9 from .common
import PostProcessor
10 from ..compat
import functools
, imghdr
15 _get_exe_version_output
,
34 EXT_TO_OUT_FORMATS
= {
48 # name: (ext, encoder, opts)
49 'mp3': ('mp3', 'libmp3lame', ()),
50 'aac': ('m4a', 'aac', ('-f', 'adts')),
51 'm4a': ('m4a', 'aac', ('-bsf:a', 'aac_adtstoasc')),
52 'opus': ('opus', 'libopus', ()),
53 'vorbis': ('ogg', 'libvorbis', ()),
54 'flac': ('flac', 'flac', ()),
55 'alac': ('m4a', None, ('-acodec', 'alac')),
56 'wav': ('wav', None, ('-f', 'wav')),
60 def create_mapping_re(supported
):
61 return re
.compile(r
'{0}(?:/{0})*$'.format(r
'(?:\s*\w+\s*>)?\s*(?:%s)\s*' % '|'.join(supported
)))
64 def resolve_mapping(source
, mapping
):
66 Get corresponding item from a mapping string like 'A>B/C>D/E'
67 @returns (target, error_message)
69 for pair
in mapping
.lower().split('/'):
70 kv
= pair
.split('>', 1)
71 if len(kv
) == 1 or kv
[0].strip() == source
:
72 target
= kv
[-1].strip()
74 return target
, f
'already is in target format {source}'
76 return None, f
'could not find a mapping for {source}'
79 class FFmpegPostProcessorError(PostProcessingError
):
83 class FFmpegPostProcessor(PostProcessor
):
84 def __init__(self
, downloader
=None):
85 PostProcessor
.__init
__(self
, downloader
)
86 self
._prefer
_ffmpeg
= self
.get_param('prefer_ffmpeg', True)
87 self
._paths
= self
._determine
_executables
()
90 def get_versions_and_features(downloader
=None):
91 pp
= FFmpegPostProcessor(downloader
)
92 return pp
._versions
, pp
._features
95 def get_versions(downloader
=None):
96 return FFmpegPostProcessor
.get_versions_and_features(downloader
)[0]
98 _ffmpeg_to_avconv
= {'ffmpeg': 'avconv', 'ffprobe': 'avprobe'}
100 def _determine_executables(self
):
101 programs
= [*self
._ffmpeg
_to
_avconv
.keys(), *self
._ffmpeg
_to
_avconv
.values()]
103 location
= self
.get_param('ffmpeg_location')
105 return {p: p for p in programs}
107 if not os
.path
.exists(location
):
108 self
.report_warning(f
'ffmpeg-location {location} does not exist! Continuing without ffmpeg')
110 elif os
.path
.isdir(location
):
111 dirname
, basename
= location
, None
113 basename
= os
.path
.splitext(os
.path
.basename(location
))[0]
114 basename
= next((p
for p
in programs
if basename
.startswith(p
)), 'ffmpeg')
115 dirname
= os
.path
.dirname(os
.path
.abspath(location
))
116 if basename
in self
._ffmpeg
_to
_avconv
.keys():
117 self
._prefer
_ffmpeg
= True
119 paths
= {p: os.path.join(dirname, p) for p in programs}
121 paths
[basename
] = location
124 _version_cache
, _features_cache
= {None: None}
, {}
126 def _get_ffmpeg_version(self
, prog
):
127 path
= self
._paths
.get(prog
)
128 if path
in self
._version
_cache
:
129 return self
._version
_cache
[path
], self
._features
_cache
.get(path
, {})
130 out
= _get_exe_version_output(path
, ['-bsfs'], to_screen
=self
.write_debug
)
131 ver
= detect_exe_version(out
) if out
else False
134 r
'(?:\d+:)?([0-9.]+)-[0-9]+ubuntu[0-9.]+$', # Ubuntu, see [1]
135 r
'n([0-9.]+)$', # Arch Linux
136 # 1. http://www.ducea.com/2006/06/17/ubuntu-package-version-naming-explanation/
139 mobj
= re
.match(regex
, ver
)
142 self
._version
_cache
[path
] = ver
143 if prog
!= 'ffmpeg' or not out
:
146 mobj
= re
.search(r
'(?m)^\s+libavformat\s+(?:[0-9. ]+)\s+/\s+(?P<runtime>[0-9. ]+)', out
)
147 lavf_runtime_version
= mobj
.group('runtime').replace(' ', '') if mobj
else None
148 self
._features
_cache
[path
] = features
= {
149 'fdk': '--enable-libfdk-aac' in out
,
150 'setts': 'setts' in out
.splitlines(),
151 'needs_adtstoasc': is_outdated_version(lavf_runtime_version
, '57.56.100', False),
157 return filter_dict({self.basename: self._version, self.probe_basename: self._probe_version}
)
159 @functools.cached_property
161 self
._version
# run property
164 @functools.cached_property
165 def probe_basename(self
):
166 self
._probe
_version
# run property
167 return self
.probe_basename
169 def _get_version(self
, kind
):
170 executables
= (kind
, self
._ffmpeg
_to
_avconv
[kind
])
171 if not self
._prefer
_ffmpeg
:
172 executables
= reversed(executables
)
173 basename
, version
, features
= next(filter(
174 lambda x
: x
[1], ((p
, *self
._get
_ffmpeg
_version
(p
)) for p
in executables
)), (None, None, {}))
176 self
.basename
, self
._features
= basename
, features
178 self
.probe_basename
= basename
179 if basename
== self
._ffmpeg
_to
_avconv
[kind
]:
180 self
.deprecation_warning(
181 f
'Support for {self._ffmpeg_to_avconv[kind]} is deprecated and may be removed in a future version. Use {kind} instead')
184 @functools.cached_property
186 return self
._get
_version
('ffmpeg')
188 @functools.cached_property
189 def _probe_version(self
):
190 return self
._get
_version
('ffprobe')
194 return self
.basename
is not None
197 def executable(self
):
198 return self
._paths
.get(self
.basename
)
201 def probe_available(self
):
202 return self
.probe_basename
is not None
205 def probe_executable(self
):
206 return self
._paths
.get(self
.probe_basename
)
209 def stream_copy_opts(copy
=True, *, ext
=None):
210 yield from ('-map', '0')
211 # Don't copy Apple TV chapters track, bin_data
212 # See https://github.com/yt-dlp/yt-dlp/issues/2, #19042, #19024, https://trac.ffmpeg.org/ticket/6016
213 yield from ('-dn', '-ignore_unknown')
215 yield from ('-c', 'copy')
216 if ext
in ('mp4', 'mov', 'm4a'):
217 yield from ('-c:s', 'mov_text')
219 def check_version(self
):
220 if not self
.available
:
221 raise FFmpegPostProcessorError('ffmpeg not found. Please install or provide the path using --ffmpeg-location')
223 required_version
= '10-0' if self
.basename
== 'avconv' else '1.0'
224 if is_outdated_version(self
._version
, required_version
):
225 self
.report_warning(f
'Your copy of {self.basename} is outdated, update {self.basename} '
226 f
'to version {required_version} or newer if you encounter any errors')
228 def get_audio_codec(self
, path
):
229 if not self
.probe_available
and not self
.available
:
230 raise PostProcessingError('ffprobe and ffmpeg not found. Please install or provide the path using --ffmpeg-location')
232 if self
.probe_available
:
234 encodeFilename(self
.probe_executable
, True),
235 encodeArgument('-show_streams')]
238 encodeFilename(self
.executable
, True),
239 encodeArgument('-i')]
240 cmd
.append(encodeFilename(self
._ffmpeg
_filename
_argument
(path
), True))
241 self
.write_debug(f
'{self.basename} command line: {shell_quote(cmd)}')
242 handle
= Popen(cmd
, stdin
=subprocess
.PIPE
, stdout
=subprocess
.PIPE
, stderr
=subprocess
.PIPE
)
243 stdout_data
, stderr_data
= handle
.communicate_or_kill()
244 expected_ret
= 0 if self
.probe_available
else 1
245 if handle
.wait() != expected_ret
:
249 output
= (stdout_data
if self
.probe_available
else stderr_data
).decode('ascii', 'ignore')
250 if self
.probe_available
:
252 for line
in output
.split('\n'):
253 if line
.startswith('codec_name='):
254 audio_codec
= line
.split('=')[1].strip()
255 elif line
.strip() == 'codec_type=audio' and audio_codec
is not None:
258 # Stream #FILE_INDEX:STREAM_INDEX[STREAM_ID](LANGUAGE): CODEC_TYPE: CODEC_NAME
260 r
'Stream\s*#\d+:\d+(?:\[0x[0-9a-f]+\])?(?:\([a-z]{3}\))?:\s*Audio:\s*([0-9a-z]+)',
266 def get_metadata_object(self
, path
, opts
=[]):
267 if self
.probe_basename
!= 'ffprobe':
268 if self
.probe_available
:
269 self
.report_warning('Only ffprobe is supported for metadata extraction')
270 raise PostProcessingError('ffprobe not found. Please install or provide the path using --ffmpeg-location')
274 encodeFilename(self
.probe_executable
, True),
275 encodeArgument('-hide_banner'),
276 encodeArgument('-show_format'),
277 encodeArgument('-show_streams'),
278 encodeArgument('-print_format'),
279 encodeArgument('json'),
283 cmd
.append(encodeFilename(self
._ffmpeg
_filename
_argument
(path
), True))
284 self
.write_debug('ffprobe command line: %s' % shell_quote(cmd
))
285 p
= Popen(cmd
, stdout
=subprocess
.PIPE
, stderr
=subprocess
.PIPE
, stdin
=subprocess
.PIPE
)
286 stdout
, stderr
= p
.communicate()
287 return json
.loads(stdout
.decode('utf-8', 'replace'))
289 def get_stream_number(self
, path
, keys
, value
):
290 streams
= self
.get_metadata_object(path
)['streams']
292 (i
for i
, stream
in enumerate(streams
) if traverse_obj(stream
, keys
, casesense
=False) == value
),
294 return num
, len(streams
)
296 def _get_real_video_duration(self
, filepath
, fatal
=True):
298 duration
= float_or_none(
299 traverse_obj(self
.get_metadata_object(filepath
), ('format', 'duration')))
301 raise PostProcessingError('ffprobe returned empty duration')
303 except PostProcessingError
as e
:
305 raise PostProcessingError(f
'Unable to determine video duration: {e.msg}')
307 def _duration_mismatch(self
, d1
, d2
, tolerance
=2):
310 # The duration is often only known to nearest second. So there can be <1sec disparity natually.
311 # Further excuse an additional <1sec difference.
312 return abs(d1
- d2
) > tolerance
314 def run_ffmpeg_multiple_files(self
, input_paths
, out_path
, opts
, **kwargs
):
315 return self
.real_run_ffmpeg(
316 [(path
, []) for path
in input_paths
],
317 [(out_path
, opts
)], **kwargs
)
319 def real_run_ffmpeg(self
, input_path_opts
, output_path_opts
, *, expected_retcodes
=(0,)):
323 os
.stat(encodeFilename(path
)).st_mtime
for path
, _
in input_path_opts
if path
)
325 cmd
= [encodeFilename(self
.executable
, True), encodeArgument('-y')]
326 # avconv does not have repeat option
327 if self
.basename
== 'ffmpeg':
328 cmd
+= [encodeArgument('-loglevel'), encodeArgument('repeat+info')]
330 def make_args(file, args
, name
, number
):
331 keys
= ['_%s%d' % (name
, number
), '_%s' % name
]
333 args
+= ['-movflags', '+faststart']
336 args
+= self
._configuration
_args
(self
.basename
, keys
)
340 [encodeArgument(arg
) for arg
in args
]
341 + [encodeFilename(self
._ffmpeg
_filename
_argument
(file), True)])
343 for arg_type
, path_opts
in (('i', input_path_opts
), ('o', output_path_opts
)):
344 cmd
+= itertools
.chain
.from_iterable(
345 make_args(path
, list(opts
), arg_type
, i
+ 1)
346 for i
, (path
, opts
) in enumerate(path_opts
) if path
)
348 self
.write_debug('ffmpeg command line: %s' % shell_quote(cmd
))
349 p
= Popen(cmd
, stdout
=subprocess
.PIPE
, stderr
=subprocess
.PIPE
, stdin
=subprocess
.PIPE
)
350 stdout
, stderr
= p
.communicate_or_kill()
351 if p
.returncode
not in variadic(expected_retcodes
):
352 stderr
= stderr
.decode('utf-8', 'replace').strip()
353 self
.write_debug(stderr
)
354 raise FFmpegPostProcessorError(stderr
.split('\n')[-1])
355 for out_path
, _
in output_path_opts
:
357 self
.try_utime(out_path
, oldest_mtime
, oldest_mtime
)
358 return stderr
.decode('utf-8', 'replace')
360 def run_ffmpeg(self
, path
, out_path
, opts
, **kwargs
):
361 return self
.run_ffmpeg_multiple_files([path
], out_path
, opts
, **kwargs
)
364 def _ffmpeg_filename_argument(fn
):
365 # Always use 'file:' because the filename may contain ':' (ffmpeg
366 # interprets that as a protocol) or can start with '-' (-- is broken in
367 # ffmpeg, see https://ffmpeg.org/trac/ffmpeg/ticket/2127 for details)
368 # Also leave '-' intact in order not to break streaming to stdout.
369 if fn
.startswith(('http://', 'https://')):
371 return 'file:' + fn
if fn
!= '-' else fn
374 def _quote_for_ffmpeg(string
):
375 # See https://ffmpeg.org/ffmpeg-utils.html#toc-Quoting-and-escaping
376 # A sequence of '' produces '\'''\'';
377 # final replace removes the empty '' between \' \'.
378 string
= string
.replace("'", r
"'\''").replace("'''", "'")
379 # Handle potential ' at string boundaries.
380 string
= string
[1:] if string
[0] == "'" else "'" + string
381 return string
[:-1] if string
[-1] == "'" else string
+ "'"
383 def force_keyframes(self
, filename
, timestamps
):
384 timestamps
= orderedSet(timestamps
)
385 if timestamps
[0] == 0:
386 timestamps
= timestamps
[1:]
387 keyframe_file
= prepend_extension(filename
, 'keyframes.temp')
388 self
.to_screen(f
'Re-encoding "{filename}" with appropriate keyframes')
389 self
.run_ffmpeg(filename
, keyframe_file
, [
390 *self
.stream_copy_opts(False, ext
=determine_ext(filename
)),
391 '-force_key_frames', ','.join(f
'{t:.6f}' for t
in timestamps
)])
394 def concat_files(self
, in_files
, out_file
, concat_opts
=None):
396 Use concat demuxer to concatenate multiple files having identical streams.
398 Only inpoint, outpoint, and duration concat options are supported.
399 See https://ffmpeg.org/ffmpeg-formats.html#concat-1 for details
401 concat_file
= f
'{out_file}.concat'
402 self
.write_debug(f
'Writing concat spec to {concat_file}')
403 with open(concat_file
, 'wt', encoding
='utf-8') as f
:
404 f
.writelines(self
._concat
_spec
(in_files
, concat_opts
))
406 out_flags
= list(self
.stream_copy_opts(ext
=determine_ext(out_file
)))
408 self
.real_run_ffmpeg(
409 [(concat_file
, ['-hide_banner', '-nostdin', '-f', 'concat', '-safe', '0'])],
410 [(out_file
, out_flags
)])
411 self
._delete
_downloaded
_files
(concat_file
)
414 def _concat_spec(cls
, in_files
, concat_opts
=None):
415 if concat_opts
is None:
416 concat_opts
= [{}] * len(in_files
)
417 yield 'ffconcat version 1.0\n'
418 for file, opts
in zip(in_files
, concat_opts
):
419 yield f
'file {cls._quote_for_ffmpeg(cls._ffmpeg_filename_argument(file))}\n'
420 # Iterate explicitly to yield the following directives in order, ignoring the rest.
421 for directive
in 'inpoint', 'outpoint', 'duration':
422 if directive
in opts
:
423 yield f
'{directive} {opts[directive]}\n'
426 class FFmpegExtractAudioPP(FFmpegPostProcessor
):
427 COMMON_AUDIO_EXTS
= ('wav', 'flac', 'm4a', 'aiff', 'mp3', 'ogg', 'mka', 'opus', 'wma')
428 SUPPORTED_EXTS
= tuple(ACODECS
.keys())
429 FORMAT_RE
= create_mapping_re(('best', *SUPPORTED_EXTS
))
431 def __init__(self
, downloader
=None, preferredcodec
=None, preferredquality
=None, nopostoverwrites
=False):
432 FFmpegPostProcessor
.__init
__(self
, downloader
)
433 self
.mapping
= preferredcodec
or 'best'
434 self
._preferredquality
= float_or_none(preferredquality
)
435 self
._nopostoverwrites
= nopostoverwrites
437 def _quality_args(self
, codec
):
438 if self
._preferredquality
is None:
440 elif self
._preferredquality
> 10:
441 return ['-b:a', f
'{self._preferredquality}k']
444 'libmp3lame': (10, 0),
445 'libvorbis': (0, 10),
446 # FFmpeg's AAC encoder does not have an upper limit for the value of -q:a.
447 # Experimentally, with values over 4, bitrate changes were minimal or non-existent
449 'libfdk_aac': (1, 5),
454 q
= limits
[1] + (limits
[0] - limits
[1]) * (self
._preferredquality
/ 10)
455 if codec
== 'libfdk_aac':
456 return ['-vbr', f
'{int(q)}']
457 return ['-q:a', f
'{q}']
459 def run_ffmpeg(self
, path
, out_path
, codec
, more_opts
):
463 acodec_opts
= ['-acodec', codec
]
464 opts
= ['-vn'] + acodec_opts
+ more_opts
466 FFmpegPostProcessor
.run_ffmpeg(self
, path
, out_path
, opts
)
467 except FFmpegPostProcessorError
as err
:
468 raise PostProcessingError(f
'audio conversion failed: {err.msg}')
470 @PostProcessor._restrict_to(images
=False)
471 def run(self
, information
):
472 orig_path
= path
= information
['filepath']
473 target_format
, _skip_msg
= resolve_mapping(information
['ext'], self
.mapping
)
474 if target_format
== 'best' and information
['ext'] in self
.COMMON_AUDIO_EXTS
:
475 target_format
, _skip_msg
= None, 'the file is already in a common audio format'
476 if not target_format
:
477 self
.to_screen(f
'Not converting audio {orig_path}; {_skip_msg}')
478 return [], information
480 filecodec
= self
.get_audio_codec(path
)
481 if filecodec
is None:
482 raise PostProcessingError('WARNING: unable to obtain file audio codec with ffprobe')
484 if filecodec
== 'aac' and target_format
in ('m4a', 'best'):
485 # Lossless, but in another container
486 extension
, _
, more_opts
, acodec
= *ACODECS
['m4a'], 'copy'
487 elif target_format
== 'best' or target_format
== filecodec
:
488 # Lossless if possible
490 extension
, _
, more_opts
, acodec
= *ACODECS
[filecodec
], 'copy'
492 extension
, acodec
, more_opts
= ACODECS
['mp3']
494 # We convert the audio (lossy if codec is lossy)
495 extension
, acodec
, more_opts
= ACODECS
[target_format
]
496 if acodec
== 'aac' and self
._features
.get('fdk'):
497 acodec
, more_opts
= 'libfdk_aac', []
499 more_opts
= list(more_opts
)
501 more_opts
= self
._quality
_args
(acodec
)
503 # not os.path.splitext, since the latter does not work on unicode in all setups
504 temp_path
= new_path
= f
'{path.rpartition(".")[0]}.{extension}'
508 self
.to_screen(f
'Not converting audio {orig_path}; file is already in target format {target_format}')
509 return [], information
510 orig_path
= prepend_extension(path
, 'orig')
511 temp_path
= prepend_extension(path
, 'temp')
512 if (self
._nopostoverwrites
and os
.path
.exists(encodeFilename(new_path
))
513 and os
.path
.exists(encodeFilename(orig_path
))):
514 self
.to_screen('Post-process file %s exists, skipping' % new_path
)
515 return [], information
517 self
.to_screen(f
'Destination: {new_path}')
518 self
.run_ffmpeg(path
, temp_path
, acodec
, more_opts
)
520 os
.replace(path
, orig_path
)
521 os
.replace(temp_path
, new_path
)
522 information
['filepath'] = new_path
523 information
['ext'] = extension
525 # Try to update the date time for extracted audio file.
526 if information
.get('filetime') is not None:
528 new_path
, time
.time(), information
['filetime'], errnote
='Cannot update utime of audio file')
530 return [orig_path
], information
533 class FFmpegVideoConvertorPP(FFmpegPostProcessor
):
534 SUPPORTED_EXTS
= ('mp4', 'mkv', 'flv', 'webm', 'mov', 'avi', 'mka', 'ogg', *FFmpegExtractAudioPP
.SUPPORTED_EXTS
)
535 FORMAT_RE
= create_mapping_re(SUPPORTED_EXTS
)
536 _ACTION
= 'converting'
538 def __init__(self
, downloader
=None, preferedformat
=None):
539 super().__init
__(downloader
)
540 self
.mapping
= preferedformat
543 def _options(target_ext
):
544 yield from FFmpegPostProcessor
.stream_copy_opts(False)
545 if target_ext
== 'avi':
546 yield from ('-c:v', 'libxvid', '-vtag', 'XVID')
548 @PostProcessor._restrict_to(images
=False)
550 filename
, source_ext
= info
['filepath'], info
['ext'].lower()
551 target_ext
, _skip_msg
= resolve_mapping(source_ext
, self
.mapping
)
553 self
.to_screen(f
'Not {self._ACTION} media file "{filename}"; {_skip_msg}')
556 outpath
= replace_extension(filename
, target_ext
, source_ext
)
557 self
.to_screen(f
'{self._ACTION.title()} video from {source_ext} to {target_ext}; Destination: {outpath}')
558 self
.run_ffmpeg(filename
, outpath
, self
._options
(target_ext
))
560 info
['filepath'] = outpath
561 info
['format'] = info
['ext'] = target_ext
562 return [filename
], info
565 class FFmpegVideoRemuxerPP(FFmpegVideoConvertorPP
):
569 def _options(target_ext
):
570 return FFmpegPostProcessor
.stream_copy_opts()
573 class FFmpegEmbedSubtitlePP(FFmpegPostProcessor
):
574 SUPPORTED_EXTS
= ('mp4', 'mov', 'm4a', 'webm', 'mkv', 'mka')
576 def __init__(self
, downloader
=None, already_have_subtitle
=False):
577 super().__init
__(downloader
)
578 self
._already
_have
_subtitle
= already_have_subtitle
580 @PostProcessor._restrict_to(images
=False)
582 if info
['ext'] not in self
.SUPPORTED_EXTS
:
583 self
.to_screen(f
'Subtitles can only be embedded in {", ".join(self.SUPPORTED_EXTS)} files')
585 subtitles
= info
.get('requested_subtitles')
587 self
.to_screen('There aren\'t any subtitles to embed')
590 filename
= info
['filepath']
592 # Disabled temporarily. There needs to be a way to overide this
593 # in case of duration actually mismatching in extractor
594 # See: https://github.com/yt-dlp/yt-dlp/issues/1870, https://github.com/yt-dlp/yt-dlp/issues/1385
596 if info.get('duration') and not info.get('__real_download') and self._duration_mismatch(
597 self._get_real_video_duration(filename, False), info['duration']):
598 self.to_screen(f'Skipping {self.pp_key()} since the real and expected durations mismatch')
603 sub_langs
, sub_names
, sub_filenames
= [], [], []
604 webm_vtt_warn
= False
607 for lang
, sub_info
in subtitles
.items():
608 if not os
.path
.exists(sub_info
.get('filepath', '')):
609 self
.report_warning(f
'Skipping embedding {lang} subtitle because the file is missing')
611 sub_ext
= sub_info
['ext']
612 if sub_ext
== 'json':
613 self
.report_warning('JSON subtitles cannot be embedded')
614 elif ext
!= 'webm' or ext
== 'webm' and sub_ext
== 'vtt':
615 sub_langs
.append(lang
)
616 sub_names
.append(sub_info
.get('name'))
617 sub_filenames
.append(sub_info
['filepath'])
619 if not webm_vtt_warn
and ext
== 'webm' and sub_ext
!= 'vtt':
621 self
.report_warning('Only WebVTT subtitles can be embedded in webm files')
622 if not mp4_ass_warn
and ext
== 'mp4' and sub_ext
== 'ass':
624 self
.report_warning('ASS subtitles cannot be properly embedded in mp4 files; expect issues')
629 input_files
= [filename
] + sub_filenames
632 *self
.stream_copy_opts(ext
=info
['ext']),
633 # Don't copy the existing subtitles, we may be running the
634 # postprocessor a second time
637 for i
, (lang
, name
) in enumerate(zip(sub_langs
, sub_names
)):
638 opts
.extend(['-map', '%d:0' % (i
+ 1)])
639 lang_code
= ISO639Utils
.short2long(lang
) or lang
640 opts
.extend(['-metadata:s:s:%d' % i
, 'language=%s' % lang_code
])
642 opts
.extend(['-metadata:s:s:%d' % i
, 'handler_name=%s' % name
,
643 '-metadata:s:s:%d' % i
, 'title=%s' % name
])
645 temp_filename
= prepend_extension(filename
, 'temp')
646 self
.to_screen('Embedding subtitles in "%s"' % filename
)
647 self
.run_ffmpeg_multiple_files(input_files
, temp_filename
, opts
)
648 os
.replace(temp_filename
, filename
)
650 files_to_delete
= [] if self
._already
_have
_subtitle
else sub_filenames
651 return files_to_delete
, info
654 class FFmpegMetadataPP(FFmpegPostProcessor
):
656 def __init__(self
, downloader
, add_metadata
=True, add_chapters
=True, add_infojson
='if_exists'):
657 FFmpegPostProcessor
.__init
__(self
, downloader
)
658 self
._add
_metadata
= add_metadata
659 self
._add
_chapters
= add_chapters
660 self
._add
_infojson
= add_infojson
663 def _options(target_ext
):
664 audio_only
= target_ext
== 'm4a'
665 yield from FFmpegPostProcessor
.stream_copy_opts(not audio_only
)
667 yield from ('-vn', '-acodec', 'copy')
669 @PostProcessor._restrict_to(images
=False)
671 filename
, metadata_filename
= info
['filepath'], None
672 files_to_delete
, options
= [], []
673 if self
._add
_chapters
and info
.get('chapters'):
674 metadata_filename
= replace_extension(filename
, 'meta')
675 options
.extend(self
._get
_chapter
_opts
(info
['chapters'], metadata_filename
))
676 files_to_delete
.append(metadata_filename
)
677 if self
._add
_metadata
:
678 options
.extend(self
._get
_metadata
_opts
(info
))
680 if self
._add
_infojson
:
681 if info
['ext'] in ('mkv', 'mka'):
682 infojson_filename
= info
.get('infojson_filename')
683 options
.extend(self
._get
_infojson
_opts
(info
, infojson_filename
))
684 if not infojson_filename
:
685 files_to_delete
.append(info
.get('infojson_filename'))
686 elif self
._add
_infojson
is True:
687 self
.to_screen('The info-json can only be attached to mkv/mka files')
690 self
.to_screen('There isn\'t any metadata to add')
693 temp_filename
= prepend_extension(filename
, 'temp')
694 self
.to_screen('Adding metadata to "%s"' % filename
)
695 self
.run_ffmpeg_multiple_files(
696 (filename
, metadata_filename
), temp_filename
,
697 itertools
.chain(self
._options
(info
['ext']), *options
))
698 self
._delete
_downloaded
_files
(*files_to_delete
)
699 os
.replace(temp_filename
, filename
)
703 def _get_chapter_opts(chapters
, metadata_filename
):
704 with open(metadata_filename
, 'wt', encoding
='utf-8') as f
:
705 def ffmpeg_escape(text
):
706 return re
.sub(r
'([\\=;#\n])', r
'\\\1', text
)
708 metadata_file_content
= ';FFMETADATA1\n'
709 for chapter
in chapters
:
710 metadata_file_content
+= '[CHAPTER]\nTIMEBASE=1/1000\n'
711 metadata_file_content
+= 'START=%d\n' % (chapter
['start_time'] * 1000)
712 metadata_file_content
+= 'END=%d\n' % (chapter
['end_time'] * 1000)
713 chapter_title
= chapter
.get('title')
715 metadata_file_content
+= 'title=%s\n' % ffmpeg_escape(chapter_title
)
716 f
.write(metadata_file_content
)
717 yield ('-map_metadata', '1')
719 def _get_metadata_opts(self
, info
):
721 metadata
= collections
.defaultdict(dict)
723 def add(meta_list
, info_list
=None):
725 str(info
[key
]) for key
in [f
'{meta_prefix}_'] + list(variadic(info_list
or meta_list
))
726 if info
.get(key
) is not None), None)
727 if value
not in ('', None):
728 value
= value
.replace('\0', '') # nul character cannot be passed in command line
729 metadata
['common'].update({meta_f: value for meta_f in variadic(meta_list)}
)
731 # See [1-4] for some info on media metadata/metadata supported
733 # 1. https://kdenlive.org/en/project/adding-meta-data-to-mp4-video/
734 # 2. https://wiki.multimedia.cx/index.php/FFmpeg_Metadata
735 # 3. https://kodi.wiki/view/Video_file_tagging
737 add('title', ('track', 'title'))
738 add('date', 'upload_date')
739 add(('description', 'synopsis'), 'description')
740 add(('purl', 'comment'), 'webpage_url')
741 add('track', 'track_number')
742 add('artist', ('artist', 'creator', 'uploader', 'uploader_id'))
746 add('disc', 'disc_number')
747 add('show', 'series')
749 add('episode_id', ('episode', 'episode_id'))
750 add('episode_sort', 'episode_number')
751 if 'embed-metadata' in self
.get_param('compat_opts', []):
752 add('comment', 'description')
753 metadata
['common'].pop('synopsis', None)
755 meta_regex
= rf
'{re.escape(meta_prefix)}(?P<i>\d+)?_(?P<key>.+)'
756 for key
, value
in info
.items():
757 mobj
= re
.fullmatch(meta_regex
, key
)
758 if value
is not None and mobj
:
759 metadata
[mobj
.group('i') or 'common'][mobj
.group('key')] = value
.replace('\0', '')
761 # Write id3v1 metadata also since Windows Explorer can't handle id3v2 tags
762 yield ('-write_id3v1', '1')
764 for name
, value
in metadata
['common'].items():
765 yield ('-metadata', f
'{name}={value}')
768 for fmt
in info
.get('requested_formats') or []:
769 stream_count
= 2 if 'none' not in (fmt
.get('vcodec'), fmt
.get('acodec')) else 1
770 lang
= ISO639Utils
.short2long(fmt
.get('language') or '') or fmt
.get('language')
771 for i
in range(stream_idx
, stream_idx
+ stream_count
):
773 metadata
[str(i
)].setdefault('language', lang
)
774 for name
, value
in metadata
[str(i
)].items():
775 yield (f
'-metadata:s:{i}', f
'{name}={value}')
776 stream_idx
+= stream_count
778 def _get_infojson_opts(self
, info
, infofn
):
779 if not infofn
or not os
.path
.exists(infofn
):
780 if self
._add
_infojson
is not True:
782 infofn
= infofn
or '%s.temp' % (
783 self
._downloader
.prepare_filename(info
, 'infojson')
784 or replace_extension(self
._downloader
.prepare_filename(info
), 'info.json', info
['ext']))
785 if not self
._downloader
._ensure
_dir
_exists
(infofn
):
787 self
.write_debug(f
'Writing info-json to: {infofn}')
788 write_json_file(self
._downloader
.sanitize_info(info
, self
.get_param('clean_infojson', True)), infofn
)
789 info
['infojson_filename'] = infofn
791 old_stream
, new_stream
= self
.get_stream_number(info
['filepath'], ('tags', 'mimetype'), 'application/json')
792 if old_stream
is not None:
793 yield ('-map', '-0:%d' % old_stream
)
798 f
'-metadata:s:{new_stream}', 'mimetype=application/json',
799 f
'-metadata:s:{new_stream}', 'filename=info.json',
803 class FFmpegMergerPP(FFmpegPostProcessor
):
804 @PostProcessor._restrict_to(images
=False)
806 filename
= info
['filepath']
807 temp_filename
= prepend_extension(filename
, 'temp')
808 args
= ['-c', 'copy']
810 for (i
, fmt
) in enumerate(info
['requested_formats']):
811 if fmt
.get('acodec') != 'none':
812 args
.extend(['-map', f
'{i}:a:0'])
813 aac_fixup
= fmt
['protocol'].startswith('m3u8') and self
.get_audio_codec(fmt
['filepath']) == 'aac'
815 args
.extend([f
'-bsf:a:{audio_streams}', 'aac_adtstoasc'])
817 if fmt
.get('vcodec') != 'none':
818 args
.extend(['-map', '%u:v:0' % (i
)])
819 self
.to_screen('Merging formats into "%s"' % filename
)
820 self
.run_ffmpeg_multiple_files(info
['__files_to_merge'], temp_filename
, args
)
821 os
.rename(encodeFilename(temp_filename
), encodeFilename(filename
))
822 return info
['__files_to_merge'], info
825 # TODO: figure out merge-capable ffmpeg version
826 if self
.basename
!= 'avconv':
829 required_version
= '10-0'
830 if is_outdated_version(
831 self
._versions
[self
.basename
], required_version
):
832 warning
= ('Your copy of %s is outdated and unable to properly mux separate video and audio files, '
833 'yt-dlp will download single file media. '
834 'Update %s to version %s or newer to fix this.') % (
835 self
.basename
, self
.basename
, required_version
)
836 self
.report_warning(warning
)
841 class FFmpegFixupPostProcessor(FFmpegPostProcessor
):
842 def _fixup(self
, msg
, filename
, options
):
843 temp_filename
= prepend_extension(filename
, 'temp')
845 self
.to_screen(f
'{msg} of "{filename}"')
846 self
.run_ffmpeg(filename
, temp_filename
, options
)
848 os
.replace(temp_filename
, filename
)
851 class FFmpegFixupStretchedPP(FFmpegFixupPostProcessor
):
852 @PostProcessor._restrict_to(images
=False, audio
=False)
854 stretched_ratio
= info
.get('stretched_ratio')
855 if stretched_ratio
not in (None, 1):
856 self
._fixup
('Fixing aspect ratio', info
['filepath'], [
857 *self
.stream_copy_opts(), '-aspect', '%f' % stretched_ratio
])
861 class FFmpegFixupM4aPP(FFmpegFixupPostProcessor
):
862 @PostProcessor._restrict_to(images
=False, video
=False)
864 if info
.get('container') == 'm4a_dash':
865 self
._fixup
('Correcting container', info
['filepath'], [*self
.stream_copy_opts(), '-f', 'mp4'])
869 class FFmpegFixupM3u8PP(FFmpegFixupPostProcessor
):
870 def _needs_fixup(self
, info
):
871 yield info
['ext'] in ('mp4', 'm4a')
872 yield info
['protocol'].startswith('m3u8')
874 metadata
= self
.get_metadata_object(info
['filepath'])
875 except PostProcessingError
as e
:
876 self
.report_warning(f
'Unable to extract metadata: {e.msg}')
879 yield traverse_obj(metadata
, ('format', 'format_name'), casesense
=False) == 'mpegts'
881 @PostProcessor._restrict_to(images
=False)
883 if all(self
._needs
_fixup
(info
)):
884 self
._fixup
('Fixing MPEG-TS in MP4 container', info
['filepath'], [
885 *self
.stream_copy_opts(), '-f', 'mp4', '-bsf:a', 'aac_adtstoasc'])
889 class FFmpegFixupTimestampPP(FFmpegFixupPostProcessor
):
891 def __init__(self
, downloader
=None, trim
=0.001):
892 # "trim" should be used when the video contains unintended packets
893 super().__init
__(downloader
)
894 assert isinstance(trim
, (int, float))
895 self
.trim
= str(trim
)
897 @PostProcessor._restrict_to(images
=False)
899 if not self
._features
.get('setts'):
901 'A re-encode is needed to fix timestamps in older versions of ffmpeg. '
902 'Please install ffmpeg 4.4 or later to fixup without re-encoding')
903 opts
= ['-vf', 'setpts=PTS-STARTPTS']
905 opts
= ['-c', 'copy', '-bsf', 'setts=ts=TS-STARTPTS']
906 self
._fixup
('Fixing frame timestamp', info
['filepath'], opts
+ [*self
.stream_copy_opts(False), '-ss', self
.trim
])
910 class FFmpegCopyStreamPP(FFmpegFixupPostProcessor
):
911 MESSAGE
= 'Copying stream'
913 @PostProcessor._restrict_to(images
=False)
915 self
._fixup
(self
.MESSAGE
, info
['filepath'], self
.stream_copy_opts())
919 class FFmpegFixupDurationPP(FFmpegCopyStreamPP
):
920 MESSAGE
= 'Fixing video duration'
923 class FFmpegFixupDuplicateMoovPP(FFmpegCopyStreamPP
):
924 MESSAGE
= 'Fixing duplicate MOOV atoms'
927 class FFmpegSubtitlesConvertorPP(FFmpegPostProcessor
):
928 SUPPORTED_EXTS
= ('srt', 'vtt', 'ass', 'lrc')
930 def __init__(self
, downloader
=None, format
=None):
931 super().__init
__(downloader
)
935 subs
= info
.get('requested_subtitles')
936 new_ext
= self
.format
938 if new_format
== 'vtt':
939 new_format
= 'webvtt'
941 self
.to_screen('There aren\'t any subtitles to convert')
943 self
.to_screen('Converting subtitles')
945 for lang
, sub
in subs
.items():
946 if not os
.path
.exists(sub
.get('filepath', '')):
947 self
.report_warning(f
'Skipping embedding {lang} subtitle because the file is missing')
951 self
.to_screen('Subtitle file for %s is already in the requested format' % new_ext
)
955 'You have requested to convert json subtitles into another format, '
956 'which is currently not possible')
958 old_file
= sub
['filepath']
959 sub_filenames
.append(old_file
)
960 new_file
= replace_extension(old_file
, new_ext
)
962 if ext
in ('dfxp', 'ttml', 'tt'):
964 'You have requested to convert dfxp (TTML) subtitles into another format, '
965 'which results in style information loss')
968 srt_file
= replace_extension(old_file
, 'srt')
970 with open(dfxp_file
, 'rb') as f
:
971 srt_data
= dfxp2srt(f
.read())
973 with open(srt_file
, 'wt', encoding
='utf-8') as f
:
980 'filepath': srt_file
,
986 sub_filenames
.append(srt_file
)
988 self
.run_ffmpeg(old_file
, new_file
, ['-f', new_format
])
990 with open(new_file
, encoding
='utf-8') as f
:
994 'filepath': new_file
,
997 info
['__files_to_move'][new_file
] = replace_extension(
998 info
['__files_to_move'][sub
['filepath']], new_ext
)
1000 return sub_filenames
, info
1003 class FFmpegSplitChaptersPP(FFmpegPostProcessor
):
1004 def __init__(self
, downloader
, force_keyframes
=False):
1005 FFmpegPostProcessor
.__init
__(self
, downloader
)
1006 self
._force
_keyframes
= force_keyframes
1008 def _prepare_filename(self
, number
, chapter
, info
):
1011 'section_number': number
,
1012 'section_title': chapter
.get('title'),
1013 'section_start': chapter
.get('start_time'),
1014 'section_end': chapter
.get('end_time'),
1016 return self
._downloader
.prepare_filename(info
, 'chapter')
1018 def _ffmpeg_args_for_chapter(self
, number
, chapter
, info
):
1019 destination
= self
._prepare
_filename
(number
, chapter
, info
)
1020 if not self
._downloader
._ensure
_dir
_exists
(encodeFilename(destination
)):
1023 chapter
['filepath'] = destination
1024 self
.to_screen('Chapter %03d; Destination: %s' % (number
, destination
))
1027 ['-ss', str(chapter
['start_time']),
1028 '-t', str(chapter
['end_time'] - chapter
['start_time'])])
1030 @PostProcessor._restrict_to(images
=False)
1031 def run(self
, info
):
1032 chapters
= info
.get('chapters') or []
1034 self
.to_screen('Chapter information is unavailable')
1037 in_file
= info
['filepath']
1038 if self
._force
_keyframes
and len(chapters
) > 1:
1039 in_file
= self
.force_keyframes(in_file
, (c
['start_time'] for c
in chapters
))
1040 self
.to_screen('Splitting video by chapters; %d chapters found' % len(chapters
))
1041 for idx
, chapter
in enumerate(chapters
):
1042 destination
, opts
= self
._ffmpeg
_args
_for
_chapter
(idx
+ 1, chapter
, info
)
1043 self
.real_run_ffmpeg([(in_file
, opts
)], [(destination
, self
.stream_copy_opts())])
1044 if in_file
!= info
['filepath']:
1045 self
._delete
_downloaded
_files
(in_file
, msg
=None)
1049 class FFmpegThumbnailsConvertorPP(FFmpegPostProcessor
):
1050 SUPPORTED_EXTS
= ('jpg', 'png', 'webp')
1051 FORMAT_RE
= create_mapping_re(SUPPORTED_EXTS
)
1053 def __init__(self
, downloader
=None, format
=None):
1054 super().__init
__(downloader
)
1055 self
.mapping
= format
1058 def is_webp(cls
, path
):
1059 write_string(f
'DeprecationWarning: {cls.__module__}.{cls.__name__}.is_webp is deprecated')
1060 return imghdr
.what(path
) == 'webp'
1062 def fixup_webp(self
, info
, idx
=-1):
1063 thumbnail_filename
= info
['thumbnails'][idx
]['filepath']
1064 _
, thumbnail_ext
= os
.path
.splitext(thumbnail_filename
)
1066 if thumbnail_ext
.lower() != '.webp' and imghdr
.what(thumbnail_filename
) == 'webp':
1067 self
.to_screen('Correcting thumbnail "%s" extension to webp' % thumbnail_filename
)
1068 webp_filename
= replace_extension(thumbnail_filename
, 'webp')
1069 os
.replace(thumbnail_filename
, webp_filename
)
1070 info
['thumbnails'][idx
]['filepath'] = webp_filename
1071 info
['__files_to_move'][webp_filename
] = replace_extension(
1072 info
['__files_to_move'].pop(thumbnail_filename
), 'webp')
1075 def _options(target_ext
):
1076 if target_ext
== 'jpg':
1077 return ['-bsf:v', 'mjpeg2jpeg']
1080 def convert_thumbnail(self
, thumbnail_filename
, target_ext
):
1081 thumbnail_conv_filename
= replace_extension(thumbnail_filename
, target_ext
)
1083 self
.to_screen(f
'Converting thumbnail "{thumbnail_filename}" to {target_ext}')
1084 self
.real_run_ffmpeg(
1085 [(thumbnail_filename
, ['-f', 'image2', '-pattern_type', 'none'])],
1086 [(thumbnail_conv_filename
.replace('%', '%%'), self
._options
(target_ext
))])
1087 return thumbnail_conv_filename
1089 def run(self
, info
):
1090 files_to_delete
= []
1091 has_thumbnail
= False
1093 for idx
, thumbnail_dict
in enumerate(info
.get('thumbnails') or []):
1094 original_thumbnail
= thumbnail_dict
.get('filepath')
1095 if not original_thumbnail
:
1097 has_thumbnail
= True
1098 self
.fixup_webp(info
, idx
)
1099 thumbnail_ext
= os
.path
.splitext(original_thumbnail
)[1][1:].lower()
1100 if thumbnail_ext
== 'jpeg':
1101 thumbnail_ext
= 'jpg'
1102 target_ext
, _skip_msg
= resolve_mapping(thumbnail_ext
, self
.mapping
)
1104 self
.to_screen(f
'Not converting thumbnail "{original_thumbnail}"; {_skip_msg}')
1106 thumbnail_dict
['filepath'] = self
.convert_thumbnail(original_thumbnail
, target_ext
)
1107 files_to_delete
.append(original_thumbnail
)
1108 info
['__files_to_move'][thumbnail_dict
['filepath']] = replace_extension(
1109 info
['__files_to_move'][original_thumbnail
], target_ext
)
1111 if not has_thumbnail
:
1112 self
.to_screen('There aren\'t any thumbnails to convert')
1113 return files_to_delete
, info
1116 class FFmpegConcatPP(FFmpegPostProcessor
):
1117 def __init__(self
, downloader
, only_multi_video
=False):
1118 self
._only
_multi
_video
= only_multi_video
1119 super().__init
__(downloader
)
1121 def _get_codecs(self
, file):
1122 codecs
= traverse_obj(self
.get_metadata_object(file), ('streams', ..., 'codec_name'))
1123 self
.write_debug(f
'Codecs = {", ".join(codecs)}')
1124 return tuple(codecs
)
1126 def concat_files(self
, in_files
, out_file
):
1127 if not self
._downloader
._ensure
_dir
_exists
(out_file
):
1129 if len(in_files
) == 1:
1130 if os
.path
.realpath(in_files
[0]) != os
.path
.realpath(out_file
):
1131 self
.to_screen(f
'Moving "{in_files[0]}" to "{out_file}"')
1132 os
.replace(in_files
[0], out_file
)
1135 if len(set(map(self
._get
_codecs
, in_files
))) > 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')
1140 self
.to_screen(f
'Concatenating {len(in_files)} files; Destination: {out_file}')
1141 super().concat_files(in_files
, out_file
)
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'):
1149 elif traverse_obj(entries
, (..., lambda k
, v
: k
== 'requested_downloads' and len(v
) > 1)):
1150 raise PostProcessingError('Concatenation is not supported when downloading multiple separate formats')
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')
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')
1161 files_to_delete
= self
.concat_files(in_files
, out_file
)
1163 info
['requested_downloads'] = [{
1164 'filepath': out_file
,
1165 'ext': ie_copy
['ext'],
1167 return files_to_delete
, info