[ie/youtube] Extract live adaptive incomplete formats (#15937)

Closes #10148
Authored by: CanOfSocks, bashonly

Co-authored-by: bashonly <88596187+bashonly@users.noreply.github.com>

README.md

@@ -1864,13 +1864,13 @@ The following extractors use this feature:
 * `player_skip`: Skip some network requests that are generally needed for robust extraction. One or more of `configs` (skip client configs), `webpage` (skip initial webpage), `js` (skip js player), `initial_data` (skip initial data/next ep request). While these options can help reduce the number of requests needed or avoid some rate-limiting, they could cause issues such as missing formats or metadata.  See [#860](https://github.com/yt-dlp/yt-dlp/pull/860) and [#12826](https://github.com/yt-dlp/yt-dlp/issues/12826) for more details
 * `webpage_skip`: Skip extraction of embedded webpage data. One or both of `player_response`, `initial_data`. These options are for testing purposes and don't skip any network requests
 * `player_params`: YouTube player parameters to use for player requests. Will overwrite any default ones set by yt-dlp.
-* `player_js_variant`: The player javascript variant to use for n/sig deciphering. The known variants are: `main`, `tcc`, `tce`, `es5`, `es6`, `tv`, `tv_es6`, `phone`. The default is `tv`, and the others are for debugging purposes. You can use `actual` to go with what is prescribed by the site
+* `player_js_variant`: The player javascript variant to use for n/sig deciphering. The known variants are: `main`, `tcc`, `tce`, `es5`, `es6`, `es6_tcc`, `es6_tce`, `tv`, `tv_es6`, `phone`, `house`. The default is `tv`, and the others are for debugging purposes. You can use `actual` to go with what is prescribed by the site
 * `player_js_version`: The player javascript version to use for n/sig deciphering, in the format of `signature_timestamp@hash` (e.g. `20348@0004de42`). The default is to use what is prescribed by the site, and can be selected with `actual`
 * `comment_sort`: `top` or `new` (default) - choose comment sorting mode (on YouTube's side)
 * `max_comments`: Limit the amount of comments to gather. Comma-separated list of integers representing `max-comments,max-parents,max-replies,max-replies-per-thread,max-depth`. Default is `all,all,all,all,all`
     * A `max-depth` value of `1` will discard all replies, regardless of the `max-replies` or `max-replies-per-thread` values given
     * E.g. `all,all,1000,10,2` will get a maximum of 1000 replies total, with up to 10 replies per thread, and only 2 levels of depth (i.e. top-level comments plus their immediate replies). `1000,all,100` will get a maximum of 1000 comments, with a maximum of 100 replies total
-* `formats`: Change the types of formats to return. `dashy` (convert HTTP to DASH), `duplicate` (identical content but different URLs or protocol; includes `dashy`), `incomplete` (cannot be downloaded completely - live dash and post-live m3u8), `missing_pot` (include formats that require a PO Token but are missing one)
+* `formats`: Change the types of formats to return. `dashy` (convert HTTP to DASH), `duplicate` (identical content but different URLs or protocol; includes `dashy`), `incomplete` (cannot be downloaded completely - live dash, live adaptive https, and post-live m3u8), `missing_pot` (include formats that require a PO Token but are missing one)
 * `innertube_host`: Innertube API host to use for all API requests; e.g. `studio.youtube.com`, `youtubei.googleapis.com`. Note that cookies exported from one subdomain will not work on others
 * `innertube_key`: Innertube API key to use for all API requests. By default, no API key is used
 * `raise_incomplete_data`: `Incomplete Data Received` raises an error instead of reporting a warning

test/test_jsc/test_ejs_integration.py

@@ -33,9 +33,12 @@ class Variant(enum.Enum):
     tce = 'player_ias_tce.vflset/en_US/base.js'
     es5 = 'player_es5.vflset/en_US/base.js'
     es6 = 'player_es6.vflset/en_US/base.js'
+    es6_tcc = 'player_es6_tcc.vflset/en_US/base.js'
+    es6_tce = 'player_es6_tce.vflset/en_US/base.js'
     tv = 'tv-player-ias.vflset/tv-player-ias.js'
     tv_es6 = 'tv-player-es6.vflset/tv-player-es6.js'
     phone = 'player-plasma-ias-phone-en_US.vflset/base.js'
+    house = 'house_brand_player.vflset/en_US/base.js'
 
 
 @dataclasses.dataclass

yt_dlp/extractor/odnoklassniki.py

@@ -9,13 +9,13 @@ from ..utils import (
     int_or_none,
     qualities,
     smuggle_url,
-    traverse_obj,
     unescapeHTML,
     unified_strdate,
     unsmuggle_url,
     url_or_none,
     urlencode_postdata,
 )
+from ..utils.traversal import find_element, traverse_obj
 
 
 class OdnoklassnikiIE(InfoExtractor):
@@ -264,9 +264,7 @@ class OdnoklassnikiIE(InfoExtractor):
             note='Downloading desktop webpage',
             headers={'Referer': smuggled['referrer']} if smuggled.get('referrer') else {})
 
-        error = self._search_regex(
-            r'[^>]+class="vp_video_stub_txt"[^>]*>([^<]+)<',
-            webpage, 'error', default=None)
+        error = traverse_obj(webpage, {find_element(cls='vp_video_stub_txt')})
         # Direct link from boosty
         if (error == 'The author of this video has not been found or is blocked'
                 and not smuggled.get('referrer') and mode == 'videoembed'):

yt_dlp/extractor/twitter.py

@@ -131,11 +131,16 @@ class TwitterBaseIE(InfoExtractor):
             video_id, headers=headers, query=query, expected_status=allowed_status,
             note=f'Downloading {"GraphQL" if graphql else "legacy API"} JSON')
 
-        if result.get('errors'):
-            errors = ', '.join(set(traverse_obj(result, ('errors', ..., 'message', {str}))))
-            if errors and 'not authorized' in errors:
-                self.raise_login_required(remove_end(errors, '.'))
-            raise ExtractorError(f'Error(s) while querying API: {errors or "Unknown error"}')
+        if errors := traverse_obj(result, ('errors', ..., {dict})):
+            error_msg = ', '.join(set(traverse_obj(errors, (..., 'message', {str}))))
+            # An error with the message 'Dependency: Unspecified' is a false positive
+            # See https://github.com/yt-dlp/yt-dlp/issues/15963
+            if len(errors) == 1 and 'dependency: unspecified' in error_msg.lower():
+                self.write_debug(f'Ignoring error message: "{error_msg}"')
+            elif 'not authorized' in error_msg.lower():
+                self.raise_login_required(remove_end(error_msg, '.'))
+            else:
+                raise ExtractorError(f'Error(s) while querying API: {error_msg or "Unknown error"}')
 
         return result
 
@@ -1078,7 +1083,7 @@ class TwitterIE(TwitterBaseIE):
             raise ExtractorError(f'Twitter API says: {cause or "Unknown error"}', expected=True)
         elif typename == 'TweetUnavailable':
             reason = result.get('reason')
-            if reason == 'NsfwLoggedOut':
+            if reason in ('NsfwLoggedOut', 'NsfwViewerHasNoStatedAge'):
                 self.raise_login_required('NSFW tweet requires authentication')
             elif reason == 'Protected':
                 self.raise_login_required('You are not authorized to view this protected tweet')

yt_dlp/extractor/youtube/_video.py

@@ -139,11 +139,6 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
     ]
     _RETURN_TYPE = 'video'  # XXX: How to handle multifeed?
 
-    _PLAYER_INFO_RE = (
-        r'/s/player/(?P<id>[a-zA-Z0-9_-]{8,})/(?:tv-)?player',
-        r'/(?P<id>[a-zA-Z0-9_-]{8,})/player(?:_ias\.vflset(?:/[a-zA-Z]{2,3}_[a-zA-Z]{2,3})?|-plasma-ias-(?:phone|tablet)-[a-z]{2}_[A-Z]{2}\.vflset)/base\.js$',
-        r'\b(?P<id>vfl[a-zA-Z0-9_-]+)\b.*?\.js$',
-    )
     _SUBTITLE_FORMATS = ('json3', 'srv1', 'srv2', 'srv3', 'ttml', 'srt', 'vtt')
     _DEFAULT_CLIENTS = ('android_vr', 'web', 'web_safari')
     _DEFAULT_JSLESS_CLIENTS = ('android_vr',)
@@ -1886,10 +1881,12 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
         'tce': 'player_ias_tce.vflset/en_US/base.js',
         'es5': 'player_es5.vflset/en_US/base.js',
         'es6': 'player_es6.vflset/en_US/base.js',
+        'es6_tcc': 'player_es6_tcc.vflset/en_US/base.js',
+        'es6_tce': 'player_es6_tce.vflset/en_US/base.js',
         'tv': 'tv-player-ias.vflset/tv-player-ias.js',
         'tv_es6': 'tv-player-es6.vflset/tv-player-es6.js',
         'phone': 'player-plasma-ias-phone-en_US.vflset/base.js',
-        'tablet': 'player-plasma-ias-tablet-en_US.vflset/base.js',  # Dead since 19712d96 (2025.11.06)
+        'house': 'house_brand_player.vflset/en_US/base.js',  # Used by Google Drive
     }
     _INVERSE_PLAYER_JS_VARIANT_MAP = {v: k for k, v in _PLAYER_JS_VARIANT_MAP.items()}
 
@@ -2179,13 +2176,9 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
 
     @classmethod
     def _extract_player_info(cls, player_url):
-        for player_re in cls._PLAYER_INFO_RE:
-            id_m = re.search(player_re, player_url)
-            if id_m:
-                break
-        else:
+        if m := re.search(r'/s/player/(?P<id>[a-fA-F0-9]{8,})/', player_url):
+            return m.group('id')
         raise ExtractorError(f'Cannot identify player {player_url!r}')
-        return id_m.group('id')
 
     def _load_player(self, video_id, player_url, fatal=True):
         player_js_key = self._player_js_cache_key(player_url)
@@ -3219,6 +3212,7 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
         ])
         skip_player_js = 'js' in self._configuration_arg('player_skip')
         format_types = self._configuration_arg('formats')
+        skip_bad_formats = 'incomplete' not in format_types
         all_formats = 'duplicate' in format_types
         if self._configuration_arg('include_duplicate_formats'):
             all_formats = True
@@ -3464,7 +3458,8 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
                 https_fmts = []
 
                 for fmt_stream in streaming_formats:
-                    if fmt_stream.get('targetDurationSec'):
+                    # Live adaptive https formats are not supported: skip unless extractor-arg given
+                    if fmt_stream.get('targetDurationSec') and skip_bad_formats:
                         continue
 
                     # FORMAT_STREAM_TYPE_OTF(otf=1) requires downloading the init fragment
@@ -3576,7 +3571,6 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
             yield from process_https_formats()
 
             needs_live_processing = self._needs_live_processing(live_status, duration)
-            skip_bad_formats = 'incomplete' not in format_types
 
             skip_manifests = set(self._configuration_arg('skip'))
             if (needs_live_processing == 'is_live'  # These will be filtered out by YoutubeDL anyway
@@ -4086,16 +4080,33 @@ class YoutubeIE(YoutubeBaseInfoExtractor):
 
         needs_live_processing = self._needs_live_processing(live_status, duration)
 
-        def is_bad_format(fmt):
-            if needs_live_processing and not fmt.get('is_from_start'):
-                return True
-            elif (live_status == 'is_live' and needs_live_processing != 'is_live'
-                    and fmt.get('protocol') == 'http_dash_segments'):
-                return True
+        def adjust_incomplete_format(fmt, note_suffix='(Last 2 hours)', pref_adjustment=-10):
+            fmt['preference'] = (fmt.get('preference') or -1) + pref_adjustment
+            fmt['format_note'] = join_nonempty(fmt.get('format_note'), note_suffix, delim=' ')
 
-        for fmt in filter(is_bad_format, formats):
-            fmt['preference'] = (fmt.get('preference') or -1) - 10
-            fmt['format_note'] = join_nonempty(fmt.get('format_note'), '(Last 2 hours)', delim=' ')
+        # Adjust preference and format note for incomplete live/post-live formats
+        if live_status in ('is_live', 'post_live'):
+            for fmt in formats:
+                protocol = fmt.get('protocol')
+                # Currently, protocol isn't set for adaptive https formats, but this could change
+                is_adaptive = protocol in (None, 'http', 'https')
+                if live_status == 'post_live' and is_adaptive:
+                    # Post-live adaptive formats cause HttpFD to raise "Did not get any data blocks"
+                    # These formats are *only* useful to external applications, so we can hide them
+                    # Set their preference <= -1000 so that FormatSorter flags them as 'hidden'
+                    adjust_incomplete_format(fmt, note_suffix='(ended)', pref_adjustment=-5000)
+                # Is it live with --live-from-start? Or is it post-live and its duration is >2hrs?
+                elif needs_live_processing:
+                    if not fmt.get('is_from_start'):
+                        # Post-live m3u8 formats for >2hr streams
+                        adjust_incomplete_format(fmt)
+                elif live_status == 'is_live':
+                    if protocol == 'http_dash_segments':
+                        # Live DASH formats without --live-from-start
+                        adjust_incomplete_format(fmt)
+                    elif is_adaptive:
+                        # Incomplete live adaptive https formats
+                        adjust_incomplete_format(fmt, note_suffix='(incomplete)', pref_adjustment=-20)
 
         if needs_live_processing:
             self._prepare_live_from_start_formats(