]> jfr.im git - yt-dlp.git/blob - yt_dlp/postprocessor/ffmpeg.py
4b61693a206cd1a7001599da4459248010f08845
[yt-dlp.git] / yt_dlp / postprocessor / ffmpeg.py
1 import collections
2 import itertools
3 import json
4 import os
5 import re
6 import subprocess
7 import time
8
9 from .common import AudioConversionError, PostProcessor
10 from ..compat import compat_str
11 from ..utils import (
12 ISO639Utils,
13 Popen,
14 PostProcessingError,
15 _get_exe_version_output,
16 detect_exe_version,
17 determine_ext,
18 dfxp2srt,
19 encodeArgument,
20 encodeFilename,
21 float_or_none,
22 is_outdated_version,
23 orderedSet,
24 prepend_extension,
25 replace_extension,
26 shell_quote,
27 traverse_obj,
28 variadic,
29 write_json_file,
30 )
31
32 EXT_TO_OUT_FORMATS = {
33 'aac': 'adts',
34 'flac': 'flac',
35 'm4a': 'ipod',
36 'mka': 'matroska',
37 'mkv': 'matroska',
38 'mpg': 'mpeg',
39 'ogv': 'ogg',
40 'ts': 'mpegts',
41 'wma': 'asf',
42 'wmv': 'asf',
43 'vtt': 'webvtt',
44 }
45 ACODECS = {
46 'mp3': 'libmp3lame',
47 'aac': 'aac',
48 'flac': 'flac',
49 'm4a': 'aac',
50 'opus': 'libopus',
51 'vorbis': 'libvorbis',
52 'wav': None,
53 'alac': None,
54 }
55
56
57 class FFmpegPostProcessorError(PostProcessingError):
58 pass
59
60
61 class FFmpegPostProcessor(PostProcessor):
62 def __init__(self, downloader=None):
63 PostProcessor.__init__(self, downloader)
64 self._determine_executables()
65
66 def check_version(self):
67 if not self.available:
68 raise FFmpegPostProcessorError('ffmpeg not found. Please install or provide the path using --ffmpeg-location')
69
70 required_version = '10-0' if self.basename == 'avconv' else '1.0'
71 if is_outdated_version(self._versions[self.basename], required_version):
72 self.report_warning(f'Your copy of {self.basename} is outdated, update {self.basename} '
73 f'to version {required_version} or newer if you encounter any errors')
74
75 @staticmethod
76 def get_versions_and_features(downloader=None):
77 pp = FFmpegPostProcessor(downloader)
78 return pp._versions, pp._features
79
80 @staticmethod
81 def get_versions(downloader=None):
82 return FFmpegPostProcessor.get_versions_and_features(downloader)[0]
83
84 _version_cache, _features_cache = {}, {}
85
86 def _determine_executables(self):
87 programs = ['avprobe', 'avconv', 'ffmpeg', 'ffprobe']
88
89 def get_ffmpeg_version(path, prog):
90 if path in self._version_cache:
91 self._versions[prog], self._features = self._version_cache[path], self._features_cache.get(path, {})
92 return
93 out = _get_exe_version_output(path, ['-bsfs'], to_screen=self.write_debug)
94 ver = detect_exe_version(out) if out else False
95 if ver:
96 regexs = [
97 r'(?:\d+:)?([0-9.]+)-[0-9]+ubuntu[0-9.]+$', # Ubuntu, see [1]
98 r'n([0-9.]+)$', # Arch Linux
99 # 1. http://www.ducea.com/2006/06/17/ubuntu-package-version-naming-explanation/
100 ]
101 for regex in regexs:
102 mobj = re.match(regex, ver)
103 if mobj:
104 ver = mobj.group(1)
105 self._versions[prog] = self._version_cache[path] = ver
106 if prog != 'ffmpeg' or not out:
107 return
108
109 mobj = re.search(r'(?m)^\s+libavformat\s+(?:[0-9. ]+)\s+/\s+(?P<runtime>[0-9. ]+)', out)
110 lavf_runtime_version = mobj.group('runtime').replace(' ', '') if mobj else None
111 self._features = self._features_cache[path] = {
112 'fdk': '--enable-libfdk-aac' in out,
113 'setts': 'setts' in out.splitlines(),
114 'needs_adtstoasc': is_outdated_version(lavf_runtime_version, '57.56.100', False),
115 }
116
117 self.basename = None
118 self.probe_basename = None
119 self._paths = None
120 self._versions = None
121 self._features = {}
122
123 prefer_ffmpeg = self.get_param('prefer_ffmpeg', True)
124 location = self.get_param('ffmpeg_location')
125 if location is None:
126 self._paths = {p: p for p in programs}
127 else:
128 if not os.path.exists(location):
129 self.report_warning(
130 'ffmpeg-location %s does not exist! '
131 'Continuing without ffmpeg.' % (location))
132 self._versions = {}
133 return
134 elif os.path.isdir(location):
135 dirname, basename = location, None
136 else:
137 basename = os.path.splitext(os.path.basename(location))[0]
138 basename = next((p for p in programs if basename.startswith(p)), 'ffmpeg')
139 dirname = os.path.dirname(os.path.abspath(location))
140 if basename in ('ffmpeg', 'ffprobe'):
141 prefer_ffmpeg = True
142
143 self._paths = {
144 p: os.path.join(dirname, p) for p in programs}
145 if basename:
146 self._paths[basename] = location
147
148 self._versions = {}
149 executables = {'basename': ('ffmpeg', 'avconv'), 'probe_basename': ('ffprobe', 'avprobe')}
150 if prefer_ffmpeg is False:
151 executables = {k: v[::-1] for k, v in executables.items()}
152 for var, prefs in executables.items():
153 for p in prefs:
154 get_ffmpeg_version(self._paths[p], p)
155 if self._versions[p]:
156 setattr(self, var, p)
157 break
158
159 if self.basename == 'avconv':
160 self.deprecation_warning(
161 'Support for avconv is deprecated and may be removed in a future version. Use ffmpeg instead')
162 if self.probe_basename == 'avprobe':
163 self.deprecation_warning(
164 'Support for avprobe is deprecated and may be removed in a future version. Use ffprobe instead')
165
166 @property
167 def available(self):
168 return self.basename is not None
169
170 @property
171 def executable(self):
172 return self._paths[self.basename]
173
174 @property
175 def probe_available(self):
176 return self.probe_basename is not None
177
178 @property
179 def probe_executable(self):
180 return self._paths[self.probe_basename]
181
182 @staticmethod
183 def stream_copy_opts(copy=True, *, ext=None):
184 yield from ('-map', '0')
185 # Don't copy Apple TV chapters track, bin_data
186 # See https://github.com/yt-dlp/yt-dlp/issues/2, #19042, #19024, https://trac.ffmpeg.org/ticket/6016
187 yield from ('-dn', '-ignore_unknown')
188 if copy:
189 yield from ('-c', 'copy')
190 # For some reason, '-c copy -map 0' is not enough to copy subtitles
191 if ext in ('mp4', 'mov'):
192 yield from ('-c:s', 'mov_text')
193
194 def get_audio_codec(self, path):
195 if not self.probe_available and not self.available:
196 raise PostProcessingError('ffprobe and ffmpeg not found. Please install or provide the path using --ffmpeg-location')
197 try:
198 if self.probe_available:
199 cmd = [
200 encodeFilename(self.probe_executable, True),
201 encodeArgument('-show_streams')]
202 else:
203 cmd = [
204 encodeFilename(self.executable, True),
205 encodeArgument('-i')]
206 cmd.append(encodeFilename(self._ffmpeg_filename_argument(path), True))
207 self.write_debug(f'{self.basename} command line: {shell_quote(cmd)}')
208 handle = Popen(cmd, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
209 stdout_data, stderr_data = handle.communicate_or_kill()
210 expected_ret = 0 if self.probe_available else 1
211 if handle.wait() != expected_ret:
212 return None
213 except OSError:
214 return None
215 output = (stdout_data if self.probe_available else stderr_data).decode('ascii', 'ignore')
216 if self.probe_available:
217 audio_codec = None
218 for line in output.split('\n'):
219 if line.startswith('codec_name='):
220 audio_codec = line.split('=')[1].strip()
221 elif line.strip() == 'codec_type=audio' and audio_codec is not None:
222 return audio_codec
223 else:
224 # Stream #FILE_INDEX:STREAM_INDEX[STREAM_ID](LANGUAGE): CODEC_TYPE: CODEC_NAME
225 mobj = re.search(
226 r'Stream\s*#\d+:\d+(?:\[0x[0-9a-f]+\])?(?:\([a-z]{3}\))?:\s*Audio:\s*([0-9a-z]+)',
227 output)
228 if mobj:
229 return mobj.group(1)
230 return None
231
232 def get_metadata_object(self, path, opts=[]):
233 if self.probe_basename != 'ffprobe':
234 if self.probe_available:
235 self.report_warning('Only ffprobe is supported for metadata extraction')
236 raise PostProcessingError('ffprobe not found. Please install or provide the path using --ffmpeg-location')
237 self.check_version()
238
239 cmd = [
240 encodeFilename(self.probe_executable, True),
241 encodeArgument('-hide_banner'),
242 encodeArgument('-show_format'),
243 encodeArgument('-show_streams'),
244 encodeArgument('-print_format'),
245 encodeArgument('json'),
246 ]
247
248 cmd += opts
249 cmd.append(encodeFilename(self._ffmpeg_filename_argument(path), True))
250 self.write_debug('ffprobe command line: %s' % shell_quote(cmd))
251 p = Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE)
252 stdout, stderr = p.communicate()
253 return json.loads(stdout.decode('utf-8', 'replace'))
254
255 def get_stream_number(self, path, keys, value):
256 streams = self.get_metadata_object(path)['streams']
257 num = next(
258 (i for i, stream in enumerate(streams) if traverse_obj(stream, keys, casesense=False) == value),
259 None)
260 return num, len(streams)
261
262 def _get_real_video_duration(self, filepath, fatal=True):
263 try:
264 duration = float_or_none(
265 traverse_obj(self.get_metadata_object(filepath), ('format', 'duration')))
266 if not duration:
267 raise PostProcessingError('ffprobe returned empty duration')
268 return duration
269 except PostProcessingError as e:
270 if fatal:
271 raise PostProcessingError(f'Unable to determine video duration: {e.msg}')
272
273 def _duration_mismatch(self, d1, d2):
274 if not d1 or not d2:
275 return None
276 # The duration is often only known to nearest second. So there can be <1sec disparity natually.
277 # Further excuse an additional <1sec difference.
278 return abs(d1 - d2) > 2
279
280 def run_ffmpeg_multiple_files(self, input_paths, out_path, opts, **kwargs):
281 return self.real_run_ffmpeg(
282 [(path, []) for path in input_paths],
283 [(out_path, opts)], **kwargs)
284
285 def real_run_ffmpeg(self, input_path_opts, output_path_opts, *, expected_retcodes=(0,)):
286 self.check_version()
287
288 oldest_mtime = min(
289 os.stat(encodeFilename(path)).st_mtime for path, _ in input_path_opts if path)
290
291 cmd = [encodeFilename(self.executable, True), encodeArgument('-y')]
292 # avconv does not have repeat option
293 if self.basename == 'ffmpeg':
294 cmd += [encodeArgument('-loglevel'), encodeArgument('repeat+info')]
295
296 def make_args(file, args, name, number):
297 keys = ['_%s%d' % (name, number), '_%s' % name]
298 if name == 'o':
299 args += ['-movflags', '+faststart']
300 if number == 1:
301 keys.append('')
302 args += self._configuration_args(self.basename, keys)
303 if name == 'i':
304 args.append('-i')
305 return (
306 [encodeArgument(arg) for arg in args]
307 + [encodeFilename(self._ffmpeg_filename_argument(file), True)])
308
309 for arg_type, path_opts in (('i', input_path_opts), ('o', output_path_opts)):
310 cmd += itertools.chain.from_iterable(
311 make_args(path, list(opts), arg_type, i + 1)
312 for i, (path, opts) in enumerate(path_opts) if path)
313
314 self.write_debug('ffmpeg command line: %s' % shell_quote(cmd))
315 p = Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE)
316 stdout, stderr = p.communicate_or_kill()
317 if p.returncode not in variadic(expected_retcodes):
318 stderr = stderr.decode('utf-8', 'replace').strip()
319 self.write_debug(stderr)
320 raise FFmpegPostProcessorError(stderr.split('\n')[-1])
321 for out_path, _ in output_path_opts:
322 if out_path:
323 self.try_utime(out_path, oldest_mtime, oldest_mtime)
324 return stderr.decode('utf-8', 'replace')
325
326 def run_ffmpeg(self, path, out_path, opts, **kwargs):
327 return self.run_ffmpeg_multiple_files([path], out_path, opts, **kwargs)
328
329 @staticmethod
330 def _ffmpeg_filename_argument(fn):
331 # Always use 'file:' because the filename may contain ':' (ffmpeg
332 # interprets that as a protocol) or can start with '-' (-- is broken in
333 # ffmpeg, see https://ffmpeg.org/trac/ffmpeg/ticket/2127 for details)
334 # Also leave '-' intact in order not to break streaming to stdout.
335 if fn.startswith(('http://', 'https://')):
336 return fn
337 return 'file:' + fn if fn != '-' else fn
338
339 @staticmethod
340 def _quote_for_ffmpeg(string):
341 # See https://ffmpeg.org/ffmpeg-utils.html#toc-Quoting-and-escaping
342 # A sequence of '' produces '\'''\'';
343 # final replace removes the empty '' between \' \'.
344 string = string.replace("'", r"'\''").replace("'''", "'")
345 # Handle potential ' at string boundaries.
346 string = string[1:] if string[0] == "'" else "'" + string
347 return string[:-1] if string[-1] == "'" else string + "'"
348
349 def force_keyframes(self, filename, timestamps):
350 timestamps = orderedSet(timestamps)
351 if timestamps[0] == 0:
352 timestamps = timestamps[1:]
353 keyframe_file = prepend_extension(filename, 'keyframes.temp')
354 self.to_screen(f'Re-encoding "{filename}" with appropriate keyframes')
355 self.run_ffmpeg(filename, keyframe_file, [
356 *self.stream_copy_opts(False, ext=determine_ext(filename)),
357 '-force_key_frames', ','.join(f'{t:.6f}' for t in timestamps)])
358 return keyframe_file
359
360 def concat_files(self, in_files, out_file, concat_opts=None):
361 """
362 Use concat demuxer to concatenate multiple files having identical streams.
363
364 Only inpoint, outpoint, and duration concat options are supported.
365 See https://ffmpeg.org/ffmpeg-formats.html#concat-1 for details
366 """
367 concat_file = f'{out_file}.concat'
368 self.write_debug(f'Writing concat spec to {concat_file}')
369 with open(concat_file, 'wt', encoding='utf-8') as f:
370 f.writelines(self._concat_spec(in_files, concat_opts))
371
372 out_flags = list(self.stream_copy_opts(ext=determine_ext(out_file)))
373
374 self.real_run_ffmpeg(
375 [(concat_file, ['-hide_banner', '-nostdin', '-f', 'concat', '-safe', '0'])],
376 [(out_file, out_flags)])
377 os.remove(concat_file)
378
379 @classmethod
380 def _concat_spec(cls, in_files, concat_opts=None):
381 if concat_opts is None:
382 concat_opts = [{}] * len(in_files)
383 yield 'ffconcat version 1.0\n'
384 for file, opts in zip(in_files, concat_opts):
385 yield f'file {cls._quote_for_ffmpeg(cls._ffmpeg_filename_argument(file))}\n'
386 # Iterate explicitly to yield the following directives in order, ignoring the rest.
387 for directive in 'inpoint', 'outpoint', 'duration':
388 if directive in opts:
389 yield f'{directive} {opts[directive]}\n'
390
391
392 class FFmpegExtractAudioPP(FFmpegPostProcessor):
393 COMMON_AUDIO_EXTS = ('wav', 'flac', 'm4a', 'aiff', 'mp3', 'ogg', 'mka', 'opus', 'wma')
394 SUPPORTED_EXTS = ('aac', 'flac', 'mp3', 'm4a', 'opus', 'vorbis', 'wav', 'alac')
395
396 def __init__(self, downloader=None, preferredcodec=None, preferredquality=None, nopostoverwrites=False):
397 FFmpegPostProcessor.__init__(self, downloader)
398 self._preferredcodec = preferredcodec or 'best'
399 self._preferredquality = float_or_none(preferredquality)
400 self._nopostoverwrites = nopostoverwrites
401
402 def _quality_args(self, codec):
403 if self._preferredquality is None:
404 return []
405 elif self._preferredquality > 10:
406 return ['-b:a', f'{self._preferredquality}k']
407
408 limits = {
409 'libmp3lame': (10, 0),
410 'libvorbis': (0, 10),
411 # FFmpeg's AAC encoder does not have an upper limit for the value of -q:a.
412 # Experimentally, with values over 4, bitrate changes were minimal or non-existent
413 'aac': (0.1, 4),
414 'libfdk_aac': (1, 5),
415 }.get(codec)
416 if not limits:
417 return []
418
419 q = limits[1] + (limits[0] - limits[1]) * (self._preferredquality / 10)
420 if codec == 'libfdk_aac':
421 return ['-vbr', f'{int(q)}']
422 return ['-q:a', f'{q}']
423
424 def run_ffmpeg(self, path, out_path, codec, more_opts):
425 if codec is None:
426 acodec_opts = []
427 else:
428 acodec_opts = ['-acodec', codec]
429 opts = ['-vn'] + acodec_opts + more_opts
430 try:
431 FFmpegPostProcessor.run_ffmpeg(self, path, out_path, opts)
432 except FFmpegPostProcessorError as err:
433 raise AudioConversionError(err.msg)
434
435 @PostProcessor._restrict_to(images=False)
436 def run(self, information):
437 orig_path = path = information['filepath']
438 orig_ext = information['ext']
439
440 if self._preferredcodec == 'best' and orig_ext in self.COMMON_AUDIO_EXTS:
441 self.to_screen('Skipping audio extraction since the file is already in a common audio format')
442 return [], information
443
444 filecodec = self.get_audio_codec(path)
445 if filecodec is None:
446 raise PostProcessingError('WARNING: unable to obtain file audio codec with ffprobe')
447
448 more_opts = []
449 if self._preferredcodec == 'best' or self._preferredcodec == filecodec or (self._preferredcodec == 'm4a' and filecodec == 'aac'):
450 if filecodec == 'aac' and self._preferredcodec in ['m4a', 'best']:
451 # Lossless, but in another container
452 acodec = 'copy'
453 extension = 'm4a'
454 more_opts = ['-bsf:a', 'aac_adtstoasc']
455 elif filecodec in ['aac', 'flac', 'mp3', 'vorbis', 'opus']:
456 # Lossless if possible
457 acodec = 'copy'
458 extension = filecodec
459 if filecodec == 'aac':
460 more_opts = ['-f', 'adts']
461 if filecodec == 'vorbis':
462 extension = 'ogg'
463 elif filecodec == 'alac':
464 acodec = None
465 extension = 'm4a'
466 more_opts += ['-acodec', 'alac']
467 else:
468 # MP3 otherwise.
469 acodec = 'libmp3lame'
470 extension = 'mp3'
471 more_opts = self._quality_args(acodec)
472 else:
473 # We convert the audio (lossy if codec is lossy)
474 acodec = ACODECS[self._preferredcodec]
475 if acodec == 'aac' and self._features.get('fdk'):
476 acodec = 'libfdk_aac'
477 extension = self._preferredcodec
478 more_opts = self._quality_args(acodec)
479 if self._preferredcodec == 'aac':
480 more_opts += ['-f', 'adts']
481 elif self._preferredcodec == 'm4a':
482 more_opts += ['-bsf:a', 'aac_adtstoasc']
483 elif self._preferredcodec == 'vorbis':
484 extension = 'ogg'
485 elif self._preferredcodec == 'wav':
486 extension = 'wav'
487 more_opts += ['-f', 'wav']
488 elif self._preferredcodec == 'alac':
489 extension = 'm4a'
490 more_opts += ['-acodec', 'alac']
491
492 prefix, sep, ext = path.rpartition('.') # not os.path.splitext, since the latter does not work on unicode in all setups
493 temp_path = new_path = prefix + sep + extension
494
495 if new_path == path:
496 if acodec == 'copy':
497 self.to_screen(f'File is already in target format {self._preferredcodec}, skipping')
498 return [], information
499 orig_path = prepend_extension(path, 'orig')
500 temp_path = prepend_extension(path, 'temp')
501 if (self._nopostoverwrites and os.path.exists(encodeFilename(new_path))
502 and os.path.exists(encodeFilename(orig_path))):
503 self.to_screen('Post-process file %s exists, skipping' % new_path)
504 return [], information
505
506 try:
507 self.to_screen(f'Destination: {new_path}')
508 self.run_ffmpeg(path, temp_path, acodec, more_opts)
509 except AudioConversionError as e:
510 raise PostProcessingError(
511 'audio conversion failed: ' + e.msg)
512 except Exception:
513 raise PostProcessingError('error running ' + self.basename)
514
515 os.replace(path, orig_path)
516 os.replace(temp_path, new_path)
517 information['filepath'] = new_path
518 information['ext'] = extension
519
520 # Try to update the date time for extracted audio file.
521 if information.get('filetime') is not None:
522 self.try_utime(
523 new_path, time.time(), information['filetime'],
524 errnote='Cannot update utime of audio file')
525
526 return [orig_path], information
527
528
529 class FFmpegVideoConvertorPP(FFmpegPostProcessor):
530 SUPPORTED_EXTS = ('mp4', 'mkv', 'flv', 'webm', 'mov', 'avi', 'mka', 'ogg', *FFmpegExtractAudioPP.SUPPORTED_EXTS)
531 FORMAT_RE = re.compile(r'{0}(?:/{0})*$'.format(r'(?:\w+>)?(?:%s)' % '|'.join(SUPPORTED_EXTS)))
532 _ACTION = 'converting'
533
534 def __init__(self, downloader=None, preferedformat=None):
535 super().__init__(downloader)
536 self._preferedformats = preferedformat.lower().split('/')
537
538 def _target_ext(self, source_ext):
539 for pair in self._preferedformats:
540 kv = pair.split('>')
541 if len(kv) == 1 or kv[0].strip() == source_ext:
542 return kv[-1].strip()
543
544 @staticmethod
545 def _options(target_ext):
546 yield from FFmpegPostProcessor.stream_copy_opts(False)
547 if target_ext == 'avi':
548 yield from ('-c:v', 'libxvid', '-vtag', 'XVID')
549
550 @PostProcessor._restrict_to(images=False)
551 def run(self, info):
552 filename, source_ext = info['filepath'], info['ext'].lower()
553 target_ext = self._target_ext(source_ext)
554 _skip_msg = (
555 f'could not find a mapping for {source_ext}' if not target_ext
556 else f'already is in target format {source_ext}' if source_ext == target_ext
557 else None)
558 if _skip_msg:
559 self.to_screen(f'Not {self._ACTION} media file "{filename}"; {_skip_msg}')
560 return [], info
561
562 outpath = replace_extension(filename, target_ext, source_ext)
563 self.to_screen(f'{self._ACTION.title()} video from {source_ext} to {target_ext}; Destination: {outpath}')
564 self.run_ffmpeg(filename, outpath, self._options(target_ext))
565
566 info['filepath'] = outpath
567 info['format'] = info['ext'] = target_ext
568 return [filename], info
569
570
571 class FFmpegVideoRemuxerPP(FFmpegVideoConvertorPP):
572 _ACTION = 'remuxing'
573
574 @staticmethod
575 def _options(target_ext):
576 return FFmpegPostProcessor.stream_copy_opts()
577
578
579 class FFmpegEmbedSubtitlePP(FFmpegPostProcessor):
580 SUPPORTED_EXTS = ('mp4', 'mov', 'm4a', 'webm', 'mkv', 'mka')
581
582 def __init__(self, downloader=None, already_have_subtitle=False):
583 super().__init__(downloader)
584 self._already_have_subtitle = already_have_subtitle
585
586 @PostProcessor._restrict_to(images=False)
587 def run(self, info):
588 if info['ext'] not in self.SUPPORTED_EXTS:
589 self.to_screen(f'Subtitles can only be embedded in {", ".join(self.SUPPORTED_EXTS)} files')
590 return [], info
591 subtitles = info.get('requested_subtitles')
592 if not subtitles:
593 self.to_screen('There aren\'t any subtitles to embed')
594 return [], info
595
596 filename = info['filepath']
597
598 # Disabled temporarily. There needs to be a way to overide this
599 # in case of duration actually mismatching in extractor
600 # See: https://github.com/yt-dlp/yt-dlp/issues/1870, https://github.com/yt-dlp/yt-dlp/issues/1385
601 '''
602 if info.get('duration') and not info.get('__real_download') and self._duration_mismatch(
603 self._get_real_video_duration(filename, False), info['duration']):
604 self.to_screen(f'Skipping {self.pp_key()} since the real and expected durations mismatch')
605 return [], info
606 '''
607
608 ext = info['ext']
609 sub_langs, sub_names, sub_filenames = [], [], []
610 webm_vtt_warn = False
611 mp4_ass_warn = False
612
613 for lang, sub_info in subtitles.items():
614 if not os.path.exists(sub_info.get('filepath', '')):
615 self.report_warning(f'Skipping embedding {lang} subtitle because the file is missing')
616 continue
617 sub_ext = sub_info['ext']
618 if sub_ext == 'json':
619 self.report_warning('JSON subtitles cannot be embedded')
620 elif ext != 'webm' or ext == 'webm' and sub_ext == 'vtt':
621 sub_langs.append(lang)
622 sub_names.append(sub_info.get('name'))
623 sub_filenames.append(sub_info['filepath'])
624 else:
625 if not webm_vtt_warn and ext == 'webm' and sub_ext != 'vtt':
626 webm_vtt_warn = True
627 self.report_warning('Only WebVTT subtitles can be embedded in webm files')
628 if not mp4_ass_warn and ext == 'mp4' and sub_ext == 'ass':
629 mp4_ass_warn = True
630 self.report_warning('ASS subtitles cannot be properly embedded in mp4 files; expect issues')
631
632 if not sub_langs:
633 return [], info
634
635 input_files = [filename] + sub_filenames
636
637 opts = [
638 *self.stream_copy_opts(ext=info['ext']),
639 # Don't copy the existing subtitles, we may be running the
640 # postprocessor a second time
641 '-map', '-0:s',
642 ]
643 for i, (lang, name) in enumerate(zip(sub_langs, sub_names)):
644 opts.extend(['-map', '%d:0' % (i + 1)])
645 lang_code = ISO639Utils.short2long(lang) or lang
646 opts.extend(['-metadata:s:s:%d' % i, 'language=%s' % lang_code])
647 if name:
648 opts.extend(['-metadata:s:s:%d' % i, 'handler_name=%s' % name,
649 '-metadata:s:s:%d' % i, 'title=%s' % name])
650
651 temp_filename = prepend_extension(filename, 'temp')
652 self.to_screen('Embedding subtitles in "%s"' % filename)
653 self.run_ffmpeg_multiple_files(input_files, temp_filename, opts)
654 os.replace(temp_filename, filename)
655
656 files_to_delete = [] if self._already_have_subtitle else sub_filenames
657 return files_to_delete, info
658
659
660 class FFmpegMetadataPP(FFmpegPostProcessor):
661
662 def __init__(self, downloader, add_metadata=True, add_chapters=True, add_infojson='if_exists'):
663 FFmpegPostProcessor.__init__(self, downloader)
664 self._add_metadata = add_metadata
665 self._add_chapters = add_chapters
666 self._add_infojson = add_infojson
667
668 @staticmethod
669 def _options(target_ext):
670 audio_only = target_ext == 'm4a'
671 yield from FFmpegPostProcessor.stream_copy_opts(not audio_only)
672 if audio_only:
673 yield from ('-vn', '-acodec', 'copy')
674
675 @PostProcessor._restrict_to(images=False)
676 def run(self, info):
677 filename, metadata_filename = info['filepath'], None
678 files_to_delete, options = [], []
679 if self._add_chapters and info.get('chapters'):
680 metadata_filename = replace_extension(filename, 'meta')
681 options.extend(self._get_chapter_opts(info['chapters'], metadata_filename))
682 files_to_delete.append(metadata_filename)
683 if self._add_metadata:
684 options.extend(self._get_metadata_opts(info))
685
686 if self._add_infojson:
687 if info['ext'] in ('mkv', 'mka'):
688 infojson_filename = info.get('infojson_filename')
689 options.extend(self._get_infojson_opts(info, infojson_filename))
690 if not infojson_filename:
691 files_to_delete.append(info.get('infojson_filename'))
692 elif self._add_infojson is True:
693 self.to_screen('The info-json can only be attached to mkv/mka files')
694
695 if not options:
696 self.to_screen('There isn\'t any metadata to add')
697 return [], info
698
699 temp_filename = prepend_extension(filename, 'temp')
700 self.to_screen('Adding metadata to "%s"' % filename)
701 self.run_ffmpeg_multiple_files(
702 (filename, metadata_filename), temp_filename,
703 itertools.chain(self._options(info['ext']), *options))
704 for file in filter(None, files_to_delete):
705 os.remove(file) # Don't obey --keep-files
706 os.replace(temp_filename, filename)
707 return [], info
708
709 @staticmethod
710 def _get_chapter_opts(chapters, metadata_filename):
711 with open(metadata_filename, 'wt', encoding='utf-8') as f:
712 def ffmpeg_escape(text):
713 return re.sub(r'([\\=;#\n])', r'\\\1', text)
714
715 metadata_file_content = ';FFMETADATA1\n'
716 for chapter in chapters:
717 metadata_file_content += '[CHAPTER]\nTIMEBASE=1/1000\n'
718 metadata_file_content += 'START=%d\n' % (chapter['start_time'] * 1000)
719 metadata_file_content += 'END=%d\n' % (chapter['end_time'] * 1000)
720 chapter_title = chapter.get('title')
721 if chapter_title:
722 metadata_file_content += 'title=%s\n' % ffmpeg_escape(chapter_title)
723 f.write(metadata_file_content)
724 yield ('-map_metadata', '1')
725
726 def _get_metadata_opts(self, info):
727 meta_prefix = 'meta'
728 metadata = collections.defaultdict(dict)
729
730 def add(meta_list, info_list=None):
731 value = next((
732 str(info[key]) for key in [f'{meta_prefix}_'] + list(variadic(info_list or meta_list))
733 if info.get(key) is not None), None)
734 if value not in ('', None):
735 value = value.replace('\0', '') # nul character cannot be passed in command line
736 metadata['common'].update({meta_f: value for meta_f in variadic(meta_list)})
737
738 # See [1-4] for some info on media metadata/metadata supported
739 # by ffmpeg.
740 # 1. https://kdenlive.org/en/project/adding-meta-data-to-mp4-video/
741 # 2. https://wiki.multimedia.cx/index.php/FFmpeg_Metadata
742 # 3. https://kodi.wiki/view/Video_file_tagging
743
744 add('title', ('track', 'title'))
745 add('date', 'upload_date')
746 add(('description', 'synopsis'), 'description')
747 add(('purl', 'comment'), 'webpage_url')
748 add('track', 'track_number')
749 add('artist', ('artist', 'creator', 'uploader', 'uploader_id'))
750 add('genre')
751 add('album')
752 add('album_artist')
753 add('disc', 'disc_number')
754 add('show', 'series')
755 add('season_number')
756 add('episode_id', ('episode', 'episode_id'))
757 add('episode_sort', 'episode_number')
758 if 'embed-metadata' in self.get_param('compat_opts', []):
759 add('comment', 'description')
760 metadata['common'].pop('synopsis', None)
761
762 meta_regex = rf'{re.escape(meta_prefix)}(?P<i>\d+)?_(?P<key>.+)'
763 for key, value in info.items():
764 mobj = re.fullmatch(meta_regex, key)
765 if value is not None and mobj:
766 metadata[mobj.group('i') or 'common'][mobj.group('key')] = value
767
768 # Write id3v1 metadata also since Windows Explorer can't handle id3v2 tags
769 yield ('-write_id3v1', '1')
770
771 for name, value in metadata['common'].items():
772 yield ('-metadata', f'{name}={value}')
773
774 stream_idx = 0
775 for fmt in info.get('requested_formats') or []:
776 stream_count = 2 if 'none' not in (fmt.get('vcodec'), fmt.get('acodec')) else 1
777 lang = ISO639Utils.short2long(fmt.get('language') or '') or fmt.get('language')
778 for i in range(stream_idx, stream_idx + stream_count):
779 if lang:
780 metadata[str(i)].setdefault('language', lang)
781 for name, value in metadata[str(i)].items():
782 yield (f'-metadata:s:{i}', f'{name}={value}')
783 stream_idx += stream_count
784
785 def _get_infojson_opts(self, info, infofn):
786 if not infofn or not os.path.exists(infofn):
787 if self._add_infojson is not True:
788 return
789 infofn = infofn or '%s.temp' % (
790 self._downloader.prepare_filename(info, 'infojson')
791 or replace_extension(self._downloader.prepare_filename(info), 'info.json', info['ext']))
792 if not self._downloader._ensure_dir_exists(infofn):
793 return
794 self.write_debug(f'Writing info-json to: {infofn}')
795 write_json_file(self._downloader.sanitize_info(info, self.get_param('clean_infojson', True)), infofn)
796 info['infojson_filename'] = infofn
797
798 old_stream, new_stream = self.get_stream_number(info['filepath'], ('tags', 'mimetype'), 'application/json')
799 if old_stream is not None:
800 yield ('-map', '-0:%d' % old_stream)
801 new_stream -= 1
802
803 yield ('-attach', infofn,
804 '-metadata:s:%d' % new_stream, 'mimetype=application/json')
805
806
807 class FFmpegMergerPP(FFmpegPostProcessor):
808 @PostProcessor._restrict_to(images=False)
809 def run(self, info):
810 filename = info['filepath']
811 temp_filename = prepend_extension(filename, 'temp')
812 args = ['-c', 'copy']
813 audio_streams = 0
814 for (i, fmt) in enumerate(info['requested_formats']):
815 if fmt.get('acodec') != 'none':
816 args.extend(['-map', f'{i}:a:0'])
817 aac_fixup = fmt['protocol'].startswith('m3u8') and self.get_audio_codec(fmt['filepath']) == 'aac'
818 if aac_fixup:
819 args.extend([f'-bsf:a:{audio_streams}', 'aac_adtstoasc'])
820 audio_streams += 1
821 if fmt.get('vcodec') != 'none':
822 args.extend(['-map', '%u:v:0' % (i)])
823 self.to_screen('Merging formats into "%s"' % filename)
824 self.run_ffmpeg_multiple_files(info['__files_to_merge'], temp_filename, args)
825 os.rename(encodeFilename(temp_filename), encodeFilename(filename))
826 return info['__files_to_merge'], info
827
828 def can_merge(self):
829 # TODO: figure out merge-capable ffmpeg version
830 if self.basename != 'avconv':
831 return True
832
833 required_version = '10-0'
834 if is_outdated_version(
835 self._versions[self.basename], required_version):
836 warning = ('Your copy of %s is outdated and unable to properly mux separate video and audio files, '
837 'yt-dlp will download single file media. '
838 'Update %s to version %s or newer to fix this.') % (
839 self.basename, self.basename, required_version)
840 self.report_warning(warning)
841 return False
842 return True
843
844
845 class FFmpegFixupPostProcessor(FFmpegPostProcessor):
846 def _fixup(self, msg, filename, options):
847 temp_filename = prepend_extension(filename, 'temp')
848
849 self.to_screen(f'{msg} of "{filename}"')
850 self.run_ffmpeg(filename, temp_filename, options)
851
852 os.replace(temp_filename, filename)
853
854
855 class FFmpegFixupStretchedPP(FFmpegFixupPostProcessor):
856 @PostProcessor._restrict_to(images=False, audio=False)
857 def run(self, info):
858 stretched_ratio = info.get('stretched_ratio')
859 if stretched_ratio not in (None, 1):
860 self._fixup('Fixing aspect ratio', info['filepath'], [
861 *self.stream_copy_opts(), '-aspect', '%f' % stretched_ratio])
862 return [], info
863
864
865 class FFmpegFixupM4aPP(FFmpegFixupPostProcessor):
866 @PostProcessor._restrict_to(images=False, video=False)
867 def run(self, info):
868 if info.get('container') == 'm4a_dash':
869 self._fixup('Correcting container', info['filepath'], [*self.stream_copy_opts(), '-f', 'mp4'])
870 return [], info
871
872
873 class FFmpegFixupM3u8PP(FFmpegFixupPostProcessor):
874 def _needs_fixup(self, info):
875 yield info['ext'] in ('mp4', 'm4a')
876 yield info['protocol'].startswith('m3u8')
877 try:
878 metadata = self.get_metadata_object(info['filepath'])
879 except PostProcessingError as e:
880 self.report_warning(f'Unable to extract metadata: {e.msg}')
881 yield True
882 else:
883 yield traverse_obj(metadata, ('format', 'format_name'), casesense=False) == 'mpegts'
884
885 @PostProcessor._restrict_to(images=False)
886 def run(self, info):
887 if all(self._needs_fixup(info)):
888 self._fixup('Fixing MPEG-TS in MP4 container', info['filepath'], [
889 *self.stream_copy_opts(), '-f', 'mp4', '-bsf:a', 'aac_adtstoasc'])
890 return [], info
891
892
893 class FFmpegFixupTimestampPP(FFmpegFixupPostProcessor):
894
895 def __init__(self, downloader=None, trim=0.001):
896 # "trim" should be used when the video contains unintended packets
897 super().__init__(downloader)
898 assert isinstance(trim, (int, float))
899 self.trim = str(trim)
900
901 @PostProcessor._restrict_to(images=False)
902 def run(self, info):
903 if not self._features.get('setts'):
904 self.report_warning(
905 'A re-encode is needed to fix timestamps in older versions of ffmpeg. '
906 'Please install ffmpeg 4.4 or later to fixup without re-encoding')
907 opts = ['-vf', 'setpts=PTS-STARTPTS']
908 else:
909 opts = ['-c', 'copy', '-bsf', 'setts=ts=TS-STARTPTS']
910 self._fixup('Fixing frame timestamp', info['filepath'], opts + [*self.stream_copy_opts(False), '-ss', self.trim])
911 return [], info
912
913
914 class FFmpegCopyStreamPP(FFmpegFixupPostProcessor):
915 MESSAGE = 'Copying stream'
916
917 @PostProcessor._restrict_to(images=False)
918 def run(self, info):
919 self._fixup(self.MESSAGE, info['filepath'], self.stream_copy_opts())
920 return [], info
921
922
923 class FFmpegFixupDurationPP(FFmpegCopyStreamPP):
924 MESSAGE = 'Fixing video duration'
925
926
927 class FFmpegFixupDuplicateMoovPP(FFmpegCopyStreamPP):
928 MESSAGE = 'Fixing duplicate MOOV atoms'
929
930
931 class FFmpegSubtitlesConvertorPP(FFmpegPostProcessor):
932 SUPPORTED_EXTS = ('srt', 'vtt', 'ass', 'lrc')
933
934 def __init__(self, downloader=None, format=None):
935 super().__init__(downloader)
936 self.format = format
937
938 def run(self, info):
939 subs = info.get('requested_subtitles')
940 new_ext = self.format
941 new_format = new_ext
942 if new_format == 'vtt':
943 new_format = 'webvtt'
944 if subs is None:
945 self.to_screen('There aren\'t any subtitles to convert')
946 return [], info
947 self.to_screen('Converting subtitles')
948 sub_filenames = []
949 for lang, sub in subs.items():
950 if not os.path.exists(sub.get('filepath', '')):
951 self.report_warning(f'Skipping embedding {lang} subtitle because the file is missing')
952 continue
953 ext = sub['ext']
954 if ext == new_ext:
955 self.to_screen('Subtitle file for %s is already in the requested format' % new_ext)
956 continue
957 elif ext == 'json':
958 self.to_screen(
959 'You have requested to convert json subtitles into another format, '
960 'which is currently not possible')
961 continue
962 old_file = sub['filepath']
963 sub_filenames.append(old_file)
964 new_file = replace_extension(old_file, new_ext)
965
966 if ext in ('dfxp', 'ttml', 'tt'):
967 self.report_warning(
968 'You have requested to convert dfxp (TTML) subtitles into another format, '
969 'which results in style information loss')
970
971 dfxp_file = old_file
972 srt_file = replace_extension(old_file, 'srt')
973
974 with open(dfxp_file, 'rb') as f:
975 srt_data = dfxp2srt(f.read())
976
977 with open(srt_file, 'wt', encoding='utf-8') as f:
978 f.write(srt_data)
979 old_file = srt_file
980
981 subs[lang] = {
982 'ext': 'srt',
983 'data': srt_data,
984 'filepath': srt_file,
985 }
986
987 if new_ext == 'srt':
988 continue
989 else:
990 sub_filenames.append(srt_file)
991
992 self.run_ffmpeg(old_file, new_file, ['-f', new_format])
993
994 with open(new_file, encoding='utf-8') as f:
995 subs[lang] = {
996 'ext': new_ext,
997 'data': f.read(),
998 'filepath': new_file,
999 }
1000
1001 info['__files_to_move'][new_file] = replace_extension(
1002 info['__files_to_move'][sub['filepath']], new_ext)
1003
1004 return sub_filenames, info
1005
1006
1007 class FFmpegSplitChaptersPP(FFmpegPostProcessor):
1008 def __init__(self, downloader, force_keyframes=False):
1009 FFmpegPostProcessor.__init__(self, downloader)
1010 self._force_keyframes = force_keyframes
1011
1012 def _prepare_filename(self, number, chapter, info):
1013 info = info.copy()
1014 info.update({
1015 'section_number': number,
1016 'section_title': chapter.get('title'),
1017 'section_start': chapter.get('start_time'),
1018 'section_end': chapter.get('end_time'),
1019 })
1020 return self._downloader.prepare_filename(info, 'chapter')
1021
1022 def _ffmpeg_args_for_chapter(self, number, chapter, info):
1023 destination = self._prepare_filename(number, chapter, info)
1024 if not self._downloader._ensure_dir_exists(encodeFilename(destination)):
1025 return
1026
1027 chapter['filepath'] = destination
1028 self.to_screen('Chapter %03d; Destination: %s' % (number, destination))
1029 return (
1030 destination,
1031 ['-ss', compat_str(chapter['start_time']),
1032 '-t', compat_str(chapter['end_time'] - chapter['start_time'])])
1033
1034 @PostProcessor._restrict_to(images=False)
1035 def run(self, info):
1036 chapters = info.get('chapters') or []
1037 if not chapters:
1038 self.to_screen('Chapter information is unavailable')
1039 return [], info
1040
1041 in_file = info['filepath']
1042 if self._force_keyframes and len(chapters) > 1:
1043 in_file = self.force_keyframes(in_file, (c['start_time'] for c in chapters))
1044 self.to_screen('Splitting video by chapters; %d chapters found' % len(chapters))
1045 for idx, chapter in enumerate(chapters):
1046 destination, opts = self._ffmpeg_args_for_chapter(idx + 1, chapter, info)
1047 self.real_run_ffmpeg([(in_file, opts)], [(destination, self.stream_copy_opts())])
1048 if in_file != info['filepath']:
1049 os.remove(in_file)
1050 return [], info
1051
1052
1053 class FFmpegThumbnailsConvertorPP(FFmpegPostProcessor):
1054 SUPPORTED_EXTS = ('jpg', 'png', 'webp')
1055
1056 def __init__(self, downloader=None, format=None):
1057 super().__init__(downloader)
1058 self.format = format
1059
1060 @staticmethod
1061 def is_webp(path):
1062 with open(encodeFilename(path), 'rb') as f:
1063 b = f.read(12)
1064 return b[0:4] == b'RIFF' and b[8:] == b'WEBP'
1065
1066 def fixup_webp(self, info, idx=-1):
1067 thumbnail_filename = info['thumbnails'][idx]['filepath']
1068 _, thumbnail_ext = os.path.splitext(thumbnail_filename)
1069 if thumbnail_ext:
1070 thumbnail_ext = thumbnail_ext[1:].lower()
1071 if thumbnail_ext != 'webp' and self.is_webp(thumbnail_filename):
1072 self.to_screen('Correcting thumbnail "%s" extension to webp' % thumbnail_filename)
1073 webp_filename = replace_extension(thumbnail_filename, 'webp')
1074 os.replace(thumbnail_filename, webp_filename)
1075 info['thumbnails'][idx]['filepath'] = webp_filename
1076 info['__files_to_move'][webp_filename] = replace_extension(
1077 info['__files_to_move'].pop(thumbnail_filename), 'webp')
1078
1079 @staticmethod
1080 def _options(target_ext):
1081 if target_ext == 'jpg':
1082 return ['-bsf:v', 'mjpeg2jpeg']
1083 return []
1084
1085 def convert_thumbnail(self, thumbnail_filename, target_ext):
1086 thumbnail_conv_filename = replace_extension(thumbnail_filename, target_ext)
1087
1088 self.to_screen(f'Converting thumbnail "{thumbnail_filename}" to {target_ext}')
1089 self.real_run_ffmpeg(
1090 [(thumbnail_filename, ['-f', 'image2', '-pattern_type', 'none'])],
1091 [(thumbnail_conv_filename.replace('%', '%%'), self._options(target_ext))])
1092 return thumbnail_conv_filename
1093
1094 def run(self, info):
1095 files_to_delete = []
1096 has_thumbnail = False
1097
1098 for idx, thumbnail_dict in enumerate(info.get('thumbnails') or []):
1099 original_thumbnail = thumbnail_dict.get('filepath')
1100 if not original_thumbnail:
1101 continue
1102 has_thumbnail = True
1103 self.fixup_webp(info, idx)
1104 _, thumbnail_ext = os.path.splitext(original_thumbnail)
1105 if thumbnail_ext:
1106 thumbnail_ext = thumbnail_ext[1:].lower()
1107 if thumbnail_ext == 'jpeg':
1108 thumbnail_ext = 'jpg'
1109 if thumbnail_ext == self.format:
1110 self.to_screen('Thumbnail "%s" is already in the requested format' % original_thumbnail)
1111 continue
1112 thumbnail_dict['filepath'] = self.convert_thumbnail(original_thumbnail, self.format)
1113 files_to_delete.append(original_thumbnail)
1114 info['__files_to_move'][thumbnail_dict['filepath']] = replace_extension(
1115 info['__files_to_move'][original_thumbnail], self.format)
1116
1117 if not has_thumbnail:
1118 self.to_screen('There aren\'t any thumbnails to convert')
1119 return files_to_delete, info
1120
1121
1122 class FFmpegConcatPP(FFmpegPostProcessor):
1123 def __init__(self, downloader, only_multi_video=False):
1124 self._only_multi_video = only_multi_video
1125 super().__init__(downloader)
1126
1127 def _get_codecs(self, file):
1128 codecs = traverse_obj(self.get_metadata_object(file), ('streams', ..., 'codec_name'))
1129 self.write_debug(f'Codecs = {", ".join(codecs)}')
1130 return tuple(codecs)
1131
1132 def concat_files(self, in_files, out_file):
1133 if not self._downloader._ensure_dir_exists(out_file):
1134 return
1135 if len(in_files) == 1:
1136 if os.path.realpath(in_files[0]) != os.path.realpath(out_file):
1137 self.to_screen(f'Moving "{in_files[0]}" to "{out_file}"')
1138 os.replace(in_files[0], out_file)
1139 return []
1140
1141 if len(set(map(self._get_codecs, in_files))) > 1:
1142 raise PostProcessingError(
1143 'The files have different streams/codecs and cannot be concatenated. '
1144 'Either select different formats or --recode-video them to a common format')
1145
1146 self.to_screen(f'Concatenating {len(in_files)} files; Destination: {out_file}')
1147 super().concat_files(in_files, out_file)
1148 return in_files
1149
1150 @PostProcessor._restrict_to(images=False, simulated=False)
1151 def run(self, info):
1152 entries = info.get('entries') or []
1153 if not any(entries) or (self._only_multi_video and info['_type'] != 'multi_video'):
1154 return [], info
1155 elif traverse_obj(entries, (..., 'requested_downloads', lambda _, v: len(v) > 1)):
1156 raise PostProcessingError('Concatenation is not supported when downloading multiple separate formats')
1157
1158 in_files = traverse_obj(entries, (..., 'requested_downloads', 0, 'filepath')) or []
1159 if len(in_files) < len(entries):
1160 raise PostProcessingError('Aborting concatenation because some downloads failed')
1161
1162 ie_copy = self._downloader._playlist_infodict(info)
1163 exts = traverse_obj(entries, (..., 'requested_downloads', 0, 'ext'), (..., 'ext'))
1164 ie_copy['ext'] = exts[0] if len(set(exts)) == 1 else 'mkv'
1165 out_file = self._downloader.prepare_filename(ie_copy, 'pl_video')
1166
1167 files_to_delete = self.concat_files(in_files, out_file)
1168
1169 info['requested_downloads'] = [{
1170 'filepath': out_file,
1171 'ext': ie_copy['ext'],
1172 }]
1173 return files_to_delete, info