From 2415e66f889f38503b73e8ebc5f43ca342390e5c Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Mon, 15 Apr 2024 18:49:24 +0200 Subject: Adding upstream version 2024.03.10. Signed-off-by: Daniel Baumann --- yt_dlp/postprocessor/__init__.py | 47 + yt_dlp/postprocessor/common.py | 215 +++++ yt_dlp/postprocessor/embedthumbnail.py | 227 +++++ yt_dlp/postprocessor/exec.py | 41 + yt_dlp/postprocessor/ffmpeg.py | 1192 ++++++++++++++++++++++++ yt_dlp/postprocessor/metadataparser.py | 125 +++ yt_dlp/postprocessor/modify_chapters.py | 336 +++++++ yt_dlp/postprocessor/movefilesafterdownload.py | 53 ++ yt_dlp/postprocessor/sponskrub.py | 98 ++ yt_dlp/postprocessor/sponsorblock.py | 104 +++ yt_dlp/postprocessor/xattrpp.py | 63 ++ 11 files changed, 2501 insertions(+) create mode 100644 yt_dlp/postprocessor/__init__.py create mode 100644 yt_dlp/postprocessor/common.py create mode 100644 yt_dlp/postprocessor/embedthumbnail.py create mode 100644 yt_dlp/postprocessor/exec.py create mode 100644 yt_dlp/postprocessor/ffmpeg.py create mode 100644 yt_dlp/postprocessor/metadataparser.py create mode 100644 yt_dlp/postprocessor/modify_chapters.py create mode 100644 yt_dlp/postprocessor/movefilesafterdownload.py create mode 100644 yt_dlp/postprocessor/sponskrub.py create mode 100644 yt_dlp/postprocessor/sponsorblock.py create mode 100644 yt_dlp/postprocessor/xattrpp.py (limited to 'yt_dlp/postprocessor') diff --git a/yt_dlp/postprocessor/__init__.py b/yt_dlp/postprocessor/__init__.py new file mode 100644 index 0000000..bfe9df7 --- /dev/null +++ b/yt_dlp/postprocessor/__init__.py @@ -0,0 +1,47 @@ +# flake8: noqa: F401 + +from .common import PostProcessor +from .embedthumbnail import EmbedThumbnailPP +from .exec import ExecAfterDownloadPP, ExecPP +from .ffmpeg import ( + FFmpegConcatPP, + FFmpegCopyStreamPP, + FFmpegEmbedSubtitlePP, + FFmpegExtractAudioPP, + FFmpegFixupDuplicateMoovPP, + FFmpegFixupDurationPP, + FFmpegFixupM3u8PP, + FFmpegFixupM4aPP, + FFmpegFixupStretchedPP, + FFmpegFixupTimestampPP, + FFmpegMergerPP, + FFmpegMetadataPP, + FFmpegPostProcessor, + FFmpegSplitChaptersPP, + FFmpegSubtitlesConvertorPP, + FFmpegThumbnailsConvertorPP, + FFmpegVideoConvertorPP, + FFmpegVideoRemuxerPP, +) +from .metadataparser import ( + MetadataFromFieldPP, + MetadataFromTitlePP, + MetadataParserPP, +) +from .modify_chapters import ModifyChaptersPP +from .movefilesafterdownload import MoveFilesAfterDownloadPP +from .sponskrub import SponSkrubPP +from .sponsorblock import SponsorBlockPP +from .xattrpp import XAttrMetadataPP +from ..plugins import load_plugins + +_PLUGIN_CLASSES = load_plugins('postprocessor', 'PP') + + +def get_postprocessor(key): + return globals()[key + 'PP'] + + +globals().update(_PLUGIN_CLASSES) +__all__ = [name for name in globals().keys() if name.endswith('PP')] +__all__.extend(('PostProcessor', 'FFmpegPostProcessor')) diff --git a/yt_dlp/postprocessor/common.py b/yt_dlp/postprocessor/common.py new file mode 100644 index 0000000..8cef86c --- /dev/null +++ b/yt_dlp/postprocessor/common.py @@ -0,0 +1,215 @@ +import functools +import json +import os + +from ..networking import Request +from ..networking.exceptions import HTTPError, network_exceptions +from ..utils import ( + PostProcessingError, + RetryManager, + _configuration_args, + deprecation_warning, + encodeFilename, +) + + +class PostProcessorMetaClass(type): + @staticmethod + def run_wrapper(func): + @functools.wraps(func) + def run(self, info, *args, **kwargs): + info_copy = self._copy_infodict(info) + self._hook_progress({'status': 'started'}, info_copy) + ret = func(self, info, *args, **kwargs) + if ret is not None: + _, info = ret + self._hook_progress({'status': 'finished'}, info_copy) + return ret + return run + + def __new__(cls, name, bases, attrs): + if 'run' in attrs: + attrs['run'] = cls.run_wrapper(attrs['run']) + return type.__new__(cls, name, bases, attrs) + + +class PostProcessor(metaclass=PostProcessorMetaClass): + """Post Processor class. + + PostProcessor objects can be added to downloaders with their + add_post_processor() method. When the downloader has finished a + successful download, it will take its internal chain of PostProcessors + and start calling the run() method on each one of them, first with + an initial argument and then with the returned value of the previous + PostProcessor. + + PostProcessor objects follow a "mutual registration" process similar + to InfoExtractor objects. + + Optionally PostProcessor can use a list of additional command-line arguments + with self._configuration_args. + """ + + _downloader = None + + def __init__(self, downloader=None): + self._progress_hooks = [] + self.add_progress_hook(self.report_progress) + self.set_downloader(downloader) + self.PP_NAME = self.pp_key() + + @classmethod + def pp_key(cls): + name = cls.__name__[:-2] + return name[6:] if name[:6].lower() == 'ffmpeg' else name + + def to_screen(self, text, prefix=True, *args, **kwargs): + if self._downloader: + tag = '[%s] ' % self.PP_NAME if prefix else '' + return self._downloader.to_screen(f'{tag}{text}', *args, **kwargs) + + def report_warning(self, text, *args, **kwargs): + if self._downloader: + return self._downloader.report_warning(text, *args, **kwargs) + + def deprecation_warning(self, msg): + warn = getattr(self._downloader, 'deprecation_warning', deprecation_warning) + return warn(msg, stacklevel=1) + + def deprecated_feature(self, msg): + if self._downloader: + return self._downloader.deprecated_feature(msg) + return deprecation_warning(msg, stacklevel=1) + + def report_error(self, text, *args, **kwargs): + self.deprecation_warning('"yt_dlp.postprocessor.PostProcessor.report_error" is deprecated. ' + 'raise "yt_dlp.utils.PostProcessingError" instead') + if self._downloader: + return self._downloader.report_error(text, *args, **kwargs) + + def write_debug(self, text, *args, **kwargs): + if self._downloader: + return self._downloader.write_debug(text, *args, **kwargs) + + def _delete_downloaded_files(self, *files_to_delete, **kwargs): + if self._downloader: + return self._downloader._delete_downloaded_files(*files_to_delete, **kwargs) + for filename in set(filter(None, files_to_delete)): + os.remove(filename) + + def get_param(self, name, default=None, *args, **kwargs): + if self._downloader: + return self._downloader.params.get(name, default, *args, **kwargs) + return default + + def set_downloader(self, downloader): + """Sets the downloader for this PP.""" + self._downloader = downloader + for ph in getattr(downloader, '_postprocessor_hooks', []): + self.add_progress_hook(ph) + + def _copy_infodict(self, info_dict): + return getattr(self._downloader, '_copy_infodict', dict)(info_dict) + + @staticmethod + def _restrict_to(*, video=True, audio=True, images=True, simulated=True): + allowed = {'video': video, 'audio': audio, 'images': images} + + def decorator(func): + @functools.wraps(func) + def wrapper(self, info): + if not simulated and (self.get_param('simulate') or self.get_param('skip_download')): + return [], info + format_type = ( + 'video' if info.get('vcodec') != 'none' + else 'audio' if info.get('acodec') != 'none' + else 'images') + if allowed[format_type]: + return func(self, info) + else: + self.to_screen('Skipping %s' % format_type) + return [], info + return wrapper + return decorator + + def run(self, information): + """Run the PostProcessor. + + The "information" argument is a dictionary like the ones + composed by InfoExtractors. The only difference is that this + one has an extra field called "filepath" that points to the + downloaded file. + + This method returns a tuple, the first element is a list of the files + that can be deleted, and the second of which is the updated + information. + + In addition, this method may raise a PostProcessingError + exception if post processing fails. + """ + return [], information # by default, keep file and do nothing + + def try_utime(self, path, atime, mtime, errnote='Cannot update utime of file'): + try: + os.utime(encodeFilename(path), (atime, mtime)) + except Exception: + self.report_warning(errnote) + + def _configuration_args(self, exe, *args, **kwargs): + return _configuration_args( + self.pp_key(), self.get_param('postprocessor_args'), exe, *args, **kwargs) + + def _hook_progress(self, status, info_dict): + if not self._progress_hooks: + return + status.update({ + 'info_dict': info_dict, + 'postprocessor': self.pp_key(), + }) + for ph in self._progress_hooks: + ph(status) + + def add_progress_hook(self, ph): + # See YoutubeDl.py (search for postprocessor_hooks) for a description of this interface + self._progress_hooks.append(ph) + + def report_progress(self, s): + s['_default_template'] = '%(postprocessor)s %(status)s' % s + if not self._downloader: + return + + progress_dict = s.copy() + progress_dict.pop('info_dict') + progress_dict = {'info': s['info_dict'], 'progress': progress_dict} + + progress_template = self.get_param('progress_template', {}) + tmpl = progress_template.get('postprocess') + if tmpl: + self._downloader.to_screen( + self._downloader.evaluate_outtmpl(tmpl, progress_dict), quiet=False) + + self._downloader.to_console_title(self._downloader.evaluate_outtmpl( + progress_template.get('postprocess-title') or 'yt-dlp %(progress._default_template)s', + progress_dict)) + + def _retry_download(self, err, count, retries): + # While this is not an extractor, it behaves similar to one and + # so obey extractor_retries and "--retry-sleep extractor" + RetryManager.report_retry(err, count, retries, info=self.to_screen, warn=self.report_warning, + sleep_func=self.get_param('retry_sleep_functions', {}).get('extractor')) + + def _download_json(self, url, *, expected_http_errors=(404,)): + self.write_debug(f'{self.PP_NAME} query: {url}') + for retry in RetryManager(self.get_param('extractor_retries', 3), self._retry_download): + try: + rsp = self._downloader.urlopen(Request(url)) + except network_exceptions as e: + if isinstance(e, HTTPError) and e.status in expected_http_errors: + return None + retry.error = PostProcessingError(f'Unable to communicate with {self.PP_NAME} API: {e}') + continue + return json.loads(rsp.read().decode(rsp.headers.get_param('charset') or 'utf-8')) + + +class AudioConversionError(PostProcessingError): # Deprecated + pass diff --git a/yt_dlp/postprocessor/embedthumbnail.py b/yt_dlp/postprocessor/embedthumbnail.py new file mode 100644 index 0000000..9c53729 --- /dev/null +++ b/yt_dlp/postprocessor/embedthumbnail.py @@ -0,0 +1,227 @@ +import base64 +import os +import re +import subprocess + +from .common import PostProcessor +from .ffmpeg import FFmpegPostProcessor, FFmpegThumbnailsConvertorPP +from ..compat import imghdr +from ..dependencies import mutagen +from ..utils import ( + Popen, + PostProcessingError, + check_executable, + encodeArgument, + encodeFilename, + error_to_compat_str, + prepend_extension, + shell_quote, +) + +if mutagen: + from mutagen.flac import FLAC, Picture + from mutagen.mp4 import MP4, MP4Cover + from mutagen.oggopus import OggOpus + from mutagen.oggvorbis import OggVorbis + + +class EmbedThumbnailPPError(PostProcessingError): + pass + + +class EmbedThumbnailPP(FFmpegPostProcessor): + + def __init__(self, downloader=None, already_have_thumbnail=False): + FFmpegPostProcessor.__init__(self, downloader) + self._already_have_thumbnail = already_have_thumbnail + + def _get_thumbnail_resolution(self, filename, thumbnail_dict): + def guess(): + width, height = thumbnail_dict.get('width'), thumbnail_dict.get('height') + if width and height: + return width, height + + try: + size_regex = r',\s*(?P\d+)x(?P\d+)\s*[,\[]' + size_result = self.run_ffmpeg(filename, None, ['-hide_banner'], expected_retcodes=(1,)) + mobj = re.search(size_regex, size_result) + if mobj is None: + return guess() + except PostProcessingError as err: + self.report_warning('unable to find the thumbnail resolution; %s' % error_to_compat_str(err)) + return guess() + return int(mobj.group('w')), int(mobj.group('h')) + + def _report_run(self, exe, filename): + self.to_screen(f'{exe}: Adding thumbnail to "{filename}"') + + @PostProcessor._restrict_to(images=False) + def run(self, info): + filename = info['filepath'] + temp_filename = prepend_extension(filename, 'temp') + + if not info.get('thumbnails'): + self.to_screen('There aren\'t any thumbnails to embed') + return [], info + + idx = next((-i for i, t in enumerate(info['thumbnails'][::-1], 1) if t.get('filepath')), None) + if idx is None: + self.to_screen('There are no thumbnails on disk') + return [], info + thumbnail_filename = info['thumbnails'][idx]['filepath'] + if not os.path.exists(encodeFilename(thumbnail_filename)): + self.report_warning('Skipping embedding the thumbnail because the file is missing.') + return [], info + + # Correct extension for WebP file with wrong extension (see #25687, #25717) + convertor = FFmpegThumbnailsConvertorPP(self._downloader) + convertor.fixup_webp(info, idx) + + original_thumbnail = thumbnail_filename = info['thumbnails'][idx]['filepath'] + + # Convert unsupported thumbnail formats (see #25687, #25717) + # PNG is preferred since JPEG is lossy + thumbnail_ext = os.path.splitext(thumbnail_filename)[1][1:] + if info['ext'] not in ('mkv', 'mka') and thumbnail_ext not in ('jpg', 'jpeg', 'png'): + thumbnail_filename = convertor.convert_thumbnail(thumbnail_filename, 'png') + thumbnail_ext = 'png' + + mtime = os.stat(encodeFilename(filename)).st_mtime + + success = True + if info['ext'] == 'mp3': + options = [ + '-c', 'copy', '-map', '0:0', '-map', '1:0', '-write_id3v1', '1', '-id3v2_version', '3', + '-metadata:s:v', 'title="Album cover"', '-metadata:s:v', 'comment=Cover (front)'] + + self._report_run('ffmpeg', filename) + self.run_ffmpeg_multiple_files([filename, thumbnail_filename], temp_filename, options) + + elif info['ext'] in ['mkv', 'mka']: + options = list(self.stream_copy_opts()) + + mimetype = f'image/{thumbnail_ext.replace("jpg", "jpeg")}' + old_stream, new_stream = self.get_stream_number( + filename, ('tags', 'mimetype'), mimetype) + if old_stream is not None: + options.extend(['-map', '-0:%d' % old_stream]) + new_stream -= 1 + options.extend([ + '-attach', self._ffmpeg_filename_argument(thumbnail_filename), + '-metadata:s:%d' % new_stream, 'mimetype=%s' % mimetype, + '-metadata:s:%d' % new_stream, 'filename=cover.%s' % thumbnail_ext]) + + self._report_run('ffmpeg', filename) + self.run_ffmpeg(filename, temp_filename, options) + + elif info['ext'] in ['m4a', 'mp4', 'm4v', 'mov']: + prefer_atomicparsley = 'embed-thumbnail-atomicparsley' in self.get_param('compat_opts', []) + # Method 1: Use mutagen + if not mutagen or prefer_atomicparsley: + success = False + else: + try: + self._report_run('mutagen', filename) + meta = MP4(filename) + # NOTE: the 'covr' atom is a non-standard MPEG-4 atom, + # Apple iTunes 'M4A' files include the 'moov.udta.meta.ilst' atom. + f = {'jpeg': MP4Cover.FORMAT_JPEG, 'png': MP4Cover.FORMAT_PNG}[imghdr.what(thumbnail_filename)] + with open(thumbnail_filename, 'rb') as thumbfile: + thumb_data = thumbfile.read() + meta.tags['covr'] = [MP4Cover(data=thumb_data, imageformat=f)] + meta.save() + temp_filename = filename + except Exception as err: + self.report_warning('unable to embed using mutagen; %s' % error_to_compat_str(err)) + success = False + + # Method 2: Use AtomicParsley + if not success: + success = True + atomicparsley = next(( + # libatomicparsley.so : See https://github.com/xibr/ytdlp-lazy/issues/1 + x for x in ['AtomicParsley', 'atomicparsley', 'libatomicparsley.so'] + if check_executable(x, ['-v'])), None) + if atomicparsley is None: + self.to_screen('Neither mutagen nor AtomicParsley was found. Falling back to ffmpeg') + success = False + else: + if not prefer_atomicparsley: + self.to_screen('mutagen was not found. Falling back to AtomicParsley') + cmd = [encodeFilename(atomicparsley, True), + encodeFilename(filename, True), + encodeArgument('--artwork'), + encodeFilename(thumbnail_filename, True), + encodeArgument('-o'), + encodeFilename(temp_filename, True)] + cmd += [encodeArgument(o) for o in self._configuration_args('AtomicParsley')] + + self._report_run('atomicparsley', filename) + self.write_debug('AtomicParsley command line: %s' % shell_quote(cmd)) + stdout, stderr, returncode = Popen.run(cmd, text=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE) + if returncode: + self.report_warning(f'Unable to embed thumbnails using AtomicParsley; {stderr.strip()}') + # for formats that don't support thumbnails (like 3gp) AtomicParsley + # won't create to the temporary file + if 'No changes' in stdout: + self.report_warning('The file format doesn\'t support embedding a thumbnail') + success = False + + # Method 3: Use ffmpeg+ffprobe + # Thumbnails attached using this method doesn't show up as cover in some cases + # See https://github.com/yt-dlp/yt-dlp/issues/2125, https://github.com/yt-dlp/yt-dlp/issues/411 + if not success: + success = True + try: + options = [*self.stream_copy_opts(), '-map', '1'] + + old_stream, new_stream = self.get_stream_number( + filename, ('disposition', 'attached_pic'), 1) + if old_stream is not None: + options.extend(['-map', '-0:%d' % old_stream]) + new_stream -= 1 + options.extend(['-disposition:%s' % new_stream, 'attached_pic']) + + self._report_run('ffmpeg', filename) + self.run_ffmpeg_multiple_files([filename, thumbnail_filename], temp_filename, options) + except PostProcessingError as err: + success = False + raise EmbedThumbnailPPError(f'Unable to embed using ffprobe & ffmpeg; {err}') + + elif info['ext'] in ['ogg', 'opus', 'flac']: + if not mutagen: + raise EmbedThumbnailPPError('module mutagen was not found. Please install using `python3 -m pip install mutagen`') + + self._report_run('mutagen', filename) + f = {'opus': OggOpus, 'flac': FLAC, 'ogg': OggVorbis}[info['ext']](filename) + + pic = Picture() + pic.mime = 'image/%s' % imghdr.what(thumbnail_filename) + with open(thumbnail_filename, 'rb') as thumbfile: + pic.data = thumbfile.read() + pic.type = 3 # front cover + res = self._get_thumbnail_resolution(thumbnail_filename, info['thumbnails'][idx]) + if res is not None: + pic.width, pic.height = res + + if info['ext'] == 'flac': + f.add_picture(pic) + else: + # https://wiki.xiph.org/VorbisComment#METADATA_BLOCK_PICTURE + f['METADATA_BLOCK_PICTURE'] = base64.b64encode(pic.write()).decode('ascii') + f.save() + temp_filename = filename + + else: + raise EmbedThumbnailPPError('Supported filetypes for thumbnail embedding are: mp3, mkv/mka, ogg/opus/flac, m4a/mp4/m4v/mov') + + if success and temp_filename != filename: + os.replace(temp_filename, filename) + + self.try_utime(filename, mtime, mtime) + converted = original_thumbnail != thumbnail_filename + self._delete_downloaded_files( + thumbnail_filename if converted or not self._already_have_thumbnail else None, + original_thumbnail if converted and not self._already_have_thumbnail else None, + info=info) + return [], info diff --git a/yt_dlp/postprocessor/exec.py b/yt_dlp/postprocessor/exec.py new file mode 100644 index 0000000..c2e73fb --- /dev/null +++ b/yt_dlp/postprocessor/exec.py @@ -0,0 +1,41 @@ +from .common import PostProcessor +from ..compat import compat_shlex_quote +from ..utils import Popen, PostProcessingError, variadic + + +class ExecPP(PostProcessor): + + def __init__(self, downloader, exec_cmd): + PostProcessor.__init__(self, downloader) + self.exec_cmd = variadic(exec_cmd) + + def parse_cmd(self, cmd, info): + tmpl, tmpl_dict = self._downloader.prepare_outtmpl(cmd, info) + if tmpl_dict: # if there are no replacements, tmpl_dict = {} + return self._downloader.escape_outtmpl(tmpl) % tmpl_dict + + filepath = info.get('filepath', info.get('_filename')) + # If video, and no replacements are found, replace {} for backard compatibility + if filepath: + if '{}' not in cmd: + cmd += ' {}' + cmd = cmd.replace('{}', compat_shlex_quote(filepath)) + return cmd + + def run(self, info): + for tmpl in self.exec_cmd: + cmd = self.parse_cmd(tmpl, info) + self.to_screen(f'Executing command: {cmd}') + _, _, return_code = Popen.run(cmd, shell=True) + if return_code != 0: + raise PostProcessingError(f'Command returned error code {return_code}') + return [], info + + +# Deprecated +class ExecAfterDownloadPP(ExecPP): + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + self.deprecation_warning( + 'yt_dlp.postprocessor.ExecAfterDownloadPP is deprecated ' + 'and may be removed in a future version. Use yt_dlp.postprocessor.ExecPP instead') diff --git a/yt_dlp/postprocessor/ffmpeg.py b/yt_dlp/postprocessor/ffmpeg.py new file mode 100644 index 0000000..7d7f3f0 --- /dev/null +++ b/yt_dlp/postprocessor/ffmpeg.py @@ -0,0 +1,1192 @@ +import collections +import contextvars +import itertools +import json +import os +import re +import subprocess +import time + +from .common import PostProcessor +from ..compat import functools, imghdr +from ..utils import ( + MEDIA_EXTENSIONS, + ISO639Utils, + Popen, + PostProcessingError, + _get_exe_version_output, + deprecation_warning, + detect_exe_version, + determine_ext, + dfxp2srt, + encodeArgument, + encodeFilename, + filter_dict, + float_or_none, + is_outdated_version, + orderedSet, + prepend_extension, + replace_extension, + shell_quote, + traverse_obj, + variadic, + write_json_file, +) + +EXT_TO_OUT_FORMATS = { + 'aac': 'adts', + 'flac': 'flac', + 'm4a': 'ipod', + 'mka': 'matroska', + 'mkv': 'matroska', + 'mpg': 'mpeg', + 'ogv': 'ogg', + 'ts': 'mpegts', + 'wma': 'asf', + 'wmv': 'asf', + 'weba': 'webm', + 'vtt': 'webvtt', +} +ACODECS = { + # name: (ext, encoder, opts) + 'mp3': ('mp3', 'libmp3lame', ()), + 'aac': ('m4a', 'aac', ('-f', 'adts')), + 'm4a': ('m4a', 'aac', ('-bsf:a', 'aac_adtstoasc')), + 'opus': ('opus', 'libopus', ()), + 'vorbis': ('ogg', 'libvorbis', ()), + 'flac': ('flac', 'flac', ()), + 'alac': ('m4a', None, ('-acodec', 'alac')), + 'wav': ('wav', None, ('-f', 'wav')), +} + + +def create_mapping_re(supported): + return re.compile(r'{0}(?:/{0})*$'.format(r'(?:\s*\w+\s*>)?\s*(?:%s)\s*' % '|'.join(supported))) + + +def resolve_mapping(source, mapping): + """ + Get corresponding item from a mapping string like 'A>B/C>D/E' + @returns (target, error_message) + """ + for pair in mapping.lower().split('/'): + kv = pair.split('>', 1) + if len(kv) == 1 or kv[0].strip() == source: + target = kv[-1].strip() + if target == source: + return target, f'already is in target format {source}' + return target, None + return None, f'could not find a mapping for {source}' + + +class FFmpegPostProcessorError(PostProcessingError): + pass + + +class FFmpegPostProcessor(PostProcessor): + _ffmpeg_location = contextvars.ContextVar('ffmpeg_location', default=None) + + def __init__(self, downloader=None): + PostProcessor.__init__(self, downloader) + self._prefer_ffmpeg = self.get_param('prefer_ffmpeg', True) + self._paths = self._determine_executables() + + @staticmethod + def get_versions_and_features(downloader=None): + pp = FFmpegPostProcessor(downloader) + return pp._versions, pp._features + + @staticmethod + def get_versions(downloader=None): + return FFmpegPostProcessor.get_versions_and_features(downloader)[0] + + _ffmpeg_to_avconv = {'ffmpeg': 'avconv', 'ffprobe': 'avprobe'} + + def _determine_executables(self): + programs = [*self._ffmpeg_to_avconv.keys(), *self._ffmpeg_to_avconv.values()] + + location = self.get_param('ffmpeg_location', self._ffmpeg_location.get()) + if location is None: + return {p: p for p in programs} + + if not os.path.exists(location): + self.report_warning( + f'ffmpeg-location {location} does not exist! Continuing without ffmpeg', only_once=True) + return {} + elif os.path.isdir(location): + dirname, basename, filename = location, None, None + else: + filename = os.path.basename(location) + basename = next((p for p in programs if p in filename), 'ffmpeg') + dirname = os.path.dirname(os.path.abspath(location)) + if basename in self._ffmpeg_to_avconv.keys(): + self._prefer_ffmpeg = True + + paths = {p: os.path.join(dirname, p) for p in programs} + if basename and basename in filename: + for p in programs: + path = os.path.join(dirname, filename.replace(basename, p)) + if os.path.exists(path): + paths[p] = path + if basename: + paths[basename] = location + return paths + + _version_cache, _features_cache = {None: None}, {} + + def _get_ffmpeg_version(self, prog): + path = self._paths.get(prog) + if path in self._version_cache: + return self._version_cache[path], self._features_cache.get(path, {}) + out = _get_exe_version_output(path, ['-bsfs']) + ver = detect_exe_version(out) if out else False + if ver: + regexs = [ + r'(?:\d+:)?([0-9.]+)-[0-9]+ubuntu[0-9.]+$', # Ubuntu, see [1] + r'n([0-9.]+)$', # Arch Linux + # 1. http://www.ducea.com/2006/06/17/ubuntu-package-version-naming-explanation/ + ] + for regex in regexs: + mobj = re.match(regex, ver) + if mobj: + ver = mobj.group(1) + self._version_cache[path] = ver + if prog != 'ffmpeg' or not out: + return ver, {} + + mobj = re.search(r'(?m)^\s+libavformat\s+(?:[0-9. ]+)\s+/\s+(?P[0-9. ]+)', out) + lavf_runtime_version = mobj.group('runtime').replace(' ', '') if mobj else None + self._features_cache[path] = features = { + 'fdk': '--enable-libfdk-aac' in out, + 'setts': 'setts' in out.splitlines(), + 'needs_adtstoasc': is_outdated_version(lavf_runtime_version, '57.56.100', False), + } + return ver, features + + @property + def _versions(self): + return filter_dict({self.basename: self._version, self.probe_basename: self._probe_version}) + + @functools.cached_property + def basename(self): + self._version # run property + return self.basename + + @functools.cached_property + def probe_basename(self): + self._probe_version # run property + return self.probe_basename + + def _get_version(self, kind): + executables = (kind, ) + if not self._prefer_ffmpeg: + executables = (kind, self._ffmpeg_to_avconv[kind]) + basename, version, features = next(filter( + lambda x: x[1], ((p, *self._get_ffmpeg_version(p)) for p in executables)), (None, None, {})) + if kind == 'ffmpeg': + self.basename, self._features = basename, features + else: + self.probe_basename = basename + if basename == self._ffmpeg_to_avconv[kind]: + self.deprecated_feature(f'Support for {self._ffmpeg_to_avconv[kind]} is deprecated and ' + f'may be removed in a future version. Use {kind} instead') + return version + + @functools.cached_property + def _version(self): + return self._get_version('ffmpeg') + + @functools.cached_property + def _probe_version(self): + return self._get_version('ffprobe') + + @property + def available(self): + return self.basename is not None + + @property + def executable(self): + return self._paths.get(self.basename) + + @property + def probe_available(self): + return self.probe_basename is not None + + @property + def probe_executable(self): + return self._paths.get(self.probe_basename) + + @staticmethod + def stream_copy_opts(copy=True, *, ext=None): + yield from ('-map', '0') + # Don't copy Apple TV chapters track, bin_data + # See https://github.com/yt-dlp/yt-dlp/issues/2, #19042, #19024, https://trac.ffmpeg.org/ticket/6016 + yield from ('-dn', '-ignore_unknown') + if copy: + yield from ('-c', 'copy') + if ext in ('mp4', 'mov', 'm4a'): + yield from ('-c:s', 'mov_text') + + def check_version(self): + if not self.available: + raise FFmpegPostProcessorError('ffmpeg not found. Please install or provide the path using --ffmpeg-location') + + required_version = '10-0' if self.basename == 'avconv' else '1.0' + if is_outdated_version(self._version, required_version): + self.report_warning(f'Your copy of {self.basename} is outdated, update {self.basename} ' + f'to version {required_version} or newer if you encounter any errors') + + def get_audio_codec(self, path): + if not self.probe_available and not self.available: + raise PostProcessingError('ffprobe and ffmpeg not found. Please install or provide the path using --ffmpeg-location') + try: + if self.probe_available: + cmd = [ + encodeFilename(self.probe_executable, True), + encodeArgument('-show_streams')] + else: + cmd = [ + encodeFilename(self.executable, True), + encodeArgument('-i')] + cmd.append(encodeFilename(self._ffmpeg_filename_argument(path), True)) + self.write_debug(f'{self.basename} command line: {shell_quote(cmd)}') + stdout, stderr, returncode = Popen.run( + cmd, text=True, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE) + if returncode != (0 if self.probe_available else 1): + return None + except OSError: + return None + output = stdout if self.probe_available else stderr + if self.probe_available: + audio_codec = None + for line in output.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 + else: + # Stream #FILE_INDEX:STREAM_INDEX[STREAM_ID](LANGUAGE): CODEC_TYPE: CODEC_NAME + mobj = re.search( + r'Stream\s*#\d+:\d+(?:\[0x[0-9a-f]+\])?(?:\([a-z]{3}\))?:\s*Audio:\s*([0-9a-z]+)', + output) + if mobj: + return mobj.group(1) + return None + + def get_metadata_object(self, path, opts=[]): + if self.probe_basename != 'ffprobe': + if self.probe_available: + self.report_warning('Only ffprobe is supported for metadata extraction') + raise PostProcessingError('ffprobe not found. Please install or provide the path using --ffmpeg-location') + self.check_version() + + cmd = [ + encodeFilename(self.probe_executable, True), + encodeArgument('-hide_banner'), + encodeArgument('-show_format'), + encodeArgument('-show_streams'), + encodeArgument('-print_format'), + encodeArgument('json'), + ] + + cmd += opts + cmd.append(self._ffmpeg_filename_argument(path)) + self.write_debug(f'ffprobe command line: {shell_quote(cmd)}') + stdout, _, _ = Popen.run(cmd, text=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE) + return json.loads(stdout) + + def get_stream_number(self, path, keys, value): + streams = self.get_metadata_object(path)['streams'] + num = next( + (i for i, stream in enumerate(streams) if traverse_obj(stream, keys, casesense=False) == value), + None) + return num, len(streams) + + def _fixup_chapters(self, info): + last_chapter = traverse_obj(info, ('chapters', -1)) + if last_chapter and not last_chapter.get('end_time'): + last_chapter['end_time'] = self._get_real_video_duration(info['filepath']) + + def _get_real_video_duration(self, filepath, fatal=True): + try: + duration = float_or_none( + traverse_obj(self.get_metadata_object(filepath), ('format', 'duration'))) + if not duration: + raise PostProcessingError('ffprobe returned empty duration') + return duration + except PostProcessingError as e: + if fatal: + raise PostProcessingError(f'Unable to determine video duration: {e.msg}') + + def _duration_mismatch(self, d1, d2, tolerance=2): + if not d1 or not d2: + return None + # The duration is often only known to nearest second. So there can be <1sec disparity natually. + # Further excuse an additional <1sec difference. + return abs(d1 - d2) > tolerance + + def run_ffmpeg_multiple_files(self, input_paths, out_path, opts, **kwargs): + return self.real_run_ffmpeg( + [(path, []) for path in input_paths], + [(out_path, opts)], **kwargs) + + def real_run_ffmpeg(self, input_path_opts, output_path_opts, *, expected_retcodes=(0,)): + self.check_version() + + oldest_mtime = min( + os.stat(encodeFilename(path)).st_mtime for path, _ in input_path_opts if path) + + cmd = [encodeFilename(self.executable, True), encodeArgument('-y')] + # avconv does not have repeat option + if self.basename == 'ffmpeg': + cmd += [encodeArgument('-loglevel'), encodeArgument('repeat+info')] + + def make_args(file, args, name, number): + keys = ['_%s%d' % (name, number), '_%s' % name] + if name == 'o': + args += ['-movflags', '+faststart'] + if number == 1: + keys.append('') + args += self._configuration_args(self.basename, keys) + if name == 'i': + args.append('-i') + return ( + [encodeArgument(arg) for arg in args] + + [encodeFilename(self._ffmpeg_filename_argument(file), True)]) + + for arg_type, path_opts in (('i', input_path_opts), ('o', output_path_opts)): + cmd += itertools.chain.from_iterable( + make_args(path, list(opts), arg_type, i + 1) + for i, (path, opts) in enumerate(path_opts) if path) + + self.write_debug('ffmpeg command line: %s' % shell_quote(cmd)) + _, stderr, returncode = Popen.run( + cmd, text=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE) + if returncode not in variadic(expected_retcodes): + self.write_debug(stderr) + raise FFmpegPostProcessorError(stderr.strip().splitlines()[-1]) + for out_path, _ in output_path_opts: + if out_path: + self.try_utime(out_path, oldest_mtime, oldest_mtime) + return stderr + + def run_ffmpeg(self, path, out_path, opts, **kwargs): + return self.run_ffmpeg_multiple_files([path], out_path, opts, **kwargs) + + @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) + # Also leave '-' intact in order not to break streaming to stdout. + if fn.startswith(('http://', 'https://')): + 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, [ + *self.stream_copy_opts(False, ext=determine_ext(filename)), + '-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, 'w', encoding='utf-8') as f: + f.writelines(self._concat_spec(in_files, concat_opts)) + + out_flags = list(self.stream_copy_opts(ext=determine_ext(out_file))) + + self.real_run_ffmpeg( + [(concat_file, ['-hide_banner', '-nostdin', '-f', 'concat', '-safe', '0'])], + [(out_file, out_flags)]) + self._delete_downloaded_files(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 = MEDIA_EXTENSIONS.common_audio + ('wma', ) + SUPPORTED_EXTS = tuple(ACODECS.keys()) + FORMAT_RE = create_mapping_re(('best', *SUPPORTED_EXTS)) + + def __init__(self, downloader=None, preferredcodec=None, preferredquality=None, nopostoverwrites=False): + FFmpegPostProcessor.__init__(self, downloader) + self.mapping = preferredcodec or 'best' + self._preferredquality = float_or_none(preferredquality) + self._nopostoverwrites = nopostoverwrites + + def _quality_args(self, codec): + if self._preferredquality is None: + return [] + elif self._preferredquality > 10: + return ['-b:a', f'{self._preferredquality}k'] + + limits = { + 'libmp3lame': (10, 0), + 'libvorbis': (0, 10), + # FFmpeg's AAC encoder does not have an upper limit for the value of -q:a. + # Experimentally, with values over 4, bitrate changes were minimal or non-existent + 'aac': (0.1, 4), + 'libfdk_aac': (1, 5), + }.get(codec) + if not limits: + return [] + + q = limits[1] + (limits[0] - limits[1]) * (self._preferredquality / 10) + if codec == 'libfdk_aac': + return ['-vbr', f'{int(q)}'] + return ['-q:a', f'{q}'] + + 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 PostProcessingError(f'audio conversion failed: {err.msg}') + + @PostProcessor._restrict_to(images=False) + def run(self, information): + orig_path = path = information['filepath'] + target_format, _skip_msg = resolve_mapping(information['ext'], self.mapping) + if target_format == 'best' and information['ext'] in self.COMMON_AUDIO_EXTS: + target_format, _skip_msg = None, 'the file is already in a common audio format' + if not target_format: + self.to_screen(f'Not converting audio {orig_path}; {_skip_msg}') + return [], information + + filecodec = self.get_audio_codec(path) + if filecodec is None: + raise PostProcessingError('WARNING: unable to obtain file audio codec with ffprobe') + + if filecodec == 'aac' and target_format in ('m4a', 'best'): + # Lossless, but in another container + extension, _, more_opts, acodec = *ACODECS['m4a'], 'copy' + elif target_format == 'best' or target_format == filecodec: + # Lossless if possible + try: + extension, _, more_opts, acodec = *ACODECS[filecodec], 'copy' + except KeyError: + extension, acodec, more_opts = ACODECS['mp3'] + else: + # We convert the audio (lossy if codec is lossy) + extension, acodec, more_opts = ACODECS[target_format] + if acodec == 'aac' and self._features.get('fdk'): + acodec, more_opts = 'libfdk_aac', [] + + more_opts = list(more_opts) + if acodec != 'copy': + more_opts = self._quality_args(acodec) + + temp_path = new_path = replace_extension(path, extension, information['ext']) + + if new_path == path: + if acodec == 'copy': + self.to_screen(f'Not converting audio {orig_path}; file is already in target format {target_format}') + return [], information + orig_path = prepend_extension(path, 'orig') + temp_path = prepend_extension(path, 'temp') + if (self._nopostoverwrites and os.path.exists(encodeFilename(new_path)) + and os.path.exists(encodeFilename(orig_path))): + self.to_screen('Post-process file %s exists, skipping' % new_path) + return [], information + + self.to_screen(f'Destination: {new_path}') + self.run_ffmpeg(path, temp_path, acodec, more_opts) + + os.replace(path, orig_path) + os.replace(temp_path, new_path) + information['filepath'] = new_path + information['ext'] = extension + + # 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') + + return [orig_path], information + + +class FFmpegVideoConvertorPP(FFmpegPostProcessor): + SUPPORTED_EXTS = ( + *sorted((*MEDIA_EXTENSIONS.common_video, 'gif')), + *sorted((*MEDIA_EXTENSIONS.common_audio, 'aac', 'vorbis')), + ) + FORMAT_RE = create_mapping_re(SUPPORTED_EXTS) + _ACTION = 'converting' + + def __init__(self, downloader=None, preferedformat=None): + super().__init__(downloader) + self.mapping = preferedformat + + @staticmethod + def _options(target_ext): + yield from FFmpegPostProcessor.stream_copy_opts(False) + if target_ext == 'avi': + yield from ('-c:v', 'libxvid', '-vtag', 'XVID') + + @PostProcessor._restrict_to(images=False) + def run(self, info): + filename, source_ext = info['filepath'], info['ext'].lower() + target_ext, _skip_msg = resolve_mapping(source_ext, self.mapping) + if _skip_msg: + self.to_screen(f'Not {self._ACTION} media file "{filename}"; {_skip_msg}') + return [], info + + 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)) + + info['filepath'] = outpath + info['format'] = info['ext'] = target_ext + return [filename], info + + +class FFmpegVideoRemuxerPP(FFmpegVideoConvertorPP): + _ACTION = 'remuxing' + + @staticmethod + def _options(target_ext): + return FFmpegPostProcessor.stream_copy_opts() + + +class FFmpegEmbedSubtitlePP(FFmpegPostProcessor): + SUPPORTED_EXTS = ('mp4', 'mov', 'm4a', 'webm', 'mkv', 'mka') + + def __init__(self, downloader=None, already_have_subtitle=False): + super().__init__(downloader) + self._already_have_subtitle = already_have_subtitle + + @PostProcessor._restrict_to(images=False) + def run(self, info): + if info['ext'] not in self.SUPPORTED_EXTS: + self.to_screen(f'Subtitles can only be embedded in {", ".join(self.SUPPORTED_EXTS)} files') + return [], info + subtitles = info.get('requested_subtitles') + if not subtitles: + self.to_screen('There aren\'t any subtitles to embed') + return [], info + + filename = info['filepath'] + + # Disabled temporarily. There needs to be a way to override this + # in case of duration actually mismatching in extractor + # See: https://github.com/yt-dlp/yt-dlp/issues/1870, https://github.com/yt-dlp/yt-dlp/issues/1385 + ''' + if info.get('duration') and not info.get('__real_download') and self._duration_mismatch( + self._get_real_video_duration(filename, False), info['duration']): + self.to_screen(f'Skipping {self.pp_key()} since the real and expected durations mismatch') + return [], info + ''' + + ext = info['ext'] + sub_langs, sub_names, sub_filenames = [], [], [] + webm_vtt_warn = False + mp4_ass_warn = False + + for lang, sub_info in subtitles.items(): + if not os.path.exists(sub_info.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') + elif ext != 'webm' or ext == 'webm' and sub_ext == 'vtt': + sub_langs.append(lang) + sub_names.append(sub_info.get('name')) + sub_filenames.append(sub_info['filepath']) + else: + if not webm_vtt_warn and ext == 'webm' and sub_ext != 'vtt': + webm_vtt_warn = True + self.report_warning('Only WebVTT subtitles can be embedded in webm files') + if not mp4_ass_warn and ext == 'mp4' and sub_ext == 'ass': + mp4_ass_warn = True + self.report_warning('ASS subtitles cannot be properly embedded in mp4 files; expect issues') + + if not sub_langs: + return [], info + + input_files = [filename] + sub_filenames + + opts = [ + *self.stream_copy_opts(ext=info['ext']), + # Don't copy the existing subtitles, we may be running the + # postprocessor a second time + '-map', '-0:s', + ] + for i, (lang, name) in enumerate(zip(sub_langs, sub_names)): + opts.extend(['-map', '%d:0' % (i + 1)]) + lang_code = ISO639Utils.short2long(lang) or lang + opts.extend(['-metadata:s:s:%d' % i, 'language=%s' % lang_code]) + if name: + opts.extend(['-metadata:s:s:%d' % i, 'handler_name=%s' % name, + '-metadata:s:s:%d' % i, 'title=%s' % name]) + + 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.replace(temp_filename, filename) + + files_to_delete = [] if self._already_have_subtitle else sub_filenames + return files_to_delete, info + + +class FFmpegMetadataPP(FFmpegPostProcessor): + + def __init__(self, downloader, add_metadata=True, add_chapters=True, add_infojson='if_exists'): + FFmpegPostProcessor.__init__(self, downloader) + self._add_metadata = add_metadata + self._add_chapters = add_chapters + self._add_infojson = add_infojson + + @staticmethod + def _options(target_ext): + audio_only = target_ext == 'm4a' + yield from FFmpegPostProcessor.stream_copy_opts(not audio_only) + if audio_only: + yield from ('-vn', '-acodec', 'copy') + + @PostProcessor._restrict_to(images=False) + def run(self, info): + self._fixup_chapters(info) + filename, metadata_filename = info['filepath'], None + files_to_delete, 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)) + files_to_delete.append(metadata_filename) + if self._add_metadata: + options.extend(self._get_metadata_opts(info)) + + if self._add_infojson: + if info['ext'] in ('mkv', 'mka'): + infojson_filename = info.get('infojson_filename') + options.extend(self._get_infojson_opts(info, infojson_filename)) + if not infojson_filename: + files_to_delete.append(info.get('infojson_filename')) + elif self._add_infojson is True: + self.to_screen('The info-json can only be attached to mkv/mka files') + + 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)) + self._delete_downloaded_files(*files_to_delete) + os.replace(temp_filename, filename) + return [], info + + @staticmethod + def _get_chapter_opts(chapters, metadata_filename): + with open(metadata_filename, 'w', 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): + meta_prefix = 'meta' + metadata = collections.defaultdict(dict) + + def add(meta_list, info_list=None): + value = next(( + info[key] for key in [f'{meta_prefix}_'] + list(variadic(info_list or meta_list)) + if info.get(key) is not None), None) + if value not in ('', None): + value = ', '.join(map(str, variadic(value))) + value = value.replace('\0', '') # nul character cannot be passed in command line + metadata['common'].update({meta_f: value for meta_f in variadic(meta_list)}) + + # Info on media metadata/metadata supported by ffmpeg: + # https://wiki.multimedia.cx/index.php/FFmpeg_Metadata + # https://kdenlive.org/en/project/adding-meta-data-to-mp4-video/ + # https://kodi.wiki/view/Video_file_tagging + + add('title', ('track', 'title')) + add('date', 'upload_date') + add(('description', 'synopsis'), 'description') + add(('purl', 'comment'), 'webpage_url') + add('track', 'track_number') + add('artist', ('artist', 'artists', 'creator', 'creators', 'uploader', 'uploader_id')) + add('composer', ('composer', 'composers')) + add('genre', ('genre', 'genres')) + add('album') + add('album_artist', ('album_artist', 'album_artists')) + add('disc', 'disc_number') + add('show', 'series') + add('season_number') + add('episode_id', ('episode', 'episode_id')) + add('episode_sort', 'episode_number') + if 'embed-metadata' in self.get_param('compat_opts', []): + add('comment', 'description') + metadata['common'].pop('synopsis', None) + + meta_regex = rf'{re.escape(meta_prefix)}(?P\d+)?_(?P.+)' + for key, value in info.items(): + mobj = re.fullmatch(meta_regex, key) + if value is not None and mobj: + metadata[mobj.group('i') or 'common'][mobj.group('key')] = value.replace('\0', '') + + # Write id3v1 metadata also since Windows Explorer can't handle id3v2 tags + yield ('-write_id3v1', '1') + + for name, value in metadata['common'].items(): + yield ('-metadata', f'{name}={value}') + + stream_idx = 0 + for fmt in info.get('requested_formats') or [info]: + stream_count = 2 if 'none' not in (fmt.get('vcodec'), fmt.get('acodec')) else 1 + lang = ISO639Utils.short2long(fmt.get('language') or '') or fmt.get('language') + for i in range(stream_idx, stream_idx + stream_count): + if lang: + metadata[str(i)].setdefault('language', lang) + for name, value in metadata[str(i)].items(): + yield (f'-metadata:s:{i}', f'{name}={value}') + stream_idx += stream_count + + def _get_infojson_opts(self, info, infofn): + if not infofn or not os.path.exists(infofn): + if self._add_infojson is not True: + return + infofn = infofn or '%s.temp' % ( + self._downloader.prepare_filename(info, 'infojson') + or replace_extension(self._downloader.prepare_filename(info), 'info.json', info['ext'])) + if not self._downloader._ensure_dir_exists(infofn): + return + self.write_debug(f'Writing info-json to: {infofn}') + write_json_file(self._downloader.sanitize_info(info, self.get_param('clean_infojson', True)), infofn) + info['infojson_filename'] = infofn + + old_stream, new_stream = self.get_stream_number(info['filepath'], ('tags', 'mimetype'), 'application/json') + if old_stream is not None: + yield ('-map', '-0:%d' % old_stream) + new_stream -= 1 + + yield ( + '-attach', self._ffmpeg_filename_argument(infofn), + f'-metadata:s:{new_stream}', 'mimetype=application/json', + f'-metadata:s:{new_stream}', 'filename=info.json', + ) + + +class FFmpegMergerPP(FFmpegPostProcessor): + SUPPORTED_EXTS = MEDIA_EXTENSIONS.common_video + + @PostProcessor._restrict_to(images=False) + def run(self, info): + 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', f'{i}:a:0']) + aac_fixup = fmt['protocol'].startswith('m3u8') and self.get_audio_codec(fmt['filepath']) == 'aac' + if aac_fixup: + 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.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, ' + 'yt-dlp will download single file media. ' + 'Update %s to version %s or newer to fix this.') % ( + self.basename, self.basename, required_version) + self.report_warning(warning) + return False + return True + + +class FFmpegFixupPostProcessor(FFmpegPostProcessor): + def _fixup(self, msg, filename, options): + temp_filename = prepend_extension(filename, 'temp') + + self.to_screen(f'{msg} of "{filename}"') + self.run_ffmpeg(filename, temp_filename, options) + + os.replace(temp_filename, filename) + + +class FFmpegFixupStretchedPP(FFmpegFixupPostProcessor): + @PostProcessor._restrict_to(images=False, audio=False) + def run(self, info): + stretched_ratio = info.get('stretched_ratio') + if stretched_ratio not in (None, 1): + self._fixup('Fixing aspect ratio', info['filepath'], [ + *self.stream_copy_opts(), '-aspect', '%f' % stretched_ratio]) + return [], info + + +class FFmpegFixupM4aPP(FFmpegFixupPostProcessor): + @PostProcessor._restrict_to(images=False, video=False) + def run(self, info): + if info.get('container') == 'm4a_dash': + self._fixup('Correcting container', info['filepath'], [*self.stream_copy_opts(), '-f', 'mp4']) + return [], info + + +class FFmpegFixupM3u8PP(FFmpegFixupPostProcessor): + def _needs_fixup(self, info): + yield info['ext'] in ('mp4', 'm4a') + yield info['protocol'].startswith('m3u8') + try: + metadata = self.get_metadata_object(info['filepath']) + except PostProcessingError as e: + self.report_warning(f'Unable to extract metadata: {e.msg}') + yield True + else: + yield traverse_obj(metadata, ('format', 'format_name'), casesense=False) == 'mpegts' + + @PostProcessor._restrict_to(images=False) + def run(self, info): + if all(self._needs_fixup(info)): + args = ['-f', 'mp4'] + if self.get_audio_codec(info['filepath']) == 'aac': + args.extend(['-bsf:a', 'aac_adtstoasc']) + self._fixup('Fixing MPEG-TS in MP4 container', info['filepath'], [ + *self.stream_copy_opts(), *args]) + return [], info + + +class FFmpegFixupTimestampPP(FFmpegFixupPostProcessor): + + def __init__(self, downloader=None, trim=0.001): + # "trim" should be used when the video contains unintended packets + super().__init__(downloader) + assert isinstance(trim, (int, float)) + self.trim = str(trim) + + @PostProcessor._restrict_to(images=False) + def run(self, info): + if not self._features.get('setts'): + self.report_warning( + 'A re-encode is needed to fix timestamps in older versions of ffmpeg. ' + 'Please install ffmpeg 4.4 or later to fixup without re-encoding') + opts = ['-vf', 'setpts=PTS-STARTPTS'] + else: + opts = ['-c', 'copy', '-bsf', 'setts=ts=TS-STARTPTS'] + self._fixup('Fixing frame timestamp', info['filepath'], opts + [*self.stream_copy_opts(False), '-ss', self.trim]) + return [], info + + +class FFmpegCopyStreamPP(FFmpegFixupPostProcessor): + MESSAGE = 'Copying stream' + + @PostProcessor._restrict_to(images=False) + def run(self, info): + self._fixup(self.MESSAGE, info['filepath'], self.stream_copy_opts()) + return [], info + + +class FFmpegFixupDurationPP(FFmpegCopyStreamPP): + MESSAGE = 'Fixing video duration' + + +class FFmpegFixupDuplicateMoovPP(FFmpegCopyStreamPP): + MESSAGE = 'Fixing duplicate MOOV atoms' + + +class FFmpegSubtitlesConvertorPP(FFmpegPostProcessor): + SUPPORTED_EXTS = MEDIA_EXTENSIONS.subtitles + + def __init__(self, downloader=None, format=None): + super().__init__(downloader) + self.format = format + + def run(self, info): + subs = info.get('requested_subtitles') + new_ext = self.format + new_format = new_ext + if new_format == 'vtt': + new_format = 'webvtt' + if subs is None: + self.to_screen('There aren\'t any subtitles to convert') + return [], info + self.to_screen('Converting subtitles') + sub_filenames = [] + for lang, sub in subs.items(): + if not os.path.exists(sub.get('filepath', '')): + self.report_warning(f'Skipping embedding {lang} subtitle because the file is missing') + continue + ext = sub['ext'] + if ext == new_ext: + self.to_screen('Subtitle file for %s is already in the requested format' % new_ext) + continue + elif ext == 'json': + self.to_screen( + 'You have requested to convert json subtitles into another format, ' + 'which is currently not possible') + continue + old_file = sub['filepath'] + sub_filenames.append(old_file) + new_file = replace_extension(old_file, new_ext) + + if ext in ('dfxp', 'ttml', 'tt'): + self.report_warning( + 'You have requested to convert dfxp (TTML) subtitles into another format, ' + 'which results in style information loss') + + dfxp_file = old_file + srt_file = replace_extension(old_file, 'srt') + + with open(dfxp_file, 'rb') as f: + srt_data = dfxp2srt(f.read()) + + with open(srt_file, 'w', encoding='utf-8') as f: + f.write(srt_data) + old_file = srt_file + + subs[lang] = { + 'ext': 'srt', + 'data': srt_data, + 'filepath': srt_file, + } + + if new_ext == 'srt': + continue + else: + sub_filenames.append(srt_file) + + self.run_ffmpeg(old_file, new_file, ['-f', new_format]) + + with open(new_file, encoding='utf-8') as f: + subs[lang] = { + 'ext': new_ext, + 'data': f.read(), + 'filepath': new_file, + } + + info['__files_to_move'][new_file] = replace_extension( + 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() + info.update({ + 'section_number': number, + 'section_title': chapter.get('title'), + 'section_start': chapter.get('start_time'), + 'section_end': chapter.get('end_time'), + }) + return self._downloader.prepare_filename(info, 'chapter') + + def _ffmpeg_args_for_chapter(self, number, chapter, info): + destination = self._prepare_filename(number, chapter, info) + if not self._downloader._ensure_dir_exists(encodeFilename(destination)): + return + + chapter['filepath'] = destination + self.to_screen('Chapter %03d; Destination: %s' % (number, destination)) + return ( + destination, + ['-ss', str(chapter['start_time']), + '-t', str(chapter['end_time'] - chapter['start_time'])]) + + @PostProcessor._restrict_to(images=False) + def run(self, info): + self._fixup_chapters(info) + chapters = info.get('chapters') or [] + if not chapters: + 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([(in_file, opts)], [(destination, self.stream_copy_opts())]) + if in_file != info['filepath']: + self._delete_downloaded_files(in_file, msg=None) + return [], info + + +class FFmpegThumbnailsConvertorPP(FFmpegPostProcessor): + SUPPORTED_EXTS = MEDIA_EXTENSIONS.thumbnails + FORMAT_RE = create_mapping_re(SUPPORTED_EXTS) + + def __init__(self, downloader=None, format=None): + super().__init__(downloader) + self.mapping = format + + @classmethod + def is_webp(cls, path): + deprecation_warning(f'{cls.__module__}.{cls.__name__}.is_webp is deprecated') + return imghdr.what(path) == 'webp' + + def fixup_webp(self, info, idx=-1): + thumbnail_filename = info['thumbnails'][idx]['filepath'] + _, thumbnail_ext = os.path.splitext(thumbnail_filename) + if thumbnail_ext: + if thumbnail_ext.lower() != '.webp' and imghdr.what(thumbnail_filename) == 'webp': + self.to_screen('Correcting thumbnail "%s" extension to webp' % thumbnail_filename) + webp_filename = replace_extension(thumbnail_filename, 'webp') + 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') + + @staticmethod + def _options(target_ext): + yield from ('-update', '1') + if target_ext == 'jpg': + yield from ('-bsf:v', 'mjpeg2jpeg') + + def convert_thumbnail(self, thumbnail_filename, target_ext): + thumbnail_conv_filename = replace_extension(thumbnail_filename, target_ext) + + self.to_screen(f'Converting thumbnail "{thumbnail_filename}" to {target_ext}') + _, source_ext = os.path.splitext(thumbnail_filename) + self.real_run_ffmpeg( + [(thumbnail_filename, [] if source_ext == '.gif' else ['-f', 'image2', '-pattern_type', 'none'])], + [(thumbnail_conv_filename, self._options(target_ext))]) + return thumbnail_conv_filename + + def run(self, info): + files_to_delete = [] + has_thumbnail = False + + for idx, thumbnail_dict in enumerate(info.get('thumbnails') or []): + original_thumbnail = thumbnail_dict.get('filepath') + if not original_thumbnail: + continue + has_thumbnail = True + self.fixup_webp(info, idx) + original_thumbnail = thumbnail_dict['filepath'] # Path can change during fixup + thumbnail_ext = os.path.splitext(original_thumbnail)[1][1:].lower() + if thumbnail_ext == 'jpeg': + thumbnail_ext = 'jpg' + target_ext, _skip_msg = resolve_mapping(thumbnail_ext, self.mapping) + if _skip_msg: + self.to_screen(f'Not converting thumbnail "{original_thumbnail}"; {_skip_msg}') + continue + thumbnail_dict['filepath'] = self.convert_thumbnail(original_thumbnail, target_ext) + files_to_delete.append(original_thumbnail) + info['__files_to_move'][thumbnail_dict['filepath']] = replace_extension( + info['__files_to_move'][original_thumbnail], target_ext) + + if not has_thumbnail: + self.to_screen('There aren\'t any thumbnails to convert') + return files_to_delete, info + + +class FFmpegConcatPP(FFmpegPostProcessor): + def __init__(self, downloader, only_multi_video=False): + self._only_multi_video = only_multi_video + super().__init__(downloader) + + def _get_codecs(self, file): + codecs = traverse_obj(self.get_metadata_object(file), ('streams', ..., 'codec_name')) + self.write_debug(f'Codecs = {", ".join(codecs)}') + return tuple(codecs) + + def concat_files(self, in_files, out_file): + if not self._downloader._ensure_dir_exists(out_file): + return + if len(in_files) == 1: + if os.path.realpath(in_files[0]) != os.path.realpath(out_file): + self.to_screen(f'Moving "{in_files[0]}" to "{out_file}"') + os.replace(in_files[0], out_file) + return [] + + if len(set(map(self._get_codecs, in_files))) > 1: + raise PostProcessingError( + 'The files have different streams/codecs and cannot be concatenated. ' + 'Either select different formats or --recode-video them to a common format') + + self.to_screen(f'Concatenating {len(in_files)} files; Destination: {out_file}') + super().concat_files(in_files, out_file) + return in_files + + @PostProcessor._restrict_to(images=False, simulated=False) + def run(self, info): + entries = info.get('entries') or [] + if not any(entries) or (self._only_multi_video and info['_type'] != 'multi_video'): + return [], info + elif traverse_obj(entries, (..., lambda k, v: k == 'requested_downloads' and len(v) > 1)): + raise PostProcessingError('Concatenation is not supported when downloading multiple separate formats') + + in_files = traverse_obj(entries, (..., 'requested_downloads', 0, 'filepath')) or [] + if len(in_files) < len(entries): + raise PostProcessingError('Aborting concatenation because some downloads failed') + + exts = traverse_obj(entries, (..., 'requested_downloads', 0, 'ext'), (..., 'ext')) + ie_copy = collections.ChainMap({'ext': exts[0] if len(set(exts)) == 1 else 'mkv'}, + info, self._downloader._playlist_infodict(info)) + out_file = self._downloader.prepare_filename(ie_copy, 'pl_video') + + files_to_delete = self.concat_files(in_files, out_file) + + info['requested_downloads'] = [{ + 'filepath': out_file, + 'ext': ie_copy['ext'], + }] + return files_to_delete, info diff --git a/yt_dlp/postprocessor/metadataparser.py b/yt_dlp/postprocessor/metadataparser.py new file mode 100644 index 0000000..1d60542 --- /dev/null +++ b/yt_dlp/postprocessor/metadataparser.py @@ -0,0 +1,125 @@ +import re + +from .common import PostProcessor +from ..utils import Namespace, filter_dict, function_with_repr + + +class MetadataParserPP(PostProcessor): + def __init__(self, downloader, actions): + super().__init__(downloader) + self._actions = [] + for f in actions: + action, *args = f + assert action in self.Actions + self._actions.append(action(self, *args)) + + @classmethod + def validate_action(cls, action, *data): + """Each action can be: + (Actions.INTERPRET, from, to) OR + (Actions.REPLACE, field, search, replace) + """ + if action not in cls.Actions: + raise ValueError(f'{action!r} is not a valid action') + action(cls, *data) # So this can raise error to validate + + @staticmethod + def field_to_template(tmpl): + if re.match(r'[a-zA-Z_]+$', tmpl): + return f'%({tmpl})s' + + from ..YoutubeDL import YoutubeDL + err = YoutubeDL.validate_outtmpl(tmpl) + if err: + raise err + return tmpl + + @staticmethod + def format_to_regex(fmt): + r""" + Converts a string like + '%(title)s - %(artist)s' + to a regex like + '(?P.+)\ \-\ (?P<artist>.+)' + """ + if not re.search(r'%\(\w+\)s', fmt): + return fmt + lastpos = 0 + regex = '' + # replace %(..)s with regex group and escape other string parts + for match in re.finditer(r'%\((\w+)\)s', fmt): + regex += re.escape(fmt[lastpos:match.start()]) + regex += rf'(?P<{match.group(1)}>.+)' + lastpos = match.end() + if lastpos < len(fmt): + regex += re.escape(fmt[lastpos:]) + return regex + + def run(self, info): + for f in self._actions: + f(info) + return [], info + + @function_with_repr + def interpretter(self, inp, out): + def f(info): + data_to_parse = self._downloader.evaluate_outtmpl(template, info) + self.write_debug(f'Searching for {out_re.pattern!r} in {template!r}') + match = out_re.search(data_to_parse) + if match is None: + self.to_screen(f'Could not interpret {inp!r} as {out!r}') + return + for attribute, value in filter_dict(match.groupdict()).items(): + info[attribute] = value + self.to_screen(f'Parsed {attribute} from {template!r}: {value!r}') + + template = self.field_to_template(inp) + out_re = re.compile(self.format_to_regex(out)) + return f + + @function_with_repr + def replacer(self, field, search, replace): + def f(info): + val = info.get(field) + if val is None: + self.to_screen(f'Video does not have a {field}') + return + elif not isinstance(val, str): + self.report_warning(f'Cannot replace in field {field} since it is a {type(val).__name__}') + return + self.write_debug(f'Replacing all {search!r} in {field} with {replace!r}') + info[field], n = search_re.subn(replace, val) + if n: + self.to_screen(f'Changed {field} to: {info[field]}') + else: + self.to_screen(f'Did not find {search!r} in {field}') + + search_re = re.compile(search) + return f + + Actions = Namespace(INTERPRET=interpretter, REPLACE=replacer) + + +class MetadataFromFieldPP(MetadataParserPP): + @classmethod + def to_action(cls, f): + match = re.match(r'(?s)(?P<in>.*?)(?<!\\):(?P<out>.+)$', f) + if match is None: + raise ValueError(f'it should be FROM:TO, not {f!r}') + return ( + cls.Actions.INTERPRET, + match.group('in').replace('\\:', ':'), + match.group('out'), + ) + + def __init__(self, downloader, formats): + super().__init__(downloader, [self.to_action(f) for f in formats]) + + +# Deprecated +class MetadataFromTitlePP(MetadataParserPP): + def __init__(self, downloader, titleformat): + super().__init__(downloader, [(self.Actions.INTERPRET, 'title', titleformat)]) + self.deprecation_warning( + 'yt_dlp.postprocessor.MetadataFromTitlePP is deprecated ' + 'and may be removed in a future version. Use yt_dlp.postprocessor.MetadataFromFieldPP instead') diff --git a/yt_dlp/postprocessor/modify_chapters.py b/yt_dlp/postprocessor/modify_chapters.py new file mode 100644 index 0000000..f521986 --- /dev/null +++ b/yt_dlp/postprocessor/modify_chapters.py @@ -0,0 +1,336 @@ +import copy +import heapq +import os + +from .common import PostProcessor +from .ffmpeg import FFmpegPostProcessor, FFmpegSubtitlesConvertorPP +from .sponsorblock import SponsorBlockPP +from ..utils import PostProcessingError, orderedSet, prepend_extension + +_TINY_CHAPTER_DURATION = 1 +DEFAULT_SPONSORBLOCK_CHAPTER_TITLE = '[SponsorBlock]: %(category_names)l' + + +class ModifyChaptersPP(FFmpegPostProcessor): + def __init__(self, downloader, remove_chapters_patterns=None, remove_sponsor_segments=None, remove_ranges=None, + *, sponsorblock_chapter_title=DEFAULT_SPONSORBLOCK_CHAPTER_TITLE, force_keyframes=False): + FFmpegPostProcessor.__init__(self, downloader) + self._remove_chapters_patterns = set(remove_chapters_patterns or []) + self._remove_sponsor_segments = set(remove_sponsor_segments or []) - set(SponsorBlockPP.NON_SKIPPABLE_CATEGORIES.keys()) + self._ranges_to_remove = set(remove_ranges or []) + self._sponsorblock_chapter_title = sponsorblock_chapter_title + self._force_keyframes = force_keyframes + + @PostProcessor._restrict_to(images=False) + def run(self, info): + self._fixup_chapters(info) + # Chapters must be preserved intact when downloading multiple formats of the same video. + chapters, sponsor_chapters = self._mark_chapters_to_remove( + copy.deepcopy(info.get('chapters')) or [], + copy.deepcopy(info.get('sponsorblock_chapters')) or []) + if not chapters and not sponsor_chapters: + return [], info + + real_duration = self._get_real_video_duration(info['filepath']) + if not chapters: + chapters = [{'start_time': 0, 'end_time': info.get('duration') or real_duration, 'title': info['title']}] + + info['chapters'], cuts = self._remove_marked_arrange_sponsors(chapters + sponsor_chapters) + if not cuts: + return [], info + elif not info['chapters']: + self.report_warning('You have requested to remove the entire video, which is not possible') + return [], info + + original_duration, info['duration'] = info.get('duration'), info['chapters'][-1]['end_time'] + if self._duration_mismatch(real_duration, original_duration, 1): + if not self._duration_mismatch(real_duration, info['duration']): + self.to_screen(f'Skipping {self.pp_key()} since the video appears to be already cut') + return [], info + if not info.get('__real_download'): + raise PostProcessingError('Cannot cut video since the real and expected durations mismatch. ' + 'Different chapters may have already been removed') + else: + self.write_debug('Expected and actual durations mismatch') + + concat_opts = self._make_concat_opts(cuts, real_duration) + self.write_debug('Concat spec = %s' % ', '.join(f'{c.get("inpoint", 0.0)}-{c.get("outpoint", "inf")}' for c in concat_opts)) + + def remove_chapters(file, is_sub): + return file, self.remove_chapters(file, cuts, concat_opts, self._force_keyframes and not is_sub) + + in_out_files = [remove_chapters(info['filepath'], False)] + in_out_files.extend(remove_chapters(in_file, True) for in_file in self._get_supported_subs(info)) + + # Renaming should only happen after all files are processed + files_to_remove = [] + for in_file, out_file in in_out_files: + mtime = os.stat(in_file).st_mtime + uncut_file = prepend_extension(in_file, 'uncut') + os.replace(in_file, uncut_file) + os.replace(out_file, in_file) + self.try_utime(in_file, mtime, mtime) + files_to_remove.append(uncut_file) + + return files_to_remove, info + + def _mark_chapters_to_remove(self, chapters, sponsor_chapters): + if self._remove_chapters_patterns: + warn_no_chapter_to_remove = True + if not chapters: + self.to_screen('Chapter information is unavailable') + warn_no_chapter_to_remove = False + for c in chapters: + if any(regex.search(c['title']) for regex in self._remove_chapters_patterns): + c['remove'] = True + warn_no_chapter_to_remove = False + if warn_no_chapter_to_remove: + self.to_screen('There are no chapters matching the regex') + + if self._remove_sponsor_segments: + warn_no_chapter_to_remove = True + if not sponsor_chapters: + self.to_screen('SponsorBlock information is unavailable') + warn_no_chapter_to_remove = False + for c in sponsor_chapters: + if c['category'] in self._remove_sponsor_segments: + c['remove'] = True + warn_no_chapter_to_remove = False + if warn_no_chapter_to_remove: + self.to_screen('There are no matching SponsorBlock chapters') + + sponsor_chapters.extend({ + 'start_time': start, + 'end_time': end, + 'category': 'manually_removed', + '_categories': [('manually_removed', start, end, 'Manually removed')], + 'remove': True, + } for start, end in self._ranges_to_remove) + + return chapters, sponsor_chapters + + def _get_supported_subs(self, info): + for sub in (info.get('requested_subtitles') or {}).values(): + sub_file = sub.get('filepath') + # The file might have been removed by --embed-subs + if not sub_file or not os.path.exists(sub_file): + continue + ext = sub['ext'] + if ext not in FFmpegSubtitlesConvertorPP.SUPPORTED_EXTS: + self.report_warning(f'Cannot remove chapters from external {ext} subtitles; "{sub_file}" is now out of sync') + continue + # TODO: create __real_download for subs? + yield sub_file + + def _remove_marked_arrange_sponsors(self, chapters): + # Store cuts separately, since adjacent and overlapping cuts must be merged. + cuts = [] + + def append_cut(c): + assert 'remove' in c, 'Not a cut is appended to cuts' + last_to_cut = cuts[-1] if cuts else None + if last_to_cut and last_to_cut['end_time'] >= c['start_time']: + last_to_cut['end_time'] = max(last_to_cut['end_time'], c['end_time']) + else: + cuts.append(c) + return len(cuts) - 1 + + def excess_duration(c): + # Cuts that are completely within the chapter reduce chapters' duration. + # Since cuts can overlap, excess duration may be less that the sum of cuts' durations. + # To avoid that, chapter stores the index to the fist cut within the chapter, + # instead of storing excess duration. append_cut ensures that subsequent cuts (if any) + # will be merged with previous ones (if necessary). + cut_idx, excess = c.pop('cut_idx', len(cuts)), 0 + while cut_idx < len(cuts): + cut = cuts[cut_idx] + if cut['start_time'] >= c['end_time']: + break + if cut['end_time'] > c['start_time']: + excess += min(cut['end_time'], c['end_time']) + excess -= max(cut['start_time'], c['start_time']) + cut_idx += 1 + return excess + + new_chapters = [] + + def append_chapter(c): + assert 'remove' not in c, 'Cut is appended to chapters' + length = c['end_time'] - c['start_time'] - excess_duration(c) + # Chapter is completely covered by cuts or sponsors. + if length <= 0: + return + start = new_chapters[-1]['end_time'] if new_chapters else 0 + c.update(start_time=start, end_time=start + length) + new_chapters.append(c) + + # Turn into a priority queue, index is a tie breaker. + # Plain stack sorted by start_time is not enough: after splitting the chapter, + # the part returned to the stack is not guaranteed to have start_time + # less than or equal to the that of the stack's head. + chapters = [(c['start_time'], i, c) for i, c in enumerate(chapters)] + heapq.heapify(chapters) + + _, cur_i, cur_chapter = heapq.heappop(chapters) + while chapters: + _, i, c = heapq.heappop(chapters) + # Non-overlapping chapters or cuts can be appended directly. However, + # adjacent non-overlapping cuts must be merged, which is handled by append_cut. + if cur_chapter['end_time'] <= c['start_time']: + (append_chapter if 'remove' not in cur_chapter else append_cut)(cur_chapter) + cur_i, cur_chapter = i, c + continue + + # Eight possibilities for overlapping chapters: (cut, cut), (cut, sponsor), + # (cut, normal), (sponsor, cut), (normal, cut), (sponsor, sponsor), + # (sponsor, normal), and (normal, sponsor). There is no (normal, normal): + # normal chapters are assumed not to overlap. + if 'remove' in cur_chapter: + # (cut, cut): adjust end_time. + if 'remove' in c: + cur_chapter['end_time'] = max(cur_chapter['end_time'], c['end_time']) + # (cut, sponsor/normal): chop the beginning of the later chapter + # (if it's not completely hidden by the cut). Push to the priority queue + # to restore sorting by start_time: with beginning chopped, c may actually + # start later than the remaining chapters from the queue. + elif cur_chapter['end_time'] < c['end_time']: + c['start_time'] = cur_chapter['end_time'] + c['_was_cut'] = True + heapq.heappush(chapters, (c['start_time'], i, c)) + # (sponsor/normal, cut). + elif 'remove' in c: + cur_chapter['_was_cut'] = True + # Chop the end of the current chapter if the cut is not contained within it. + # Chopping the end doesn't break start_time sorting, no PQ push is necessary. + if cur_chapter['end_time'] <= c['end_time']: + cur_chapter['end_time'] = c['start_time'] + append_chapter(cur_chapter) + cur_i, cur_chapter = i, c + continue + # Current chapter contains the cut within it. If the current chapter is + # a sponsor chapter, check whether the categories before and after the cut differ. + if '_categories' in cur_chapter: + after_c = dict(cur_chapter, start_time=c['end_time'], _categories=[]) + cur_cats = [] + for cat_start_end in cur_chapter['_categories']: + if cat_start_end[1] < c['start_time']: + cur_cats.append(cat_start_end) + if cat_start_end[2] > c['end_time']: + after_c['_categories'].append(cat_start_end) + cur_chapter['_categories'] = cur_cats + if cur_chapter['_categories'] != after_c['_categories']: + # Categories before and after the cut differ: push the after part to PQ. + heapq.heappush(chapters, (after_c['start_time'], cur_i, after_c)) + cur_chapter['end_time'] = c['start_time'] + append_chapter(cur_chapter) + cur_i, cur_chapter = i, c + continue + # Either sponsor categories before and after the cut are the same or + # we're dealing with a normal chapter. Just register an outstanding cut: + # subsequent append_chapter will reduce the duration. + cur_chapter.setdefault('cut_idx', append_cut(c)) + # (sponsor, normal): if a normal chapter is not completely overlapped, + # chop the beginning of it and push it to PQ. + elif '_categories' in cur_chapter and '_categories' not in c: + if cur_chapter['end_time'] < c['end_time']: + c['start_time'] = cur_chapter['end_time'] + c['_was_cut'] = True + heapq.heappush(chapters, (c['start_time'], i, c)) + # (normal, sponsor) and (sponsor, sponsor) + else: + assert '_categories' in c, 'Normal chapters overlap' + cur_chapter['_was_cut'] = True + c['_was_cut'] = True + # Push the part after the sponsor to PQ. + if cur_chapter['end_time'] > c['end_time']: + # deepcopy to make categories in after_c and cur_chapter/c refer to different lists. + after_c = dict(copy.deepcopy(cur_chapter), start_time=c['end_time']) + heapq.heappush(chapters, (after_c['start_time'], cur_i, after_c)) + # Push the part after the overlap to PQ. + elif c['end_time'] > cur_chapter['end_time']: + after_cur = dict(copy.deepcopy(c), start_time=cur_chapter['end_time']) + heapq.heappush(chapters, (after_cur['start_time'], cur_i, after_cur)) + c['end_time'] = cur_chapter['end_time'] + # (sponsor, sponsor): merge categories in the overlap. + if '_categories' in cur_chapter: + c['_categories'] = cur_chapter['_categories'] + c['_categories'] + # Inherit the cuts that the current chapter has accumulated within it. + if 'cut_idx' in cur_chapter: + c['cut_idx'] = cur_chapter['cut_idx'] + cur_chapter['end_time'] = c['start_time'] + append_chapter(cur_chapter) + cur_i, cur_chapter = i, c + (append_chapter if 'remove' not in cur_chapter else append_cut)(cur_chapter) + return self._remove_tiny_rename_sponsors(new_chapters), cuts + + def _remove_tiny_rename_sponsors(self, chapters): + new_chapters = [] + for i, c in enumerate(chapters): + # Merge with the previous/next if the chapter is tiny. + # Only tiny chapters resulting from a cut can be skipped. + # Chapters that were already tiny in the original list will be preserved. + if (('_was_cut' in c or '_categories' in c) + and c['end_time'] - c['start_time'] < _TINY_CHAPTER_DURATION): + if not new_chapters: + # Prepend tiny chapter to the next one if possible. + if i < len(chapters) - 1: + chapters[i + 1]['start_time'] = c['start_time'] + continue + else: + old_c = new_chapters[-1] + if i < len(chapters) - 1: + next_c = chapters[i + 1] + # Not a typo: key names in old_c and next_c are really different. + prev_is_sponsor = 'categories' in old_c + next_is_sponsor = '_categories' in next_c + # Preferentially prepend tiny normals to normals and sponsors to sponsors. + if (('_categories' not in c and prev_is_sponsor and not next_is_sponsor) + or ('_categories' in c and not prev_is_sponsor and next_is_sponsor)): + next_c['start_time'] = c['start_time'] + continue + old_c['end_time'] = c['end_time'] + continue + + c.pop('_was_cut', None) + cats = c.pop('_categories', None) + if cats: + category, _, _, category_name = min(cats, key=lambda c: c[2] - c[1]) + c.update({ + 'category': category, + 'categories': orderedSet(x[0] for x in cats), + 'name': category_name, + 'category_names': orderedSet(x[3] for x in cats), + }) + c['title'] = self._downloader.evaluate_outtmpl(self._sponsorblock_chapter_title, c.copy()) + # Merge identically named sponsors. + if (new_chapters and 'categories' in new_chapters[-1] + and new_chapters[-1]['title'] == c['title']): + new_chapters[-1]['end_time'] = c['end_time'] + continue + new_chapters.append(c) + return new_chapters + + def remove_chapters(self, filename, ranges_to_cut, concat_opts, force_keyframes=False): + in_file = filename + out_file = prepend_extension(in_file, 'temp') + if force_keyframes: + in_file = self.force_keyframes(in_file, (t for c in ranges_to_cut for t in (c['start_time'], c['end_time']))) + self.to_screen(f'Removing chapters from {filename}') + self.concat_files([in_file] * len(concat_opts), out_file, concat_opts) + if in_file != filename: + self._delete_downloaded_files(in_file, msg=None) + return out_file + + @staticmethod + def _make_concat_opts(chapters_to_remove, duration): + opts = [{}] + for s in chapters_to_remove: + # Do not create 0 duration chunk at the beginning. + if s['start_time'] == 0: + opts[-1]['inpoint'] = f'{s["end_time"]:.6f}' + continue + opts[-1]['outpoint'] = f'{s["start_time"]:.6f}' + # Do not create 0 duration chunk at the end. + if s['end_time'] < duration: + opts.append({'inpoint': f'{s["end_time"]:.6f}'}) + return opts diff --git a/yt_dlp/postprocessor/movefilesafterdownload.py b/yt_dlp/postprocessor/movefilesafterdownload.py new file mode 100644 index 0000000..23b0924 --- /dev/null +++ b/yt_dlp/postprocessor/movefilesafterdownload.py @@ -0,0 +1,53 @@ +import os + +from .common import PostProcessor +from ..compat import shutil +from ..utils import ( + PostProcessingError, + decodeFilename, + encodeFilename, + make_dir, +) + + +class MoveFilesAfterDownloadPP(PostProcessor): + + def __init__(self, downloader=None, downloaded=True): + PostProcessor.__init__(self, downloader) + self._downloaded = downloaded + + @classmethod + def pp_key(cls): + return 'MoveFiles' + + def run(self, info): + dl_path, dl_name = os.path.split(encodeFilename(info['filepath'])) + finaldir = info.get('__finaldir', dl_path) + finalpath = os.path.join(finaldir, dl_name) + if self._downloaded: + info['__files_to_move'][info['filepath']] = decodeFilename(finalpath) + + make_newfilename = lambda old: decodeFilename(os.path.join(finaldir, os.path.basename(encodeFilename(old)))) + for oldfile, newfile in info['__files_to_move'].items(): + if not newfile: + newfile = make_newfilename(oldfile) + if os.path.abspath(encodeFilename(oldfile)) == os.path.abspath(encodeFilename(newfile)): + continue + if not os.path.exists(encodeFilename(oldfile)): + self.report_warning('File "%s" cannot be found' % oldfile) + continue + if os.path.exists(encodeFilename(newfile)): + if self.get_param('overwrites', True): + self.report_warning('Replacing existing file "%s"' % newfile) + os.remove(encodeFilename(newfile)) + else: + self.report_warning( + 'Cannot move file "%s" out of temporary directory since "%s" already exists. ' + % (oldfile, newfile)) + continue + make_dir(newfile, PostProcessingError) + self.to_screen(f'Moving file "{oldfile}" to "{newfile}"') + shutil.move(oldfile, newfile) # os.rename cannot move between volumes + + info['filepath'] = finalpath + return [], info diff --git a/yt_dlp/postprocessor/sponskrub.py b/yt_dlp/postprocessor/sponskrub.py new file mode 100644 index 0000000..ff50d5b --- /dev/null +++ b/yt_dlp/postprocessor/sponskrub.py @@ -0,0 +1,98 @@ +import os +import shlex +import subprocess + +from .common import PostProcessor +from ..utils import ( + Popen, + PostProcessingError, + check_executable, + cli_option, + encodeArgument, + encodeFilename, + prepend_extension, + shell_quote, + str_or_none, +) + + +# Deprecated in favor of the native implementation +class SponSkrubPP(PostProcessor): + _temp_ext = 'spons' + _exe_name = 'sponskrub' + + def __init__(self, downloader, path='', args=None, ignoreerror=False, cut=False, force=False, _from_cli=False): + PostProcessor.__init__(self, downloader) + self.force = force + self.cutout = cut + self.args = str_or_none(args) or '' # For backward compatibility + self.path = self.get_exe(path) + + if not _from_cli: + self.deprecation_warning( + 'yt_dlp.postprocessor.SponSkrubPP support is deprecated and may be removed in a future version. ' + 'Use yt_dlp.postprocessor.SponsorBlock and yt_dlp.postprocessor.ModifyChaptersPP instead') + + if not ignoreerror and self.path is None: + if path: + raise PostProcessingError('sponskrub not found in "%s"' % path) + else: + raise PostProcessingError('sponskrub not found. Please install or provide the path using --sponskrub-path') + + def get_exe(self, path=''): + if not path or not check_executable(path, ['-h']): + path = os.path.join(path, self._exe_name) + if not check_executable(path, ['-h']): + return None + return path + + @PostProcessor._restrict_to(images=False) + def run(self, information): + if self.path is None: + return [], information + + filename = information['filepath'] + if not os.path.exists(encodeFilename(filename)): # no download + return [], information + + if information['extractor_key'].lower() != 'youtube': + self.to_screen('Skipping sponskrub since it is not a YouTube video') + return [], information + if self.cutout and not self.force and not information.get('__real_download', False): + self.report_warning( + 'Skipping sponskrub since the video was already downloaded. ' + 'Use --sponskrub-force to run sponskrub anyway') + return [], information + + self.to_screen('Trying to %s sponsor sections' % ('remove' if self.cutout else 'mark')) + if self.cutout: + self.report_warning('Cutting out sponsor segments will cause the subtitles to go out of sync.') + if not information.get('__real_download', False): + self.report_warning('If sponskrub is run multiple times, unintended parts of the video could be cut out.') + + temp_filename = prepend_extension(filename, self._temp_ext) + if os.path.exists(encodeFilename(temp_filename)): + os.remove(encodeFilename(temp_filename)) + + cmd = [self.path] + if not self.cutout: + cmd += ['-chapter'] + cmd += cli_option(self._downloader.params, '-proxy', 'proxy') + cmd += shlex.split(self.args) # For backward compatibility + cmd += self._configuration_args(self._exe_name, use_compat=False) + cmd += ['--', information['id'], filename, temp_filename] + cmd = [encodeArgument(i) for i in cmd] + + self.write_debug('sponskrub command line: %s' % shell_quote(cmd)) + stdout, _, returncode = Popen.run(cmd, text=True, stdout=None if self.get_param('verbose') else subprocess.PIPE) + + if not returncode: + os.replace(temp_filename, filename) + self.to_screen('Sponsor sections have been %s' % ('removed' if self.cutout else 'marked')) + elif returncode == 3: + self.to_screen('No segments in the SponsorBlock database') + else: + raise PostProcessingError( + stdout.strip().splitlines()[0 if stdout.strip().lower().startswith('unrecognised') else -1] + or f'sponskrub failed with error code {returncode}') + return [], information diff --git a/yt_dlp/postprocessor/sponsorblock.py b/yt_dlp/postprocessor/sponsorblock.py new file mode 100644 index 0000000..6ba87cd --- /dev/null +++ b/yt_dlp/postprocessor/sponsorblock.py @@ -0,0 +1,104 @@ +import hashlib +import json +import re +import urllib.parse + +from .ffmpeg import FFmpegPostProcessor + + +class SponsorBlockPP(FFmpegPostProcessor): + # https://wiki.sponsor.ajay.app/w/Types + EXTRACTORS = { + 'Youtube': 'YouTube', + } + POI_CATEGORIES = { + 'poi_highlight': 'Highlight', + } + NON_SKIPPABLE_CATEGORIES = { + **POI_CATEGORIES, + 'chapter': 'Chapter', + } + CATEGORIES = { + 'sponsor': 'Sponsor', + 'intro': 'Intermission/Intro Animation', + 'outro': 'Endcards/Credits', + 'selfpromo': 'Unpaid/Self Promotion', + 'preview': 'Preview/Recap', + 'filler': 'Filler Tangent', + 'interaction': 'Interaction Reminder', + 'music_offtopic': 'Non-Music Section', + **NON_SKIPPABLE_CATEGORIES + } + + def __init__(self, downloader, categories=None, api='https://sponsor.ajay.app'): + FFmpegPostProcessor.__init__(self, downloader) + self._categories = tuple(categories or self.CATEGORIES.keys()) + self._API_URL = api if re.match('^https?://', api) else 'https://' + api + + def run(self, info): + extractor = info['extractor_key'] + if extractor not in self.EXTRACTORS: + self.to_screen(f'SponsorBlock is not supported for {extractor}') + return [], info + + self.to_screen('Fetching SponsorBlock segments') + info['sponsorblock_chapters'] = self._get_sponsor_chapters(info, info.get('duration')) + return [], info + + def _get_sponsor_chapters(self, info, duration): + segments = self._get_sponsor_segments(info['id'], self.EXTRACTORS[info['extractor_key']]) + + def duration_filter(s): + start_end = s['segment'] + # Ignore entire video segments (https://wiki.sponsor.ajay.app/w/Types). + if start_end == (0, 0): + return False + # Ignore milliseconds difference at the start. + if start_end[0] <= 1: + start_end[0] = 0 + # Make POI chapters 1 sec so that we can properly mark them + if s['category'] in self.POI_CATEGORIES.keys(): + start_end[1] += 1 + # Ignore milliseconds difference at the end. + # Never allow the segment to exceed the video. + if duration and duration - start_end[1] <= 1: + start_end[1] = duration + # SponsorBlock duration may be absent or it may deviate from the real one. + diff = abs(duration - s['videoDuration']) if s['videoDuration'] else 0 + return diff < 1 or (diff < 5 and diff / (start_end[1] - start_end[0]) < 0.05) + + duration_match = [s for s in segments if duration_filter(s)] + if len(duration_match) != len(segments): + self.report_warning('Some SponsorBlock segments are from a video of different duration, maybe from an old version of this video') + + def to_chapter(s): + (start, end), cat = s['segment'], s['category'] + title = s['description'] if cat == 'chapter' else self.CATEGORIES[cat] + return { + 'start_time': start, + 'end_time': end, + 'category': cat, + 'title': title, + 'type': s['actionType'], + '_categories': [(cat, start, end, title)], + } + + sponsor_chapters = [to_chapter(s) for s in duration_match] + if not sponsor_chapters: + self.to_screen('No matching segments were found in the SponsorBlock database') + else: + self.to_screen(f'Found {len(sponsor_chapters)} segments in the SponsorBlock database') + return sponsor_chapters + + def _get_sponsor_segments(self, video_id, service): + hash = hashlib.sha256(video_id.encode('ascii')).hexdigest() + # SponsorBlock API recommends using first 4 hash characters. + url = f'{self._API_URL}/api/skipSegments/{hash[:4]}?' + urllib.parse.urlencode({ + 'service': service, + 'categories': json.dumps(self._categories), + 'actionTypes': json.dumps(['skip', 'poi', 'chapter']) + }) + for d in self._download_json(url) or []: + if d['videoID'] == video_id: + return d['segments'] + return [] diff --git a/yt_dlp/postprocessor/xattrpp.py b/yt_dlp/postprocessor/xattrpp.py new file mode 100644 index 0000000..f822eff --- /dev/null +++ b/yt_dlp/postprocessor/xattrpp.py @@ -0,0 +1,63 @@ +import os + +from .common import PostProcessor +from ..compat import compat_os_name +from ..utils import ( + PostProcessingError, + XAttrMetadataError, + XAttrUnavailableError, + hyphenate_date, + write_xattr, +) + + +class XAttrMetadataPP(PostProcessor): + """Set extended attributes on downloaded file (if xattr support is found) + + More info about extended attributes for media: + http://freedesktop.org/wiki/CommonExtendedAttributes/ + http://www.freedesktop.org/wiki/PhreedomDraft/ + http://dublincore.org/documents/usageguide/elements.shtml + + TODO: + * capture youtube keywords and put them in 'user.dublincore.subject' (comma-separated) + * figure out which xattrs can be used for 'duration', 'thumbnail', 'resolution' + """ + + XATTR_MAPPING = { + 'user.xdg.referrer.url': 'webpage_url', + # 'user.xdg.comment': 'description', + 'user.dublincore.title': 'title', + 'user.dublincore.date': 'upload_date', + 'user.dublincore.description': 'description', + 'user.dublincore.contributor': 'uploader', + 'user.dublincore.format': 'format', + } + + def run(self, info): + mtime = os.stat(info['filepath']).st_mtime + self.to_screen('Writing metadata to file\'s xattrs') + try: + for xattrname, infoname in self.XATTR_MAPPING.items(): + value = info.get(infoname) + if value: + if infoname == 'upload_date': + value = hyphenate_date(value) + write_xattr(info['filepath'], xattrname, value.encode()) + + except XAttrUnavailableError as e: + raise PostProcessingError(str(e)) + except XAttrMetadataError as e: + if e.reason == 'NO_SPACE': + self.report_warning( + 'There\'s no disk space left, disk quota exceeded or filesystem xattr limit exceeded. ' + 'Some extended attributes are not written') + elif e.reason == 'VALUE_TOO_LONG': + self.report_warning('Unable to write extended attributes due to too long values.') + else: + tip = ('You need to use NTFS' if compat_os_name == 'nt' + else 'You may have to enable them in your "/etc/fstab"') + raise PostProcessingError(f'This filesystem doesn\'t support extended attributes. {tip}') + + self.try_utime(info['filepath'], mtime, mtime) + return [], info -- cgit v1.2.3