report_warning('Signature extraction failed: Some formats may be missing',
video_id=video_id, only_once=True)
self.write_debug(error_to_compat_str(e), only_once=True)
continue
if itag: language_preference = (
itags.append(itag) 10 if audio_track.get('audioIsDefault')
tbr = float_or_none( else -10 if 'descriptive' in (traverse_obj(audio_track, ('displayName', T(lower))) or '')
fmt.get('averageBitrate') or fmt.get('bitrate'), 1000) else -1)
name = (
traverse_obj(fmt, ('qualityLabel', T(txt_or_none)))
or quality.replace('audio_quality_', ''))
dct = { dct = {
'asr': int_or_none(fmt.get('audioSampleRate')), 'format_id': join_nonempty(itag, fmt.get('isDrc') and 'drc'),
'filesize': int_or_none(fmt.get('contentLength')),
'format_id': itag,
'format_note': fmt.get('qualityLabel') or quality,
'fps': int_or_none(fmt.get('fps')),
'height': int_or_none(fmt.get('height')),
'quality': q(quality),
'tbr': tbr,
'url': fmt_url, 'url': fmt_url,
'width': fmt.get('width'), # Format 22 is likely to be damaged: see https://github.com/yt-dlp/yt-dlp/issues/3372
'source_preference': ((-5 if itag == '22' else -1)
+ (100 if 'Premium' in name else 0)),
'quality': q(quality),
'language': join_nonempty(audio_track.get('id', '').split('.')[0],
'desc' if language_preference < -1 else '') or None,
'language_preference': language_preference,
# Strictly de-prioritize 3gp formats
'preference': -2 if itag == '17' else None,
} }
mimetype = fmt.get('mimeType') if itag:
if mimetype: itags[itag].add(('https', dct.get('language')))
mobj = re.match( self._unthrottle_format_urls(video_id, player_url, dct)
r'((?:[^/]+)/(?:[^;]+))(?:;\s*codecs="([^"]+)")?', mimetype) dct.update(traverse_obj(fmt, {
if mobj: 'asr': ('audioSampleRate', T(int_or_none)),
dct['ext'] = mimetype2ext(mobj.group(1)) 'filesize': ('contentLength', T(int_or_none)),
dct.update(parse_codecs(mobj.group(2))) 'format_note': ('qualityLabel', T(lambda x: x or quality)),
# for some formats, fps is wrongly returned as 1
'fps': ('fps', T(int_or_none), T(lambda f: f if f > 1 else None)),
'audio_channels': ('audioChannels', T(int_or_none)),
'height': ('height', T(int_or_none)),
'has_drm': ('drmFamilies', T(bool)),
'tbr': (('averageBitrate', 'bitrate'), T(lambda t: float_or_none(t, 1000))),
'width': ('width', T(int_or_none)),
'_duration_ms': ('approxDurationMs', T(int_or_none)),
}, get_all=False))
mime_mobj = re.match(
r'((?:[^/]+)/(?:[^;]+))(?:;\s*codecs="([^"]+)")?', fmt.get('mimeType') or '')
if mime_mobj:
dct['ext'] = mimetype2ext(mime_mobj.group(1))
dct.update(parse_codecs(mime_mobj.group(2)))
single_stream = 'none' in (dct.get(c) for c in ('acodec', 'vcodec')) single_stream = 'none' in (dct.get(c) for c in ('acodec', 'vcodec'))
if single_stream and dct.get('ext'): if single_stream and dct.get('ext'):
dct['container'] = dct['ext'] + '_dash' dct['container'] = dct['ext'] + '_dash'
@ -2069,14 +2136,42 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
formats.append(dct) formats.append(dct)
def process_manifest_format(f, proto, client_name, itag, all_formats=False):
key = (proto, f.get('language'))
if not all_formats and key in itags[itag]:
return False
itags[itag].add(key)
if itag:
f['format_id'] = (
'{0}-{1}'.format(itag, proto)
if all_formats or any(p != proto for p, _ in itags[itag])
else itag)
if f.get('source_preference') is None:
f['source_preference'] = -1
if itag in ('616', '235'):
f['format_note'] = join_nonempty(f.get('format_note'), 'Premium', delim=' ')
f['source_preference'] += 100
f['quality'] = q(traverse_obj(f, (
'format_id', T(lambda s: itag_qualities[s.split('-')[0]])), default=-1))
if try_call(lambda: f['fps'] <= 1):
del f['fps']
if proto == 'hls' and f.get('has_drm'):
f['has_drm'] = 'maybe'
f['source_preference'] -= 5
return True
hls_manifest_url = streaming_data.get('hlsManifestUrl') hls_manifest_url = streaming_data.get('hlsManifestUrl')
if hls_manifest_url: if hls_manifest_url:
for f in self._extract_m3u8_formats( for f in self._extract_m3u8_formats(
hls_manifest_url, video_id, 'mp4', fatal=False): hls_manifest_url, video_id, 'mp4', fatal=False):
itag = self._search_regex( if process_manifest_format(
r'/itag/(\d+)', f['url'], 'itag', default=None) f, 'hls', None, self._search_regex(
if itag: r'/itag/(\d+)', f['url'], 'itag', default=None)):
f['format_id'] = itag
formats.append(f) formats.append(f)
if self._downloader.params.get('youtube_include_dash_manifest', True): if self._downloader.params.get('youtube_include_dash_manifest', True):
@ -2084,18 +2179,20 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
if dash_manifest_url: if dash_manifest_url:
for f in self._extract_mpd_formats( for f in self._extract_mpd_formats(
dash_manifest_url, video_id, fatal=False): dash_manifest_url, video_id, fatal=False):
itag = f['format_id'] if process_manifest_format(
if itag in itags: f, 'dash', None, f['format_id']):
continue f['filesize'] = traverse_obj(f, (
if itag in itag_qualities: ('fragment_base_url', 'url'), T(lambda u: self._search_regex(
f['quality'] = q(itag_qualities[itag]) r'/clen/(\d+)', u, 'file size', default=None)),
filesize = int_or_none(self._search_regex( T(int_or_none)), get_all=False)
r'/clen/(\d+)', f.get('fragment_base_url')
or f['url'], 'file size', default=None))
if filesize:
f['filesize'] = filesize
formats.append(f) formats.append(f)
playable_formats = [f for f in formats if not f.get('has_drm')]
if formats and not playable_formats:
# If there are no formats that definitely don't have DRM, all have DRM
self.report_drm(video_id)
formats[:] = playable_formats
if not formats: if not formats:
if streaming_data.get('licenseInfos'): if streaming_data.get('licenseInfos'):
raise ExtractorError( raise ExtractorError(
@ -2166,6 +2263,17 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
video_details.get('lengthSeconds') video_details.get('lengthSeconds')
or microformat.get('lengthSeconds')) \ or microformat.get('lengthSeconds')) \
or parse_duration(search_meta('duration')) or parse_duration(search_meta('duration'))
for f in formats:
# Some formats may have much smaller duration than others (possibly damaged during encoding)
# but avoid false positives with small duration differences.
# Ref: https://github.com/yt-dlp/yt-dlp/issues/2823
if try_call(lambda x: float(x.pop('_duration_ms')) / duration < 500, args=(f,)):
self.report_warning(
'{0}: Some possibly damaged formats will be deprioritized'.format(video_id), only_once=True)
# Strictly de-prioritize damaged formats
f['preference'] = -10
is_live = video_details.get('isLive') is_live = video_details.get('isLive')
owner_profile_url = self._yt_urljoin(self._extract_author_var( owner_profile_url = self._yt_urljoin(self._extract_author_var(
@ -2174,10 +2282,6 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
uploader = self._extract_author_var( uploader = self._extract_author_var(
webpage, 'name', videodetails=video_details, metadata=microformat) webpage, 'name', videodetails=video_details, metadata=microformat)
if not player_url:
player_url = self._extract_player_url(webpage)
self._unthrottle_format_urls(video_id, player_url, formats)
info = { info = {
'id': video_id, 'id': video_id,
'title': self._live_title(video_title) if is_live else video_title, 'title': self._live_title(video_title) if is_live else video_title,