]> jfr.im git - yt-dlp.git/blob - yt_dlp/postprocessor/ffmpeg.py
[docs] Improvements
[yt-dlp.git] / 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78 warning = 'Your copy of %s is outdated, update %s to version %s or newer if you encounter any errors.'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_version_and_features(downloader)[0]
90
91 def _determine_executables(self):
92 programs = ['avprobe', 'avconv', 'ffmpeg', 'ffprobe']
93
94 def get_ffmpeg_version(path, prog):
95 out = _get_exe_version_output(path, ['-bsfs'])
96 ver = detect_exe_version(out) if out else False
97 if ver:
98 regexs = [
99 r'(?:\d+:)?([0-9.]+)-[0-9]+ubuntu[0-9.]+$', # Ubuntu, see [1]
100 r'n([0-9.]+)$', # Arch Linux
101 # 1. http://www.ducea.com/2006/06/17/ubuntu-package-version-naming-explanation/
102 ]
103 for regex in regexs:
104 mobj = re.match(regex, ver)
105 if mobj:
106 ver = mobj.group(1)
107 self._versions[prog] = ver
108 if prog != 'ffmpeg' or not out:
109 return
110
111 mobj = re.search(r'(?m)^\s+libavformat\s+(?:[0-9. ]+)\s+/\s+(?P<runtime>[0-9. ]+)', out)
112 lavf_runtime_version = mobj.group('runtime').replace(' ', '') if mobj else None
113 self._features = {
114 'fdk': '--enable-libfdk-aac' in out,
115 'setts': 'setts' in out.splitlines(),
116 'needs_adtstoasc': is_outdated_version(lavf_runtime_version, '57.56.100', False),
117 }
118
119 self.basename = None
120 self.probe_basename = None
121 self._paths = None
122 self._versions = None
123 self._features = {}
124
125 prefer_ffmpeg = self.get_param('prefer_ffmpeg', True)
126 location = self.get_param('ffmpeg_location')
127 if location is None:
128 self._paths = {p: p for p in programs}
129 else:
130 if not os.path.exists(location):
131 self.report_warning(
132 'ffmpeg-location %s does not exist! '
133 'Continuing without ffmpeg.' % (location))
134 self._versions = {}
135 return
136 elif os.path.isdir(location):
137 dirname, basename = location, None
138 else:
139 basename = os.path.splitext(os.path.basename(location))[0]
140 basename = next((p for p in programs if basename.startswith(p)), 'ffmpeg')
141 dirname = os.path.dirname(os.path.abspath(location))
142 if basename in ('ffmpeg', 'ffprobe'):
143 prefer_ffmpeg = True
144
145 self._paths = dict(
146 (p, os.path.join(dirname, p)) for p in programs)
147 if basename:
148 self._paths[basename] = location
149
150 self._versions = {}
151 for p in programs:
152 get_ffmpeg_version(self._paths[p], p)
153
154 if prefer_ffmpeg is False:
155 prefs = ('avconv', 'ffmpeg')
156 else:
157 prefs = ('ffmpeg', 'avconv')
158 for p in prefs:
159 if self._versions[p]:
160 self.basename = p
161 break
162
163 if prefer_ffmpeg is False:
164 prefs = ('avprobe', 'ffprobe')
165 else:
166 prefs = ('ffprobe', 'avprobe')
167 for p in prefs:
168 if self._versions[p]:
169 self.probe_basename = p
170 break
171
172 if self.basename == 'avconv':
173 self.deprecation_warning(
174 'Support for avconv is deprecated and may be removed in a future version. Use ffmpeg instead')
175 if self.probe_basename == 'avprobe':
176 self.deprecation_warning(
177 'Support for avprobe is deprecated and may be removed in a future version. Use ffprobe instead')
178
179 @property
180 def available(self):
181 return self.basename is not None
182
183 @property
184 def executable(self):
185 return self._paths[self.basename]
186
187 @property
188 def probe_available(self):
189 return self.probe_basename is not None
190
191 @property
192 def probe_executable(self):
193 return self._paths[self.probe_basename]
194
195 @staticmethod
196 def stream_copy_opts(copy=True, *, ext=None):
197 yield from ('-map', '0')
198 # Don't copy Apple TV chapters track, bin_data
199 # See https://github.com/yt-dlp/yt-dlp/issues/2, #19042, #19024, https://trac.ffmpeg.org/ticket/6016
200 yield from ('-dn', '-ignore_unknown')
201 if copy:
202 yield from ('-c', 'copy')
203 # For some reason, '-c copy -map 0' is not enough to copy subtitles
204 if ext in ('mp4', 'mov'):
205 yield from ('-c:s', 'mov_text')
206
207 def get_audio_codec(self, path):
208 if not self.probe_available and not self.available:
209 raise PostProcessingError('ffprobe and ffmpeg not found. Please install or provide the path using --ffmpeg-location')
210 try:
211 if self.probe_available:
212 cmd = [
213 encodeFilename(self.probe_executable, True),
214 encodeArgument('-show_streams')]
215 else:
216 cmd = [
217 encodeFilename(self.executable, True),
218 encodeArgument('-i')]
219 cmd.append(encodeFilename(self._ffmpeg_filename_argument(path), True))
220 self.write_debug('%s command line: %s' % (self.basename, shell_quote(cmd)))
221 handle = Popen(cmd, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
222 stdout_data, stderr_data = handle.communicate_or_kill()
223 expected_ret = 0 if self.probe_available else 1
224 if handle.wait() != expected_ret:
225 return None
226 except (IOError, OSError):
227 return None
228 output = (stdout_data if self.probe_available else stderr_data).decode('ascii', 'ignore')
229 if self.probe_available:
230 audio_codec = None
231 for line in output.split('\n'):
232 if line.startswith('codec_name='):
233 audio_codec = line.split('=')[1].strip()
234 elif line.strip() == 'codec_type=audio' and audio_codec is not None:
235 return audio_codec
236 else:
237 # Stream #FILE_INDEX:STREAM_INDEX[STREAM_ID](LANGUAGE): CODEC_TYPE: CODEC_NAME
238 mobj = re.search(
239 r'Stream\s*#\d+:\d+(?:\[0x[0-9a-f]+\])?(?:\([a-z]{3}\))?:\s*Audio:\s*([0-9a-z]+)',
240 output)
241 if mobj:
242 return mobj.group(1)
243 return None
244
245 def get_metadata_object(self, path, opts=[]):
246 if self.probe_basename != 'ffprobe':
247 if self.probe_available:
248 self.report_warning('Only ffprobe is supported for metadata extraction')
249 raise PostProcessingError('ffprobe not found. Please install or provide the path using --ffmpeg-location')
250 self.check_version()
251
252 cmd = [
253 encodeFilename(self.probe_executable, True),
254 encodeArgument('-hide_banner'),
255 encodeArgument('-show_format'),
256 encodeArgument('-show_streams'),
257 encodeArgument('-print_format'),
258 encodeArgument('json'),
259 ]
260
261 cmd += opts
262 cmd.append(encodeFilename(self._ffmpeg_filename_argument(path), True))
263 self.write_debug('ffprobe command line: %s' % shell_quote(cmd))
264 p = Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE)
265 stdout, stderr = p.communicate()
266 return json.loads(stdout.decode('utf-8', 'replace'))
267
268 def get_stream_number(self, path, keys, value):
269 streams = self.get_metadata_object(path)['streams']
270 num = next(
271 (i for i, stream in enumerate(streams) if traverse_obj(stream, keys, casesense=False) == value),
272 None)
273 return num, len(streams)
274
275 def _get_real_video_duration(self, filepath, fatal=True):
276 try:
277 duration = float_or_none(
278 traverse_obj(self.get_metadata_object(filepath), ('format', 'duration')))
279 if not duration:
280 raise PostProcessingError('ffprobe returned empty duration')
281 return duration
282 except PostProcessingError as e:
283 if fatal:
284 raise PostProcessingError(f'Unable to determine video duration: {e.msg}')
285
286 def _duration_mismatch(self, d1, d2):
287 if not d1 or not d2:
288 return None
289 # The duration is often only known to nearest second. So there can be <1sec disparity natually.
290 # Further excuse an additional <1sec difference.
291 return abs(d1 - d2) > 2
292
293 def run_ffmpeg_multiple_files(self, input_paths, out_path, opts, **kwargs):
294 return self.real_run_ffmpeg(
295 [(path, []) for path in input_paths],
296 [(out_path, opts)], **kwargs)
297
298 def real_run_ffmpeg(self, input_path_opts, output_path_opts, *, expected_retcodes=(0,)):
299 self.check_version()
300
301 oldest_mtime = min(
302 os.stat(encodeFilename(path)).st_mtime for path, _ in input_path_opts if path)
303
304 cmd = [encodeFilename(self.executable, True), encodeArgument('-y')]
305 # avconv does not have repeat option
306 if self.basename == 'ffmpeg':
307 cmd += [encodeArgument('-loglevel'), encodeArgument('repeat+info')]
308
309 def make_args(file, args, name, number):
310 keys = ['_%s%d' % (name, number), '_%s' % name]
311 if name == 'o':
312 args += ['-movflags', '+faststart']
313 if number == 1:
314 keys.append('')
315 args += self._configuration_args(self.basename, keys)
316 if name == 'i':
317 args.append('-i')
318 return (
319 [encodeArgument(arg) for arg in args]
320 + [encodeFilename(self._ffmpeg_filename_argument(file), True)])
321
322 for arg_type, path_opts in (('i', input_path_opts), ('o', output_path_opts)):
323 cmd += itertools.chain.from_iterable(
324 make_args(path, list(opts), arg_type, i + 1)
325 for i, (path, opts) in enumerate(path_opts) if path)
326
327 self.write_debug('ffmpeg command line: %s' % shell_quote(cmd))
328 p = Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE)
329 stdout, stderr = p.communicate_or_kill()
330 if p.returncode not in variadic(expected_retcodes):
331 stderr = stderr.decode('utf-8', 'replace').strip()
332 self.write_debug(stderr)
333 raise FFmpegPostProcessorError(stderr.split('\n')[-1])
334 for out_path, _ in output_path_opts:
335 if out_path:
336 self.try_utime(out_path, oldest_mtime, oldest_mtime)
337 return stderr.decode('utf-8', 'replace')
338
339 def run_ffmpeg(self, path, out_path, opts, **kwargs):
340 return self.run_ffmpeg_multiple_files([path], out_path, opts, **kwargs)
341
342 @staticmethod
343 def _ffmpeg_filename_argument(fn):
344 # Always use 'file:' because the filename may contain ':' (ffmpeg
345 # interprets that as a protocol) or can start with '-' (-- is broken in
346 # ffmpeg, see https://ffmpeg.org/trac/ffmpeg/ticket/2127 for details)
347 # Also leave '-' intact in order not to break streaming to stdout.
348 if fn.startswith(('http://', 'https://')):
349 return fn
350 return 'file:' + fn if fn != '-' else fn
351
352 @staticmethod
353 def _quote_for_ffmpeg(string):
354 # See https://ffmpeg.org/ffmpeg-utils.html#toc-Quoting-and-escaping
355 # A sequence of '' produces '\'''\'';
356 # final replace removes the empty '' between \' \'.
357 string = string.replace("'", r"'\''").replace("'''", "'")
358 # Handle potential ' at string boundaries.
359 string = string[1:] if string[0] == "'" else "'" + string
360 return string[:-1] if string[-1] == "'" else string + "'"
361
362 def force_keyframes(self, filename, timestamps):
363 timestamps = orderedSet(timestamps)
364 if timestamps[0] == 0:
365 timestamps = timestamps[1:]
366 keyframe_file = prepend_extension(filename, 'keyframes.temp')
367 self.to_screen(f'Re-encoding "{filename}" with appropriate keyframes')
368 self.run_ffmpeg(filename, keyframe_file, [
369 *self.stream_copy_opts(False, ext=determine_ext(filename)),
370 '-force_key_frames', ','.join(f'{t:.6f}' for t in timestamps)])
371 return keyframe_file
372
373 def concat_files(self, in_files, out_file, concat_opts=None):
374 """
375 Use concat demuxer to concatenate multiple files having identical streams.
376
377 Only inpoint, outpoint, and duration concat options are supported.
378 See https://ffmpeg.org/ffmpeg-formats.html#concat-1 for details
379 """
380 concat_file = f'{out_file}.concat'
381 self.write_debug(f'Writing concat spec to {concat_file}')
382 with open(concat_file, 'wt', encoding='utf-8') as f:
383 f.writelines(self._concat_spec(in_files, concat_opts))
384
385 out_flags = list(self.stream_copy_opts(ext=determine_ext(out_file)))
386
387 try:
388 self.real_run_ffmpeg(
389 [(concat_file, ['-hide_banner', '-nostdin', '-f', 'concat', '-safe', '0'])],
390 [(out_file, out_flags)])
391 finally:
392 os.remove(concat_file)
393
394 @classmethod
395 def _concat_spec(cls, in_files, concat_opts=None):
396 if concat_opts is None:
397 concat_opts = [{}] * len(in_files)
398 yield 'ffconcat version 1.0\n'
399 for file, opts in zip(in_files, concat_opts):
400 yield f'file {cls._quote_for_ffmpeg(cls._ffmpeg_filename_argument(file))}\n'
401 # Iterate explicitly to yield the following directives in order, ignoring the rest.
402 for directive in 'inpoint', 'outpoint', 'duration':
403 if directive in opts:
404 yield f'{directive} {opts[directive]}\n'
405
406
407 class FFmpegExtractAudioPP(FFmpegPostProcessor):
408 COMMON_AUDIO_EXTS = ('wav', 'flac', 'm4a', 'aiff', 'mp3', 'ogg', 'mka', 'opus', 'wma')
409 SUPPORTED_EXTS = ('best', 'aac', 'flac', 'mp3', 'm4a', 'opus', 'vorbis', 'wav', 'alac')
410
411 def __init__(self, downloader=None, preferredcodec=None, preferredquality=None, nopostoverwrites=False):
412 FFmpegPostProcessor.__init__(self, downloader)
413 self._preferredcodec = preferredcodec or 'best'
414 self._preferredquality = float_or_none(preferredquality)
415 self._nopostoverwrites = nopostoverwrites
416
417 def _quality_args(self, codec):
418 if self._preferredquality is None:
419 return []
420 elif self._preferredquality > 10:
421 return ['-b:a', f'{self._preferredquality}k']
422
423 limits = {
424 'libmp3lame': (10, 0),
425 'libvorbis': (0, 10),
426 # FFmpeg's AAC encoder does not have an upper limit for the value of -q:a.
427 # Experimentally, with values over 4, bitrate changes were minimal or non-existent
428 'aac': (0.1, 4),
429 'libfdk_aac': (1, 5),
430 }.get(codec)
431 if not limits:
432 return []
433
434 q = limits[1] + (limits[0] - limits[1]) * (self._preferredquality / 10)
435 if codec == 'libfdk_aac':
436 return ['-vbr', f'{int(q)}']
437 return ['-q:a', f'{q}']
438
439 def run_ffmpeg(self, path, out_path, codec, more_opts):
440 if codec is None:
441 acodec_opts = []
442 else:
443 acodec_opts = ['-acodec', codec]
444 opts = ['-vn'] + acodec_opts + more_opts
445 try:
446 FFmpegPostProcessor.run_ffmpeg(self, path, out_path, opts)
447 except FFmpegPostProcessorError as err:
448 raise AudioConversionError(err.msg)
449
450 @PostProcessor._restrict_to(images=False)
451 def run(self, information):
452 orig_path = path = information['filepath']
453 orig_ext = information['ext']
454
455 if self._preferredcodec == 'best' and orig_ext in self.COMMON_AUDIO_EXTS:
456 self.to_screen('Skipping audio extraction since the file is already in a common audio format')
457 return [], information
458
459 filecodec = self.get_audio_codec(path)
460 if filecodec is None:
461 raise PostProcessingError('WARNING: unable to obtain file audio codec with ffprobe')
462
463 more_opts = []
464 if self._preferredcodec == 'best' or self._preferredcodec == filecodec or (self._preferredcodec == 'm4a' and filecodec == 'aac'):
465 if filecodec == 'aac' and self._preferredcodec in ['m4a', 'best']:
466 # Lossless, but in another container
467 acodec = 'copy'
468 extension = 'm4a'
469 more_opts = ['-bsf:a', 'aac_adtstoasc']
470 elif filecodec in ['aac', 'flac', 'mp3', 'vorbis', 'opus']:
471 # Lossless if possible
472 acodec = 'copy'
473 extension = filecodec
474 if filecodec == 'aac':
475 more_opts = ['-f', 'adts']
476 if filecodec == 'vorbis':
477 extension = 'ogg'
478 elif filecodec == 'alac':
479 acodec = None
480 extension = 'm4a'
481 more_opts += ['-acodec', 'alac']
482 else:
483 # MP3 otherwise.
484 acodec = 'libmp3lame'
485 extension = 'mp3'
486 more_opts = self._quality_args(acodec)
487 else:
488 # We convert the audio (lossy if codec is lossy)
489 acodec = ACODECS[self._preferredcodec]
490 if acodec == 'aac' and self._features.get('fdk'):
491 acodec = 'libfdk_aac'
492 extension = self._preferredcodec
493 more_opts = self._quality_args(acodec)
494 if self._preferredcodec == 'aac':
495 more_opts += ['-f', 'adts']
496 elif self._preferredcodec == 'm4a':
497 more_opts += ['-bsf:a', 'aac_adtstoasc']
498 elif self._preferredcodec == 'vorbis':
499 extension = 'ogg'
500 elif self._preferredcodec == 'wav':
501 extension = 'wav'
502 more_opts += ['-f', 'wav']
503 elif self._preferredcodec == 'alac':
504 extension = 'm4a'
505 more_opts += ['-acodec', 'alac']
506
507 prefix, sep, ext = path.rpartition('.') # not os.path.splitext, since the latter does not work on unicode in all setups
508 temp_path = new_path = prefix + sep + extension
509
510 if new_path == path:
511 orig_path = prepend_extension(path, 'orig')
512 temp_path = prepend_extension(path, 'temp')
513 if (self._nopostoverwrites and os.path.exists(encodeFilename(new_path))
514 and os.path.exists(encodeFilename(orig_path))):
515 self.to_screen('Post-process file %s exists, skipping' % new_path)
516 return [], information
517
518 try:
519 self.to_screen(f'Destination: {new_path}')
520 self.run_ffmpeg(path, temp_path, acodec, more_opts)
521 except AudioConversionError as e:
522 raise PostProcessingError(
523 'audio conversion failed: ' + e.msg)
524 except Exception:
525 raise PostProcessingError('error running ' + self.basename)
526
527 os.replace(path, orig_path)
528 os.replace(temp_path, new_path)
529 information['filepath'] = new_path
530 information['ext'] = extension
531
532 # Try to update the date time for extracted audio file.
533 if information.get('filetime') is not None:
534 self.try_utime(
535 new_path, time.time(), information['filetime'],
536 errnote='Cannot update utime of audio file')
537
538 return [orig_path], information
539
540
541 class FFmpegVideoConvertorPP(FFmpegPostProcessor):
542 SUPPORTED_EXTS = ('mp4', 'mkv', 'flv', 'webm', 'mov', 'avi', 'mp3', 'mka', 'm4a', 'ogg', 'opus')
543 FORMAT_RE = re.compile(r'{0}(?:/{0})*$'.format(r'(?:\w+>)?(?:%s)' % '|'.join(SUPPORTED_EXTS)))
544 _ACTION = 'converting'
545
546 def __init__(self, downloader=None, preferedformat=None):
547 super(FFmpegVideoConvertorPP, self).__init__(downloader)
548 self._preferedformats = preferedformat.lower().split('/')
549
550 def _target_ext(self, source_ext):
551 for pair in self._preferedformats:
552 kv = pair.split('>')
553 if len(kv) == 1 or kv[0].strip() == source_ext:
554 return kv[-1].strip()
555
556 @staticmethod
557 def _options(target_ext):
558 if target_ext == 'avi':
559 return ['-c:v', 'libxvid', '-vtag', 'XVID']
560 return []
561
562 @PostProcessor._restrict_to(images=False)
563 def run(self, info):
564 filename, source_ext = info['filepath'], info['ext'].lower()
565 target_ext = self._target_ext(source_ext)
566 _skip_msg = (
567 f'could not find a mapping for {source_ext}' if not target_ext
568 else f'already is in target format {source_ext}' if source_ext == target_ext
569 else None)
570 if _skip_msg:
571 self.to_screen(f'Not {self._ACTION} media file {filename!r}; {_skip_msg}')
572 return [], info
573
574 outpath = replace_extension(filename, target_ext, source_ext)
575 self.to_screen(f'{self._ACTION.title()} video from {source_ext} to {target_ext}; Destination: {outpath}')
576 self.run_ffmpeg(filename, outpath, self._options(target_ext))
577
578 info['filepath'] = outpath
579 info['format'] = info['ext'] = target_ext
580 return [filename], info
581
582
583 class FFmpegVideoRemuxerPP(FFmpegVideoConvertorPP):
584 _ACTION = 'remuxing'
585
586 @staticmethod
587 def _options(target_ext):
588 return FFmpegPostProcessor.stream_copy_opts()
589
590
591 class FFmpegEmbedSubtitlePP(FFmpegPostProcessor):
592 def __init__(self, downloader=None, already_have_subtitle=False):
593 super(FFmpegEmbedSubtitlePP, self).__init__(downloader)
594 self._already_have_subtitle = already_have_subtitle
595
596 @PostProcessor._restrict_to(images=False)
597 def run(self, info):
598 if info['ext'] not in ('mp4', 'webm', 'mkv'):
599 self.to_screen('Subtitles can only be embedded in mp4, webm or mkv files')
600 return [], info
601 subtitles = info.get('requested_subtitles')
602 if not subtitles:
603 self.to_screen('There aren\'t any subtitles to embed')
604 return [], info
605
606 filename = info['filepath']
607
608 # Disabled temporarily. There needs to be a way to overide this
609 # in case of duration actually mismatching in extractor
610 # See: https://github.com/yt-dlp/yt-dlp/issues/1870, https://github.com/yt-dlp/yt-dlp/issues/1385
611 '''
612 if info.get('duration') and not info.get('__real_download') and self._duration_mismatch(
613 self._get_real_video_duration(filename, False), info['duration']):
614 self.to_screen(f'Skipping {self.pp_key()} since the real and expected durations mismatch')
615 return [], info
616 '''
617
618 ext = info['ext']
619 sub_langs, sub_names, sub_filenames = [], [], []
620 webm_vtt_warn = False
621 mp4_ass_warn = False
622
623 for lang, sub_info in subtitles.items():
624 if not os.path.exists(sub_info.get('filepath', '')):
625 self.report_warning(f'Skipping embedding {lang} subtitle because the file is missing')
626 continue
627 sub_ext = sub_info['ext']
628 if sub_ext == 'json':
629 self.report_warning('JSON subtitles cannot be embedded')
630 elif ext != 'webm' or ext == 'webm' and sub_ext == 'vtt':
631 sub_langs.append(lang)
632 sub_names.append(sub_info.get('name'))
633 sub_filenames.append(sub_info['filepath'])
634 else:
635 if not webm_vtt_warn and ext == 'webm' and sub_ext != 'vtt':
636 webm_vtt_warn = True
637 self.report_warning('Only WebVTT subtitles can be embedded in webm files')
638 if not mp4_ass_warn and ext == 'mp4' and sub_ext == 'ass':
639 mp4_ass_warn = True
640 self.report_warning('ASS subtitles cannot be properly embedded in mp4 files; expect issues')
641
642 if not sub_langs:
643 return [], info
644
645 input_files = [filename] + sub_filenames
646
647 opts = [
648 *self.stream_copy_opts(ext=info['ext']),
649 # Don't copy the existing subtitles, we may be running the
650 # postprocessor a second time
651 '-map', '-0:s',
652 ]
653 for i, (lang, name) in enumerate(zip(sub_langs, sub_names)):
654 opts.extend(['-map', '%d:0' % (i + 1)])
655 lang_code = ISO639Utils.short2long(lang) or lang
656 opts.extend(['-metadata:s:s:%d' % i, 'language=%s' % lang_code])
657 if name:
658 opts.extend(['-metadata:s:s:%d' % i, 'handler_name=%s' % name,
659 '-metadata:s:s:%d' % i, 'title=%s' % name])
660
661 temp_filename = prepend_extension(filename, 'temp')
662 self.to_screen('Embedding subtitles in "%s"' % filename)
663 self.run_ffmpeg_multiple_files(input_files, temp_filename, opts)
664 os.replace(temp_filename, filename)
665
666 files_to_delete = [] if self._already_have_subtitle else sub_filenames
667 return files_to_delete, info
668
669
670 class FFmpegMetadataPP(FFmpegPostProcessor):
671
672 def __init__(self, downloader, add_metadata=True, add_chapters=True, add_infojson='if_exists'):
673 FFmpegPostProcessor.__init__(self, downloader)
674 self._add_metadata = add_metadata
675 self._add_chapters = add_chapters
676 self._add_infojson = add_infojson
677
678 @staticmethod
679 def _options(target_ext):
680 audio_only = target_ext == 'm4a'
681 yield from FFmpegPostProcessor.stream_copy_opts(not audio_only)
682 if audio_only:
683 yield from ('-vn', '-acodec', 'copy')
684
685 @PostProcessor._restrict_to(images=False)
686 def run(self, info):
687 filename, metadata_filename = info['filepath'], None
688 files_to_delete, options = [], []
689 if self._add_chapters and info.get('chapters'):
690 metadata_filename = replace_extension(filename, 'meta')
691 options.extend(self._get_chapter_opts(info['chapters'], metadata_filename))
692 files_to_delete.append(metadata_filename)
693 if self._add_metadata:
694 options.extend(self._get_metadata_opts(info))
695
696 if self._add_infojson:
697 if info['ext'] in ('mkv', 'mka'):
698 infojson_filename = info.get('infojson_filename')
699 options.extend(self._get_infojson_opts(info, infojson_filename))
700 if not infojson_filename:
701 files_to_delete.append(info.get('infojson_filename'))
702 elif self._add_infojson is True:
703 self.to_screen('The info-json can only be attached to mkv/mka files')
704
705 if not options:
706 self.to_screen('There isn\'t any metadata to add')
707 return [], info
708
709 temp_filename = prepend_extension(filename, 'temp')
710 self.to_screen('Adding metadata to "%s"' % filename)
711 self.run_ffmpeg_multiple_files(
712 (filename, metadata_filename), temp_filename,
713 itertools.chain(self._options(info['ext']), *options))
714 for file in filter(None, files_to_delete):
715 os.remove(file) # Don't obey --keep-files
716 os.replace(temp_filename, filename)
717 return [], info
718
719 @staticmethod
720 def _get_chapter_opts(chapters, metadata_filename):
721 with io.open(metadata_filename, 'wt', encoding='utf-8') as f:
722 def ffmpeg_escape(text):
723 return re.sub(r'([\\=;#\n])', r'\\\1', text)
724
725 metadata_file_content = ';FFMETADATA1\n'
726 for chapter in chapters:
727 metadata_file_content += '[CHAPTER]\nTIMEBASE=1/1000\n'
728 metadata_file_content += 'START=%d\n' % (chapter['start_time'] * 1000)
729 metadata_file_content += 'END=%d\n' % (chapter['end_time'] * 1000)
730 chapter_title = chapter.get('title')
731 if chapter_title:
732 metadata_file_content += 'title=%s\n' % ffmpeg_escape(chapter_title)
733 f.write(metadata_file_content)
734 yield ('-map_metadata', '1')
735
736 def _get_metadata_opts(self, info):
737 meta_prefix = 'meta'
738 metadata = collections.defaultdict(dict)
739
740 def add(meta_list, info_list=None):
741 value = next((
742 str(info[key]) for key in [f'{meta_prefix}_'] + list(variadic(info_list or meta_list))
743 if info.get(key) is not None), None)
744 if value not in ('', None):
745 metadata['common'].update({meta_f: value for meta_f in variadic(meta_list)})
746
747 # See [1-4] for some info on media metadata/metadata supported
748 # by ffmpeg.
749 # 1. https://kdenlive.org/en/project/adding-meta-data-to-mp4-video/
750 # 2. https://wiki.multimedia.cx/index.php/FFmpeg_Metadata
751 # 3. https://kodi.wiki/view/Video_file_tagging
752
753 add('title', ('track', 'title'))
754 add('date', 'upload_date')
755 add(('description', 'synopsis'), 'description')
756 add(('purl', 'comment'), 'webpage_url')
757 add('track', 'track_number')
758 add('artist', ('artist', 'creator', 'uploader', 'uploader_id'))
759 add('genre')
760 add('album')
761 add('album_artist')
762 add('disc', 'disc_number')
763 add('show', 'series')
764 add('season_number')
765 add('episode_id', ('episode', 'episode_id'))
766 add('episode_sort', 'episode_number')
767 if 'embed-metadata' in self.get_param('compat_opts', []):
768 add('comment', 'description')
769 metadata['common'].pop('synopsis', None)
770
771 meta_regex = rf'{re.escape(meta_prefix)}(?P<i>\d+)?_(?P<key>.+)'
772 for key, value in info.items():
773 mobj = re.fullmatch(meta_regex, key)
774 if value is not None and mobj:
775 metadata[mobj.group('i') or 'common'][mobj.group('key')] = value
776
777 for name, value in metadata['common'].items():
778 yield ('-metadata', f'{name}={value}')
779
780 stream_idx = 0
781 for fmt in info.get('requested_formats') or []:
782 stream_count = 2 if 'none' not in (fmt.get('vcodec'), fmt.get('acodec')) else 1
783 lang = ISO639Utils.short2long(fmt.get('language') or '') or fmt.get('language')
784 for i in range(stream_idx, stream_idx + stream_count):
785 if lang:
786 metadata[str(i)].setdefault('language', lang)
787 for name, value in metadata[str(i)].items():
788 yield (f'-metadata:s:{i}', f'{name}={value}')
789 stream_idx += stream_count
790
791 def _get_infojson_opts(self, info, infofn):
792 if not infofn or not os.path.exists(infofn):
793 if self._add_infojson is not True:
794 return
795 infofn = infofn or '%s.temp' % (
796 self._downloader.prepare_filename(info, 'infojson')
797 or replace_extension(self._downloader.prepare_filename(info), 'info.json', info['ext']))
798 if not self._downloader._ensure_dir_exists(infofn):
799 return
800 self.write_debug(f'Writing info-json to: {infofn}')
801 write_json_file(self._downloader.sanitize_info(info, self.get_param('clean_infojson', True)), infofn)
802 info['infojson_filename'] = infofn
803
804 old_stream, new_stream = self.get_stream_number(info['filepath'], ('tags', 'mimetype'), 'application/json')
805 if old_stream is not None:
806 yield ('-map', '-0:%d' % old_stream)
807 new_stream -= 1
808
809 yield ('-attach', infofn,
810 '-metadata:s:%d' % new_stream, 'mimetype=application/json')
811
812
813 class FFmpegMergerPP(FFmpegPostProcessor):
814 @PostProcessor._restrict_to(images=False)
815 def run(self, info):
816 filename = info['filepath']
817 temp_filename = prepend_extension(filename, 'temp')
818 args = ['-c', 'copy']
819 audio_streams = 0
820 for (i, fmt) in enumerate(info['requested_formats']):
821 if fmt.get('acodec') != 'none':
822 args.extend(['-map', f'{i}:a:0'])
823 aac_fixup = fmt['protocol'].startswith('m3u8') and self.get_audio_codec(fmt['filepath']) == 'aac'
824 if aac_fixup:
825 args.extend([f'-bsf:a:{audio_streams}', 'aac_adtstoasc'])
826 audio_streams += 1
827 if fmt.get('vcodec') != 'none':
828 args.extend(['-map', '%u:v:0' % (i)])
829 self.to_screen('Merging formats into "%s"' % filename)
830 self.run_ffmpeg_multiple_files(info['__files_to_merge'], temp_filename, args)
831 os.rename(encodeFilename(temp_filename), encodeFilename(filename))
832 return info['__files_to_merge'], info
833
834 def can_merge(self):
835 # TODO: figure out merge-capable ffmpeg version
836 if self.basename != 'avconv':
837 return True
838
839 required_version = '10-0'
840 if is_outdated_version(
841 self._versions[self.basename], required_version):
842 warning = ('Your copy of %s is outdated and unable to properly mux separate video and audio files, '
843 'yt-dlp will download single file media. '
844 'Update %s to version %s or newer to fix this.') % (
845 self.basename, self.basename, required_version)
846 self.report_warning(warning)
847 return False
848 return True
849
850
851 class FFmpegFixupPostProcessor(FFmpegPostProcessor):
852 def _fixup(self, msg, filename, options):
853 temp_filename = prepend_extension(filename, 'temp')
854
855 self.to_screen(f'{msg} of "{filename}"')
856 self.run_ffmpeg(filename, temp_filename, options)
857
858 os.replace(temp_filename, filename)
859
860
861 class FFmpegFixupStretchedPP(FFmpegFixupPostProcessor):
862 @PostProcessor._restrict_to(images=False, audio=False)
863 def run(self, info):
864 stretched_ratio = info.get('stretched_ratio')
865 if stretched_ratio not in (None, 1):
866 self._fixup('Fixing aspect ratio', info['filepath'], [
867 *self.stream_copy_opts(), '-aspect', '%f' % stretched_ratio])
868 return [], info
869
870
871 class FFmpegFixupM4aPP(FFmpegFixupPostProcessor):
872 @PostProcessor._restrict_to(images=False, video=False)
873 def run(self, info):
874 if info.get('container') == 'm4a_dash':
875 self._fixup('Correcting container', info['filepath'], [*self.stream_copy_opts(), '-f', 'mp4'])
876 return [], info
877
878
879 class FFmpegFixupM3u8PP(FFmpegFixupPostProcessor):
880 def _needs_fixup(self, info):
881 yield info['ext'] in ('mp4', 'm4a')
882 yield info['protocol'].startswith('m3u8')
883 try:
884 metadata = self.get_metadata_object(info['filepath'])
885 except PostProcessingError as e:
886 self.report_warning(f'Unable to extract metadata: {e.msg}')
887 yield True
888 else:
889 yield traverse_obj(metadata, ('format', 'format_name'), casesense=False) == 'mpegts'
890
891 @PostProcessor._restrict_to(images=False)
892 def run(self, info):
893 if all(self._needs_fixup(info)):
894 self._fixup('Fixing MPEG-TS in MP4 container', info['filepath'], [
895 *self.stream_copy_opts(), '-f', 'mp4', '-bsf:a', 'aac_adtstoasc'])
896 return [], info
897
898
899 class FFmpegFixupTimestampPP(FFmpegFixupPostProcessor):
900
901 def __init__(self, downloader=None, trim=0.001):
902 # "trim" should be used when the video contains unintended packets
903 super(FFmpegFixupTimestampPP, self).__init__(downloader)
904 assert isinstance(trim, (int, float))
905 self.trim = str(trim)
906
907 @PostProcessor._restrict_to(images=False)
908 def run(self, info):
909 if not self._features.get('setts'):
910 self.report_warning(
911 'A re-encode is needed to fix timestamps in older versions of ffmpeg. '
912 'Please install ffmpeg 4.4 or later to fixup without re-encoding')
913 opts = ['-vf', 'setpts=PTS-STARTPTS']
914 else:
915 opts = ['-c', 'copy', '-bsf', 'setts=ts=TS-STARTPTS']
916 self._fixup('Fixing frame timestamp', info['filepath'], opts + [*self.stream_copy_opts(False), '-ss', self.trim])
917 return [], info
918
919
920 class FFmpegCopyStreamPostProcessor(FFmpegFixupPostProcessor):
921 MESSAGE = 'Copying stream'
922
923 @PostProcessor._restrict_to(images=False)
924 def run(self, info):
925 self._fixup(self.MESSAGE, info['filepath'], self.stream_copy_opts())
926 return [], info
927
928
929 class FFmpegFixupDurationPP(FFmpegCopyStreamPostProcessor):
930 MESSAGE = 'Fixing video duration'
931
932
933 class FFmpegFixupDuplicateMoovPP(FFmpegCopyStreamPostProcessor):
934 MESSAGE = 'Fixing duplicate MOOV atoms'
935
936
937 class FFmpegSubtitlesConvertorPP(FFmpegPostProcessor):
938 SUPPORTED_EXTS = ('srt', 'vtt', 'ass', 'lrc')
939
940 def __init__(self, downloader=None, format=None):
941 super(FFmpegSubtitlesConvertorPP, self).__init__(downloader)
942 self.format = format
943
944 def run(self, info):
945 subs = info.get('requested_subtitles')
946 new_ext = self.format
947 new_format = new_ext
948 if new_format == 'vtt':
949 new_format = 'webvtt'
950 if subs is None:
951 self.to_screen('There aren\'t any subtitles to convert')
952 return [], info
953 self.to_screen('Converting subtitles')
954 sub_filenames = []
955 for lang, sub in subs.items():
956 if not os.path.exists(sub.get('filepath', '')):
957 self.report_warning(f'Skipping embedding {lang} subtitle because the file is missing')
958 continue
959 ext = sub['ext']
960 if ext == new_ext:
961 self.to_screen('Subtitle file for %s is already in the requested format' % new_ext)
962 continue
963 elif ext == 'json':
964 self.to_screen(
965 'You have requested to convert json subtitles into another format, '
966 'which is currently not possible')
967 continue
968 old_file = sub['filepath']
969 sub_filenames.append(old_file)
970 new_file = replace_extension(old_file, new_ext)
971
972 if ext in ('dfxp', 'ttml', 'tt'):
973 self.report_warning(
974 'You have requested to convert dfxp (TTML) subtitles into another format, '
975 'which results in style information loss')
976
977 dfxp_file = old_file
978 srt_file = replace_extension(old_file, 'srt')
979
980 with open(dfxp_file, 'rb') as f:
981 srt_data = dfxp2srt(f.read())
982
983 with io.open(srt_file, 'wt', encoding='utf-8') as f:
984 f.write(srt_data)
985 old_file = srt_file
986
987 subs[lang] = {
988 'ext': 'srt',
989 'data': srt_data,
990 'filepath': srt_file,
991 }
992
993 if new_ext == 'srt':
994 continue
995 else:
996 sub_filenames.append(srt_file)
997
998 self.run_ffmpeg(old_file, new_file, ['-f', new_format])
999
1000 with io.open(new_file, 'rt', encoding='utf-8') as f:
1001 subs[lang] = {
1002 'ext': new_ext,
1003 'data': f.read(),
1004 'filepath': new_file,
1005 }
1006
1007 info['__files_to_move'][new_file] = replace_extension(
1008 info['__files_to_move'][sub['filepath']], new_ext)
1009
1010 return sub_filenames, info
1011
1012
1013 class FFmpegSplitChaptersPP(FFmpegPostProcessor):
1014 def __init__(self, downloader, force_keyframes=False):
1015 FFmpegPostProcessor.__init__(self, downloader)
1016 self._force_keyframes = force_keyframes
1017
1018 def _prepare_filename(self, number, chapter, info):
1019 info = info.copy()
1020 info.update({
1021 'section_number': number,
1022 'section_title': chapter.get('title'),
1023 'section_start': chapter.get('start_time'),
1024 'section_end': chapter.get('end_time'),
1025 })
1026 return self._downloader.prepare_filename(info, 'chapter')
1027
1028 def _ffmpeg_args_for_chapter(self, number, chapter, info):
1029 destination = self._prepare_filename(number, chapter, info)
1030 if not self._downloader._ensure_dir_exists(encodeFilename(destination)):
1031 return
1032
1033 chapter['filepath'] = destination
1034 self.to_screen('Chapter %03d; Destination: %s' % (number, destination))
1035 return (
1036 destination,
1037 ['-ss', compat_str(chapter['start_time']),
1038 '-t', compat_str(chapter['end_time'] - chapter['start_time'])])
1039
1040 @PostProcessor._restrict_to(images=False)
1041 def run(self, info):
1042 chapters = info.get('chapters') or []
1043 if not chapters:
1044 self.to_screen('Chapter information is unavailable')
1045 return [], info
1046
1047 in_file = info['filepath']
1048 if self._force_keyframes and len(chapters) > 1:
1049 in_file = self.force_keyframes(in_file, (c['start_time'] for c in chapters))
1050 self.to_screen('Splitting video by chapters; %d chapters found' % len(chapters))
1051 for idx, chapter in enumerate(chapters):
1052 destination, opts = self._ffmpeg_args_for_chapter(idx + 1, chapter, info)
1053 self.real_run_ffmpeg([(in_file, opts)], [(destination, self.stream_copy_opts())])
1054 if in_file != info['filepath']:
1055 os.remove(in_file)
1056 return [], info
1057
1058
1059 class FFmpegThumbnailsConvertorPP(FFmpegPostProcessor):
1060 SUPPORTED_EXTS = ('jpg', 'png')
1061
1062 def __init__(self, downloader=None, format=None):
1063 super(FFmpegThumbnailsConvertorPP, self).__init__(downloader)
1064 self.format = format
1065
1066 @staticmethod
1067 def is_webp(path):
1068 with open(encodeFilename(path), 'rb') as f:
1069 b = f.read(12)
1070 return b[0:4] == b'RIFF' and b[8:] == b'WEBP'
1071
1072 def fixup_webp(self, info, idx=-1):
1073 thumbnail_filename = info['thumbnails'][idx]['filepath']
1074 _, thumbnail_ext = os.path.splitext(thumbnail_filename)
1075 if thumbnail_ext:
1076 thumbnail_ext = thumbnail_ext[1:].lower()
1077 if thumbnail_ext != 'webp' and self.is_webp(thumbnail_filename):
1078 self.to_screen('Correcting thumbnail "%s" extension to webp' % thumbnail_filename)
1079 webp_filename = replace_extension(thumbnail_filename, 'webp')
1080 os.replace(thumbnail_filename, webp_filename)
1081 info['thumbnails'][idx]['filepath'] = webp_filename
1082 info['__files_to_move'][webp_filename] = replace_extension(
1083 info['__files_to_move'].pop(thumbnail_filename), 'webp')
1084
1085 @staticmethod
1086 def _options(target_ext):
1087 if target_ext == 'jpg':
1088 return ['-bsf:v', 'mjpeg2jpeg']
1089 return []
1090
1091 def convert_thumbnail(self, thumbnail_filename, target_ext):
1092 thumbnail_conv_filename = replace_extension(thumbnail_filename, target_ext)
1093
1094 self.to_screen('Converting thumbnail "%s" to %s' % (thumbnail_filename, target_ext))
1095 self.real_run_ffmpeg(
1096 [(thumbnail_filename, ['-f', 'image2', '-pattern_type', 'none'])],
1097 [(thumbnail_conv_filename.replace('%', '%%'), self._options(target_ext))])
1098 return thumbnail_conv_filename
1099
1100 def run(self, info):
1101 files_to_delete = []
1102 has_thumbnail = False
1103
1104 for idx, thumbnail_dict in enumerate(info.get('thumbnails') or []):
1105 original_thumbnail = thumbnail_dict.get('filepath')
1106 if not original_thumbnail:
1107 continue
1108 has_thumbnail = True
1109 self.fixup_webp(info, idx)
1110 _, thumbnail_ext = os.path.splitext(original_thumbnail)
1111 if thumbnail_ext:
1112 thumbnail_ext = thumbnail_ext[1:].lower()
1113 if thumbnail_ext == 'jpeg':
1114 thumbnail_ext = 'jpg'
1115 if thumbnail_ext == self.format:
1116 self.to_screen('Thumbnail "%s" is already in the requested format' % original_thumbnail)
1117 continue
1118 thumbnail_dict['filepath'] = self.convert_thumbnail(original_thumbnail, self.format)
1119 files_to_delete.append(original_thumbnail)
1120 info['__files_to_move'][thumbnail_dict['filepath']] = replace_extension(
1121 info['__files_to_move'][original_thumbnail], self.format)
1122
1123 if not has_thumbnail:
1124 self.to_screen('There aren\'t any thumbnails to convert')
1125 return files_to_delete, info
1126
1127
1128 class FFmpegConcatPP(FFmpegPostProcessor):
1129 def __init__(self, downloader, only_multi_video=False):
1130 self._only_multi_video = only_multi_video
1131 super().__init__(downloader)
1132
1133 def concat_files(self, in_files, out_file):
1134 if len(in_files) == 1:
1135 os.replace(in_files[0], out_file)
1136 return
1137
1138 codecs = [traverse_obj(self.get_metadata_object(file), ('streams', ..., 'codec_name')) for file in in_files]
1139 if len(set(map(tuple, codecs))) > 1:
1140 raise PostProcessingError(
1141 'The files have different streams/codecs and cannot be concatenated. '
1142 'Either select different formats or --recode-video them to a common format')
1143 super().concat_files(in_files, out_file)
1144
1145 @PostProcessor._restrict_to(images=False)
1146 def run(self, info):
1147 if not info.get('entries') or self._only_multi_video and info['_type'] != 'multi_video':
1148 return [], info
1149 elif None in info['entries']:
1150 raise PostProcessingError('Aborting concatenation because some downloads failed')
1151 elif any(len(entry) > 1 for entry in traverse_obj(info, ('entries', ..., 'requested_downloads')) or []):
1152 raise PostProcessingError('Concatenation is not supported when downloading multiple separate formats')
1153
1154 in_files = traverse_obj(info, ('entries', ..., 'requested_downloads', 0, 'filepath'))
1155 if not in_files:
1156 self.to_screen('There are no files to concatenate')
1157 return [], info
1158
1159 ie_copy = self._downloader._playlist_infodict(info)
1160 exts = [traverse_obj(entry, ('requested_downloads', 0, 'ext'), 'ext') for entry in info['entries']]
1161 ie_copy['ext'] = exts[0] if len(set(exts)) == 1 else 'mkv'
1162 out_file = self._downloader.prepare_filename(ie_copy, 'pl_video')
1163
1164 self.concat_files(in_files, out_file)
1165
1166 info['requested_downloads'] = [{
1167 'filepath': out_file,
1168 'ext': ie_copy['ext'],
1169 }]
1170 return in_files, info