yt-dlc/youtube_dl/postprocessor/ffmpeg.py

524 lines
20 KiB
Python

from __future__ import unicode_literals
import io
import os
import subprocess
import time
from .common import AudioConversionError, PostProcessor
from ..compat import (
compat_subprocess_get_DEVNULL,
)
from ..utils import (
encodeArgument,
encodeFilename,
get_exe_version,
is_outdated_version,
PostProcessingError,
prepend_extension,
shell_quote,
subtitles_filename,
dfxp2srt,
ISO639Utils,
)
class FFmpegPostProcessorError(PostProcessingError):
pass
class FFmpegPostProcessor(PostProcessor):
def __init__(self, downloader=None):
PostProcessor.__init__(self, downloader)
self._determine_executables()
def check_version(self):
if not self.available:
raise FFmpegPostProcessorError('ffmpeg or avconv not found. Please install one.')
required_version = '10-0' if self.basename == 'avconv' else '1.0'
if is_outdated_version(
self._versions[self.basename], required_version):
warning = 'Your copy of %s is outdated, update %s to version %s or newer if you encounter any errors.' % (
self.basename, self.basename, required_version)
if self._downloader:
self._downloader.report_warning(warning)
@staticmethod
def get_versions(downloader=None):
return FFmpegPostProcessor(downloader)._versions
def _determine_executables(self):
programs = ['avprobe', 'avconv', 'ffmpeg', 'ffprobe']
prefer_ffmpeg = self._downloader.params.get('prefer_ffmpeg', False)
self.basename = None
self.probe_basename = None
self._paths = None
self._versions = None
if self._downloader:
location = self._downloader.params.get('ffmpeg_location')
if location is not None:
if not os.path.exists(location):
self._downloader.report_warning(
'ffmpeg-location %s does not exist! '
'Continuing without avconv/ffmpeg.' % (location))
self._versions = {}
return
elif not os.path.isdir(location):
basename = os.path.splitext(os.path.basename(location))[0]
if basename not in programs:
self._downloader.report_warning(
'Cannot identify executable %s, its basename should be one of %s. '
'Continuing without avconv/ffmpeg.' %
(location, ', '.join(programs)))
self._versions = {}
return None
location = 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)
self._versions = dict(
(p, get_exe_version(self._paths[p], args=['-version']))
for p in programs)
if self._versions is None:
self._versions = dict(
(p, get_exe_version(p, args=['-version'])) for p in programs)
self._paths = dict((p, p) for p in programs)
if prefer_ffmpeg:
prefs = ('ffmpeg', 'avconv')
else:
prefs = ('avconv', 'ffmpeg')
for p in prefs:
if self._versions[p]:
self.basename = p
break
if prefer_ffmpeg:
prefs = ('ffprobe', 'avprobe')
else:
prefs = ('avprobe', 'ffprobe')
for p in prefs:
if self._versions[p]:
self.probe_basename = p
break
@property
def available(self):
return self.basename is not None
@property
def executable(self):
return self._paths[self.basename]
@property
def probe_available(self):
return self.probe_basename is not None
@property
def probe_executable(self):
return self._paths[self.probe_basename]
def run_ffmpeg_multiple_files(self, input_paths, out_path, opts):
self.check_version()
oldest_mtime = min(
os.stat(encodeFilename(path)).st_mtime for path in input_paths)
opts += self._configuration_args()
files_cmd = []
for path in input_paths:
files_cmd.extend([
encodeArgument('-i'),
encodeFilename(self._ffmpeg_filename_argument(path), True)
])
cmd = ([encodeFilename(self.executable, True), encodeArgument('-y')] +
files_cmd +
[encodeArgument(o) for o in opts] +
[encodeFilename(self._ffmpeg_filename_argument(out_path), True)])
if self._downloader.params.get('verbose', False):
self._downloader.to_screen('[debug] ffmpeg command line: %s' % shell_quote(cmd))
p = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE)
stdout, stderr = p.communicate()
if p.returncode != 0:
stderr = stderr.decode('utf-8', 'replace')
msg = stderr.strip().split('\n')[-1]
raise FFmpegPostProcessorError(msg)
self.try_utime(out_path, oldest_mtime, oldest_mtime)
def run_ffmpeg(self, path, out_path, opts):
self.run_ffmpeg_multiple_files([path], out_path, opts)
def _ffmpeg_filename_argument(self, 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 'file:' + fn
class FFmpegExtractAudioPP(FFmpegPostProcessor):
def __init__(self, downloader=None, preferredcodec=None, preferredquality=None, nopostoverwrites=False):
FFmpegPostProcessor.__init__(self, downloader)
if preferredcodec is None:
preferredcodec = 'best'
self._preferredcodec = preferredcodec
self._preferredquality = preferredquality
self._nopostoverwrites = nopostoverwrites
def get_audio_codec(self, path):
if not self.probe_available:
raise PostProcessingError('ffprobe or avprobe not found. Please install one.')
try:
cmd = [
encodeFilename(self.probe_executable, True),
encodeArgument('-show_streams'),
encodeFilename(self._ffmpeg_filename_argument(path), True)]
if self._downloader.params.get('verbose', False):
self._downloader.to_screen('[debug] %s command line: %s' % (self.basename, shell_quote(cmd)))
handle = subprocess.Popen(cmd, stderr=compat_subprocess_get_DEVNULL(), stdout=subprocess.PIPE, stdin=subprocess.PIPE)
output = handle.communicate()[0]
if handle.wait() != 0:
return None
except (IOError, OSError):
return None
audio_codec = None
for line in output.decode('ascii', 'ignore').split('\n'):
if line.startswith('codec_name='):
audio_codec = line.split('=')[1].strip()
elif line.strip() == 'codec_type=audio' and audio_codec is not None:
return audio_codec
return None
def run_ffmpeg(self, path, out_path, codec, more_opts):
if codec is None:
acodec_opts = []
else:
acodec_opts = ['-acodec', codec]
opts = ['-vn'] + acodec_opts + more_opts
try:
FFmpegPostProcessor.run_ffmpeg(self, path, out_path, opts)
except FFmpegPostProcessorError as err:
raise AudioConversionError(err.msg)
def run(self, information):
path = information['filepath']
filecodec = self.get_audio_codec(path)
if filecodec is None:
raise PostProcessingError('WARNING: unable to obtain file audio codec with ffprobe')
more_opts = []
if self._preferredcodec == 'best' or self._preferredcodec == filecodec or (self._preferredcodec == 'm4a' and filecodec == 'aac'):
if filecodec == 'aac' and self._preferredcodec in ['m4a', 'best']:
# Lossless, but in another container
acodec = 'copy'
extension = 'm4a'
more_opts = ['-bsf:a', 'aac_adtstoasc']
elif filecodec in ['aac', 'mp3', 'vorbis', 'opus']:
# Lossless if possible
acodec = 'copy'
extension = filecodec
if filecodec == 'aac':
more_opts = ['-f', 'adts']
if filecodec == 'vorbis':
extension = 'ogg'
else:
# MP3 otherwise.
acodec = 'libmp3lame'
extension = 'mp3'
more_opts = []
if self._preferredquality is not None:
if int(self._preferredquality) < 10:
more_opts += ['-q:a', self._preferredquality]
else:
more_opts += ['-b:a', self._preferredquality + 'k']
else:
# We convert the audio (lossy)
acodec = {'mp3': 'libmp3lame', 'aac': 'aac', 'm4a': 'aac', 'opus': 'opus', 'vorbis': 'libvorbis', 'wav': None}[self._preferredcodec]
extension = self._preferredcodec
more_opts = []
if self._preferredquality is not None:
# The opus codec doesn't support the -aq option
if int(self._preferredquality) < 10 and extension != 'opus':
more_opts += ['-q:a', self._preferredquality]
else:
more_opts += ['-b:a', self._preferredquality + 'k']
if self._preferredcodec == 'aac':
more_opts += ['-f', 'adts']
if self._preferredcodec == 'm4a':
more_opts += ['-bsf:a', 'aac_adtstoasc']
if self._preferredcodec == 'vorbis':
extension = 'ogg'
if self._preferredcodec == 'wav':
extension = 'wav'
more_opts += ['-f', 'wav']
prefix, sep, ext = path.rpartition('.') # not os.path.splitext, since the latter does not work on unicode in all setups
new_path = prefix + sep + extension
# If we download foo.mp3 and convert it to... foo.mp3, then don't delete foo.mp3, silly.
if (new_path == path or
(self._nopostoverwrites and os.path.exists(encodeFilename(new_path)))):
self._downloader.to_screen('[ffmpeg] Post-process file %s exists, skipping' % new_path)
return [], information
try:
self._downloader.to_screen('[ffmpeg] Destination: ' + new_path)
self.run_ffmpeg(path, new_path, acodec, more_opts)
except AudioConversionError as e:
raise PostProcessingError(
'audio conversion failed: ' + e.msg)
except Exception:
raise PostProcessingError('error running ' + self.basename)
# Try to update the date time for extracted audio file.
if information.get('filetime') is not None:
self.try_utime(
new_path, time.time(), information['filetime'],
errnote='Cannot update utime of audio file')
information['filepath'] = new_path
information['ext'] = extension
return [path], information
class FFmpegVideoConvertorPP(FFmpegPostProcessor):
def __init__(self, downloader=None, preferedformat=None):
super(FFmpegVideoConvertorPP, self).__init__(downloader)
self._preferedformat = preferedformat
def run(self, information):
path = information['filepath']
if information['ext'] == self._preferedformat:
self._downloader.to_screen('[ffmpeg] Not converting video file %s - already is in target format %s' % (path, self._preferedformat))
return [], information
options = []
if self._preferedformat == 'avi':
options.extend(['-c:v', 'libxvid', '-vtag', 'XVID'])
prefix, sep, ext = path.rpartition('.')
outpath = prefix + sep + self._preferedformat
self._downloader.to_screen('[' + 'ffmpeg' + '] Converting video from %s to %s, Destination: ' % (information['ext'], self._preferedformat) + outpath)
self.run_ffmpeg(path, outpath, options)
information['filepath'] = outpath
information['format'] = self._preferedformat
information['ext'] = self._preferedformat
return [path], information
class FFmpegEmbedSubtitlePP(FFmpegPostProcessor):
def run(self, information):
if information['ext'] not in ['mp4', 'mkv']:
self._downloader.to_screen('[ffmpeg] Subtitles can only be embedded in mp4 or mkv files')
return [], information
subtitles = information.get('requested_subtitles')
if not subtitles:
self._downloader.to_screen('[ffmpeg] There aren\'t any subtitles to embed')
return [], information
sub_langs = list(subtitles.keys())
filename = information['filepath']
sub_filenames = [subtitles_filename(filename, lang, sub_info['ext']) for lang, sub_info in subtitles.items()]
input_files = [filename] + sub_filenames
opts = [
'-map', '0',
'-c', 'copy',
# Don't copy the existing subtitles, we may be running the
# postprocessor a second time
'-map', '-0:s',
]
if information['ext'] == 'mp4':
opts += ['-c:s', 'mov_text']
for (i, lang) in enumerate(sub_langs):
opts.extend(['-map', '%d:0' % (i + 1)])
lang_code = ISO639Utils.short2long(lang)
if lang_code is not None:
opts.extend(['-metadata:s:s:%d' % i, 'language=%s' % lang_code])
temp_filename = prepend_extension(filename, 'temp')
self._downloader.to_screen('[ffmpeg] 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))
return sub_filenames, information
class FFmpegMetadataPP(FFmpegPostProcessor):
def run(self, info):
metadata = {}
if info.get('title') is not None:
metadata['title'] = info['title']
if info.get('upload_date') is not None:
metadata['date'] = info['upload_date']
if info.get('artist') is not None:
metadata['artist'] = info['artist']
elif info.get('uploader') is not None:
metadata['artist'] = info['uploader']
elif info.get('uploader_id') is not None:
metadata['artist'] = info['uploader_id']
if info.get('description') is not None:
metadata['description'] = info['description']
metadata['comment'] = info['description']
if info.get('webpage_url') is not None:
metadata['purl'] = info['webpage_url']
if info.get('album') is not None:
metadata['album'] = info['album']
if not metadata:
self._downloader.to_screen('[ffmpeg] There isn\'t any metadata to add')
return [], info
filename = info['filepath']
temp_filename = prepend_extension(filename, 'temp')
if info['ext'] == 'm4a':
options = ['-vn', '-acodec', 'copy']
else:
options = ['-c', 'copy']
for (name, value) in metadata.items():
options.extend(['-metadata', '%s=%s' % (name, value)])
self._downloader.to_screen('[ffmpeg] Adding metadata to \'%s\'' % filename)
self.run_ffmpeg(filename, temp_filename, options)
os.remove(encodeFilename(filename))
os.rename(encodeFilename(temp_filename), encodeFilename(filename))
return [], info
class FFmpegMergerPP(FFmpegPostProcessor):
def run(self, info):
filename = info['filepath']
temp_filename = prepend_extension(filename, 'temp')
args = ['-c', 'copy', '-map', '0:v:0', '-map', '1:a:0']
self._downloader.to_screen('[ffmpeg] Merging formats into "%s"' % filename)
self.run_ffmpeg_multiple_files(info['__files_to_merge'], temp_filename, args)
os.rename(encodeFilename(temp_filename), encodeFilename(filename))
return info['__files_to_merge'], info
def can_merge(self):
# TODO: figure out merge-capable ffmpeg version
if self.basename != 'avconv':
return True
required_version = '10-0'
if is_outdated_version(
self._versions[self.basename], required_version):
warning = ('Your copy of %s is outdated and unable to properly mux separate video and audio files, '
'youtube-dl will download single file media. '
'Update %s to version %s or newer to fix this.') % (
self.basename, self.basename, required_version)
if self._downloader:
self._downloader.report_warning(warning)
return False
return True
class FFmpegFixupStretchedPP(FFmpegPostProcessor):
def run(self, info):
stretched_ratio = info.get('stretched_ratio')
if stretched_ratio is None or stretched_ratio == 1:
return [], info
filename = info['filepath']
temp_filename = prepend_extension(filename, 'temp')
options = ['-c', 'copy', '-aspect', '%f' % stretched_ratio]
self._downloader.to_screen('[ffmpeg] Fixing aspect ratio in "%s"' % filename)
self.run_ffmpeg(filename, temp_filename, options)
os.remove(encodeFilename(filename))
os.rename(encodeFilename(temp_filename), encodeFilename(filename))
return [], info
class FFmpegFixupM4aPP(FFmpegPostProcessor):
def run(self, info):
if info.get('container') != 'm4a_dash':
return [], info
filename = info['filepath']
temp_filename = prepend_extension(filename, 'temp')
options = ['-c', 'copy', '-f', 'mp4']
self._downloader.to_screen('[ffmpeg] Correcting container in "%s"' % filename)
self.run_ffmpeg(filename, temp_filename, options)
os.remove(encodeFilename(filename))
os.rename(encodeFilename(temp_filename), encodeFilename(filename))
return [], info
class FFmpegSubtitlesConvertorPP(FFmpegPostProcessor):
def __init__(self, downloader=None, format=None):
super(FFmpegSubtitlesConvertorPP, self).__init__(downloader)
self.format = format
def run(self, info):
subs = info.get('requested_subtitles')
filename = info['filepath']
new_ext = self.format
new_format = new_ext
if new_format == 'vtt':
new_format = 'webvtt'
if subs is None:
self._downloader.to_screen('[ffmpeg] There aren\'t any subtitles to convert')
return [], info
self._downloader.to_screen('[ffmpeg] Converting subtitles')
for lang, sub in subs.items():
ext = sub['ext']
if ext == new_ext:
self._downloader.to_screen(
'[ffmpeg] Subtitle file for %s is already in the requested'
'format' % new_ext)
continue
new_file = subtitles_filename(filename, lang, new_ext)
if ext == 'dfxp' or ext == 'ttml':
self._downloader.report_warning(
'You have requested to convert dfxp (TTML) subtitles into another format, '
'which results in style information loss')
dfxp_file = subtitles_filename(filename, lang, ext)
srt_file = subtitles_filename(filename, lang, 'srt')
with io.open(dfxp_file, 'rt', encoding='utf-8') as f:
srt_data = dfxp2srt(f.read())
with io.open(srt_file, 'wt', encoding='utf-8') as f:
f.write(srt_data)
ext = 'srt'
subs[lang] = {
'ext': 'srt',
'data': srt_data
}
if new_ext == 'srt':
continue
self.run_ffmpeg(
subtitles_filename(filename, lang, ext),
new_file, ['-f', new_format])
with io.open(new_file, 'rt', encoding='utf-8') as f:
subs[lang] = {
'ext': ext,
'data': f.read(),
}
return [], info