yt_dlp/postprocessor/ffmpeg.py

   1 import collections
   2 import contextvars
   3 import functools
   4 import itertools
   5 import json
   6 import os
   7 import re
   8 import subprocess
   9 import time
  10
  11 from .common import PostProcessor
  12 from ..compat import imghdr
  13 from ..utils import (
  14     MEDIA_EXTENSIONS,
  15     ISO639Utils,
  16     Popen,
  17     PostProcessingError,
  18     _get_exe_version_output,
  19     deprecation_warning,
  20     detect_exe_version,
  21     determine_ext,
  22     dfxp2srt,
  23     encodeArgument,
  24     encodeFilename,
  25     filter_dict,
  26     float_or_none,
  27     is_outdated_version,
  28     orderedSet,
  29     prepend_extension,
  30     replace_extension,
  31     shell_quote,
  32     traverse_obj,
  33     variadic,
  34     write_json_file,
  35 )
  36
  37 EXT_TO_OUT_FORMATS = {
  38     'aac': 'adts',
  39     'flac': 'flac',
  40     'm4a': 'ipod',
  41     'mka': 'matroska',
  42     'mkv': 'matroska',
  43     'mpg': 'mpeg',
  44     'ogv': 'ogg',
  45     'ts': 'mpegts',
  46     'wma': 'asf',
  47     'wmv': 'asf',
  48     'weba': 'webm',
  49     'vtt': 'webvtt',
  50 }
  51 ACODECS = {
  52     # name: (ext, encoder, opts)
  53     'mp3': ('mp3', 'libmp3lame', ()),
  54     'aac': ('m4a', 'aac', ('-f', 'adts')),
  55     'm4a': ('m4a', 'aac', ('-bsf:a', 'aac_adtstoasc')),
  56     'opus': ('opus', 'libopus', ()),
  57     'vorbis': ('ogg', 'libvorbis', ()),
  58     'flac': ('flac', 'flac', ()),
  59     'alac': ('m4a', None, ('-acodec', 'alac')),
  60     'wav': ('wav', None, ('-f', 'wav')),
  61 }
  62
  63
  64 def create_mapping_re(supported):
  65     return re.compile(r'{0}(?:/{0})*$'.format(r'(?:\s*\w+\s*>)?\s*(?:{})\s*'.format('|'.join(supported))))
  66
  67
  68 def resolve_mapping(source, mapping):
  69     """
  70     Get corresponding item from a mapping string like 'A>B/C>D/E'
  71     @returns    (target, error_message)
  72     """
  73     for pair in mapping.lower().split('/'):
  74         kv = pair.split('>', 1)
  75         if len(kv) == 1 or kv[0].strip() == source:
  76             target = kv[-1].strip()
  77             if target == source:
  78                 return target, f'already is in target format {source}'
  79             return target, None
  80     return None, f'could not find a mapping for {source}'
  81
  82
  83 class FFmpegPostProcessorError(PostProcessingError):
  84     pass
  85
  86
  87 class FFmpegPostProcessor(PostProcessor):
  88     _ffmpeg_location = contextvars.ContextVar('ffmpeg_location', default=None)
  89
  90     def __init__(self, downloader=None):
  91         PostProcessor.__init__(self, downloader)
  92         self._prefer_ffmpeg = self.get_param('prefer_ffmpeg', True)
  93         self._paths = self._determine_executables()
  94
  95     @staticmethod
  96     def get_versions_and_features(downloader=None):
  97         pp = FFmpegPostProcessor(downloader)
  98         return pp._versions, pp._features
  99
 100     @staticmethod
 101     def get_versions(downloader=None):
 102         return FFmpegPostProcessor.get_versions_and_features(downloader)[0]
 103
 104     _ffmpeg_to_avconv = {'ffmpeg': 'avconv', 'ffprobe': 'avprobe'}
 105
 106     def _determine_executables(self):
 107         programs = [*self._ffmpeg_to_avconv.keys(), *self._ffmpeg_to_avconv.values()]
 108
 109         location = self.get_param('ffmpeg_location', self._ffmpeg_location.get())
 110         if location is None:
 111             return {p: p for p in programs}
 112
 113         if not os.path.exists(location):
 114             self.report_warning(
 115                 f'ffmpeg-location {location} does not exist! Continuing without ffmpeg', only_once=True)
 116             return {}
 117         elif os.path.isdir(location):
 118             dirname, basename, filename = location, None, None
 119         else:
 120             filename = os.path.basename(location)
 121             basename = next((p for p in programs if p in filename), 'ffmpeg')
 122             dirname = os.path.dirname(os.path.abspath(location))
 123             if basename in self._ffmpeg_to_avconv:
 124                 self._prefer_ffmpeg = True
 125
 126         paths = {p: os.path.join(dirname, p) for p in programs}
 127         if basename and basename in filename:
 128             for p in programs:
 129                 path = os.path.join(dirname, filename.replace(basename, p))
 130                 if os.path.exists(path):
 131                     paths[p] = path
 132         if basename:
 133             paths[basename] = location
 134         return paths
 135
 136     _version_cache, _features_cache = {None: None}, {}
 137
 138     def _get_ffmpeg_version(self, prog):
 139         path = self._paths.get(prog)
 140         if path in self._version_cache:
 141             return self._version_cache[path], self._features_cache.get(path, {})
 142         out = _get_exe_version_output(path, ['-bsfs'])
 143         ver = detect_exe_version(out) if out else False
 144         if ver:
 145             regexs = [
 146                 r'(?:\d+:)?([0-9.]+)-[0-9]+ubuntu[0-9.]+$',  # Ubuntu, see [1]
 147                 r'n([0-9.]+)$',  # Arch Linux
 148                 # 1. http://www.ducea.com/2006/06/17/ubuntu-package-version-naming-explanation/
 149             ]
 150             for regex in regexs:
 151                 mobj = re.match(regex, ver)
 152                 if mobj:
 153                     ver = mobj.group(1)
 154         self._version_cache[path] = ver
 155         if prog != 'ffmpeg' or not out:
 156             return ver, {}
 157
 158         mobj = re.search(r'(?m)^\s+libavformat\s+(?:[0-9. ]+)\s+/\s+(?P<runtime>[0-9. ]+)', out)
 159         lavf_runtime_version = mobj.group('runtime').replace(' ', '') if mobj else None
 160         self._features_cache[path] = features = {
 161             'fdk': '--enable-libfdk-aac' in out,
 162             'setts': 'setts' in out.splitlines(),
 163             'needs_adtstoasc': is_outdated_version(lavf_runtime_version, '57.56.100', False),
 164         }
 165         return ver, features
 166
 167     @property
 168     def _versions(self):
 169         return filter_dict({self.basename: self._version, self.probe_basename: self._probe_version})
 170
 171     @functools.cached_property
 172     def basename(self):
 173         _ = self._version  # run property
 174         return self.basename
 175
 176     @functools.cached_property
 177     def probe_basename(self):
 178         _ = self._probe_version  # run property
 179         return self.probe_basename
 180
 181     def _get_version(self, kind):
 182         executables = (kind, )
 183         if not self._prefer_ffmpeg:
 184             executables = (kind, self._ffmpeg_to_avconv[kind])
 185         basename, version, features = next(filter(
 186             lambda x: x[1], ((p, *self._get_ffmpeg_version(p)) for p in executables)), (None, None, {}))
 187         if kind == 'ffmpeg':
 188             self.basename, self._features = basename, features
 189         else:
 190             self.probe_basename = basename
 191         if basename == self._ffmpeg_to_avconv[kind]:
 192             self.deprecated_feature(f'Support for {self._ffmpeg_to_avconv[kind]} is deprecated and '
 193                                     f'may be removed in a future version. Use {kind} instead')
 194         return version
 195
 196     @functools.cached_property
 197     def _version(self):
 198         return self._get_version('ffmpeg')
 199
 200     @functools.cached_property
 201     def _probe_version(self):
 202         return self._get_version('ffprobe')
 203
 204     @property
 205     def available(self):
 206         return self.basename is not None
 207
 208     @property
 209     def executable(self):
 210         return self._paths.get(self.basename)
 211
 212     @property
 213     def probe_available(self):
 214         return self.probe_basename is not None
 215
 216     @property
 217     def probe_executable(self):
 218         return self._paths.get(self.probe_basename)
 219
 220     @staticmethod
 221     def stream_copy_opts(copy=True, *, ext=None):
 222         yield from ('-map', '0')
 223         # Don't copy Apple TV chapters track, bin_data
 224         # See https://github.com/yt-dlp/yt-dlp/issues/2, #19042, #19024, https://trac.ffmpeg.org/ticket/6016
 225         yield from ('-dn', '-ignore_unknown')
 226         if copy:
 227             yield from ('-c', 'copy')
 228         if ext in ('mp4', 'mov', 'm4a'):
 229             yield from ('-c:s', 'mov_text')
 230
 231     def check_version(self):
 232         if not self.available:
 233             raise FFmpegPostProcessorError('ffmpeg not found. Please install or provide the path using --ffmpeg-location')
 234
 235         required_version = '10-0' if self.basename == 'avconv' else '1.0'
 236         if is_outdated_version(self._version, required_version):
 237             self.report_warning(f'Your copy of {self.basename} is outdated, update {self.basename} '
 238                                 f'to version {required_version} or newer if you encounter any errors')
 239
 240     def get_audio_codec(self, path):
 241         if not self.probe_available and not self.available:
 242             raise PostProcessingError('ffprobe and ffmpeg not found. Please install or provide the path using --ffmpeg-location')
 243         try:
 244             if self.probe_available:
 245                 cmd = [
 246                     encodeFilename(self.probe_executable, True),
 247                     encodeArgument('-show_streams')]
 248             else:
 249                 cmd = [
 250                     encodeFilename(self.executable, True),
 251                     encodeArgument('-i')]
 252             cmd.append(encodeFilename(self._ffmpeg_filename_argument(path), True))
 253             self.write_debug(f'{self.basename} command line: {shell_quote(cmd)}')
 254             stdout, stderr, returncode = Popen.run(
 255                 cmd, text=True, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
 256             if returncode != (0 if self.probe_available else 1):
 257                 return None
 258         except OSError:
 259             return None
 260         output = stdout if self.probe_available else stderr
 261         if self.probe_available:
 262             audio_codec = None
 263             for line in output.split('\n'):
 264                 if line.startswith('codec_name='):
 265                     audio_codec = line.split('=')[1].strip()
 266                 elif line.strip() == 'codec_type=audio' and audio_codec is not None:
 267                     return audio_codec
 268         else:
 269             # Stream #FILE_INDEX:STREAM_INDEX[STREAM_ID](LANGUAGE): CODEC_TYPE: CODEC_NAME
 270             mobj = re.search(
 271                 r'Stream\s*#\d+:\d+(?:\[0x[0-9a-f]+\])?(?:\([a-z]{3}\))?:\s*Audio:\s*([0-9a-z]+)',
 272                 output)
 273             if mobj:
 274                 return mobj.group(1)
 275         return None
 276
 277     def get_metadata_object(self, path, opts=[]):
 278         if self.probe_basename != 'ffprobe':
 279             if self.probe_available:
 280                 self.report_warning('Only ffprobe is supported for metadata extraction')
 281             raise PostProcessingError('ffprobe not found. Please install or provide the path using --ffmpeg-location')
 282         self.check_version()
 283
 284         cmd = [
 285             encodeFilename(self.probe_executable, True),
 286             encodeArgument('-hide_banner'),
 287             encodeArgument('-show_format'),
 288             encodeArgument('-show_streams'),
 289             encodeArgument('-print_format'),
 290             encodeArgument('json'),
 291         ]
 292
 293         cmd += opts
 294         cmd.append(self._ffmpeg_filename_argument(path))
 295         self.write_debug(f'ffprobe command line: {shell_quote(cmd)}')
 296         stdout, _, _ = Popen.run(cmd, text=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE)
 297         return json.loads(stdout)
 298
 299     def get_stream_number(self, path, keys, value):
 300         streams = self.get_metadata_object(path)['streams']
 301         num = next(
 302             (i for i, stream in enumerate(streams) if traverse_obj(stream, keys, casesense=False) == value),
 303             None)
 304         return num, len(streams)
 305
 306     def _fixup_chapters(self, info):
 307         last_chapter = traverse_obj(info, ('chapters', -1))
 308         if last_chapter and not last_chapter.get('end_time'):
 309             last_chapter['end_time'] = self._get_real_video_duration(info['filepath'])
 310
 311     def _get_real_video_duration(self, filepath, fatal=True):
 312         try:
 313             duration = float_or_none(
 314                 traverse_obj(self.get_metadata_object(filepath), ('format', 'duration')))
 315             if not duration:
 316                 raise PostProcessingError('ffprobe returned empty duration')
 317             return duration
 318         except PostProcessingError as e:
 319             if fatal:
 320                 raise PostProcessingError(f'Unable to determine video duration: {e.msg}')
 321
 322     def _duration_mismatch(self, d1, d2, tolerance=2):
 323         if not d1 or not d2:
 324             return None
 325         # The duration is often only known to nearest second. So there can be <1sec disparity natually.
 326         # Further excuse an additional <1sec difference.
 327         return abs(d1 - d2) > tolerance
 328
 329     def run_ffmpeg_multiple_files(self, input_paths, out_path, opts, **kwargs):
 330         return self.real_run_ffmpeg(
 331             [(path, []) for path in input_paths],
 332             [(out_path, opts)], **kwargs)
 333
 334     def real_run_ffmpeg(self, input_path_opts, output_path_opts, *, expected_retcodes=(0,)):
 335         self.check_version()
 336
 337         oldest_mtime = min(
 338             os.stat(encodeFilename(path)).st_mtime for path, _ in input_path_opts if path)
 339
 340         cmd = [encodeFilename(self.executable, True), encodeArgument('-y')]
 341         # avconv does not have repeat option
 342         if self.basename == 'ffmpeg':
 343             cmd += [encodeArgument('-loglevel'), encodeArgument('repeat+info')]
 344
 345         def make_args(file, args, name, number):
 346             keys = [f'_{name}{number}', f'_{name}']
 347             if name == 'o':
 348                 args += ['-movflags', '+faststart']
 349                 if number == 1:
 350                     keys.append('')
 351             args += self._configuration_args(self.basename, keys)
 352             if name == 'i':
 353                 args.append('-i')
 354             return (
 355                 [encodeArgument(arg) for arg in args]
 356                 + [encodeFilename(self._ffmpeg_filename_argument(file), True)])
 357
 358         for arg_type, path_opts in (('i', input_path_opts), ('o', output_path_opts)):
 359             cmd += itertools.chain.from_iterable(
 360                 make_args(path, list(opts), arg_type, i + 1)
 361                 for i, (path, opts) in enumerate(path_opts) if path)
 362
 363         self.write_debug(f'ffmpeg command line: {shell_quote(cmd)}')
 364         _, stderr, returncode = Popen.run(
 365             cmd, text=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE)
 366         if returncode not in variadic(expected_retcodes):
 367             self.write_debug(stderr)
 368             raise FFmpegPostProcessorError(stderr.strip().splitlines()[-1])
 369         for out_path, _ in output_path_opts:
 370             if out_path:
 371                 self.try_utime(out_path, oldest_mtime, oldest_mtime)
 372         return stderr
 373
 374     def run_ffmpeg(self, path, out_path, opts, **kwargs):
 375         return self.run_ffmpeg_multiple_files([path], out_path, opts, **kwargs)
 376
 377     @staticmethod
 378     def _ffmpeg_filename_argument(fn):
 379         # Always use 'file:' because the filename may contain ':' (ffmpeg
 380         # interprets that as a protocol) or can start with '-' (-- is broken in
 381         # ffmpeg, see https://ffmpeg.org/trac/ffmpeg/ticket/2127 for details)
 382         # Also leave '-' intact in order not to break streaming to stdout.
 383         if fn.startswith(('http://', 'https://')):
 384             return fn
 385         return 'file:' + fn if fn != '-' else fn
 386
 387     @staticmethod
 388     def _quote_for_ffmpeg(string):
 389         # See https://ffmpeg.org/ffmpeg-utils.html#toc-Quoting-and-escaping
 390         # A sequence of '' produces '\'''\'';
 391         # final replace removes the empty '' between \' \'.
 392         string = string.replace("'", r"'\''").replace("'''", "'")
 393         # Handle potential ' at string boundaries.
 394         string = string[1:] if string[0] == "'" else "'" + string
 395         return string[:-1] if string[-1] == "'" else string + "'"
 396
 397     def force_keyframes(self, filename, timestamps):
 398         timestamps = orderedSet(timestamps)
 399         if timestamps[0] == 0:
 400             timestamps = timestamps[1:]
 401         keyframe_file = prepend_extension(filename, 'keyframes.temp')
 402         self.to_screen(f'Re-encoding "{filename}" with appropriate keyframes')
 403         self.run_ffmpeg(filename, keyframe_file, [
 404             *self.stream_copy_opts(False, ext=determine_ext(filename)),
 405             '-force_key_frames', ','.join(f'{t:.6f}' for t in timestamps)])
 406         return keyframe_file
 407
 408     def concat_files(self, in_files, out_file, concat_opts=None):
 409         """
 410         Use concat demuxer to concatenate multiple files having identical streams.
 411
 412         Only inpoint, outpoint, and duration concat options are supported.
 413         See https://ffmpeg.org/ffmpeg-formats.html#concat-1 for details
 414         """
 415         concat_file = f'{out_file}.concat'
 416         self.write_debug(f'Writing concat spec to {concat_file}')
 417         with open(concat_file, 'w', encoding='utf-8') as f:
 418             f.writelines(self._concat_spec(in_files, concat_opts))
 419
 420         out_flags = list(self.stream_copy_opts(ext=determine_ext(out_file)))
 421
 422         self.real_run_ffmpeg(
 423             [(concat_file, ['-hide_banner', '-nostdin', '-f', 'concat', '-safe', '0'])],
 424             [(out_file, out_flags)])
 425         self._delete_downloaded_files(concat_file)
 426
 427     @classmethod
 428     def _concat_spec(cls, in_files, concat_opts=None):
 429         if concat_opts is None:
 430             concat_opts = [{}] * len(in_files)
 431         yield 'ffconcat version 1.0\n'
 432         for file, opts in zip(in_files, concat_opts):
 433             yield f'file {cls._quote_for_ffmpeg(cls._ffmpeg_filename_argument(file))}\n'
 434             # Iterate explicitly to yield the following directives in order, ignoring the rest.
 435             for directive in 'inpoint', 'outpoint', 'duration':
 436                 if directive in opts:
 437                     yield f'{directive} {opts[directive]}\n'
 438
 439
 440 class FFmpegExtractAudioPP(FFmpegPostProcessor):
 441     COMMON_AUDIO_EXTS = (*MEDIA_EXTENSIONS.common_audio, 'wma')
 442     SUPPORTED_EXTS = tuple(ACODECS.keys())
 443     FORMAT_RE = create_mapping_re(('best', *SUPPORTED_EXTS))
 444
 445     def __init__(self, downloader=None, preferredcodec=None, preferredquality=None, nopostoverwrites=False):
 446         FFmpegPostProcessor.__init__(self, downloader)
 447         self.mapping = preferredcodec or 'best'
 448         self._preferredquality = float_or_none(preferredquality)
 449         self._nopostoverwrites = nopostoverwrites
 450
 451     def _quality_args(self, codec):
 452         if self._preferredquality is None:
 453             return []
 454         elif self._preferredquality > 10:
 455             return ['-b:a', f'{self._preferredquality}k']
 456
 457         limits = {
 458             'libmp3lame': (10, 0),
 459             'libvorbis': (0, 10),
 460             # FFmpeg's AAC encoder does not have an upper limit for the value of -q:a.
 461             # Experimentally, with values over 4, bitrate changes were minimal or non-existent
 462             'aac': (0.1, 4),
 463             'libfdk_aac': (1, 5),
 464         }.get(codec)
 465         if not limits:
 466             return []
 467
 468         q = limits[1] + (limits[0] - limits[1]) * (self._preferredquality / 10)
 469         if codec == 'libfdk_aac':
 470             return ['-vbr', f'{int(q)}']
 471         return ['-q:a', f'{q}']
 472
 473     def run_ffmpeg(self, path, out_path, codec, more_opts):
 474         if codec is None:
 475             acodec_opts = []
 476         else:
 477             acodec_opts = ['-acodec', codec]
 478         opts = ['-vn', *acodec_opts, *more_opts]
 479         try:
 480             FFmpegPostProcessor.run_ffmpeg(self, path, out_path, opts)
 481         except FFmpegPostProcessorError as err:
 482             raise PostProcessingError(f'audio conversion failed: {err.msg}')
 483
 484     @PostProcessor._restrict_to(images=False)
 485     def run(self, information):
 486         orig_path = path = information['filepath']
 487         target_format, _skip_msg = resolve_mapping(information['ext'], self.mapping)
 488         if target_format == 'best' and information['ext'] in self.COMMON_AUDIO_EXTS:
 489             target_format, _skip_msg = None, 'the file is already in a common audio format'
 490         if not target_format:
 491             self.to_screen(f'Not converting audio {orig_path}; {_skip_msg}')
 492             return [], information
 493
 494         filecodec = self.get_audio_codec(path)
 495         if filecodec is None:
 496             raise PostProcessingError('WARNING: unable to obtain file audio codec with ffprobe')
 497
 498         if filecodec == 'aac' and target_format in ('m4a', 'best'):
 499             # Lossless, but in another container
 500             extension, _, more_opts, acodec = *ACODECS['m4a'], 'copy'
 501         elif target_format == 'best' or target_format == filecodec:
 502             # Lossless if possible
 503             try:
 504                 extension, _, more_opts, acodec = *ACODECS[filecodec], 'copy'
 505             except KeyError:
 506                 extension, acodec, more_opts = ACODECS['mp3']
 507         else:
 508             # We convert the audio (lossy if codec is lossy)
 509             extension, acodec, more_opts = ACODECS[target_format]
 510             if acodec == 'aac' and self._features.get('fdk'):
 511                 acodec, more_opts = 'libfdk_aac', []
 512
 513         more_opts = list(more_opts)
 514         if acodec != 'copy':
 515             more_opts = self._quality_args(acodec)
 516
 517         temp_path = new_path = replace_extension(path, extension, information['ext'])
 518
 519         if new_path == path:
 520             if acodec == 'copy':
 521                 self.to_screen(f'Not converting audio {orig_path}; file is already in target format {target_format}')
 522                 return [], information
 523             orig_path = prepend_extension(path, 'orig')
 524             temp_path = prepend_extension(path, 'temp')
 525         if (self._nopostoverwrites and os.path.exists(encodeFilename(new_path))
 526                 and os.path.exists(encodeFilename(orig_path))):
 527             self.to_screen(f'Post-process file {new_path} exists, skipping')
 528             return [], information
 529
 530         self.to_screen(f'Destination: {new_path}')
 531         self.run_ffmpeg(path, temp_path, acodec, more_opts)
 532
 533         os.replace(path, orig_path)
 534         os.replace(temp_path, new_path)
 535         information['filepath'] = new_path
 536         information['ext'] = extension
 537
 538         # Try to update the date time for extracted audio file.
 539         if information.get('filetime') is not None:
 540             self.try_utime(
 541                 new_path, time.time(), information['filetime'], errnote='Cannot update utime of audio file')
 542
 543         return [orig_path], information
 544
 545
 546 class FFmpegVideoConvertorPP(FFmpegPostProcessor):
 547     SUPPORTED_EXTS = (
 548         *sorted((*MEDIA_EXTENSIONS.common_video, 'gif')),
 549         *sorted((*MEDIA_EXTENSIONS.common_audio, 'aac', 'vorbis')),
 550     )
 551     FORMAT_RE = create_mapping_re(SUPPORTED_EXTS)
 552     _ACTION = 'converting'
 553
 554     def __init__(self, downloader=None, preferedformat=None):
 555         super().__init__(downloader)
 556         self.mapping = preferedformat
 557
 558     @staticmethod
 559     def _options(target_ext):
 560         yield from FFmpegPostProcessor.stream_copy_opts(False)
 561         if target_ext == 'avi':
 562             yield from ('-c:v', 'libxvid', '-vtag', 'XVID')
 563
 564     @PostProcessor._restrict_to(images=False)
 565     def run(self, info):
 566         filename, source_ext = info['filepath'], info['ext'].lower()
 567         target_ext, _skip_msg = resolve_mapping(source_ext, self.mapping)
 568         if _skip_msg:
 569             self.to_screen(f'Not {self._ACTION} media file "{filename}"; {_skip_msg}')
 570             return [], info
 571
 572         outpath = replace_extension(filename, target_ext, source_ext)
 573         self.to_screen(f'{self._ACTION.title()} video from {source_ext} to {target_ext}; Destination: {outpath}')
 574         self.run_ffmpeg(filename, outpath, self._options(target_ext))
 575
 576         info['filepath'] = outpath
 577         info['format'] = info['ext'] = target_ext
 578         return [filename], info
 579
 580
 581 class FFmpegVideoRemuxerPP(FFmpegVideoConvertorPP):
 582     _ACTION = 'remuxing'
 583
 584     @staticmethod
 585     def _options(target_ext):
 586         return FFmpegPostProcessor.stream_copy_opts()
 587
 588
 589 class FFmpegEmbedSubtitlePP(FFmpegPostProcessor):
 590     SUPPORTED_EXTS = ('mp4', 'mov', 'm4a', 'webm', 'mkv', 'mka')
 591
 592     def __init__(self, downloader=None, already_have_subtitle=False):
 593         super().__init__(downloader)
 594         self._already_have_subtitle = already_have_subtitle
 595
 596     @PostProcessor._restrict_to(images=False)
 597     def run(self, info):
 598         if info['ext'] not in self.SUPPORTED_EXTS:
 599             self.to_screen(f'Subtitles can only be embedded in {", ".join(self.SUPPORTED_EXTS)} files')
 600             return [], info
 601         subtitles = info.get('requested_subtitles')
 602         if not subtitles:
 603             self.to_screen('There aren\'t any subtitles to embed')
 604             return [], info
 605
 606         filename = info['filepath']
 607
 608         # Disabled temporarily. There needs to be a way to override this
 609         # in case of duration actually mismatching in extractor
 610         # See: https://github.com/yt-dlp/yt-dlp/issues/1870, https://github.com/yt-dlp/yt-dlp/issues/1385
 611         '''
 612         if info.get('duration') and not info.get('__real_download') and self._duration_mismatch(
 613                 self._get_real_video_duration(filename, False), info['duration']):
 614             self.to_screen(f'Skipping {self.pp_key()} since the real and expected durations mismatch')
 615             return [], info
 616         '''
 617
 618         ext = info['ext']
 619         sub_langs, sub_names, sub_filenames = [], [], []
 620         webm_vtt_warn = False
 621         mp4_ass_warn = False
 622
 623         for lang, sub_info in subtitles.items():
 624             if not os.path.exists(sub_info.get('filepath', '')):
 625                 self.report_warning(f'Skipping embedding {lang} subtitle because the file is missing')
 626                 continue
 627             sub_ext = sub_info['ext']
 628             if sub_ext == 'json':
 629                 self.report_warning('JSON subtitles cannot be embedded')
 630             elif ext != 'webm' or ext == 'webm' and sub_ext == 'vtt':
 631                 sub_langs.append(lang)
 632                 sub_names.append(sub_info.get('name'))
 633                 sub_filenames.append(sub_info['filepath'])
 634             else:
 635                 if not webm_vtt_warn and ext == 'webm' and sub_ext != 'vtt':
 636                     webm_vtt_warn = True
 637                     self.report_warning('Only WebVTT subtitles can be embedded in webm files')
 638             if not mp4_ass_warn and ext == 'mp4' and sub_ext == 'ass':
 639                 mp4_ass_warn = True
 640                 self.report_warning('ASS subtitles cannot be properly embedded in mp4 files; expect issues')
 641
 642         if not sub_langs:
 643             return [], info
 644
 645         input_files = [filename, *sub_filenames]
 646
 647         opts = [
 648             *self.stream_copy_opts(ext=info['ext']),
 649             # Don't copy the existing subtitles, we may be running the
 650             # postprocessor a second time
 651             '-map', '-0:s',
 652         ]
 653         for i, (lang, name) in enumerate(zip(sub_langs, sub_names)):
 654             opts.extend(['-map', f'{i + 1}:0'])
 655             lang_code = ISO639Utils.short2long(lang) or lang
 656             opts.extend([f'-metadata:s:s:{i}', f'language={lang_code}'])
 657             if name:
 658                 opts.extend([f'-metadata:s:s:{i}', f'handler_name={name}',
 659                              f'-metadata:s:s:{i}', f'title={name}'])
 660
 661         temp_filename = prepend_extension(filename, 'temp')
 662         self.to_screen(f'Embedding subtitles in "{filename}"')
 663         self.run_ffmpeg_multiple_files(input_files, temp_filename, opts)
 664         os.replace(temp_filename, filename)
 665
 666         files_to_delete = [] if self._already_have_subtitle else sub_filenames
 667         return files_to_delete, info
 668
 669
 670 class FFmpegMetadataPP(FFmpegPostProcessor):
 671
 672     def __init__(self, downloader, add_metadata=True, add_chapters=True, add_infojson='if_exists'):
 673         FFmpegPostProcessor.__init__(self, downloader)
 674         self._add_metadata = add_metadata
 675         self._add_chapters = add_chapters
 676         self._add_infojson = add_infojson
 677
 678     @staticmethod
 679     def _options(target_ext):
 680         audio_only = target_ext == 'm4a'
 681         yield from FFmpegPostProcessor.stream_copy_opts(not audio_only)
 682         if audio_only:
 683             yield from ('-vn', '-acodec', 'copy')
 684
 685     @PostProcessor._restrict_to(images=False)
 686     def run(self, info):
 687         self._fixup_chapters(info)
 688         filename, metadata_filename = info['filepath'], None
 689         files_to_delete, options = [], []
 690         if self._add_chapters and info.get('chapters'):
 691             metadata_filename = replace_extension(filename, 'meta')
 692             options.extend(self._get_chapter_opts(info['chapters'], metadata_filename))
 693             files_to_delete.append(metadata_filename)
 694         if self._add_metadata:
 695             options.extend(self._get_metadata_opts(info))
 696
 697         if self._add_infojson:
 698             if info['ext'] in ('mkv', 'mka'):
 699                 infojson_filename = info.get('infojson_filename')
 700                 options.extend(self._get_infojson_opts(info, infojson_filename))
 701                 if not infojson_filename:
 702                     files_to_delete.append(info.get('infojson_filename'))
 703             elif self._add_infojson is True:
 704                 self.to_screen('The info-json can only be attached to mkv/mka files')
 705
 706         if not options:
 707             self.to_screen('There isn\'t any metadata to add')
 708             return [], info
 709
 710         temp_filename = prepend_extension(filename, 'temp')
 711         self.to_screen(f'Adding metadata to "{filename}"')
 712         self.run_ffmpeg_multiple_files(
 713             (filename, metadata_filename), temp_filename,
 714             itertools.chain(self._options(info['ext']), *options))
 715         self._delete_downloaded_files(*files_to_delete)
 716         os.replace(temp_filename, filename)
 717         return [], info
 718
 719     @staticmethod
 720     def _get_chapter_opts(chapters, metadata_filename):
 721         with open(metadata_filename, 'w', encoding='utf-8') as f:
 722             def ffmpeg_escape(text):
 723                 return re.sub(r'([\\=;#\n])', r'\\\1', text)
 724
 725             metadata_file_content = ';FFMETADATA1\n'
 726             for chapter in chapters:
 727                 metadata_file_content += '[CHAPTER]\nTIMEBASE=1/1000\n'
 728                 metadata_file_content += 'START=%d\n' % (chapter['start_time'] * 1000)
 729                 metadata_file_content += 'END=%d\n' % (chapter['end_time'] * 1000)
 730                 chapter_title = chapter.get('title')
 731                 if chapter_title:
 732                     metadata_file_content += f'title={ffmpeg_escape(chapter_title)}\n'
 733             f.write(metadata_file_content)
 734         yield ('-map_metadata', '1')
 735
 736     def _get_metadata_opts(self, info):
 737         meta_prefix = 'meta'
 738         metadata = collections.defaultdict(dict)
 739
 740         def add(meta_list, info_list=None):
 741             value = next((
 742                 info[key] for key in [f'{meta_prefix}_', *variadic(info_list or meta_list)]
 743                 if info.get(key) is not None), None)
 744             if value not in ('', None):
 745                 value = ', '.join(map(str, variadic(value)))
 746                 value = value.replace('\0', '')  # nul character cannot be passed in command line
 747                 metadata['common'].update({meta_f: value for meta_f in variadic(meta_list)})
 748
 749         # Info on media metadata/metadata supported by ffmpeg:
 750         # https://wiki.multimedia.cx/index.php/FFmpeg_Metadata
 751         # https://kdenlive.org/en/project/adding-meta-data-to-mp4-video/
 752         # https://kodi.wiki/view/Video_file_tagging
 753
 754         add('title', ('track', 'title'))
 755         add('date', 'upload_date')
 756         add(('description', 'synopsis'), 'description')
 757         add(('purl', 'comment'), 'webpage_url')
 758         add('track', 'track_number')
 759         add('artist', ('artist', 'artists', 'creator', 'creators', 'uploader', 'uploader_id'))
 760         add('composer', ('composer', 'composers'))
 761         add('genre', ('genre', 'genres'))
 762         add('album')
 763         add('album_artist', ('album_artist', 'album_artists'))
 764         add('disc', 'disc_number')
 765         add('show', 'series')
 766         add('season_number')
 767         add('episode_id', ('episode', 'episode_id'))
 768         add('episode_sort', 'episode_number')
 769         if 'embed-metadata' in self.get_param('compat_opts', []):
 770             add('comment', 'description')
 771             metadata['common'].pop('synopsis', None)
 772
 773         meta_regex = rf'{re.escape(meta_prefix)}(?P<i>\d+)?_(?P<key>.+)'
 774         for key, value in info.items():
 775             mobj = re.fullmatch(meta_regex, key)
 776             if value is not None and mobj:
 777                 metadata[mobj.group('i') or 'common'][mobj.group('key')] = value.replace('\0', '')
 778
 779         # Write id3v1 metadata also since Windows Explorer can't handle id3v2 tags
 780         yield ('-write_id3v1', '1')
 781
 782         for name, value in metadata['common'].items():
 783             yield ('-metadata', f'{name}={value}')
 784
 785         stream_idx = 0
 786         for fmt in info.get('requested_formats') or [info]:
 787             stream_count = 2 if 'none' not in (fmt.get('vcodec'), fmt.get('acodec')) else 1
 788             lang = ISO639Utils.short2long(fmt.get('language') or '') or fmt.get('language')
 789             for i in range(stream_idx, stream_idx + stream_count):
 790                 if lang:
 791                     metadata[str(i)].setdefault('language', lang)
 792                 for name, value in metadata[str(i)].items():
 793                     yield (f'-metadata:s:{i}', f'{name}={value}')
 794             stream_idx += stream_count
 795
 796     def _get_infojson_opts(self, info, infofn):
 797         if not infofn or not os.path.exists(infofn):
 798             if self._add_infojson is not True:
 799                 return
 800             infofn = infofn or '%s.temp' % (
 801                 self._downloader.prepare_filename(info, 'infojson')
 802                 or replace_extension(self._downloader.prepare_filename(info), 'info.json', info['ext']))
 803             if not self._downloader._ensure_dir_exists(infofn):
 804                 return
 805             self.write_debug(f'Writing info-json to: {infofn}')
 806             write_json_file(self._downloader.sanitize_info(info, self.get_param('clean_infojson', True)), infofn)
 807             info['infojson_filename'] = infofn
 808
 809         old_stream, new_stream = self.get_stream_number(info['filepath'], ('tags', 'mimetype'), 'application/json')
 810         if old_stream is not None:
 811             yield ('-map', f'-0:{old_stream}')
 812             new_stream -= 1
 813
 814         yield (
 815             '-attach', self._ffmpeg_filename_argument(infofn),
 816             f'-metadata:s:{new_stream}', 'mimetype=application/json',
 817             f'-metadata:s:{new_stream}', 'filename=info.json',
 818         )
 819
 820
 821 class FFmpegMergerPP(FFmpegPostProcessor):
 822     SUPPORTED_EXTS = MEDIA_EXTENSIONS.common_video
 823
 824     @PostProcessor._restrict_to(images=False)
 825     def run(self, info):
 826         filename = info['filepath']
 827         temp_filename = prepend_extension(filename, 'temp')
 828         args = ['-c', 'copy']
 829         audio_streams = 0
 830         for (i, fmt) in enumerate(info['requested_formats']):
 831             if fmt.get('acodec') != 'none':
 832                 args.extend(['-map', f'{i}:a:0'])
 833                 aac_fixup = fmt['protocol'].startswith('m3u8') and self.get_audio_codec(fmt['filepath']) == 'aac'
 834                 if aac_fixup:
 835                     args.extend([f'-bsf:a:{audio_streams}', 'aac_adtstoasc'])
 836                 audio_streams += 1
 837             if fmt.get('vcodec') != 'none':
 838                 args.extend(['-map', f'{i}:v:0'])
 839         self.to_screen(f'Merging formats into "{filename}"')
 840         self.run_ffmpeg_multiple_files(info['__files_to_merge'], temp_filename, args)
 841         os.rename(encodeFilename(temp_filename), encodeFilename(filename))
 842         return info['__files_to_merge'], info
 843
 844     def can_merge(self):
 845         # TODO: figure out merge-capable ffmpeg version
 846         if self.basename != 'avconv':
 847             return True
 848
 849         required_version = '10-0'
 850         if is_outdated_version(
 851                 self._versions[self.basename], required_version):
 852             warning = (f'Your copy of {self.basename} is outdated and unable to properly mux separate video and audio files, '
 853                        'yt-dlp will download single file media. '
 854                        f'Update {self.basename} to version {required_version} or newer to fix this.')
 855             self.report_warning(warning)
 856             return False
 857         return True
 858
 859
 860 class FFmpegFixupPostProcessor(FFmpegPostProcessor):
 861     def _fixup(self, msg, filename, options):
 862         temp_filename = prepend_extension(filename, 'temp')
 863
 864         self.to_screen(f'{msg} of "{filename}"')
 865         self.run_ffmpeg(filename, temp_filename, options)
 866
 867         os.replace(temp_filename, filename)
 868
 869
 870 class FFmpegFixupStretchedPP(FFmpegFixupPostProcessor):
 871     @PostProcessor._restrict_to(images=False, audio=False)
 872     def run(self, info):
 873         stretched_ratio = info.get('stretched_ratio')
 874         if stretched_ratio not in (None, 1):
 875             self._fixup('Fixing aspect ratio', info['filepath'], [
 876                 *self.stream_copy_opts(), '-aspect', f'{stretched_ratio:f}'])
 877         return [], info
 878
 879
 880 class FFmpegFixupM4aPP(FFmpegFixupPostProcessor):
 881     @PostProcessor._restrict_to(images=False, video=False)
 882     def run(self, info):
 883         if info.get('container') == 'm4a_dash':
 884             self._fixup('Correcting container', info['filepath'], [*self.stream_copy_opts(), '-f', 'mp4'])
 885         return [], info
 886
 887
 888 class FFmpegFixupM3u8PP(FFmpegFixupPostProcessor):
 889     def _needs_fixup(self, info):
 890         yield info['ext'] in ('mp4', 'm4a')
 891         yield info['protocol'].startswith('m3u8')
 892         try:
 893             metadata = self.get_metadata_object(info['filepath'])
 894         except PostProcessingError as e:
 895             self.report_warning(f'Unable to extract metadata: {e.msg}')
 896             yield True
 897         else:
 898             yield traverse_obj(metadata, ('format', 'format_name'), casesense=False) == 'mpegts'
 899
 900     @PostProcessor._restrict_to(images=False)
 901     def run(self, info):
 902         if all(self._needs_fixup(info)):
 903             args = ['-f', 'mp4']
 904             if self.get_audio_codec(info['filepath']) == 'aac':
 905                 args.extend(['-bsf:a', 'aac_adtstoasc'])
 906             self._fixup('Fixing MPEG-TS in MP4 container', info['filepath'], [
 907                 *self.stream_copy_opts(), *args])
 908         return [], info
 909
 910
 911 class FFmpegFixupTimestampPP(FFmpegFixupPostProcessor):
 912
 913     def __init__(self, downloader=None, trim=0.001):
 914         # "trim" should be used when the video contains unintended packets
 915         super().__init__(downloader)
 916         assert isinstance(trim, (int, float))
 917         self.trim = str(trim)
 918
 919     @PostProcessor._restrict_to(images=False)
 920     def run(self, info):
 921         if not self._features.get('setts'):
 922             self.report_warning(
 923                 'A re-encode is needed to fix timestamps in older versions of ffmpeg. '
 924                 'Please install ffmpeg 4.4 or later to fixup without re-encoding')
 925             opts = ['-vf', 'setpts=PTS-STARTPTS']
 926         else:
 927             opts = ['-c', 'copy', '-bsf', 'setts=ts=TS-STARTPTS']
 928         self._fixup('Fixing frame timestamp', info['filepath'], [*opts, *self.stream_copy_opts(False), '-ss', self.trim])
 929         return [], info
 930
 931
 932 class FFmpegCopyStreamPP(FFmpegFixupPostProcessor):
 933     MESSAGE = 'Copying stream'
 934
 935     @PostProcessor._restrict_to(images=False)
 936     def run(self, info):
 937         self._fixup(self.MESSAGE, info['filepath'], self.stream_copy_opts())
 938         return [], info
 939
 940
 941 class FFmpegFixupDurationPP(FFmpegCopyStreamPP):
 942     MESSAGE = 'Fixing video duration'
 943
 944
 945 class FFmpegFixupDuplicateMoovPP(FFmpegCopyStreamPP):
 946     MESSAGE = 'Fixing duplicate MOOV atoms'
 947
 948
 949 class FFmpegSubtitlesConvertorPP(FFmpegPostProcessor):
 950     SUPPORTED_EXTS = MEDIA_EXTENSIONS.subtitles
 951
 952     def __init__(self, downloader=None, format=None):
 953         super().__init__(downloader)
 954         self.format = format
 955
 956     def run(self, info):
 957         subs = info.get('requested_subtitles')
 958         new_ext = self.format
 959         new_format = new_ext
 960         if new_format == 'vtt':
 961             new_format = 'webvtt'
 962         if subs is None:
 963             self.to_screen('There aren\'t any subtitles to convert')
 964             return [], info
 965         self.to_screen('Converting subtitles')
 966         sub_filenames = []
 967         for lang, sub in subs.items():
 968             if not os.path.exists(sub.get('filepath', '')):
 969                 self.report_warning(f'Skipping embedding {lang} subtitle because the file is missing')
 970                 continue
 971             ext = sub['ext']
 972             if ext == new_ext:
 973                 self.to_screen(f'Subtitle file for {new_ext} is already in the requested format')
 974                 continue
 975             elif ext == 'json':
 976                 self.to_screen(
 977                     'You have requested to convert json subtitles into another format, '
 978                     'which is currently not possible')
 979                 continue
 980             old_file = sub['filepath']
 981             sub_filenames.append(old_file)
 982             new_file = replace_extension(old_file, new_ext)
 983
 984             if ext in ('dfxp', 'ttml', 'tt'):
 985                 self.report_warning(
 986                     'You have requested to convert dfxp (TTML) subtitles into another format, '
 987                     'which results in style information loss')
 988
 989                 dfxp_file = old_file
 990                 srt_file = replace_extension(old_file, 'srt')
 991
 992                 with open(dfxp_file, 'rb') as f:
 993                     srt_data = dfxp2srt(f.read())
 994
 995                 with open(srt_file, 'w', encoding='utf-8') as f:
 996                     f.write(srt_data)
 997                 old_file = srt_file
 998
 999                 subs[lang] = {
1000                     'ext': 'srt',
1001                     'data': srt_data,
1002                     'filepath': srt_file,
1003                 }
1004
1005                 if new_ext == 'srt':
1006                     continue
1007                 else:
1008                     sub_filenames.append(srt_file)
1009
1010             self.run_ffmpeg(old_file, new_file, ['-f', new_format])
1011
1012             with open(new_file, encoding='utf-8') as f:
1013                 subs[lang] = {
1014                     'ext': new_ext,
1015                     'data': f.read(),
1016                     'filepath': new_file,
1017                 }
1018
1019             info['__files_to_move'][new_file] = replace_extension(
1020                 info['__files_to_move'][sub['filepath']], new_ext)
1021
1022         return sub_filenames, info
1023
1024
1025 class FFmpegSplitChaptersPP(FFmpegPostProcessor):
1026     def __init__(self, downloader, force_keyframes=False):
1027         FFmpegPostProcessor.__init__(self, downloader)
1028         self._force_keyframes = force_keyframes
1029
1030     def _prepare_filename(self, number, chapter, info):
1031         info = info.copy()
1032         info.update({
1033             'section_number': number,
1034             'section_title': chapter.get('title'),
1035             'section_start': chapter.get('start_time'),
1036             'section_end': chapter.get('end_time'),
1037         })
1038         return self._downloader.prepare_filename(info, 'chapter')
1039
1040     def _ffmpeg_args_for_chapter(self, number, chapter, info):
1041         destination = self._prepare_filename(number, chapter, info)
1042         if not self._downloader._ensure_dir_exists(encodeFilename(destination)):
1043             return
1044
1045         chapter['filepath'] = destination
1046         self.to_screen('Chapter %03d; Destination: %s' % (number, destination))
1047         return (
1048             destination,
1049             ['-ss', str(chapter['start_time']),
1050              '-t', str(chapter['end_time'] - chapter['start_time'])])
1051
1052     @PostProcessor._restrict_to(images=False)
1053     def run(self, info):
1054         self._fixup_chapters(info)
1055         chapters = info.get('chapters') or []
1056         if not chapters:
1057             self.to_screen('Chapter information is unavailable')
1058             return [], info
1059
1060         in_file = info['filepath']
1061         if self._force_keyframes and len(chapters) > 1:
1062             in_file = self.force_keyframes(in_file, (c['start_time'] for c in chapters))
1063         self.to_screen(f'Splitting video by chapters; {len(chapters)} chapters found')
1064         for idx, chapter in enumerate(chapters):
1065             destination, opts = self._ffmpeg_args_for_chapter(idx + 1, chapter, info)
1066             self.real_run_ffmpeg([(in_file, opts)], [(destination, self.stream_copy_opts())])
1067         if in_file != info['filepath']:
1068             self._delete_downloaded_files(in_file, msg=None)
1069         return [], info
1070
1071
1072 class FFmpegThumbnailsConvertorPP(FFmpegPostProcessor):
1073     SUPPORTED_EXTS = MEDIA_EXTENSIONS.thumbnails
1074     FORMAT_RE = create_mapping_re(SUPPORTED_EXTS)
1075
1076     def __init__(self, downloader=None, format=None):
1077         super().__init__(downloader)
1078         self.mapping = format
1079
1080     @classmethod
1081     def is_webp(cls, path):
1082         deprecation_warning(f'{cls.__module__}.{cls.__name__}.is_webp is deprecated')
1083         return imghdr.what(path) == 'webp'
1084
1085     def fixup_webp(self, info, idx=-1):
1086         thumbnail_filename = info['thumbnails'][idx]['filepath']
1087         _, thumbnail_ext = os.path.splitext(thumbnail_filename)
1088         if thumbnail_ext:
1089             if thumbnail_ext.lower() != '.webp' and imghdr.what(thumbnail_filename) == 'webp':
1090                 self.to_screen(f'Correcting thumbnail "{thumbnail_filename}" extension to webp')
1091                 webp_filename = replace_extension(thumbnail_filename, 'webp')
1092                 os.replace(thumbnail_filename, webp_filename)
1093                 info['thumbnails'][idx]['filepath'] = webp_filename
1094                 info['__files_to_move'][webp_filename] = replace_extension(
1095                     info['__files_to_move'].pop(thumbnail_filename), 'webp')
1096
1097     @staticmethod
1098     def _options(target_ext):
1099         yield from ('-update', '1')
1100         if target_ext == 'jpg':
1101             yield from ('-bsf:v', 'mjpeg2jpeg')
1102
1103     def convert_thumbnail(self, thumbnail_filename, target_ext):
1104         thumbnail_conv_filename = replace_extension(thumbnail_filename, target_ext)
1105
1106         self.to_screen(f'Converting thumbnail "{thumbnail_filename}" to {target_ext}')
1107         _, source_ext = os.path.splitext(thumbnail_filename)
1108         self.real_run_ffmpeg(
1109             [(thumbnail_filename, [] if source_ext == '.gif' else ['-f', 'image2', '-pattern_type', 'none'])],
1110             [(thumbnail_conv_filename, self._options(target_ext))])
1111         return thumbnail_conv_filename
1112
1113     def run(self, info):
1114         files_to_delete = []
1115         has_thumbnail = False
1116
1117         for idx, thumbnail_dict in enumerate(info.get('thumbnails') or []):
1118             original_thumbnail = thumbnail_dict.get('filepath')
1119             if not original_thumbnail:
1120                 continue
1121             has_thumbnail = True
1122             self.fixup_webp(info, idx)
1123             original_thumbnail = thumbnail_dict['filepath']  # Path can change during fixup
1124             thumbnail_ext = os.path.splitext(original_thumbnail)[1][1:].lower()
1125             if thumbnail_ext == 'jpeg':
1126                 thumbnail_ext = 'jpg'
1127             target_ext, _skip_msg = resolve_mapping(thumbnail_ext, self.mapping)
1128             if _skip_msg:
1129                 self.to_screen(f'Not converting thumbnail "{original_thumbnail}"; {_skip_msg}')
1130                 continue
1131             thumbnail_dict['filepath'] = self.convert_thumbnail(original_thumbnail, target_ext)
1132             files_to_delete.append(original_thumbnail)
1133             info['__files_to_move'][thumbnail_dict['filepath']] = replace_extension(
1134                 info['__files_to_move'][original_thumbnail], target_ext)
1135
1136         if not has_thumbnail:
1137             self.to_screen('There aren\'t any thumbnails to convert')
1138         return files_to_delete, info
1139
1140
1141 class FFmpegConcatPP(FFmpegPostProcessor):
1142     def __init__(self, downloader, only_multi_video=False):
1143         self._only_multi_video = only_multi_video
1144         super().__init__(downloader)
1145
1146     def _get_codecs(self, file):
1147         codecs = traverse_obj(self.get_metadata_object(file), ('streams', ..., 'codec_name'))
1148         self.write_debug(f'Codecs = {", ".join(codecs)}')
1149         return tuple(codecs)
1150
1151     def concat_files(self, in_files, out_file):
1152         if not self._downloader._ensure_dir_exists(out_file):
1153             return
1154         if len(in_files) == 1:
1155             if os.path.realpath(in_files[0]) != os.path.realpath(out_file):
1156                 self.to_screen(f'Moving "{in_files[0]}" to "{out_file}"')
1157             os.replace(in_files[0], out_file)
1158             return []
1159
1160         if len(set(map(self._get_codecs, in_files))) > 1:
1161             raise PostProcessingError(
1162                 'The files have different streams/codecs and cannot be concatenated. '
1163                 'Either select different formats or --recode-video them to a common format')
1164
1165         self.to_screen(f'Concatenating {len(in_files)} files; Destination: {out_file}')
1166         super().concat_files(in_files, out_file)
1167         return in_files
1168
1169     @PostProcessor._restrict_to(images=False, simulated=False)
1170     def run(self, info):
1171         entries = info.get('entries') or []
1172         if not any(entries) or (self._only_multi_video and info['_type'] != 'multi_video'):
1173             return [], info
1174         elif traverse_obj(entries, (..., lambda k, v: k == 'requested_downloads' and len(v) > 1)):
1175             raise PostProcessingError('Concatenation is not supported when downloading multiple separate formats')
1176
1177         in_files = traverse_obj(entries, (..., 'requested_downloads', 0, 'filepath')) or []
1178         if len(in_files) < len(entries):
1179             raise PostProcessingError('Aborting concatenation because some downloads failed')
1180
1181         exts = traverse_obj(entries, (..., 'requested_downloads', 0, 'ext'), (..., 'ext'))
1182         ie_copy = collections.ChainMap({'ext': exts[0] if len(set(exts)) == 1 else 'mkv'},
1183                                        info, self._downloader._playlist_infodict(info))
1184         out_file = self._downloader.prepare_filename(ie_copy, 'pl_video')
1185
1186         files_to_delete = self.concat_files(in_files, out_file)
1187
1188         info['requested_downloads'] = [{
1189             'filepath': out_file,
1190             'ext': ie_copy['ext'],
1191         }]
1192         return files_to_delete, info