import re
import json
-
from .common import AudioConversionError, PostProcessor
from ..compat import compat_str, compat_numeric_types
from ..utils import (
+ dfxp2srt,
encodeArgument,
encodeFilename,
+ float_or_none,
get_exe_version,
is_outdated_version,
+ ISO639Utils,
+ orderedSet,
PostProcessingError,
prepend_extension,
- shell_quote,
- dfxp2srt,
- ISO639Utils,
process_communicate_or_kill,
replace_extension,
+ shell_quote,
traverse_obj,
variadic,
)
'Continuing without ffmpeg.' % (location))
self._versions = {}
return
- elif not os.path.isdir(location):
+ elif os.path.isdir(location):
+ dirname, basename = location, None
+ else:
basename = os.path.splitext(os.path.basename(location))[0]
- if basename not in programs:
- self.report_warning(
- 'Cannot identify executable %s, its basename should be one of %s. '
- 'Continuing without ffmpeg.' %
- (location, ', '.join(programs)))
- self._versions = {}
- return None
- location = os.path.dirname(os.path.abspath(location))
+ basename = next((p for p in programs if basename.startswith(p)), 'ffmpeg')
+ dirname = os.path.dirname(os.path.abspath(location))
if basename in ('ffmpeg', 'ffprobe'):
prefer_ffmpeg = True
self._paths = dict(
- (p, os.path.join(location, p)) for p in programs)
+ (p, os.path.join(dirname, p)) for p in programs)
+ if basename:
+ self._paths[basename] = location
self._versions = dict(
(p, get_ffmpeg_version(self._paths[p])) for p in programs)
if self._versions is None:
None)
return num, len(streams)
+ def _get_real_video_duration(self, info, fatal=True):
+ try:
+ if '_real_duration' not in info:
+ info['_real_duration'] = float_or_none(
+ traverse_obj(self.get_metadata_object(info['filepath']), ('format', 'duration')))
+ if not info['_real_duration']:
+ raise PostProcessingError('ffprobe returned empty duration')
+ except PostProcessingError as e:
+ if fatal:
+ raise PostProcessingError(f'Unable to determine video duration; {e}')
+ return info.setdefault('_real_duration', None)
+
+ def _duration_mismatch(self, d1, d2):
+ if not d1 or not d2:
+ return None
+ return abs(d1 - d2) > 1
+
def run_ffmpeg_multiple_files(self, input_paths, out_path, opts, **kwargs):
return self.real_run_ffmpeg(
[(path, []) for path in input_paths],
oldest_mtime = min(
os.stat(encodeFilename(path)).st_mtime for path, _ in input_path_opts if path)
- cmd = [encodeFilename(self.executable, True), encodeArgument('-y')]
+ cmd = [encodeFilename(self.executable, True), encodeArgument('-y'), encodeArgument('-probesize'), encodeArgument('max')]
# avconv does not have repeat option
if self.basename == 'ffmpeg':
cmd += [encodeArgument('-loglevel'), encodeArgument('repeat+info')]
stdout, stderr = process_communicate_or_kill(p)
if p.returncode not in variadic(expected_retcodes):
stderr = stderr.decode('utf-8', 'replace').strip()
- if self.get_param('verbose', False):
- self.report_error(stderr)
+ self.write_debug(stderr)
raise FFmpegPostProcessorError(stderr.split('\n')[-1])
for out_path, _ in output_path_opts:
if out_path:
def run_ffmpeg(self, path, out_path, opts, **kwargs):
return self.run_ffmpeg_multiple_files([path], out_path, opts, **kwargs)
- def _ffmpeg_filename_argument(self, fn):
+ @staticmethod
+ def _ffmpeg_filename_argument(fn):
# Always use 'file:' because the filename may contain ':' (ffmpeg
# interprets that as a protocol) or can start with '-' (-- is broken in
# ffmpeg, see https://ffmpeg.org/trac/ffmpeg/ticket/2127 for details)
return fn
return 'file:' + fn if fn != '-' else fn
+ @staticmethod
+ def _quote_for_ffmpeg(string):
+ # See https://ffmpeg.org/ffmpeg-utils.html#toc-Quoting-and-escaping
+ # A sequence of '' produces '\'''\'';
+ # final replace removes the empty '' between \' \'.
+ string = string.replace("'", r"'\''").replace("'''", "'")
+ # Handle potential ' at string boundaries.
+ string = string[1:] if string[0] == "'" else "'" + string
+ return string[:-1] if string[-1] == "'" else string + "'"
+
+ def force_keyframes(self, filename, timestamps):
+ timestamps = orderedSet(timestamps)
+ if timestamps[0] == 0:
+ timestamps = timestamps[1:]
+ keyframe_file = prepend_extension(filename, 'keyframes.temp')
+ self.to_screen(f'Re-encoding "{filename}" with appropriate keyframes')
+ self.run_ffmpeg(filename, keyframe_file, ['-force_key_frames', ','.join(
+ f'{t:.6f}' for t in timestamps)])
+ return keyframe_file
+
+ def concat_files(self, in_files, out_file, concat_opts=None):
+ """
+ Use concat demuxer to concatenate multiple files having identical streams.
+
+ Only inpoint, outpoint, and duration concat options are supported.
+ See https://ffmpeg.org/ffmpeg-formats.html#concat-1 for details
+ """
+ concat_file = f'{out_file}.concat'
+ self.write_debug(f'Writing concat spec to {concat_file}')
+ with open(concat_file, 'wt', encoding='utf-8') as f:
+ f.writelines(self._concat_spec(in_files, concat_opts))
+
+ out_flags = ['-c', 'copy']
+ if out_file.rpartition('.')[-1] in ('mp4', 'mov'):
+ # For some reason, '-c copy' is not enough to copy subtitles
+ out_flags.extend(['-c:s', 'mov_text', '-movflags', '+faststart'])
+
+ try:
+ self.real_run_ffmpeg(
+ [(concat_file, ['-hide_banner', '-nostdin', '-f', 'concat', '-safe', '0'])],
+ [(out_file, out_flags)])
+ finally:
+ os.remove(concat_file)
+
+ @classmethod
+ def _concat_spec(cls, in_files, concat_opts=None):
+ if concat_opts is None:
+ concat_opts = [{}] * len(in_files)
+ yield 'ffconcat version 1.0\n'
+ for file, opts in zip(in_files, concat_opts):
+ yield f'file {cls._quote_for_ffmpeg(cls._ffmpeg_filename_argument(file))}\n'
+ # Iterate explicitly to yield the following directives in order, ignoring the rest.
+ for directive in 'inpoint', 'outpoint', 'duration':
+ if directive in opts:
+ yield f'{directive} {opts[directive]}\n'
+
class FFmpegExtractAudioPP(FFmpegPostProcessor):
COMMON_AUDIO_EXTS = ('wav', 'flac', 'm4a', 'aiff', 'mp3', 'ogg', 'mka', 'opus', 'wma')
class FFmpegVideoConvertorPP(FFmpegPostProcessor):
SUPPORTED_EXTS = ('mp4', 'mkv', 'flv', 'webm', 'mov', 'avi', 'mp3', 'mka', 'm4a', 'ogg', 'opus')
FORMAT_RE = re.compile(r'{0}(?:/{0})*$'.format(r'(?:\w+>)?(?:%s)' % '|'.join(SUPPORTED_EXTS)))
- _action = 'converting'
+ _ACTION = 'converting'
def __init__(self, downloader=None, preferedformat=None):
super(FFmpegVideoConvertorPP, self).__init__(downloader)
return []
@PostProcessor._restrict_to(images=False)
- def run(self, information):
- path, source_ext = information['filepath'], information['ext'].lower()
+ def run(self, info):
+ filename, source_ext = info['filepath'], info['ext'].lower()
target_ext = self._target_ext(source_ext)
_skip_msg = (
- 'could not find a mapping for %s' if not target_ext
- else 'already is in target format %s' if source_ext == target_ext
+ f'could not find a mapping for {source_ext}' if not target_ext
+ else f'already is in target format {source_ext}' if source_ext == target_ext
else None)
if _skip_msg:
- self.to_screen('Not %s media file "%s"; %s' % (self._action, path, _skip_msg % source_ext))
- return [], information
+ self.to_screen(f'Not {self._ACTION} media file {filename!r}; {_skip_msg}')
+ return [], info
- prefix, sep, oldext = path.rpartition('.')
- outpath = prefix + sep + target_ext
- self.to_screen('%s video from %s to %s; Destination: %s' % (self._action.title(), source_ext, target_ext, outpath))
- self.run_ffmpeg(path, outpath, self._options(target_ext))
+ outpath = replace_extension(filename, target_ext, source_ext)
+ self.to_screen(f'{self._ACTION.title()} video from {source_ext} to {target_ext}; Destination: {outpath}')
+ self.run_ffmpeg(filename, outpath, self._options(target_ext))
- information['filepath'] = outpath
- information['format'] = information['ext'] = target_ext
- return [path], information
+ info['filepath'] = outpath
+ info['format'] = info['ext'] = target_ext
+ return [filename], info
class FFmpegVideoRemuxerPP(FFmpegVideoConvertorPP):
- _action = 'remuxing'
+ _ACTION = 'remuxing'
@staticmethod
def _options(target_ext):
return [], information
filename = information['filepath']
+ if information.get('duration') and self._duration_mismatch(
+ self._get_real_video_duration(information, False), information['duration']):
+ self.to_screen(f'Skipping {self.pp_key()} since the real and expected durations mismatch')
+ return [], information
ext = information['ext']
sub_langs, sub_names, sub_filenames = [], [], []
mp4_ass_warn = False
for lang, sub_info in subtitles.items():
+ if not os.path.exists(information.get('filepath', '')):
+ self.report_warning(f'Skipping embedding {lang} subtitle because the file is missing')
+ continue
sub_ext = sub_info['ext']
if sub_ext == 'json':
self.report_warning('JSON subtitles cannot be embedded')
temp_filename = prepend_extension(filename, 'temp')
self.to_screen('Embedding subtitles in "%s"' % filename)
self.run_ffmpeg_multiple_files(input_files, temp_filename, opts)
- os.remove(encodeFilename(filename))
- os.rename(encodeFilename(temp_filename), encodeFilename(filename))
+ os.replace(temp_filename, filename)
files_to_delete = [] if self._already_have_subtitle else sub_filenames
return files_to_delete, information
class FFmpegMetadataPP(FFmpegPostProcessor):
+ def __init__(self, downloader, add_metadata=True, add_chapters=True):
+ FFmpegPostProcessor.__init__(self, downloader)
+ self._add_metadata = add_metadata
+ self._add_chapters = add_chapters
+
@staticmethod
def _options(target_ext):
yield from ('-map', '0', '-dn')
@PostProcessor._restrict_to(images=False)
def run(self, info):
+ filename, metadata_filename = info['filepath'], None
+ options = []
+ if self._add_chapters and info.get('chapters'):
+ metadata_filename = replace_extension(filename, 'meta')
+ options.extend(self._get_chapter_opts(info['chapters'], metadata_filename))
+ if self._add_metadata:
+ options.extend(self._get_metadata_opts(info))
+
+ if not options:
+ self.to_screen('There isn\'t any metadata to add')
+ return [], info
+
+ temp_filename = prepend_extension(filename, 'temp')
+ self.to_screen('Adding metadata to "%s"' % filename)
+ self.run_ffmpeg_multiple_files(
+ (filename, metadata_filename), temp_filename,
+ itertools.chain(self._options(info['ext']), *options))
+ if metadata_filename:
+ os.remove(metadata_filename)
+ os.replace(temp_filename, filename)
+ return [], info
+
+ @staticmethod
+ def _get_chapter_opts(chapters, metadata_filename):
+ with io.open(metadata_filename, 'wt', encoding='utf-8') as f:
+ def ffmpeg_escape(text):
+ return re.sub(r'([\\=;#\n])', r'\\\1', text)
+
+ metadata_file_content = ';FFMETADATA1\n'
+ for chapter in chapters:
+ metadata_file_content += '[CHAPTER]\nTIMEBASE=1/1000\n'
+ metadata_file_content += 'START=%d\n' % (chapter['start_time'] * 1000)
+ metadata_file_content += 'END=%d\n' % (chapter['end_time'] * 1000)
+ chapter_title = chapter.get('title')
+ if chapter_title:
+ metadata_file_content += 'title=%s\n' % ffmpeg_escape(chapter_title)
+ f.write(metadata_file_content)
+ yield ('-map_metadata', '1')
+
+ def _get_metadata_opts(self, info):
metadata = {}
def add(meta_list, info_list=None):
for key in filter(lambda k: k.startswith(prefix), info.keys()):
add(key[len(prefix):], key)
- filename, metadata_filename = info['filepath'], None
- options = [('-metadata', f'{name}={value}') for name, value in metadata.items()]
+ for name, value in metadata.items():
+ yield ('-metadata', f'{name}={value}')
stream_idx = 0
for fmt in info.get('requested_formats') or []:
stream_count = 2 if 'none' not in (fmt.get('vcodec'), fmt.get('acodec')) else 1
if fmt.get('language'):
lang = ISO639Utils.short2long(fmt['language']) or fmt['language']
- options.extend(('-metadata:s:%d' % (stream_idx + i), 'language=%s' % lang)
- for i in range(stream_count))
+ for i in range(stream_count):
+ yield ('-metadata:s:%d' % (stream_idx + i), 'language=%s' % lang)
stream_idx += stream_count
- chapters = info.get('chapters', [])
- if chapters:
- metadata_filename = replace_extension(filename, 'meta')
- with io.open(metadata_filename, 'wt', encoding='utf-8') as f:
- def ffmpeg_escape(text):
- return re.sub(r'(=|;|#|\\|\n)', r'\\\1', text)
-
- metadata_file_content = ';FFMETADATA1\n'
- for chapter in chapters:
- metadata_file_content += '[CHAPTER]\nTIMEBASE=1/1000\n'
- metadata_file_content += 'START=%d\n' % (chapter['start_time'] * 1000)
- metadata_file_content += 'END=%d\n' % (chapter['end_time'] * 1000)
- chapter_title = chapter.get('title')
- if chapter_title:
- metadata_file_content += 'title=%s\n' % ffmpeg_escape(chapter_title)
- f.write(metadata_file_content)
- options.append(('-map_metadata', '1'))
-
if ('no-attach-info-json' not in self.get_param('compat_opts', [])
and '__infojson_filename' in info and info['ext'] in ('mkv', 'mka')):
- old_stream, new_stream = self.get_stream_number(filename, ('tags', 'mimetype'), 'application/json')
+ old_stream, new_stream = self.get_stream_number(info['filepath'], ('tags', 'mimetype'), 'application/json')
if old_stream is not None:
- options.append(('-map', '-0:%d' % old_stream))
+ yield ('-map', '-0:%d' % old_stream)
new_stream -= 1
- options.append((
- '-attach', info['__infojson_filename'],
- '-metadata:s:%d' % new_stream, 'mimetype=application/json'
- ))
-
- if not options:
- self.to_screen('There isn\'t any metadata to add')
- return [], info
-
- temp_filename = prepend_extension(filename, 'temp')
- self.to_screen('Adding metadata to "%s"' % filename)
- self.run_ffmpeg_multiple_files(
- (filename, metadata_filename), temp_filename,
- itertools.chain(self._options(info['ext']), *options))
- if chapters:
- os.remove(metadata_filename)
- os.remove(encodeFilename(filename))
- os.rename(encodeFilename(temp_filename), encodeFilename(filename))
- return [], info
+ yield ('-attach', info['__infojson_filename'],
+ '-metadata:s:%d' % new_stream, 'mimetype=application/json')
class FFmpegMergerPP(FFmpegPostProcessor):
filename = info['filepath']
temp_filename = prepend_extension(filename, 'temp')
args = ['-c', 'copy']
+ audio_streams = 0
for (i, fmt) in enumerate(info['requested_formats']):
if fmt.get('acodec') != 'none':
- args.extend(['-map', '%u:a:0' % (i)])
+ args.extend(['-map', f'{i}:a:0'])
+ if self.get_audio_codec(fmt['filepath']) == 'aac':
+ args.extend([f'-bsf:a:{audio_streams}', 'aac_adtstoasc'])
+ audio_streams += 1
if fmt.get('vcodec') != 'none':
args.extend(['-map', '%u:v:0' % (i)])
self.to_screen('Merging formats into "%s"' % filename)
self.to_screen(f'{msg} of "{filename}"')
self.run_ffmpeg(filename, temp_filename, options)
- os.remove(encodeFilename(filename))
- os.rename(encodeFilename(temp_filename), encodeFilename(filename))
+ os.replace(temp_filename, filename)
class FFmpegFixupStretchedPP(FFmpegFixupPostProcessor):
}
info['__files_to_move'][new_file] = replace_extension(
- info['__files_to_move'][old_file], new_ext)
+ info['__files_to_move'][sub['filepath']], new_ext)
return sub_filenames, info
class FFmpegSplitChaptersPP(FFmpegPostProcessor):
+ def __init__(self, downloader, force_keyframes=False):
+ FFmpegPostProcessor.__init__(self, downloader)
+ self._force_keyframes = force_keyframes
def _prepare_filename(self, number, chapter, info):
info = info.copy()
def run(self, info):
chapters = info.get('chapters') or []
if not chapters:
- self.report_warning('Chapter information is unavailable')
+ self.to_screen('Chapter information is unavailable')
return [], info
+ in_file = info['filepath']
+ if self._force_keyframes and len(chapters) > 1:
+ in_file = self.force_keyframes(in_file, (c['start_time'] for c in chapters))
self.to_screen('Splitting video by chapters; %d chapters found' % len(chapters))
for idx, chapter in enumerate(chapters):
destination, opts = self._ffmpeg_args_for_chapter(idx + 1, chapter, info)
- self.real_run_ffmpeg([(info['filepath'], opts)], [(destination, ['-c', 'copy'])])
+ self.real_run_ffmpeg([(in_file, opts)], [(destination, ['-c', 'copy'])])
+ if in_file != info['filepath']:
+ os.remove(in_file)
return [], info
if thumbnail_ext != 'webp' and self.is_webp(thumbnail_filename):
self.to_screen('Correcting thumbnail "%s" extension to webp' % thumbnail_filename)
webp_filename = replace_extension(thumbnail_filename, 'webp')
- if os.path.exists(webp_filename):
- os.remove(webp_filename)
- os.rename(encodeFilename(thumbnail_filename), encodeFilename(webp_filename))
+ os.replace(thumbnail_filename, webp_filename)
info['thumbnails'][idx]['filepath'] = webp_filename
info['__files_to_move'][webp_filename] = replace_extension(
info['__files_to_move'].pop(thumbnail_filename), 'webp')