9 from .common
import AudioConversionError
, PostProcessor
10 from ..compat
import functools
, imghdr
15 _get_exe_version_output
,
34 EXT_TO_OUT_FORMATS
= {
53 'vorbis': 'libvorbis',
59 class FFmpegPostProcessorError(PostProcessingError
):
63 class FFmpegPostProcessor(PostProcessor
):
64 def __init__(self
, downloader
=None):
65 PostProcessor
.__init
__(self
, downloader
)
66 self
._prefer
_ffmpeg
= self
.get_param('prefer_ffmpeg', True)
67 self
._paths
= self
._determine
_executables
()
70 def get_versions_and_features(downloader
=None):
71 pp
= FFmpegPostProcessor(downloader
)
72 return pp
._versions
, pp
._features
75 def get_versions(downloader
=None):
76 return FFmpegPostProcessor
.get_versions_and_features(downloader
)[0]
78 _ffmpeg_to_avconv
= {'ffmpeg': 'avconv', 'ffprobe': 'avprobe'}
80 def _determine_executables(self
):
81 programs
= [*self
._ffmpeg
_to
_avconv
.keys(), *self
._ffmpeg
_to
_avconv
.values()]
83 location
= self
.get_param('ffmpeg_location')
85 return {p: p for p in programs}
87 if not os
.path
.exists(location
):
88 self
.report_warning(f
'ffmpeg-location {location} does not exist! Continuing without ffmpeg')
90 elif os
.path
.isdir(location
):
91 dirname
, basename
= location
, None
93 basename
= os
.path
.splitext(os
.path
.basename(location
))[0]
94 basename
= next((p
for p
in programs
if basename
.startswith(p
)), 'ffmpeg')
95 dirname
= os
.path
.dirname(os
.path
.abspath(location
))
96 if basename
in self
._ffmpeg
_to
_avconv
.keys():
97 self
._prefer
_ffmpeg
= True
99 paths
= {p: os.path.join(dirname, p) for p in programs}
101 paths
[basename
] = location
104 _version_cache
, _features_cache
= {None: None}
, {}
106 def _get_ffmpeg_version(self
, prog
):
107 path
= self
._paths
.get(prog
)
108 if path
in self
._version
_cache
:
109 return self
._version
_cache
[path
], self
._features
_cache
.get(path
, {})
110 out
= _get_exe_version_output(path
, ['-bsfs'], to_screen
=self
.write_debug
)
111 ver
= detect_exe_version(out
) if out
else False
114 r
'(?:\d+:)?([0-9.]+)-[0-9]+ubuntu[0-9.]+$', # Ubuntu, see [1]
115 r
'n([0-9.]+)$', # Arch Linux
116 # 1. http://www.ducea.com/2006/06/17/ubuntu-package-version-naming-explanation/
119 mobj
= re
.match(regex
, ver
)
122 self
._version
_cache
[path
] = ver
123 if prog
!= 'ffmpeg' or not out
:
126 mobj
= re
.search(r
'(?m)^\s+libavformat\s+(?:[0-9. ]+)\s+/\s+(?P<runtime>[0-9. ]+)', out
)
127 lavf_runtime_version
= mobj
.group('runtime').replace(' ', '') if mobj
else None
128 self
._features
_cache
[path
] = features
= {
129 'fdk': '--enable-libfdk-aac' in out
,
130 'setts': 'setts' in out
.splitlines(),
131 'needs_adtstoasc': is_outdated_version(lavf_runtime_version
, '57.56.100', False),
137 return filter_dict({self.basename: self._version, self.probe_basename: self._probe_version}
)
139 @functools.cached_property
141 self
._version
# run property
144 @functools.cached_property
145 def probe_basename(self
):
146 self
._probe
_version
# run property
147 return self
.probe_basename
149 def _get_version(self
, kind
):
150 executables
= (kind
, self
._ffmpeg
_to
_avconv
[kind
])
151 if not self
._prefer
_ffmpeg
:
152 executables
= reversed(executables
)
153 basename
, version
, features
= next(filter(
154 lambda x
: x
[1], ((p
, *self
._get
_ffmpeg
_version
(p
)) for p
in executables
)), (None, None, {}))
156 self
.basename
, self
._features
= basename
, features
158 self
.probe_basename
= basename
159 if basename
== self
._ffmpeg
_to
_avconv
[kind
]:
160 self
.deprecation_warning(
161 f
'Support for {self._ffmpeg_to_avconv[kind]} is deprecated and may be removed in a future version. Use {kind} instead')
164 @functools.cached_property
166 return self
._get
_version
('ffmpeg')
168 @functools.cached_property
169 def _probe_version(self
):
170 return self
._get
_version
('ffprobe')
174 return self
.basename
is not None
177 def executable(self
):
178 return self
._paths
.get(self
.basename
)
181 def probe_available(self
):
182 return self
.probe_basename
is not None
185 def probe_executable(self
):
186 return self
._paths
.get(self
.probe_basename
)
189 def stream_copy_opts(copy
=True, *, ext
=None):
190 yield from ('-map', '0')
191 # Don't copy Apple TV chapters track, bin_data
192 # See https://github.com/yt-dlp/yt-dlp/issues/2, #19042, #19024, https://trac.ffmpeg.org/ticket/6016
193 yield from ('-dn', '-ignore_unknown')
195 yield from ('-c', 'copy')
196 if ext
in ('mp4', 'mov', 'm4a'):
197 yield from ('-c:s', 'mov_text')
199 def check_version(self
):
200 if not self
.available
:
201 raise FFmpegPostProcessorError('ffmpeg not found. Please install or provide the path using --ffmpeg-location')
203 required_version
= '10-0' if self
.basename
== 'avconv' else '1.0'
204 if is_outdated_version(self
._version
, required_version
):
205 self
.report_warning(f
'Your copy of {self.basename} is outdated, update {self.basename} '
206 f
'to version {required_version} or newer if you encounter any errors')
208 def get_audio_codec(self
, path
):
209 if not self
.probe_available
and not self
.available
:
210 raise PostProcessingError('ffprobe and ffmpeg not found. Please install or provide the path using --ffmpeg-location')
212 if self
.probe_available
:
214 encodeFilename(self
.probe_executable
, True),
215 encodeArgument('-show_streams')]
218 encodeFilename(self
.executable
, True),
219 encodeArgument('-i')]
220 cmd
.append(encodeFilename(self
._ffmpeg
_filename
_argument
(path
), True))
221 self
.write_debug(f
'{self.basename} command line: {shell_quote(cmd)}')
222 handle
= Popen(cmd
, stdin
=subprocess
.PIPE
, stdout
=subprocess
.PIPE
, stderr
=subprocess
.PIPE
)
223 stdout_data
, stderr_data
= handle
.communicate_or_kill()
224 expected_ret
= 0 if self
.probe_available
else 1
225 if handle
.wait() != expected_ret
:
229 output
= (stdout_data
if self
.probe_available
else stderr_data
).decode('ascii', 'ignore')
230 if self
.probe_available
:
232 for line
in output
.split('\n'):
233 if line
.startswith('codec_name='):
234 audio_codec
= line
.split('=')[1].strip()
235 elif line
.strip() == 'codec_type=audio' and audio_codec
is not None:
238 # Stream #FILE_INDEX:STREAM_INDEX[STREAM_ID](LANGUAGE): CODEC_TYPE: CODEC_NAME
240 r
'Stream\s*#\d+:\d+(?:\[0x[0-9a-f]+\])?(?:\([a-z]{3}\))?:\s*Audio:\s*([0-9a-z]+)',
246 def get_metadata_object(self
, path
, opts
=[]):
247 if self
.probe_basename
!= 'ffprobe':
248 if self
.probe_available
:
249 self
.report_warning('Only ffprobe is supported for metadata extraction')
250 raise PostProcessingError('ffprobe not found. Please install or provide the path using --ffmpeg-location')
254 encodeFilename(self
.probe_executable
, True),
255 encodeArgument('-hide_banner'),
256 encodeArgument('-show_format'),
257 encodeArgument('-show_streams'),
258 encodeArgument('-print_format'),
259 encodeArgument('json'),
263 cmd
.append(encodeFilename(self
._ffmpeg
_filename
_argument
(path
), True))
264 self
.write_debug('ffprobe command line: %s' % shell_quote(cmd
))
265 p
= Popen(cmd
, stdout
=subprocess
.PIPE
, stderr
=subprocess
.PIPE
, stdin
=subprocess
.PIPE
)
266 stdout
, stderr
= p
.communicate()
267 return json
.loads(stdout
.decode('utf-8', 'replace'))
269 def get_stream_number(self
, path
, keys
, value
):
270 streams
= self
.get_metadata_object(path
)['streams']
272 (i
for i
, stream
in enumerate(streams
) if traverse_obj(stream
, keys
, casesense
=False) == value
),
274 return num
, len(streams
)
276 def _get_real_video_duration(self
, filepath
, fatal
=True):
278 duration
= float_or_none(
279 traverse_obj(self
.get_metadata_object(filepath
), ('format', 'duration')))
281 raise PostProcessingError('ffprobe returned empty duration')
283 except PostProcessingError
as e
:
285 raise PostProcessingError(f
'Unable to determine video duration: {e.msg}')
287 def _duration_mismatch(self
, d1
, d2
, tolerance
=2):
290 # The duration is often only known to nearest second. So there can be <1sec disparity natually.
291 # Further excuse an additional <1sec difference.
292 return abs(d1
- d2
) > tolerance
294 def run_ffmpeg_multiple_files(self
, input_paths
, out_path
, opts
, **kwargs
):
295 return self
.real_run_ffmpeg(
296 [(path
, []) for path
in input_paths
],
297 [(out_path
, opts
)], **kwargs
)
299 def real_run_ffmpeg(self
, input_path_opts
, output_path_opts
, *, expected_retcodes
=(0,)):
303 os
.stat(encodeFilename(path
)).st_mtime
for path
, _
in input_path_opts
if path
)
305 cmd
= [encodeFilename(self
.executable
, True), encodeArgument('-y')]
306 # avconv does not have repeat option
307 if self
.basename
== 'ffmpeg':
308 cmd
+= [encodeArgument('-loglevel'), encodeArgument('repeat+info')]
310 def make_args(file, args
, name
, number
):
311 keys
= ['_%s%d' % (name
, number
), '_%s' % name
]
313 args
+= ['-movflags', '+faststart']
316 args
+= self
._configuration
_args
(self
.basename
, keys
)
320 [encodeArgument(arg
) for arg
in args
]
321 + [encodeFilename(self
._ffmpeg
_filename
_argument
(file), True)])
323 for arg_type
, path_opts
in (('i', input_path_opts
), ('o', output_path_opts
)):
324 cmd
+= itertools
.chain
.from_iterable(
325 make_args(path
, list(opts
), arg_type
, i
+ 1)
326 for i
, (path
, opts
) in enumerate(path_opts
) if path
)
328 self
.write_debug('ffmpeg command line: %s' % shell_quote(cmd
))
329 p
= Popen(cmd
, stdout
=subprocess
.PIPE
, stderr
=subprocess
.PIPE
, stdin
=subprocess
.PIPE
)
330 stdout
, stderr
= p
.communicate_or_kill()
331 if p
.returncode
not in variadic(expected_retcodes
):
332 stderr
= stderr
.decode('utf-8', 'replace').strip()
333 self
.write_debug(stderr
)
334 raise FFmpegPostProcessorError(stderr
.split('\n')[-1])
335 for out_path
, _
in output_path_opts
:
337 self
.try_utime(out_path
, oldest_mtime
, oldest_mtime
)
338 return stderr
.decode('utf-8', 'replace')
340 def run_ffmpeg(self
, path
, out_path
, opts
, **kwargs
):
341 return self
.run_ffmpeg_multiple_files([path
], out_path
, opts
, **kwargs
)
344 def _ffmpeg_filename_argument(fn
):
345 # Always use 'file:' because the filename may contain ':' (ffmpeg
346 # interprets that as a protocol) or can start with '-' (-- is broken in
347 # ffmpeg, see https://ffmpeg.org/trac/ffmpeg/ticket/2127 for details)
348 # Also leave '-' intact in order not to break streaming to stdout.
349 if fn
.startswith(('http://', 'https://')):
351 return 'file:' + fn
if fn
!= '-' else fn
354 def _quote_for_ffmpeg(string
):
355 # See https://ffmpeg.org/ffmpeg-utils.html#toc-Quoting-and-escaping
356 # A sequence of '' produces '\'''\'';
357 # final replace removes the empty '' between \' \'.
358 string
= string
.replace("'", r
"'\''").replace("'''", "'")
359 # Handle potential ' at string boundaries.
360 string
= string
[1:] if string
[0] == "'" else "'" + string
361 return string
[:-1] if string
[-1] == "'" else string
+ "'"
363 def force_keyframes(self
, filename
, timestamps
):
364 timestamps
= orderedSet(timestamps
)
365 if timestamps
[0] == 0:
366 timestamps
= timestamps
[1:]
367 keyframe_file
= prepend_extension(filename
, 'keyframes.temp')
368 self
.to_screen(f
'Re-encoding "{filename}" with appropriate keyframes')
369 self
.run_ffmpeg(filename
, keyframe_file
, [
370 *self
.stream_copy_opts(False, ext
=determine_ext(filename
)),
371 '-force_key_frames', ','.join(f
'{t:.6f}' for t
in timestamps
)])
374 def concat_files(self
, in_files
, out_file
, concat_opts
=None):
376 Use concat demuxer to concatenate multiple files having identical streams.
378 Only inpoint, outpoint, and duration concat options are supported.
379 See https://ffmpeg.org/ffmpeg-formats.html#concat-1 for details
381 concat_file
= f
'{out_file}.concat'
382 self
.write_debug(f
'Writing concat spec to {concat_file}')
383 with open(concat_file
, 'wt', encoding
='utf-8') as f
:
384 f
.writelines(self
._concat
_spec
(in_files
, concat_opts
))
386 out_flags
= list(self
.stream_copy_opts(ext
=determine_ext(out_file
)))
388 self
.real_run_ffmpeg(
389 [(concat_file
, ['-hide_banner', '-nostdin', '-f', 'concat', '-safe', '0'])],
390 [(out_file
, out_flags
)])
391 self
._delete
_downloaded
_files
(concat_file
)
394 def _concat_spec(cls
, in_files
, concat_opts
=None):
395 if concat_opts
is None:
396 concat_opts
= [{}] * len(in_files
)
397 yield 'ffconcat version 1.0\n'
398 for file, opts
in zip(in_files
, concat_opts
):
399 yield f
'file {cls._quote_for_ffmpeg(cls._ffmpeg_filename_argument(file))}\n'
400 # Iterate explicitly to yield the following directives in order, ignoring the rest.
401 for directive
in 'inpoint', 'outpoint', 'duration':
402 if directive
in opts
:
403 yield f
'{directive} {opts[directive]}\n'
406 class FFmpegExtractAudioPP(FFmpegPostProcessor
):
407 COMMON_AUDIO_EXTS
= ('wav', 'flac', 'm4a', 'aiff', 'mp3', 'ogg', 'mka', 'opus', 'wma')
408 SUPPORTED_EXTS
= ('aac', 'flac', 'mp3', 'm4a', 'opus', 'vorbis', 'wav', 'alac')
410 def __init__(self
, downloader
=None, preferredcodec
=None, preferredquality
=None, nopostoverwrites
=False):
411 FFmpegPostProcessor
.__init
__(self
, downloader
)
412 self
._preferredcodec
= preferredcodec
or 'best'
413 self
._preferredquality
= float_or_none(preferredquality
)
414 self
._nopostoverwrites
= nopostoverwrites
416 def _quality_args(self
, codec
):
417 if self
._preferredquality
is None:
419 elif self
._preferredquality
> 10:
420 return ['-b:a', f
'{self._preferredquality}k']
423 'libmp3lame': (10, 0),
424 'libvorbis': (0, 10),
425 # FFmpeg's AAC encoder does not have an upper limit for the value of -q:a.
426 # Experimentally, with values over 4, bitrate changes were minimal or non-existent
428 'libfdk_aac': (1, 5),
433 q
= limits
[1] + (limits
[0] - limits
[1]) * (self
._preferredquality
/ 10)
434 if codec
== 'libfdk_aac':
435 return ['-vbr', f
'{int(q)}']
436 return ['-q:a', f
'{q}']
438 def run_ffmpeg(self
, path
, out_path
, codec
, more_opts
):
442 acodec_opts
= ['-acodec', codec
]
443 opts
= ['-vn'] + acodec_opts
+ more_opts
445 FFmpegPostProcessor
.run_ffmpeg(self
, path
, out_path
, opts
)
446 except FFmpegPostProcessorError
as err
:
447 raise AudioConversionError(err
.msg
)
449 @PostProcessor._restrict_to(images
=False)
450 def run(self
, information
):
451 orig_path
= path
= information
['filepath']
452 orig_ext
= information
['ext']
454 if self
._preferredcodec
== 'best' and orig_ext
in self
.COMMON_AUDIO_EXTS
:
455 self
.to_screen('Skipping audio extraction since the file is already in a common audio format')
456 return [], information
458 filecodec
= self
.get_audio_codec(path
)
459 if filecodec
is None:
460 raise PostProcessingError('WARNING: unable to obtain file audio codec with ffprobe')
463 if self
._preferredcodec
== 'best' or self
._preferredcodec
== filecodec
or (self
._preferredcodec
== 'm4a' and filecodec
== 'aac'):
464 if filecodec
== 'aac' and self
._preferredcodec
in ['m4a', 'best']:
465 # Lossless, but in another container
468 more_opts
= ['-bsf:a', 'aac_adtstoasc']
469 elif filecodec
in ['aac', 'flac', 'mp3', 'vorbis', 'opus']:
470 # Lossless if possible
472 extension
= filecodec
473 if filecodec
== 'aac':
474 more_opts
= ['-f', 'adts']
475 if filecodec
== 'vorbis':
477 elif filecodec
== 'alac':
480 more_opts
+= ['-acodec', 'alac']
483 acodec
= 'libmp3lame'
485 more_opts
= self
._quality
_args
(acodec
)
487 # We convert the audio (lossy if codec is lossy)
488 acodec
= ACODECS
[self
._preferredcodec
]
489 if acodec
== 'aac' and self
._features
.get('fdk'):
490 acodec
= 'libfdk_aac'
491 extension
= self
._preferredcodec
492 more_opts
= self
._quality
_args
(acodec
)
493 if self
._preferredcodec
== 'aac':
494 more_opts
+= ['-f', 'adts']
495 elif self
._preferredcodec
== 'm4a':
496 more_opts
+= ['-bsf:a', 'aac_adtstoasc']
497 elif self
._preferredcodec
== 'vorbis':
499 elif self
._preferredcodec
== 'wav':
501 more_opts
+= ['-f', 'wav']
502 elif self
._preferredcodec
== 'alac':
504 more_opts
+= ['-acodec', 'alac']
506 prefix
, sep
, ext
= path
.rpartition('.') # not os.path.splitext, since the latter does not work on unicode in all setups
507 temp_path
= new_path
= prefix
+ sep
+ extension
511 self
.to_screen(f
'File is already in target format {self._preferredcodec}, skipping')
512 return [], information
513 orig_path
= prepend_extension(path
, 'orig')
514 temp_path
= prepend_extension(path
, 'temp')
515 if (self
._nopostoverwrites
and os
.path
.exists(encodeFilename(new_path
))
516 and os
.path
.exists(encodeFilename(orig_path
))):
517 self
.to_screen('Post-process file %s exists, skipping' % new_path
)
518 return [], information
521 self
.to_screen(f
'Destination: {new_path}')
522 self
.run_ffmpeg(path
, temp_path
, acodec
, more_opts
)
523 except AudioConversionError
as e
:
524 raise PostProcessingError(
525 'audio conversion failed: ' + e
.msg
)
527 raise PostProcessingError('error running ' + self
.basename
)
529 os
.replace(path
, orig_path
)
530 os
.replace(temp_path
, new_path
)
531 information
['filepath'] = new_path
532 information
['ext'] = extension
534 # Try to update the date time for extracted audio file.
535 if information
.get('filetime') is not None:
537 new_path
, time
.time(), information
['filetime'],
538 errnote
='Cannot update utime of audio file')
540 return [orig_path
], information
543 class FFmpegVideoConvertorPP(FFmpegPostProcessor
):
544 SUPPORTED_EXTS
= ('mp4', 'mkv', 'flv', 'webm', 'mov', 'avi', 'mka', 'ogg', *FFmpegExtractAudioPP
.SUPPORTED_EXTS
)
545 FORMAT_RE
= re
.compile(r
'{0}(?:/{0})*$'.format(r
'(?:\w+>)?(?:%s)' % '|'.join(SUPPORTED_EXTS
)))
546 _ACTION
= 'converting'
548 def __init__(self
, downloader
=None, preferedformat
=None):
549 super().__init
__(downloader
)
550 self
._preferedformats
= preferedformat
.lower().split('/')
552 def _target_ext(self
, source_ext
):
553 for pair
in self
._preferedformats
:
555 if len(kv
) == 1 or kv
[0].strip() == source_ext
:
556 return kv
[-1].strip()
559 def _options(target_ext
):
560 yield from FFmpegPostProcessor
.stream_copy_opts(False)
561 if target_ext
== 'avi':
562 yield from ('-c:v', 'libxvid', '-vtag', 'XVID')
564 @PostProcessor._restrict_to(images
=False)
566 filename
, source_ext
= info
['filepath'], info
['ext'].lower()
567 target_ext
= self
._target
_ext
(source_ext
)
569 f
'could not find a mapping for {source_ext}' if not target_ext
570 else f
'already is in target format {source_ext}' if source_ext
== target_ext
573 self
.to_screen(f
'Not {self._ACTION} media file "{filename}"; {_skip_msg}')
576 outpath
= replace_extension(filename
, target_ext
, source_ext
)
577 self
.to_screen(f
'{self._ACTION.title()} video from {source_ext} to {target_ext}; Destination: {outpath}')
578 self
.run_ffmpeg(filename
, outpath
, self
._options
(target_ext
))
580 info
['filepath'] = outpath
581 info
['format'] = info
['ext'] = target_ext
582 return [filename
], info
585 class FFmpegVideoRemuxerPP(FFmpegVideoConvertorPP
):
589 def _options(target_ext
):
590 return FFmpegPostProcessor
.stream_copy_opts()
593 class FFmpegEmbedSubtitlePP(FFmpegPostProcessor
):
594 SUPPORTED_EXTS
= ('mp4', 'mov', 'm4a', 'webm', 'mkv', 'mka')
596 def __init__(self
, downloader
=None, already_have_subtitle
=False):
597 super().__init
__(downloader
)
598 self
._already
_have
_subtitle
= already_have_subtitle
600 @PostProcessor._restrict_to(images
=False)
602 if info
['ext'] not in self
.SUPPORTED_EXTS
:
603 self
.to_screen(f
'Subtitles can only be embedded in {", ".join(self.SUPPORTED_EXTS)} files')
605 subtitles
= info
.get('requested_subtitles')
607 self
.to_screen('There aren\'t any subtitles to embed')
610 filename
= info
['filepath']
612 # Disabled temporarily. There needs to be a way to overide this
613 # in case of duration actually mismatching in extractor
614 # See: https://github.com/yt-dlp/yt-dlp/issues/1870, https://github.com/yt-dlp/yt-dlp/issues/1385
616 if info.get('duration') and not info.get('__real_download') and self._duration_mismatch(
617 self._get_real_video_duration(filename, False), info['duration']):
618 self.to_screen(f'Skipping {self.pp_key()} since the real and expected durations mismatch')
623 sub_langs
, sub_names
, sub_filenames
= [], [], []
624 webm_vtt_warn
= False
627 for lang
, sub_info
in subtitles
.items():
628 if not os
.path
.exists(sub_info
.get('filepath', '')):
629 self
.report_warning(f
'Skipping embedding {lang} subtitle because the file is missing')
631 sub_ext
= sub_info
['ext']
632 if sub_ext
== 'json':
633 self
.report_warning('JSON subtitles cannot be embedded')
634 elif ext
!= 'webm' or ext
== 'webm' and sub_ext
== 'vtt':
635 sub_langs
.append(lang
)
636 sub_names
.append(sub_info
.get('name'))
637 sub_filenames
.append(sub_info
['filepath'])
639 if not webm_vtt_warn
and ext
== 'webm' and sub_ext
!= 'vtt':
641 self
.report_warning('Only WebVTT subtitles can be embedded in webm files')
642 if not mp4_ass_warn
and ext
== 'mp4' and sub_ext
== 'ass':
644 self
.report_warning('ASS subtitles cannot be properly embedded in mp4 files; expect issues')
649 input_files
= [filename
] + sub_filenames
652 *self
.stream_copy_opts(ext
=info
['ext']),
653 # Don't copy the existing subtitles, we may be running the
654 # postprocessor a second time
657 for i
, (lang
, name
) in enumerate(zip(sub_langs
, sub_names
)):
658 opts
.extend(['-map', '%d:0' % (i
+ 1)])
659 lang_code
= ISO639Utils
.short2long(lang
) or lang
660 opts
.extend(['-metadata:s:s:%d' % i
, 'language=%s' % lang_code
])
662 opts
.extend(['-metadata:s:s:%d' % i
, 'handler_name=%s' % name
,
663 '-metadata:s:s:%d' % i
, 'title=%s' % name
])
665 temp_filename
= prepend_extension(filename
, 'temp')
666 self
.to_screen('Embedding subtitles in "%s"' % filename
)
667 self
.run_ffmpeg_multiple_files(input_files
, temp_filename
, opts
)
668 os
.replace(temp_filename
, filename
)
670 files_to_delete
= [] if self
._already
_have
_subtitle
else sub_filenames
671 return files_to_delete
, info
674 class FFmpegMetadataPP(FFmpegPostProcessor
):
676 def __init__(self
, downloader
, add_metadata
=True, add_chapters
=True, add_infojson
='if_exists'):
677 FFmpegPostProcessor
.__init
__(self
, downloader
)
678 self
._add
_metadata
= add_metadata
679 self
._add
_chapters
= add_chapters
680 self
._add
_infojson
= add_infojson
683 def _options(target_ext
):
684 audio_only
= target_ext
== 'm4a'
685 yield from FFmpegPostProcessor
.stream_copy_opts(not audio_only
)
687 yield from ('-vn', '-acodec', 'copy')
689 @PostProcessor._restrict_to(images
=False)
691 filename
, metadata_filename
= info
['filepath'], None
692 files_to_delete
, options
= [], []
693 if self
._add
_chapters
and info
.get('chapters'):
694 metadata_filename
= replace_extension(filename
, 'meta')
695 options
.extend(self
._get
_chapter
_opts
(info
['chapters'], metadata_filename
))
696 files_to_delete
.append(metadata_filename
)
697 if self
._add
_metadata
:
698 options
.extend(self
._get
_metadata
_opts
(info
))
700 if self
._add
_infojson
:
701 if info
['ext'] in ('mkv', 'mka'):
702 infojson_filename
= info
.get('infojson_filename')
703 options
.extend(self
._get
_infojson
_opts
(info
, infojson_filename
))
704 if not infojson_filename
:
705 files_to_delete
.append(info
.get('infojson_filename'))
706 elif self
._add
_infojson
is True:
707 self
.to_screen('The info-json can only be attached to mkv/mka files')
710 self
.to_screen('There isn\'t any metadata to add')
713 temp_filename
= prepend_extension(filename
, 'temp')
714 self
.to_screen('Adding metadata to "%s"' % filename
)
715 self
.run_ffmpeg_multiple_files(
716 (filename
, metadata_filename
), temp_filename
,
717 itertools
.chain(self
._options
(info
['ext']), *options
))
718 self
._delete
_downloaded
_files
(*files_to_delete
)
719 os
.replace(temp_filename
, filename
)
723 def _get_chapter_opts(chapters
, metadata_filename
):
724 with open(metadata_filename
, 'wt', encoding
='utf-8') as f
:
725 def ffmpeg_escape(text
):
726 return re
.sub(r
'([\\=;#\n])', r
'\\\1', text
)
728 metadata_file_content
= ';FFMETADATA1\n'
729 for chapter
in chapters
:
730 metadata_file_content
+= '[CHAPTER]\nTIMEBASE=1/1000\n'
731 metadata_file_content
+= 'START=%d\n' % (chapter
['start_time'] * 1000)
732 metadata_file_content
+= 'END=%d\n' % (chapter
['end_time'] * 1000)
733 chapter_title
= chapter
.get('title')
735 metadata_file_content
+= 'title=%s\n' % ffmpeg_escape(chapter_title
)
736 f
.write(metadata_file_content
)
737 yield ('-map_metadata', '1')
739 def _get_metadata_opts(self
, info
):
741 metadata
= collections
.defaultdict(dict)
743 def add(meta_list
, info_list
=None):
745 str(info
[key
]) for key
in [f
'{meta_prefix}_'] + list(variadic(info_list
or meta_list
))
746 if info
.get(key
) is not None), None)
747 if value
not in ('', None):
748 value
= value
.replace('\0', '') # nul character cannot be passed in command line
749 metadata
['common'].update({meta_f: value for meta_f in variadic(meta_list)}
)
751 # See [1-4] for some info on media metadata/metadata supported
753 # 1. https://kdenlive.org/en/project/adding-meta-data-to-mp4-video/
754 # 2. https://wiki.multimedia.cx/index.php/FFmpeg_Metadata
755 # 3. https://kodi.wiki/view/Video_file_tagging
757 add('title', ('track', 'title'))
758 add('date', 'upload_date')
759 add(('description', 'synopsis'), 'description')
760 add(('purl', 'comment'), 'webpage_url')
761 add('track', 'track_number')
762 add('artist', ('artist', 'creator', 'uploader', 'uploader_id'))
766 add('disc', 'disc_number')
767 add('show', 'series')
769 add('episode_id', ('episode', 'episode_id'))
770 add('episode_sort', 'episode_number')
771 if 'embed-metadata' in self
.get_param('compat_opts', []):
772 add('comment', 'description')
773 metadata
['common'].pop('synopsis', None)
775 meta_regex
= rf
'{re.escape(meta_prefix)}(?P<i>\d+)?_(?P<key>.+)'
776 for key
, value
in info
.items():
777 mobj
= re
.fullmatch(meta_regex
, key
)
778 if value
is not None and mobj
:
779 metadata
[mobj
.group('i') or 'common'][mobj
.group('key')] = value
.replace('\0', '')
781 # Write id3v1 metadata also since Windows Explorer can't handle id3v2 tags
782 yield ('-write_id3v1', '1')
784 for name
, value
in metadata
['common'].items():
785 yield ('-metadata', f
'{name}={value}')
788 for fmt
in info
.get('requested_formats') or []:
789 stream_count
= 2 if 'none' not in (fmt
.get('vcodec'), fmt
.get('acodec')) else 1
790 lang
= ISO639Utils
.short2long(fmt
.get('language') or '') or fmt
.get('language')
791 for i
in range(stream_idx
, stream_idx
+ stream_count
):
793 metadata
[str(i
)].setdefault('language', lang
)
794 for name
, value
in metadata
[str(i
)].items():
795 yield (f
'-metadata:s:{i}', f
'{name}={value}')
796 stream_idx
+= stream_count
798 def _get_infojson_opts(self
, info
, infofn
):
799 if not infofn
or not os
.path
.exists(infofn
):
800 if self
._add
_infojson
is not True:
802 infofn
= infofn
or '%s.temp' % (
803 self
._downloader
.prepare_filename(info
, 'infojson')
804 or replace_extension(self
._downloader
.prepare_filename(info
), 'info.json', info
['ext']))
805 if not self
._downloader
._ensure
_dir
_exists
(infofn
):
807 self
.write_debug(f
'Writing info-json to: {infofn}')
808 write_json_file(self
._downloader
.sanitize_info(info
, self
.get_param('clean_infojson', True)), infofn
)
809 info
['infojson_filename'] = infofn
811 old_stream
, new_stream
= self
.get_stream_number(info
['filepath'], ('tags', 'mimetype'), 'application/json')
812 if old_stream
is not None:
813 yield ('-map', '-0:%d' % old_stream
)
818 f
'-metadata:s:{new_stream}', 'mimetype=application/json',
819 f
'-metadata:s:{new_stream}', 'filename=info.json',
823 class FFmpegMergerPP(FFmpegPostProcessor
):
824 @PostProcessor._restrict_to(images
=False)
826 filename
= info
['filepath']
827 temp_filename
= prepend_extension(filename
, 'temp')
828 args
= ['-c', 'copy']
830 for (i
, fmt
) in enumerate(info
['requested_formats']):
831 if fmt
.get('acodec') != 'none':
832 args
.extend(['-map', f
'{i}:a:0'])
833 aac_fixup
= fmt
['protocol'].startswith('m3u8') and self
.get_audio_codec(fmt
['filepath']) == 'aac'
835 args
.extend([f
'-bsf:a:{audio_streams}', 'aac_adtstoasc'])
837 if fmt
.get('vcodec') != 'none':
838 args
.extend(['-map', '%u:v:0' % (i
)])
839 self
.to_screen('Merging formats into "%s"' % filename
)
840 self
.run_ffmpeg_multiple_files(info
['__files_to_merge'], temp_filename
, args
)
841 os
.rename(encodeFilename(temp_filename
), encodeFilename(filename
))
842 return info
['__files_to_merge'], info
845 # TODO: figure out merge-capable ffmpeg version
846 if self
.basename
!= 'avconv':
849 required_version
= '10-0'
850 if is_outdated_version(
851 self
._versions
[self
.basename
], required_version
):
852 warning
= ('Your copy of %s is outdated and unable to properly mux separate video and audio files, '
853 'yt-dlp will download single file media. '
854 'Update %s to version %s or newer to fix this.') % (
855 self
.basename
, self
.basename
, required_version
)
856 self
.report_warning(warning
)
861 class FFmpegFixupPostProcessor(FFmpegPostProcessor
):
862 def _fixup(self
, msg
, filename
, options
):
863 temp_filename
= prepend_extension(filename
, 'temp')
865 self
.to_screen(f
'{msg} of "{filename}"')
866 self
.run_ffmpeg(filename
, temp_filename
, options
)
868 os
.replace(temp_filename
, filename
)
871 class FFmpegFixupStretchedPP(FFmpegFixupPostProcessor
):
872 @PostProcessor._restrict_to(images
=False, audio
=False)
874 stretched_ratio
= info
.get('stretched_ratio')
875 if stretched_ratio
not in (None, 1):
876 self
._fixup
('Fixing aspect ratio', info
['filepath'], [
877 *self
.stream_copy_opts(), '-aspect', '%f' % stretched_ratio
])
881 class FFmpegFixupM4aPP(FFmpegFixupPostProcessor
):
882 @PostProcessor._restrict_to(images
=False, video
=False)
884 if info
.get('container') == 'm4a_dash':
885 self
._fixup
('Correcting container', info
['filepath'], [*self
.stream_copy_opts(), '-f', 'mp4'])
889 class FFmpegFixupM3u8PP(FFmpegFixupPostProcessor
):
890 def _needs_fixup(self
, info
):
891 yield info
['ext'] in ('mp4', 'm4a')
892 yield info
['protocol'].startswith('m3u8')
894 metadata
= self
.get_metadata_object(info
['filepath'])
895 except PostProcessingError
as e
:
896 self
.report_warning(f
'Unable to extract metadata: {e.msg}')
899 yield traverse_obj(metadata
, ('format', 'format_name'), casesense
=False) == 'mpegts'
901 @PostProcessor._restrict_to(images
=False)
903 if all(self
._needs
_fixup
(info
)):
904 self
._fixup
('Fixing MPEG-TS in MP4 container', info
['filepath'], [
905 *self
.stream_copy_opts(), '-f', 'mp4', '-bsf:a', 'aac_adtstoasc'])
909 class FFmpegFixupTimestampPP(FFmpegFixupPostProcessor
):
911 def __init__(self
, downloader
=None, trim
=0.001):
912 # "trim" should be used when the video contains unintended packets
913 super().__init
__(downloader
)
914 assert isinstance(trim
, (int, float))
915 self
.trim
= str(trim
)
917 @PostProcessor._restrict_to(images
=False)
919 if not self
._features
.get('setts'):
921 'A re-encode is needed to fix timestamps in older versions of ffmpeg. '
922 'Please install ffmpeg 4.4 or later to fixup without re-encoding')
923 opts
= ['-vf', 'setpts=PTS-STARTPTS']
925 opts
= ['-c', 'copy', '-bsf', 'setts=ts=TS-STARTPTS']
926 self
._fixup
('Fixing frame timestamp', info
['filepath'], opts
+ [*self
.stream_copy_opts(False), '-ss', self
.trim
])
930 class FFmpegCopyStreamPP(FFmpegFixupPostProcessor
):
931 MESSAGE
= 'Copying stream'
933 @PostProcessor._restrict_to(images
=False)
935 self
._fixup
(self
.MESSAGE
, info
['filepath'], self
.stream_copy_opts())
939 class FFmpegFixupDurationPP(FFmpegCopyStreamPP
):
940 MESSAGE
= 'Fixing video duration'
943 class FFmpegFixupDuplicateMoovPP(FFmpegCopyStreamPP
):
944 MESSAGE
= 'Fixing duplicate MOOV atoms'
947 class FFmpegSubtitlesConvertorPP(FFmpegPostProcessor
):
948 SUPPORTED_EXTS
= ('srt', 'vtt', 'ass', 'lrc')
950 def __init__(self
, downloader
=None, format
=None):
951 super().__init
__(downloader
)
955 subs
= info
.get('requested_subtitles')
956 new_ext
= self
.format
958 if new_format
== 'vtt':
959 new_format
= 'webvtt'
961 self
.to_screen('There aren\'t any subtitles to convert')
963 self
.to_screen('Converting subtitles')
965 for lang
, sub
in subs
.items():
966 if not os
.path
.exists(sub
.get('filepath', '')):
967 self
.report_warning(f
'Skipping embedding {lang} subtitle because the file is missing')
971 self
.to_screen('Subtitle file for %s is already in the requested format' % new_ext
)
975 'You have requested to convert json subtitles into another format, '
976 'which is currently not possible')
978 old_file
= sub
['filepath']
979 sub_filenames
.append(old_file
)
980 new_file
= replace_extension(old_file
, new_ext
)
982 if ext
in ('dfxp', 'ttml', 'tt'):
984 'You have requested to convert dfxp (TTML) subtitles into another format, '
985 'which results in style information loss')
988 srt_file
= replace_extension(old_file
, 'srt')
990 with open(dfxp_file
, 'rb') as f
:
991 srt_data
= dfxp2srt(f
.read())
993 with open(srt_file
, 'wt', encoding
='utf-8') as f
:
1000 'filepath': srt_file
,
1003 if new_ext
== 'srt':
1006 sub_filenames
.append(srt_file
)
1008 self
.run_ffmpeg(old_file
, new_file
, ['-f', new_format
])
1010 with open(new_file
, encoding
='utf-8') as f
:
1014 'filepath': new_file
,
1017 info
['__files_to_move'][new_file
] = replace_extension(
1018 info
['__files_to_move'][sub
['filepath']], new_ext
)
1020 return sub_filenames
, info
1023 class FFmpegSplitChaptersPP(FFmpegPostProcessor
):
1024 def __init__(self
, downloader
, force_keyframes
=False):
1025 FFmpegPostProcessor
.__init
__(self
, downloader
)
1026 self
._force
_keyframes
= force_keyframes
1028 def _prepare_filename(self
, number
, chapter
, info
):
1031 'section_number': number
,
1032 'section_title': chapter
.get('title'),
1033 'section_start': chapter
.get('start_time'),
1034 'section_end': chapter
.get('end_time'),
1036 return self
._downloader
.prepare_filename(info
, 'chapter')
1038 def _ffmpeg_args_for_chapter(self
, number
, chapter
, info
):
1039 destination
= self
._prepare
_filename
(number
, chapter
, info
)
1040 if not self
._downloader
._ensure
_dir
_exists
(encodeFilename(destination
)):
1043 chapter
['filepath'] = destination
1044 self
.to_screen('Chapter %03d; Destination: %s' % (number
, destination
))
1047 ['-ss', str(chapter
['start_time']),
1048 '-t', str(chapter
['end_time'] - chapter
['start_time'])])
1050 @PostProcessor._restrict_to(images
=False)
1051 def run(self
, info
):
1052 chapters
= info
.get('chapters') or []
1054 self
.to_screen('Chapter information is unavailable')
1057 in_file
= info
['filepath']
1058 if self
._force
_keyframes
and len(chapters
) > 1:
1059 in_file
= self
.force_keyframes(in_file
, (c
['start_time'] for c
in chapters
))
1060 self
.to_screen('Splitting video by chapters; %d chapters found' % len(chapters
))
1061 for idx
, chapter
in enumerate(chapters
):
1062 destination
, opts
= self
._ffmpeg
_args
_for
_chapter
(idx
+ 1, chapter
, info
)
1063 self
.real_run_ffmpeg([(in_file
, opts
)], [(destination
, self
.stream_copy_opts())])
1064 if in_file
!= info
['filepath']:
1065 self
._delete
_downloaded
_files
(in_file
, msg
=None)
1069 class FFmpegThumbnailsConvertorPP(FFmpegPostProcessor
):
1070 SUPPORTED_EXTS
= ('jpg', 'png', 'webp')
1072 def __init__(self
, downloader
=None, format
=None):
1073 super().__init
__(downloader
)
1074 self
.format
= format
1077 def is_webp(cls
, path
):
1078 write_string(f
'DeprecationWarning: {cls.__module__}.{cls.__name__}.is_webp is deprecated')
1079 return imghdr
.what(path
) == 'webp'
1081 def fixup_webp(self
, info
, idx
=-1):
1082 thumbnail_filename
= info
['thumbnails'][idx
]['filepath']
1083 _
, thumbnail_ext
= os
.path
.splitext(thumbnail_filename
)
1085 if thumbnail_ext
.lower() != '.webp' and imghdr
.what(thumbnail_filename
) == 'webp':
1086 self
.to_screen('Correcting thumbnail "%s" extension to webp' % thumbnail_filename
)
1087 webp_filename
= replace_extension(thumbnail_filename
, 'webp')
1088 os
.replace(thumbnail_filename
, webp_filename
)
1089 info
['thumbnails'][idx
]['filepath'] = webp_filename
1090 info
['__files_to_move'][webp_filename
] = replace_extension(
1091 info
['__files_to_move'].pop(thumbnail_filename
), 'webp')
1094 def _options(target_ext
):
1095 if target_ext
== 'jpg':
1096 return ['-bsf:v', 'mjpeg2jpeg']
1099 def convert_thumbnail(self
, thumbnail_filename
, target_ext
):
1100 thumbnail_conv_filename
= replace_extension(thumbnail_filename
, target_ext
)
1102 self
.to_screen(f
'Converting thumbnail "{thumbnail_filename}" to {target_ext}')
1103 self
.real_run_ffmpeg(
1104 [(thumbnail_filename
, ['-f', 'image2', '-pattern_type', 'none'])],
1105 [(thumbnail_conv_filename
.replace('%', '%%'), self
._options
(target_ext
))])
1106 return thumbnail_conv_filename
1108 def run(self
, info
):
1109 files_to_delete
= []
1110 has_thumbnail
= False
1112 for idx
, thumbnail_dict
in enumerate(info
.get('thumbnails') or []):
1113 original_thumbnail
= thumbnail_dict
.get('filepath')
1114 if not original_thumbnail
:
1116 has_thumbnail
= True
1117 self
.fixup_webp(info
, idx
)
1118 _
, thumbnail_ext
= os
.path
.splitext(original_thumbnail
)
1120 thumbnail_ext
= thumbnail_ext
[1:].lower()
1121 if thumbnail_ext
== 'jpeg':
1122 thumbnail_ext
= 'jpg'
1123 if thumbnail_ext
== self
.format
:
1124 self
.to_screen('Thumbnail "%s" is already in the requested format' % original_thumbnail
)
1126 thumbnail_dict
['filepath'] = self
.convert_thumbnail(original_thumbnail
, self
.format
)
1127 files_to_delete
.append(original_thumbnail
)
1128 info
['__files_to_move'][thumbnail_dict
['filepath']] = replace_extension(
1129 info
['__files_to_move'][original_thumbnail
], self
.format
)
1131 if not has_thumbnail
:
1132 self
.to_screen('There aren\'t any thumbnails to convert')
1133 return files_to_delete
, info
1136 class FFmpegConcatPP(FFmpegPostProcessor
):
1137 def __init__(self
, downloader
, only_multi_video
=False):
1138 self
._only
_multi
_video
= only_multi_video
1139 super().__init
__(downloader
)
1141 def _get_codecs(self
, file):
1142 codecs
= traverse_obj(self
.get_metadata_object(file), ('streams', ..., 'codec_name'))
1143 self
.write_debug(f
'Codecs = {", ".join(codecs)}')
1144 return tuple(codecs
)
1146 def concat_files(self
, in_files
, out_file
):
1147 if not self
._downloader
._ensure
_dir
_exists
(out_file
):
1149 if len(in_files
) == 1:
1150 if os
.path
.realpath(in_files
[0]) != os
.path
.realpath(out_file
):
1151 self
.to_screen(f
'Moving "{in_files[0]}" to "{out_file}"')
1152 os
.replace(in_files
[0], out_file
)
1155 if len(set(map(self
._get
_codecs
, in_files
))) > 1:
1156 raise PostProcessingError(
1157 'The files have different streams/codecs and cannot be concatenated. '
1158 'Either select different formats or --recode-video them to a common format')
1160 self
.to_screen(f
'Concatenating {len(in_files)} files; Destination: {out_file}')
1161 super().concat_files(in_files
, out_file
)
1164 @PostProcessor._restrict_to(images
=False, simulated
=False)
1165 def run(self
, info
):
1166 entries
= info
.get('entries') or []
1167 if not any(entries
) or (self
._only
_multi
_video
and info
['_type'] != 'multi_video'):
1169 elif traverse_obj(entries
, (..., lambda k
, v
: k
== 'requested_downloads' and len(v
) > 1)):
1170 raise PostProcessingError('Concatenation is not supported when downloading multiple separate formats')
1172 in_files
= traverse_obj(entries
, (..., 'requested_downloads', 0, 'filepath')) or []
1173 if len(in_files
) < len(entries
):
1174 raise PostProcessingError('Aborting concatenation because some downloads failed')
1176 ie_copy
= self
._downloader
._playlist
_infodict
(info
)
1177 exts
= traverse_obj(entries
, (..., 'requested_downloads', 0, 'ext'), (..., 'ext'))
1178 ie_copy
['ext'] = exts
[0] if len(set(exts
)) == 1 else 'mkv'
1179 out_file
= self
._downloader
.prepare_filename(ie_copy
, 'pl_video')
1181 files_to_delete
= self
.concat_files(in_files
, out_file
)
1183 info
['requested_downloads'] = [{
1184 'filepath': out_file
,
1185 'ext': ie_copy
['ext'],
1187 return files_to_delete
, info