yt_dlp/postprocessor/ffmpeg.py

   1 from __future__ import unicode_literals
   2
   3 import collections
   4 import io
   5 import itertools
   6 import os
   7 import subprocess
   8 import time
   9 import re
  10 import json
  11
  12 from .common import AudioConversionError, PostProcessor
  13
  14 from ..compat import compat_str
  15 from ..utils import (
  16     determine_ext,
  17     dfxp2srt,
  18     encodeArgument,
  19     encodeFilename,
  20     float_or_none,
  21     _get_exe_version_output,
  22     detect_exe_version,
  23     is_outdated_version,
  24     ISO639Utils,
  25     orderedSet,
  26     Popen,
  27     PostProcessingError,
  28     prepend_extension,
  29     replace_extension,
  30     shell_quote,
  31     traverse_obj,
  32     variadic,
  33     write_json_file,
  34 )
  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     'vtt': 'webvtt',
  49 }
  50 ACODECS = {
  51     'mp3': 'libmp3lame',
  52     'aac': 'aac',
  53     'flac': 'flac',
  54     'm4a': 'aac',
  55     'opus': 'libopus',
  56     'vorbis': 'libvorbis',
  57     'wav': None,
  58     'alac': None,
  59 }
  60
  61
  62 class FFmpegPostProcessorError(PostProcessingError):
  63     pass
  64
  65
  66 class FFmpegPostProcessor(PostProcessor):
  67     def __init__(self, downloader=None):
  68         PostProcessor.__init__(self, downloader)
  69         self._determine_executables()
  70
  71     def check_version(self):
  72         if not self.available:
  73             raise FFmpegPostProcessorError('ffmpeg not found. Please install or provide the path using --ffmpeg-location')
  74
  75         required_version = '10-0' if self.basename == 'avconv' else '1.0'
  76         if is_outdated_version(
  77                 self._versions[self.basename], required_version):
  78             warning = 'Your copy of %s is outdated, update %s to version %s or newer if you encounter any errors.' % (
  79                 self.basename, self.basename, required_version)
  80             self.report_warning(warning)
  81
  82     @staticmethod
  83     def get_versions_and_features(downloader=None):
  84         pp = FFmpegPostProcessor(downloader)
  85         return pp._versions, pp._features
  86
  87     @staticmethod
  88     def get_versions(downloader=None):
  89         return FFmpegPostProcessor.get_versions_and_features(downloader)[0]
  90
  91     _version_cache, _features_cache = {}, {}
  92
  93     def _determine_executables(self):
  94         programs = ['avprobe', 'avconv', 'ffmpeg', 'ffprobe']
  95
  96         def get_ffmpeg_version(path, prog):
  97             if path in self._version_cache:
  98                 self._versions[prog], self._features = self._version_cache[path], self._features_cache.get(path, {})
  99                 return
 100             out = _get_exe_version_output(path, ['-bsfs'], to_screen=self.write_debug)
 101             ver = detect_exe_version(out) if out else False
 102             if ver:
 103                 regexs = [
 104                     r'(?:\d+:)?([0-9.]+)-[0-9]+ubuntu[0-9.]+$',  # Ubuntu, see [1]
 105                     r'n([0-9.]+)$',  # Arch Linux
 106                     # 1. http://www.ducea.com/2006/06/17/ubuntu-package-version-naming-explanation/
 107                 ]
 108                 for regex in regexs:
 109                     mobj = re.match(regex, ver)
 110                     if mobj:
 111                         ver = mobj.group(1)
 112             self._versions[prog] = self._version_cache[path] = ver
 113             if prog != 'ffmpeg' or not out:
 114                 return
 115
 116             mobj = re.search(r'(?m)^\s+libavformat\s+(?:[0-9. ]+)\s+/\s+(?P<runtime>[0-9. ]+)', out)
 117             lavf_runtime_version = mobj.group('runtime').replace(' ', '') if mobj else None
 118             self._features = self._features_cache[path] = {
 119                 'fdk': '--enable-libfdk-aac' in out,
 120                 'setts': 'setts' in out.splitlines(),
 121                 'needs_adtstoasc': is_outdated_version(lavf_runtime_version, '57.56.100', False),
 122             }
 123
 124         self.basename = None
 125         self.probe_basename = None
 126         self._paths = None
 127         self._versions = None
 128         self._features = {}
 129
 130         prefer_ffmpeg = self.get_param('prefer_ffmpeg', True)
 131         location = self.get_param('ffmpeg_location')
 132         if location is None:
 133             self._paths = {p: p for p in programs}
 134         else:
 135             if not os.path.exists(location):
 136                 self.report_warning(
 137                     'ffmpeg-location %s does not exist! '
 138                     'Continuing without ffmpeg.' % (location))
 139                 self._versions = {}
 140                 return
 141             elif os.path.isdir(location):
 142                 dirname, basename = location, None
 143             else:
 144                 basename = os.path.splitext(os.path.basename(location))[0]
 145                 basename = next((p for p in programs if basename.startswith(p)), 'ffmpeg')
 146                 dirname = os.path.dirname(os.path.abspath(location))
 147                 if basename in ('ffmpeg', 'ffprobe'):
 148                     prefer_ffmpeg = True
 149
 150             self._paths = dict(
 151                 (p, os.path.join(dirname, p)) for p in programs)
 152             if basename:
 153                 self._paths[basename] = location
 154
 155         self._versions = {}
 156         executables = {'basename': ('ffmpeg', 'avconv'), 'probe_basename': ('ffprobe', 'avprobe')}
 157         if prefer_ffmpeg is False:
 158             executables = {k: v[::-1] for k, v in executables.items()}
 159         for var, prefs in executables.items():
 160             for p in prefs:
 161                 get_ffmpeg_version(self._paths[p], p)
 162                 if self._versions[p]:
 163                     setattr(self, var, p)
 164                     break
 165
 166         if self.basename == 'avconv':
 167             self.deprecation_warning(
 168                 'Support for avconv is deprecated and may be removed in a future version. Use ffmpeg instead')
 169         if self.probe_basename == 'avprobe':
 170             self.deprecation_warning(
 171                 'Support for avprobe is deprecated and may be removed in a future version. Use ffprobe instead')
 172
 173     @property
 174     def available(self):
 175         return self.basename is not None
 176
 177     @property
 178     def executable(self):
 179         return self._paths[self.basename]
 180
 181     @property
 182     def probe_available(self):
 183         return self.probe_basename is not None
 184
 185     @property
 186     def probe_executable(self):
 187         return self._paths[self.probe_basename]
 188
 189     @staticmethod
 190     def stream_copy_opts(copy=True, *, ext=None):
 191         yield from ('-map', '0')
 192         # Don't copy Apple TV chapters track, bin_data
 193         # See https://github.com/yt-dlp/yt-dlp/issues/2, #19042, #19024, https://trac.ffmpeg.org/ticket/6016
 194         yield from ('-dn', '-ignore_unknown')
 195         if copy:
 196             yield from ('-c', 'copy')
 197         # For some reason, '-c copy -map 0' is not enough to copy subtitles
 198         if ext in ('mp4', 'mov'):
 199             yield from ('-c:s', 'mov_text')
 200
 201     def get_audio_codec(self, path):
 202         if not self.probe_available and not self.available:
 203             raise PostProcessingError('ffprobe and ffmpeg not found. Please install or provide the path using --ffmpeg-location')
 204         try:
 205             if self.probe_available:
 206                 cmd = [
 207                     encodeFilename(self.probe_executable, True),
 208                     encodeArgument('-show_streams')]
 209             else:
 210                 cmd = [
 211                     encodeFilename(self.executable, True),
 212                     encodeArgument('-i')]
 213             cmd.append(encodeFilename(self._ffmpeg_filename_argument(path), True))
 214             self.write_debug('%s command line: %s' % (self.basename, shell_quote(cmd)))
 215             handle = Popen(cmd, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
 216             stdout_data, stderr_data = handle.communicate_or_kill()
 217             expected_ret = 0 if self.probe_available else 1
 218             if handle.wait() != expected_ret:
 219                 return None
 220         except (IOError, OSError):
 221             return None
 222         output = (stdout_data if self.probe_available else stderr_data).decode('ascii', 'ignore')
 223         if self.probe_available:
 224             audio_codec = None
 225             for line in output.split('\n'):
 226                 if line.startswith('codec_name='):
 227                     audio_codec = line.split('=')[1].strip()
 228                 elif line.strip() == 'codec_type=audio' and audio_codec is not None:
 229                     return audio_codec
 230         else:
 231             # Stream #FILE_INDEX:STREAM_INDEX[STREAM_ID](LANGUAGE): CODEC_TYPE: CODEC_NAME
 232             mobj = re.search(
 233                 r'Stream\s*#\d+:\d+(?:\[0x[0-9a-f]+\])?(?:\([a-z]{3}\))?:\s*Audio:\s*([0-9a-z]+)',
 234                 output)
 235             if mobj:
 236                 return mobj.group(1)
 237         return None
 238
 239     def get_metadata_object(self, path, opts=[]):
 240         if self.probe_basename != 'ffprobe':
 241             if self.probe_available:
 242                 self.report_warning('Only ffprobe is supported for metadata extraction')
 243             raise PostProcessingError('ffprobe not found. Please install or provide the path using --ffmpeg-location')
 244         self.check_version()
 245
 246         cmd = [
 247             encodeFilename(self.probe_executable, True),
 248             encodeArgument('-hide_banner'),
 249             encodeArgument('-show_format'),
 250             encodeArgument('-show_streams'),
 251             encodeArgument('-print_format'),
 252             encodeArgument('json'),
 253         ]
 254
 255         cmd += opts
 256         cmd.append(encodeFilename(self._ffmpeg_filename_argument(path), True))
 257         self.write_debug('ffprobe command line: %s' % shell_quote(cmd))
 258         p = Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE)
 259         stdout, stderr = p.communicate()
 260         return json.loads(stdout.decode('utf-8', 'replace'))
 261
 262     def get_stream_number(self, path, keys, value):
 263         streams = self.get_metadata_object(path)['streams']
 264         num = next(
 265             (i for i, stream in enumerate(streams) if traverse_obj(stream, keys, casesense=False) == value),
 266             None)
 267         return num, len(streams)
 268
 269     def _get_real_video_duration(self, filepath, fatal=True):
 270         try:
 271             duration = float_or_none(
 272                 traverse_obj(self.get_metadata_object(filepath), ('format', 'duration')))
 273             if not duration:
 274                 raise PostProcessingError('ffprobe returned empty duration')
 275             return duration
 276         except PostProcessingError as e:
 277             if fatal:
 278                 raise PostProcessingError(f'Unable to determine video duration: {e.msg}')
 279
 280     def _duration_mismatch(self, d1, d2):
 281         if not d1 or not d2:
 282             return None
 283         # The duration is often only known to nearest second. So there can be <1sec disparity natually.
 284         # Further excuse an additional <1sec difference.
 285         return abs(d1 - d2) > 2
 286
 287     def run_ffmpeg_multiple_files(self, input_paths, out_path, opts, **kwargs):
 288         return self.real_run_ffmpeg(
 289             [(path, []) for path in input_paths],
 290             [(out_path, opts)], **kwargs)
 291
 292     def real_run_ffmpeg(self, input_path_opts, output_path_opts, *, expected_retcodes=(0,)):
 293         self.check_version()
 294
 295         oldest_mtime = min(
 296             os.stat(encodeFilename(path)).st_mtime for path, _ in input_path_opts if path)
 297
 298         cmd = [encodeFilename(self.executable, True), encodeArgument('-y')]
 299         # avconv does not have repeat option
 300         if self.basename == 'ffmpeg':
 301             cmd += [encodeArgument('-loglevel'), encodeArgument('repeat+info')]
 302
 303         def make_args(file, args, name, number):
 304             keys = ['_%s%d' % (name, number), '_%s' % name]
 305             if name == 'o':
 306                 args += ['-movflags', '+faststart']
 307                 if number == 1:
 308                     keys.append('')
 309             args += self._configuration_args(self.basename, keys)
 310             if name == 'i':
 311                 args.append('-i')
 312             return (
 313                 [encodeArgument(arg) for arg in args]
 314                 + [encodeFilename(self._ffmpeg_filename_argument(file), True)])
 315
 316         for arg_type, path_opts in (('i', input_path_opts), ('o', output_path_opts)):
 317             cmd += itertools.chain.from_iterable(
 318                 make_args(path, list(opts), arg_type, i + 1)
 319                 for i, (path, opts) in enumerate(path_opts) if path)
 320
 321         self.write_debug('ffmpeg command line: %s' % shell_quote(cmd))
 322         p = Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE)
 323         stdout, stderr = p.communicate_or_kill()
 324         if p.returncode not in variadic(expected_retcodes):
 325             stderr = stderr.decode('utf-8', 'replace').strip()
 326             self.write_debug(stderr)
 327             raise FFmpegPostProcessorError(stderr.split('\n')[-1])
 328         for out_path, _ in output_path_opts:
 329             if out_path:
 330                 self.try_utime(out_path, oldest_mtime, oldest_mtime)
 331         return stderr.decode('utf-8', 'replace')
 332
 333     def run_ffmpeg(self, path, out_path, opts, **kwargs):
 334         return self.run_ffmpeg_multiple_files([path], out_path, opts, **kwargs)
 335
 336     @staticmethod
 337     def _ffmpeg_filename_argument(fn):
 338         # Always use 'file:' because the filename may contain ':' (ffmpeg
 339         # interprets that as a protocol) or can start with '-' (-- is broken in
 340         # ffmpeg, see https://ffmpeg.org/trac/ffmpeg/ticket/2127 for details)
 341         # Also leave '-' intact in order not to break streaming to stdout.
 342         if fn.startswith(('http://', 'https://')):
 343             return fn
 344         return 'file:' + fn if fn != '-' else fn
 345
 346     @staticmethod
 347     def _quote_for_ffmpeg(string):
 348         # See https://ffmpeg.org/ffmpeg-utils.html#toc-Quoting-and-escaping
 349         # A sequence of '' produces '\'''\'';
 350         # final replace removes the empty '' between \' \'.
 351         string = string.replace("'", r"'\''").replace("'''", "'")
 352         # Handle potential ' at string boundaries.
 353         string = string[1:] if string[0] == "'" else "'" + string
 354         return string[:-1] if string[-1] == "'" else string + "'"
 355
 356     def force_keyframes(self, filename, timestamps):
 357         timestamps = orderedSet(timestamps)
 358         if timestamps[0] == 0:
 359             timestamps = timestamps[1:]
 360         keyframe_file = prepend_extension(filename, 'keyframes.temp')
 361         self.to_screen(f'Re-encoding "{filename}" with appropriate keyframes')
 362         self.run_ffmpeg(filename, keyframe_file, [
 363             *self.stream_copy_opts(False, ext=determine_ext(filename)),
 364             '-force_key_frames', ','.join(f'{t:.6f}' for t in timestamps)])
 365         return keyframe_file
 366
 367     def concat_files(self, in_files, out_file, concat_opts=None):
 368         """
 369         Use concat demuxer to concatenate multiple files having identical streams.
 370
 371         Only inpoint, outpoint, and duration concat options are supported.
 372         See https://ffmpeg.org/ffmpeg-formats.html#concat-1 for details
 373         """
 374         concat_file = f'{out_file}.concat'
 375         self.write_debug(f'Writing concat spec to {concat_file}')
 376         with open(concat_file, 'wt', encoding='utf-8') as f:
 377             f.writelines(self._concat_spec(in_files, concat_opts))
 378
 379         out_flags = list(self.stream_copy_opts(ext=determine_ext(out_file)))
 380
 381         self.real_run_ffmpeg(
 382             [(concat_file, ['-hide_banner', '-nostdin', '-f', 'concat', '-safe', '0'])],
 383             [(out_file, out_flags)])
 384         os.remove(concat_file)
 385
 386     @classmethod
 387     def _concat_spec(cls, in_files, concat_opts=None):
 388         if concat_opts is None:
 389             concat_opts = [{}] * len(in_files)
 390         yield 'ffconcat version 1.0\n'
 391         for file, opts in zip(in_files, concat_opts):
 392             yield f'file {cls._quote_for_ffmpeg(cls._ffmpeg_filename_argument(file))}\n'
 393             # Iterate explicitly to yield the following directives in order, ignoring the rest.
 394             for directive in 'inpoint', 'outpoint', 'duration':
 395                 if directive in opts:
 396                     yield f'{directive} {opts[directive]}\n'
 397
 398
 399 class FFmpegExtractAudioPP(FFmpegPostProcessor):
 400     COMMON_AUDIO_EXTS = ('wav', 'flac', 'm4a', 'aiff', 'mp3', 'ogg', 'mka', 'opus', 'wma')
 401     SUPPORTED_EXTS = ('aac', 'flac', 'mp3', 'm4a', 'opus', 'vorbis', 'wav', 'alac')
 402
 403     def __init__(self, downloader=None, preferredcodec=None, preferredquality=None, nopostoverwrites=False):
 404         FFmpegPostProcessor.__init__(self, downloader)
 405         self._preferredcodec = preferredcodec or 'best'
 406         self._preferredquality = float_or_none(preferredquality)
 407         self._nopostoverwrites = nopostoverwrites
 408
 409     def _quality_args(self, codec):
 410         if self._preferredquality is None:
 411             return []
 412         elif self._preferredquality > 10:
 413             return ['-b:a', f'{self._preferredquality}k']
 414
 415         limits = {
 416             'libmp3lame': (10, 0),
 417             'libvorbis': (0, 10),
 418             # FFmpeg's AAC encoder does not have an upper limit for the value of -q:a.
 419             # Experimentally, with values over 4, bitrate changes were minimal or non-existent
 420             'aac': (0.1, 4),
 421             'libfdk_aac': (1, 5),
 422         }.get(codec)
 423         if not limits:
 424             return []
 425
 426         q = limits[1] + (limits[0] - limits[1]) * (self._preferredquality / 10)
 427         if codec == 'libfdk_aac':
 428             return ['-vbr', f'{int(q)}']
 429         return ['-q:a', f'{q}']
 430
 431     def run_ffmpeg(self, path, out_path, codec, more_opts):
 432         if codec is None:
 433             acodec_opts = []
 434         else:
 435             acodec_opts = ['-acodec', codec]
 436         opts = ['-vn'] + acodec_opts + more_opts
 437         try:
 438             FFmpegPostProcessor.run_ffmpeg(self, path, out_path, opts)
 439         except FFmpegPostProcessorError as err:
 440             raise AudioConversionError(err.msg)
 441
 442     @PostProcessor._restrict_to(images=False)
 443     def run(self, information):
 444         orig_path = path = information['filepath']
 445         orig_ext = information['ext']
 446
 447         if self._preferredcodec == 'best' and orig_ext in self.COMMON_AUDIO_EXTS:
 448             self.to_screen('Skipping audio extraction since the file is already in a common audio format')
 449             return [], information
 450
 451         filecodec = self.get_audio_codec(path)
 452         if filecodec is None:
 453             raise PostProcessingError('WARNING: unable to obtain file audio codec with ffprobe')
 454
 455         more_opts = []
 456         if self._preferredcodec == 'best' or self._preferredcodec == filecodec or (self._preferredcodec == 'm4a' and filecodec == 'aac'):
 457             if filecodec == 'aac' and self._preferredcodec in ['m4a', 'best']:
 458                 # Lossless, but in another container
 459                 acodec = 'copy'
 460                 extension = 'm4a'
 461                 more_opts = ['-bsf:a', 'aac_adtstoasc']
 462             elif filecodec in ['aac', 'flac', 'mp3', 'vorbis', 'opus']:
 463                 # Lossless if possible
 464                 acodec = 'copy'
 465                 extension = filecodec
 466                 if filecodec == 'aac':
 467                     more_opts = ['-f', 'adts']
 468                 if filecodec == 'vorbis':
 469                     extension = 'ogg'
 470             elif filecodec == 'alac':
 471                 acodec = None
 472                 extension = 'm4a'
 473                 more_opts += ['-acodec', 'alac']
 474             else:
 475                 # MP3 otherwise.
 476                 acodec = 'libmp3lame'
 477                 extension = 'mp3'
 478                 more_opts = self._quality_args(acodec)
 479         else:
 480             # We convert the audio (lossy if codec is lossy)
 481             acodec = ACODECS[self._preferredcodec]
 482             if acodec == 'aac' and self._features.get('fdk'):
 483                 acodec = 'libfdk_aac'
 484             extension = self._preferredcodec
 485             more_opts = self._quality_args(acodec)
 486             if self._preferredcodec == 'aac':
 487                 more_opts += ['-f', 'adts']
 488             elif self._preferredcodec == 'm4a':
 489                 more_opts += ['-bsf:a', 'aac_adtstoasc']
 490             elif self._preferredcodec == 'vorbis':
 491                 extension = 'ogg'
 492             elif self._preferredcodec == 'wav':
 493                 extension = 'wav'
 494                 more_opts += ['-f', 'wav']
 495             elif self._preferredcodec == 'alac':
 496                 extension = 'm4a'
 497                 more_opts += ['-acodec', 'alac']
 498
 499         prefix, sep, ext = path.rpartition('.')  # not os.path.splitext, since the latter does not work on unicode in all setups
 500         temp_path = new_path = prefix + sep + extension
 501
 502         if new_path == path:
 503             orig_path = prepend_extension(path, 'orig')
 504             temp_path = prepend_extension(path, 'temp')
 505         if (self._nopostoverwrites and os.path.exists(encodeFilename(new_path))
 506                 and os.path.exists(encodeFilename(orig_path))):
 507             self.to_screen('Post-process file %s exists, skipping' % new_path)
 508             return [], information
 509
 510         try:
 511             self.to_screen(f'Destination: {new_path}')
 512             self.run_ffmpeg(path, temp_path, acodec, more_opts)
 513         except AudioConversionError as e:
 514             raise PostProcessingError(
 515                 'audio conversion failed: ' + e.msg)
 516         except Exception:
 517             raise PostProcessingError('error running ' + self.basename)
 518
 519         os.replace(path, orig_path)
 520         os.replace(temp_path, new_path)
 521         information['filepath'] = new_path
 522         information['ext'] = extension
 523
 524         # Try to update the date time for extracted audio file.
 525         if information.get('filetime') is not None:
 526             self.try_utime(
 527                 new_path, time.time(), information['filetime'],
 528                 errnote='Cannot update utime of audio file')
 529
 530         return [orig_path], information
 531
 532
 533 class FFmpegVideoConvertorPP(FFmpegPostProcessor):
 534     SUPPORTED_EXTS = ('mp4', 'mkv', 'flv', 'webm', 'mov', 'avi', 'mka', 'ogg', *FFmpegExtractAudioPP.SUPPORTED_EXTS)
 535     FORMAT_RE = re.compile(r'{0}(?:/{0})*$'.format(r'(?:\w+>)?(?:%s)' % '|'.join(SUPPORTED_EXTS)))
 536     _ACTION = 'converting'
 537
 538     def __init__(self, downloader=None, preferedformat=None):
 539         super(FFmpegVideoConvertorPP, self).__init__(downloader)
 540         self._preferedformats = preferedformat.lower().split('/')
 541
 542     def _target_ext(self, source_ext):
 543         for pair in self._preferedformats:
 544             kv = pair.split('>')
 545             if len(kv) == 1 or kv[0].strip() == source_ext:
 546                 return kv[-1].strip()
 547
 548     @staticmethod
 549     def _options(target_ext):
 550         yield from FFmpegPostProcessor.stream_copy_opts(False)
 551         if target_ext == 'avi':
 552             yield from ('-c:v', 'libxvid', '-vtag', 'XVID')
 553
 554     @PostProcessor._restrict_to(images=False)
 555     def run(self, info):
 556         filename, source_ext = info['filepath'], info['ext'].lower()
 557         target_ext = self._target_ext(source_ext)
 558         _skip_msg = (
 559             f'could not find a mapping for {source_ext}' if not target_ext
 560             else f'already is in target format {source_ext}' if source_ext == target_ext
 561             else None)
 562         if _skip_msg:
 563             self.to_screen(f'Not {self._ACTION} media file "{filename}"; {_skip_msg}')
 564             return [], info
 565
 566         outpath = replace_extension(filename, target_ext, source_ext)
 567         self.to_screen(f'{self._ACTION.title()} video from {source_ext} to {target_ext}; Destination: {outpath}')
 568         self.run_ffmpeg(filename, outpath, self._options(target_ext))
 569
 570         info['filepath'] = outpath
 571         info['format'] = info['ext'] = target_ext
 572         return [filename], info
 573
 574
 575 class FFmpegVideoRemuxerPP(FFmpegVideoConvertorPP):
 576     _ACTION = 'remuxing'
 577
 578     @staticmethod
 579     def _options(target_ext):
 580         return FFmpegPostProcessor.stream_copy_opts()
 581
 582
 583 class FFmpegEmbedSubtitlePP(FFmpegPostProcessor):
 584     def __init__(self, downloader=None, already_have_subtitle=False):
 585         super(FFmpegEmbedSubtitlePP, self).__init__(downloader)
 586         self._already_have_subtitle = already_have_subtitle
 587
 588     @PostProcessor._restrict_to(images=False)
 589     def run(self, info):
 590         if info['ext'] not in ('mp4', 'webm', 'mkv'):
 591             self.to_screen('Subtitles can only be embedded in mp4, webm or mkv files')
 592             return [], info
 593         subtitles = info.get('requested_subtitles')
 594         if not subtitles:
 595             self.to_screen('There aren\'t any subtitles to embed')
 596             return [], info
 597
 598         filename = info['filepath']
 599
 600         # Disabled temporarily. There needs to be a way to overide this
 601         # in case of duration actually mismatching in extractor
 602         # See: https://github.com/yt-dlp/yt-dlp/issues/1870, https://github.com/yt-dlp/yt-dlp/issues/1385
 603         '''
 604         if info.get('duration') and not info.get('__real_download') and self._duration_mismatch(
 605                 self._get_real_video_duration(filename, False), info['duration']):
 606             self.to_screen(f'Skipping {self.pp_key()} since the real and expected durations mismatch')
 607             return [], info
 608         '''
 609
 610         ext = info['ext']
 611         sub_langs, sub_names, sub_filenames = [], [], []
 612         webm_vtt_warn = False
 613         mp4_ass_warn = False
 614
 615         for lang, sub_info in subtitles.items():
 616             if not os.path.exists(sub_info.get('filepath', '')):
 617                 self.report_warning(f'Skipping embedding {lang} subtitle because the file is missing')
 618                 continue
 619             sub_ext = sub_info['ext']
 620             if sub_ext == 'json':
 621                 self.report_warning('JSON subtitles cannot be embedded')
 622             elif ext != 'webm' or ext == 'webm' and sub_ext == 'vtt':
 623                 sub_langs.append(lang)
 624                 sub_names.append(sub_info.get('name'))
 625                 sub_filenames.append(sub_info['filepath'])
 626             else:
 627                 if not webm_vtt_warn and ext == 'webm' and sub_ext != 'vtt':
 628                     webm_vtt_warn = True
 629                     self.report_warning('Only WebVTT subtitles can be embedded in webm files')
 630             if not mp4_ass_warn and ext == 'mp4' and sub_ext == 'ass':
 631                 mp4_ass_warn = True
 632                 self.report_warning('ASS subtitles cannot be properly embedded in mp4 files; expect issues')
 633
 634         if not sub_langs:
 635             return [], info
 636
 637         input_files = [filename] + sub_filenames
 638
 639         opts = [
 640             *self.stream_copy_opts(ext=info['ext']),
 641             # Don't copy the existing subtitles, we may be running the
 642             # postprocessor a second time
 643             '-map', '-0:s',
 644         ]
 645         for i, (lang, name) in enumerate(zip(sub_langs, sub_names)):
 646             opts.extend(['-map', '%d:0' % (i + 1)])
 647             lang_code = ISO639Utils.short2long(lang) or lang
 648             opts.extend(['-metadata:s:s:%d' % i, 'language=%s' % lang_code])
 649             if name:
 650                 opts.extend(['-metadata:s:s:%d' % i, 'handler_name=%s' % name,
 651                              '-metadata:s:s:%d' % i, 'title=%s' % name])
 652
 653         temp_filename = prepend_extension(filename, 'temp')
 654         self.to_screen('Embedding subtitles in "%s"' % filename)
 655         self.run_ffmpeg_multiple_files(input_files, temp_filename, opts)
 656         os.replace(temp_filename, filename)
 657
 658         files_to_delete = [] if self._already_have_subtitle else sub_filenames
 659         return files_to_delete, info
 660
 661
 662 class FFmpegMetadataPP(FFmpegPostProcessor):
 663
 664     def __init__(self, downloader, add_metadata=True, add_chapters=True, add_infojson='if_exists'):
 665         FFmpegPostProcessor.__init__(self, downloader)
 666         self._add_metadata = add_metadata
 667         self._add_chapters = add_chapters
 668         self._add_infojson = add_infojson
 669
 670     @staticmethod
 671     def _options(target_ext):
 672         audio_only = target_ext == 'm4a'
 673         yield from FFmpegPostProcessor.stream_copy_opts(not audio_only)
 674         if audio_only:
 675             yield from ('-vn', '-acodec', 'copy')
 676
 677     @PostProcessor._restrict_to(images=False)
 678     def run(self, info):
 679         filename, metadata_filename = info['filepath'], None
 680         files_to_delete, options = [], []
 681         if self._add_chapters and info.get('chapters'):
 682             metadata_filename = replace_extension(filename, 'meta')
 683             options.extend(self._get_chapter_opts(info['chapters'], metadata_filename))
 684             files_to_delete.append(metadata_filename)
 685         if self._add_metadata:
 686             options.extend(self._get_metadata_opts(info))
 687
 688         if self._add_infojson:
 689             if info['ext'] in ('mkv', 'mka'):
 690                 infojson_filename = info.get('infojson_filename')
 691                 options.extend(self._get_infojson_opts(info, infojson_filename))
 692                 if not infojson_filename:
 693                     files_to_delete.append(info.get('infojson_filename'))
 694             elif self._add_infojson is True:
 695                 self.to_screen('The info-json can only be attached to mkv/mka files')
 696
 697         if not options:
 698             self.to_screen('There isn\'t any metadata to add')
 699             return [], info
 700
 701         temp_filename = prepend_extension(filename, 'temp')
 702         self.to_screen('Adding metadata to "%s"' % filename)
 703         self.run_ffmpeg_multiple_files(
 704             (filename, metadata_filename), temp_filename,
 705             itertools.chain(self._options(info['ext']), *options))
 706         for file in filter(None, files_to_delete):
 707             os.remove(file)  # Don't obey --keep-files
 708         os.replace(temp_filename, filename)
 709         return [], info
 710
 711     @staticmethod
 712     def _get_chapter_opts(chapters, metadata_filename):
 713         with io.open(metadata_filename, 'wt', encoding='utf-8') as f:
 714             def ffmpeg_escape(text):
 715                 return re.sub(r'([\\=;#\n])', r'\\\1', text)
 716
 717             metadata_file_content = ';FFMETADATA1\n'
 718             for chapter in chapters:
 719                 metadata_file_content += '[CHAPTER]\nTIMEBASE=1/1000\n'
 720                 metadata_file_content += 'START=%d\n' % (chapter['start_time'] * 1000)
 721                 metadata_file_content += 'END=%d\n' % (chapter['end_time'] * 1000)
 722                 chapter_title = chapter.get('title')
 723                 if chapter_title:
 724                     metadata_file_content += 'title=%s\n' % ffmpeg_escape(chapter_title)
 725             f.write(metadata_file_content)
 726         yield ('-map_metadata', '1')
 727
 728     def _get_metadata_opts(self, info):
 729         meta_prefix = 'meta'
 730         metadata = collections.defaultdict(dict)
 731
 732         def add(meta_list, info_list=None):
 733             value = next((
 734                 str(info[key]) for key in [f'{meta_prefix}_'] + list(variadic(info_list or meta_list))
 735                 if info.get(key) is not None), None)
 736             if value not in ('', None):
 737                 metadata['common'].update({meta_f: value for meta_f in variadic(meta_list)})
 738
 739         # See [1-4] for some info on media metadata/metadata supported
 740         # by ffmpeg.
 741         # 1. https://kdenlive.org/en/project/adding-meta-data-to-mp4-video/
 742         # 2. https://wiki.multimedia.cx/index.php/FFmpeg_Metadata
 743         # 3. https://kodi.wiki/view/Video_file_tagging
 744
 745         add('title', ('track', 'title'))
 746         add('date', 'upload_date')
 747         add(('description', 'synopsis'), 'description')
 748         add(('purl', 'comment'), 'webpage_url')
 749         add('track', 'track_number')
 750         add('artist', ('artist', 'creator', 'uploader', 'uploader_id'))
 751         add('genre')
 752         add('album')
 753         add('album_artist')
 754         add('disc', 'disc_number')
 755         add('show', 'series')
 756         add('season_number')
 757         add('episode_id', ('episode', 'episode_id'))
 758         add('episode_sort', 'episode_number')
 759         if 'embed-metadata' in self.get_param('compat_opts', []):
 760             add('comment', 'description')
 761             metadata['common'].pop('synopsis', None)
 762
 763         meta_regex = rf'{re.escape(meta_prefix)}(?P<i>\d+)?_(?P<key>.+)'
 764         for key, value in info.items():
 765             mobj = re.fullmatch(meta_regex, key)
 766             if value is not None and mobj:
 767                 metadata[mobj.group('i') or 'common'][mobj.group('key')] = value
 768
 769         for name, value in metadata['common'].items():
 770             yield ('-metadata', f'{name}={value}')
 771
 772         stream_idx = 0
 773         for fmt in info.get('requested_formats') or []:
 774             stream_count = 2 if 'none' not in (fmt.get('vcodec'), fmt.get('acodec')) else 1
 775             lang = ISO639Utils.short2long(fmt.get('language') or '') or fmt.get('language')
 776             for i in range(stream_idx, stream_idx + stream_count):
 777                 if lang:
 778                     metadata[str(i)].setdefault('language', lang)
 779                 for name, value in metadata[str(i)].items():
 780                     yield (f'-metadata:s:{i}', f'{name}={value}')
 781             stream_idx += stream_count
 782
 783     def _get_infojson_opts(self, info, infofn):
 784         if not infofn or not os.path.exists(infofn):
 785             if self._add_infojson is not True:
 786                 return
 787             infofn = infofn or '%s.temp' % (
 788                 self._downloader.prepare_filename(info, 'infojson')
 789                 or replace_extension(self._downloader.prepare_filename(info), 'info.json', info['ext']))
 790             if not self._downloader._ensure_dir_exists(infofn):
 791                 return
 792             self.write_debug(f'Writing info-json to: {infofn}')
 793             write_json_file(self._downloader.sanitize_info(info, self.get_param('clean_infojson', True)), infofn)
 794             info['infojson_filename'] = infofn
 795
 796         old_stream, new_stream = self.get_stream_number(info['filepath'], ('tags', 'mimetype'), 'application/json')
 797         if old_stream is not None:
 798             yield ('-map', '-0:%d' % old_stream)
 799             new_stream -= 1
 800
 801         yield ('-attach', infofn,
 802                '-metadata:s:%d' % new_stream, 'mimetype=application/json')
 803
 804
 805 class FFmpegMergerPP(FFmpegPostProcessor):
 806     @PostProcessor._restrict_to(images=False)
 807     def run(self, info):
 808         filename = info['filepath']
 809         temp_filename = prepend_extension(filename, 'temp')
 810         args = ['-c', 'copy']
 811         audio_streams = 0
 812         for (i, fmt) in enumerate(info['requested_formats']):
 813             if fmt.get('acodec') != 'none':
 814                 args.extend(['-map', f'{i}:a:0'])
 815                 aac_fixup = fmt['protocol'].startswith('m3u8') and self.get_audio_codec(fmt['filepath']) == 'aac'
 816                 if aac_fixup:
 817                     args.extend([f'-bsf:a:{audio_streams}', 'aac_adtstoasc'])
 818                 audio_streams += 1
 819             if fmt.get('vcodec') != 'none':
 820                 args.extend(['-map', '%u:v:0' % (i)])
 821         self.to_screen('Merging formats into "%s"' % filename)
 822         self.run_ffmpeg_multiple_files(info['__files_to_merge'], temp_filename, args)
 823         os.rename(encodeFilename(temp_filename), encodeFilename(filename))
 824         return info['__files_to_merge'], info
 825
 826     def can_merge(self):
 827         # TODO: figure out merge-capable ffmpeg version
 828         if self.basename != 'avconv':
 829             return True
 830
 831         required_version = '10-0'
 832         if is_outdated_version(
 833                 self._versions[self.basename], required_version):
 834             warning = ('Your copy of %s is outdated and unable to properly mux separate video and audio files, '
 835                        'yt-dlp will download single file media. '
 836                        'Update %s to version %s or newer to fix this.') % (
 837                            self.basename, self.basename, required_version)
 838             self.report_warning(warning)
 839             return False
 840         return True
 841
 842
 843 class FFmpegFixupPostProcessor(FFmpegPostProcessor):
 844     def _fixup(self, msg, filename, options):
 845         temp_filename = prepend_extension(filename, 'temp')
 846
 847         self.to_screen(f'{msg} of "{filename}"')
 848         self.run_ffmpeg(filename, temp_filename, options)
 849
 850         os.replace(temp_filename, filename)
 851
 852
 853 class FFmpegFixupStretchedPP(FFmpegFixupPostProcessor):
 854     @PostProcessor._restrict_to(images=False, audio=False)
 855     def run(self, info):
 856         stretched_ratio = info.get('stretched_ratio')
 857         if stretched_ratio not in (None, 1):
 858             self._fixup('Fixing aspect ratio', info['filepath'], [
 859                 *self.stream_copy_opts(), '-aspect', '%f' % stretched_ratio])
 860         return [], info
 861
 862
 863 class FFmpegFixupM4aPP(FFmpegFixupPostProcessor):
 864     @PostProcessor._restrict_to(images=False, video=False)
 865     def run(self, info):
 866         if info.get('container') == 'm4a_dash':
 867             self._fixup('Correcting container', info['filepath'], [*self.stream_copy_opts(), '-f', 'mp4'])
 868         return [], info
 869
 870
 871 class FFmpegFixupM3u8PP(FFmpegFixupPostProcessor):
 872     def _needs_fixup(self, info):
 873         yield info['ext'] in ('mp4', 'm4a')
 874         yield info['protocol'].startswith('m3u8')
 875         try:
 876             metadata = self.get_metadata_object(info['filepath'])
 877         except PostProcessingError as e:
 878             self.report_warning(f'Unable to extract metadata: {e.msg}')
 879             yield True
 880         else:
 881             yield traverse_obj(metadata, ('format', 'format_name'), casesense=False) == 'mpegts'
 882
 883     @PostProcessor._restrict_to(images=False)
 884     def run(self, info):
 885         if all(self._needs_fixup(info)):
 886             self._fixup('Fixing MPEG-TS in MP4 container', info['filepath'], [
 887                 *self.stream_copy_opts(), '-f', 'mp4', '-bsf:a', 'aac_adtstoasc'])
 888         return [], info
 889
 890
 891 class FFmpegFixupTimestampPP(FFmpegFixupPostProcessor):
 892
 893     def __init__(self, downloader=None, trim=0.001):
 894         # "trim" should be used when the video contains unintended packets
 895         super(FFmpegFixupTimestampPP, self).__init__(downloader)
 896         assert isinstance(trim, (int, float))
 897         self.trim = str(trim)
 898
 899     @PostProcessor._restrict_to(images=False)
 900     def run(self, info):
 901         if not self._features.get('setts'):
 902             self.report_warning(
 903                 'A re-encode is needed to fix timestamps in older versions of ffmpeg. '
 904                 'Please install ffmpeg 4.4 or later to fixup without re-encoding')
 905             opts = ['-vf', 'setpts=PTS-STARTPTS']
 906         else:
 907             opts = ['-c', 'copy', '-bsf', 'setts=ts=TS-STARTPTS']
 908         self._fixup('Fixing frame timestamp', info['filepath'], opts + [*self.stream_copy_opts(False), '-ss', self.trim])
 909         return [], info
 910
 911
 912 class FFmpegCopyStreamPP(FFmpegFixupPostProcessor):
 913     MESSAGE = 'Copying stream'
 914
 915     @PostProcessor._restrict_to(images=False)
 916     def run(self, info):
 917         self._fixup(self.MESSAGE, info['filepath'], self.stream_copy_opts())
 918         return [], info
 919
 920
 921 class FFmpegFixupDurationPP(FFmpegCopyStreamPP):
 922     MESSAGE = 'Fixing video duration'
 923
 924
 925 class FFmpegFixupDuplicateMoovPP(FFmpegCopyStreamPP):
 926     MESSAGE = 'Fixing duplicate MOOV atoms'
 927
 928
 929 class FFmpegSubtitlesConvertorPP(FFmpegPostProcessor):
 930     SUPPORTED_EXTS = ('srt', 'vtt', 'ass', 'lrc')
 931
 932     def __init__(self, downloader=None, format=None):
 933         super(FFmpegSubtitlesConvertorPP, self).__init__(downloader)
 934         self.format = format
 935
 936     def run(self, info):
 937         subs = info.get('requested_subtitles')
 938         new_ext = self.format
 939         new_format = new_ext
 940         if new_format == 'vtt':
 941             new_format = 'webvtt'
 942         if subs is None:
 943             self.to_screen('There aren\'t any subtitles to convert')
 944             return [], info
 945         self.to_screen('Converting subtitles')
 946         sub_filenames = []
 947         for lang, sub in subs.items():
 948             if not os.path.exists(sub.get('filepath', '')):
 949                 self.report_warning(f'Skipping embedding {lang} subtitle because the file is missing')
 950                 continue
 951             ext = sub['ext']
 952             if ext == new_ext:
 953                 self.to_screen('Subtitle file for %s is already in the requested format' % new_ext)
 954                 continue
 955             elif ext == 'json':
 956                 self.to_screen(
 957                     'You have requested to convert json subtitles into another format, '
 958                     'which is currently not possible')
 959                 continue
 960             old_file = sub['filepath']
 961             sub_filenames.append(old_file)
 962             new_file = replace_extension(old_file, new_ext)
 963
 964             if ext in ('dfxp', 'ttml', 'tt'):
 965                 self.report_warning(
 966                     'You have requested to convert dfxp (TTML) subtitles into another format, '
 967                     'which results in style information loss')
 968
 969                 dfxp_file = old_file
 970                 srt_file = replace_extension(old_file, 'srt')
 971
 972                 with open(dfxp_file, 'rb') as f:
 973                     srt_data = dfxp2srt(f.read())
 974
 975                 with io.open(srt_file, 'wt', encoding='utf-8') as f:
 976                     f.write(srt_data)
 977                 old_file = srt_file
 978
 979                 subs[lang] = {
 980                     'ext': 'srt',
 981                     'data': srt_data,
 982                     'filepath': srt_file,
 983                 }
 984
 985                 if new_ext == 'srt':
 986                     continue
 987                 else:
 988                     sub_filenames.append(srt_file)
 989
 990             self.run_ffmpeg(old_file, new_file, ['-f', new_format])
 991
 992             with io.open(new_file, 'rt', encoding='utf-8') as f:
 993                 subs[lang] = {
 994                     'ext': new_ext,
 995                     'data': f.read(),
 996                     'filepath': new_file,
 997                 }
 998
 999             info['__files_to_move'][new_file] = replace_extension(
1000                 info['__files_to_move'][sub['filepath']], new_ext)
1001
1002         return sub_filenames, info
1003
1004
1005 class FFmpegSplitChaptersPP(FFmpegPostProcessor):
1006     def __init__(self, downloader, force_keyframes=False):
1007         FFmpegPostProcessor.__init__(self, downloader)
1008         self._force_keyframes = force_keyframes
1009
1010     def _prepare_filename(self, number, chapter, info):
1011         info = info.copy()
1012         info.update({
1013             'section_number': number,
1014             'section_title': chapter.get('title'),
1015             'section_start': chapter.get('start_time'),
1016             'section_end': chapter.get('end_time'),
1017         })
1018         return self._downloader.prepare_filename(info, 'chapter')
1019
1020     def _ffmpeg_args_for_chapter(self, number, chapter, info):
1021         destination = self._prepare_filename(number, chapter, info)
1022         if not self._downloader._ensure_dir_exists(encodeFilename(destination)):
1023             return
1024
1025         chapter['filepath'] = destination
1026         self.to_screen('Chapter %03d; Destination: %s' % (number, destination))
1027         return (
1028             destination,
1029             ['-ss', compat_str(chapter['start_time']),
1030              '-t', compat_str(chapter['end_time'] - chapter['start_time'])])
1031
1032     @PostProcessor._restrict_to(images=False)
1033     def run(self, info):
1034         chapters = info.get('chapters') or []
1035         if not chapters:
1036             self.to_screen('Chapter information is unavailable')
1037             return [], info
1038
1039         in_file = info['filepath']
1040         if self._force_keyframes and len(chapters) > 1:
1041             in_file = self.force_keyframes(in_file, (c['start_time'] for c in chapters))
1042         self.to_screen('Splitting video by chapters; %d chapters found' % len(chapters))
1043         for idx, chapter in enumerate(chapters):
1044             destination, opts = self._ffmpeg_args_for_chapter(idx + 1, chapter, info)
1045             self.real_run_ffmpeg([(in_file, opts)], [(destination, self.stream_copy_opts())])
1046         if in_file != info['filepath']:
1047             os.remove(in_file)
1048         return [], info
1049
1050
1051 class FFmpegThumbnailsConvertorPP(FFmpegPostProcessor):
1052     SUPPORTED_EXTS = ('jpg', 'png', 'webp')
1053
1054     def __init__(self, downloader=None, format=None):
1055         super(FFmpegThumbnailsConvertorPP, self).__init__(downloader)
1056         self.format = format
1057
1058     @staticmethod
1059     def is_webp(path):
1060         with open(encodeFilename(path), 'rb') as f:
1061             b = f.read(12)
1062         return b[0:4] == b'RIFF' and b[8:] == b'WEBP'
1063
1064     def fixup_webp(self, info, idx=-1):
1065         thumbnail_filename = info['thumbnails'][idx]['filepath']
1066         _, thumbnail_ext = os.path.splitext(thumbnail_filename)
1067         if thumbnail_ext:
1068             thumbnail_ext = thumbnail_ext[1:].lower()
1069             if thumbnail_ext != 'webp' and self.is_webp(thumbnail_filename):
1070                 self.to_screen('Correcting thumbnail "%s" extension to webp' % thumbnail_filename)
1071                 webp_filename = replace_extension(thumbnail_filename, 'webp')
1072                 os.replace(thumbnail_filename, webp_filename)
1073                 info['thumbnails'][idx]['filepath'] = webp_filename
1074                 info['__files_to_move'][webp_filename] = replace_extension(
1075                     info['__files_to_move'].pop(thumbnail_filename), 'webp')
1076
1077     @staticmethod
1078     def _options(target_ext):
1079         if target_ext == 'jpg':
1080             return ['-bsf:v', 'mjpeg2jpeg']
1081         return []
1082
1083     def convert_thumbnail(self, thumbnail_filename, target_ext):
1084         thumbnail_conv_filename = replace_extension(thumbnail_filename, target_ext)
1085
1086         self.to_screen('Converting thumbnail "%s" to %s' % (thumbnail_filename, target_ext))
1087         self.real_run_ffmpeg(
1088             [(thumbnail_filename, ['-f', 'image2', '-pattern_type', 'none'])],
1089             [(thumbnail_conv_filename.replace('%', '%%'), self._options(target_ext))])
1090         return thumbnail_conv_filename
1091
1092     def run(self, info):
1093         files_to_delete = []
1094         has_thumbnail = False
1095
1096         for idx, thumbnail_dict in enumerate(info.get('thumbnails') or []):
1097             original_thumbnail = thumbnail_dict.get('filepath')
1098             if not original_thumbnail:
1099                 continue
1100             has_thumbnail = True
1101             self.fixup_webp(info, idx)
1102             _, thumbnail_ext = os.path.splitext(original_thumbnail)
1103             if thumbnail_ext:
1104                 thumbnail_ext = thumbnail_ext[1:].lower()
1105             if thumbnail_ext == 'jpeg':
1106                 thumbnail_ext = 'jpg'
1107             if thumbnail_ext == self.format:
1108                 self.to_screen('Thumbnail "%s" is already in the requested format' % original_thumbnail)
1109                 continue
1110             thumbnail_dict['filepath'] = self.convert_thumbnail(original_thumbnail, self.format)
1111             files_to_delete.append(original_thumbnail)
1112             info['__files_to_move'][thumbnail_dict['filepath']] = replace_extension(
1113                 info['__files_to_move'][original_thumbnail], self.format)
1114
1115         if not has_thumbnail:
1116             self.to_screen('There aren\'t any thumbnails to convert')
1117         return files_to_delete, info
1118
1119
1120 class FFmpegConcatPP(FFmpegPostProcessor):
1121     def __init__(self, downloader, only_multi_video=False):
1122         self._only_multi_video = only_multi_video
1123         super().__init__(downloader)
1124
1125     def concat_files(self, in_files, out_file):
1126         if not self._downloader._ensure_dir_exists(out_file):
1127             return
1128         if len(in_files) == 1:
1129             if os.path.realpath(in_files[0]) != os.path.realpath(out_file):
1130                 self.to_screen(f'Moving "{in_files[0]}" to "{out_file}"')
1131             os.replace(in_files[0], out_file)
1132             return []
1133
1134         codecs = [traverse_obj(self.get_metadata_object(file), ('streams', ..., 'codec_name')) for file in in_files]
1135         if len(set(map(tuple, codecs))) > 1:
1136             raise PostProcessingError(
1137                 'The files have different streams/codecs and cannot be concatenated. '
1138                 'Either select different formats or --recode-video them to a common format')
1139
1140         self.to_screen(f'Concatenating {len(in_files)} files; Destination: {out_file}')
1141         super().concat_files(in_files, out_file)
1142         return in_files
1143
1144     @PostProcessor._restrict_to(images=False, simulated=False)
1145     def run(self, info):
1146         entries = info.get('entries') or []
1147         if not any(entries) or (self._only_multi_video and info['_type'] != 'multi_video'):
1148             return [], info
1149         elif any(len(entry) > 1 for entry in traverse_obj(entries, (..., 'requested_downloads')) or []):
1150             raise PostProcessingError('Concatenation is not supported when downloading multiple separate formats')
1151
1152         in_files = traverse_obj(entries, (..., 'requested_downloads', 0, 'filepath')) or []
1153         if len(in_files) < len(entries):
1154             raise PostProcessingError('Aborting concatenation because some downloads failed')
1155
1156         ie_copy = self._downloader._playlist_infodict(info)
1157         exts = traverse_obj(entries, (..., 'requested_downloads', 0, 'ext'), (..., 'ext'))
1158         ie_copy['ext'] = exts[0] if len(set(exts)) == 1 else 'mkv'
1159         out_file = self._downloader.prepare_filename(ie_copy, 'pl_video')
1160
1161         files_to_delete = self.concat_files(in_files, out_file)
1162
1163         info['requested_downloads'] = [{
1164             'filepath': out_file,
1165             'ext': ie_copy['ext'],
1166         }]
1167         return files_to_delete, info