]> jfr.im git - yt-dlp.git/blobdiff - yt_dlp/postprocessor/ffmpeg.py
[XAttrMetadata] Refactor and document dependencies
[yt-dlp.git] / yt_dlp / postprocessor / ffmpeg.py
index 1bde170ce62fd14b7596879abefffda4618dd43b..d1d8e1687c2c422430849d80c70d1692d85259f5 100644 (file)
@@ -1,28 +1,26 @@
-from __future__ import unicode_literals
-
-import io
+import collections
 import itertools
+import json
 import os
+import re
 import subprocess
 import time
-import re
-import json
 
 from .common import AudioConversionError, PostProcessor
-
 from ..compat import compat_str
 from ..utils import (
+    ISO639Utils,
+    Popen,
+    PostProcessingError,
+    _get_exe_version_output,
+    detect_exe_version,
+    determine_ext,
     dfxp2srt,
     encodeArgument,
     encodeFilename,
     float_or_none,
-    _get_exe_version_output,
-    detect_exe_version,
     is_outdated_version,
-    ISO639Utils,
     orderedSet,
-    Popen,
-    PostProcessingError,
     prepend_extension,
     replace_extension,
     shell_quote,
@@ -31,7 +29,6 @@
     write_json_file,
 )
 
-
 EXT_TO_OUT_FORMATS = {
     'aac': 'adts',
     'flac': 'flac',
@@ -71,11 +68,9 @@ def check_version(self):
             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._versions[self.basename], required_version):
-            warning = 'Your copy of %s is outdated, update %s to version %s or newer if you encounter any errors.' % (
-                self.basename, self.basename, required_version)
-            self.report_warning(warning)
+        if is_outdated_version(self._versions[self.basename], 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+            self.report_warning(f'Your copy of {self.basename} is outdated, update {self.basename} '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')
 
     @staticmethod
     def get_versions_and_features(downloader=None):
@@ -84,13 +79,18 @@ def get_versions_and_features(downloader=None):
 
     @staticmethod
     def get_versions(downloader=None):
-        return FFmpegPostProcessor.get_version_and_features(downloader)[0]
+        return FFmpegPostProcessor.get_versions_and_features(downloader)[0]
+
+    _version_cache, _features_cache = {}, {}
 
     def _determine_executables(self):
         programs = ['avprobe', 'avconv', 'ffmpeg', 'ffprobe']
 
         def get_ffmpeg_version(path, prog):
-            out = _get_exe_version_output(path, ['-bsfs'])
+            if path in self._version_cache:
+                self._versions[prog], self._features = self._version_cache[path], self._features_cache.get(path, {})
+                return
+            out = _get_exe_version_output(path, ['-bsfs'], to_screen=self.write_debug)
             ver = detect_exe_version(out) if out else False
             if ver:
                 regexs = [
@@ -102,13 +102,13 @@ def get_ffmpeg_version(path, prog):
                     mobj = re.match(regex, ver)
                     if mobj:
                         ver = mobj.group(1)
-            self._versions[prog] = ver
+            self._versions[prog] = self._version_cache[path] = ver
             if prog != 'ffmpeg' or not out:
                 return
 
             mobj = re.search(r'(?m)^\s+libavformat\s+(?:[0-9. ]+)\s+/\s+(?P<runtime>[0-9. ]+)', out)
             lavf_runtime_version = mobj.group('runtime').replace(' ', '') if mobj else None
-            self._features = {
+            self._features = self._features_cache[path] = {
                 'fdk': '--enable-libfdk-aac' in out,
                 'setts': 'setts' in out.splitlines(),
                 'needs_adtstoasc': is_outdated_version(lavf_runtime_version, '57.56.100', False),
@@ -140,32 +140,29 @@ def get_ffmpeg_version(path, prog):
                 if basename in ('ffmpeg', 'ffprobe'):
                     prefer_ffmpeg = True
 
-            self._paths = dict(
-                (p, os.path.join(dirname, p)) for p in programs)
+            self._paths = {
+                p: os.path.join(dirname, p) for p in programs}
             if basename:
                 self._paths[basename] = location
 
         self._versions = {}
-        for p in programs:
-            get_ffmpeg_version(self._paths[p], p)
-
-        if prefer_ffmpeg is False:
-            prefs = ('avconv', 'ffmpeg')
-        else:
-            prefs = ('ffmpeg', 'avconv')
-        for p in prefs:
-            if self._versions[p]:
-                self.basename = p
-                break
-
+        # NB: probe must be first for _features to be poulated correctly
+        executables = {'probe_basename': ('ffprobe', 'avprobe'), 'basename': ('ffmpeg', 'avconv')}
         if prefer_ffmpeg is False:
-            prefs = ('avprobe', 'ffprobe')
-        else:
-            prefs = ('ffprobe', 'avprobe')
-        for p in prefs:
-            if self._versions[p]:
-                self.probe_basename = p
-                break
+            executables = {k: v[::-1] for k, v in executables.items()}
+        for var, prefs in executables.items():
+            for p in prefs:
+                get_ffmpeg_version(self._paths[p], p)
+                if self._versions[p]:
+                    setattr(self, var, p)
+                    break
+
+        if self.basename == 'avconv':
+            self.deprecation_warning(
+                'Support for avconv is deprecated and may be removed in a future version. Use ffmpeg instead')
+        if self.probe_basename == 'avprobe':
+            self.deprecation_warning(
+                'Support for avprobe is deprecated and may be removed in a future version. Use ffprobe instead')
 
     @property
     def available(self):
@@ -183,6 +180,17 @@ def probe_available(self):
     def probe_executable(self):
         return self._paths[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 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')
@@ -196,13 +204,13 @@ def get_audio_codec(self, path):
                     encodeFilename(self.executable, True),
                     encodeArgument('-i')]
             cmd.append(encodeFilename(self._ffmpeg_filename_argument(path), True))
-            self.write_debug('%s command line: %s' % (self.basename, shell_quote(cmd)))
+            self.write_debug(f'{self.basename} command line: {shell_quote(cmd)}')
             handle = Popen(cmd, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
             stdout_data, stderr_data = handle.communicate_or_kill()
             expected_ret = 0 if self.probe_available else 1
             if handle.wait() != expected_ret:
                 return None
-        except (IOError, OSError):
+        except OSError:
             return None
         output = (stdout_data if self.probe_available else stderr_data).decode('ascii', 'ignore')
         if self.probe_available:
@@ -251,22 +259,23 @@ def get_stream_number(self, path, keys, value):
             None)
         return num, len(streams)
 
-    def _get_real_video_duration(self, info, fatal=True):
+    def _get_real_video_duration(self, filepath, fatal=True):
         try:
-            if '_real_duration' not in info:
-                info['_real_duration'] = float_or_none(
-                    traverse_obj(self.get_metadata_object(info['filepath']), ('format', 'duration')))
-            if not info['_real_duration']:
+            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}')
-        return info.setdefault('_real_duration', None)
+                raise PostProcessingError(f'Unable to determine video duration: {e.msg}')
 
     def _duration_mismatch(self, d1, d2):
         if not d1 or not d2:
             return None
-        return abs(d1 - d2) > 1
+        # 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) > 2
 
     def run_ffmpeg_multiple_files(self, input_paths, out_path, opts, **kwargs):
         return self.real_run_ffmpeg(
@@ -286,8 +295,10 @@ def real_run_ffmpeg(self, input_path_opts, output_path_opts, *, expected_retcode
 
         def make_args(file, args, name, number):
             keys = ['_%s%d' % (name, number), '_%s' % name]
-            if name == 'o' and number == 1:
-                keys.append('')
+            if name == 'o':
+                args += ['-movflags', '+faststart']
+                if number == 1:
+                    keys.append('')
             args += self._configuration_args(self.basename, keys)
             if name == 'i':
                 args.append('-i')
@@ -341,8 +352,9 @@ def force_keyframes(self, filename, timestamps):
             timestamps = timestamps[1:]
         keyframe_file = prepend_extension(filename, 'keyframes.temp')
         self.to_screen(f'Re-encoding "{filename}" with appropriate keyframes')
-        self.run_ffmpeg(filename, keyframe_file, ['-force_key_frames', ','.join(
-            f'{t:.6f}' for t in timestamps)])
+        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):
@@ -357,17 +369,12 @@ def concat_files(self, in_files, out_file, concat_opts=None):
         with open(concat_file, 'wt', encoding='utf-8') as f:
             f.writelines(self._concat_spec(in_files, concat_opts))
 
-        out_flags = ['-c', 'copy']
-        if out_file.rpartition('.')[-1] in ('mp4', 'mov'):
-            # For some reason, '-c copy' is not enough to copy subtitles
-            out_flags.extend(['-c:s', 'mov_text', '-movflags', '+faststart'])
+        out_flags = list(self.stream_copy_opts(ext=determine_ext(out_file)))
 
-        try:
-            self.real_run_ffmpeg(
-                [(concat_file, ['-hide_banner', '-nostdin', '-f', 'concat', '-safe', '0'])],
-                [(out_file, out_flags)])
-        finally:
-            os.remove(concat_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):
@@ -384,7 +391,7 @@ def _concat_spec(cls, in_files, concat_opts=None):
 
 class FFmpegExtractAudioPP(FFmpegPostProcessor):
     COMMON_AUDIO_EXTS = ('wav', 'flac', 'm4a', 'aiff', 'mp3', 'ogg', 'mka', 'opus', 'wma')
-    SUPPORTED_EXTS = ('best', 'aac', 'flac', 'mp3', 'm4a', 'opus', 'vorbis', 'wav', 'alac')
+    SUPPORTED_EXTS = ('aac', 'flac', 'mp3', 'm4a', 'opus', 'vorbis', 'wav', 'alac')
 
     def __init__(self, downloader=None, preferredcodec=None, preferredquality=None, nopostoverwrites=False):
         FFmpegPostProcessor.__init__(self, downloader)
@@ -486,6 +493,9 @@ def run(self, information):
         temp_path = new_path = prefix + sep + extension
 
         if new_path == path:
+            if acodec == 'copy':
+                self.to_screen(f'File is already in target format {self._preferredcodec}, skipping')
+                return [], information
             orig_path = prepend_extension(path, 'orig')
             temp_path = prepend_extension(path, 'temp')
         if (self._nopostoverwrites and os.path.exists(encodeFilename(new_path))
@@ -517,12 +527,12 @@ def run(self, information):
 
 
 class FFmpegVideoConvertorPP(FFmpegPostProcessor):
-    SUPPORTED_EXTS = ('mp4', 'mkv', 'flv', 'webm', 'mov', 'avi', 'mp3', 'mka', 'm4a', 'ogg', 'opus')
+    SUPPORTED_EXTS = ('mp4', 'mkv', 'flv', 'webm', 'mov', 'avi', 'mka', 'ogg', *FFmpegExtractAudioPP.SUPPORTED_EXTS)
     FORMAT_RE = re.compile(r'{0}(?:/{0})*$'.format(r'(?:\w+>)?(?:%s)' % '|'.join(SUPPORTED_EXTS)))
     _ACTION = 'converting'
 
     def __init__(self, downloader=None, preferedformat=None):
-        super(FFmpegVideoConvertorPP, self).__init__(downloader)
+        super().__init__(downloader)
         self._preferedformats = preferedformat.lower().split('/')
 
     def _target_ext(self, source_ext):
@@ -533,9 +543,9 @@ def _target_ext(self, source_ext):
 
     @staticmethod
     def _options(target_ext):
+        yield from FFmpegPostProcessor.stream_copy_opts(False)
         if target_ext == 'avi':
-            return ['-c:v', 'libxvid', '-vtag', 'XVID']
-        return []
+            yield from ('-c:v', 'libxvid', '-vtag', 'XVID')
 
     @PostProcessor._restrict_to(images=False)
     def run(self, info):
@@ -546,7 +556,7 @@ def run(self, info):
             else f'already is in target format {source_ext}' if source_ext == target_ext
             else None)
         if _skip_msg:
-            self.to_screen(f'Not {self._ACTION} media file {filename!r}; {_skip_msg}')
+            self.to_screen(f'Not {self._ACTION} media file "{filename}"; {_skip_msg}')
             return [], info
 
         outpath = replace_extension(filename, target_ext, source_ext)
@@ -563,34 +573,39 @@ class FFmpegVideoRemuxerPP(FFmpegVideoConvertorPP):
 
     @staticmethod
     def _options(target_ext):
-        options = ['-c', 'copy', '-map', '0', '-dn']
-        if target_ext in ['mp4', 'm4a', 'mov']:
-            options.extend(['-movflags', '+faststart'])
-        return options
+        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(FFmpegEmbedSubtitlePP, self).__init__(downloader)
+        super().__init__(downloader)
         self._already_have_subtitle = already_have_subtitle
 
     @PostProcessor._restrict_to(images=False)
-    def run(self, information):
-        if information['ext'] not in ('mp4', 'webm', 'mkv'):
-            self.to_screen('Subtitles can only be embedded in mp4, webm or mkv files')
-            return [], information
-        subtitles = information.get('requested_subtitles')
+    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 [], information
+            return [], info
 
-        filename = information['filepath']
-        if information.get('duration') and self._duration_mismatch(
-                self._get_real_video_duration(information, False), information['duration']):
+        filename = info['filepath']
+
+        # Disabled temporarily. There needs to be a way to overide 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 [], information
+            return [], info
+        '''
 
-        ext = information['ext']
+        ext = info['ext']
         sub_langs, sub_names, sub_filenames = [], [], []
         webm_vtt_warn = False
         mp4_ass_warn = False
@@ -615,21 +630,16 @@ def run(self, information):
                 self.report_warning('ASS subtitles cannot be properly embedded in mp4 files; expect issues')
 
         if not sub_langs:
-            return [], information
+            return [], info
 
         input_files = [filename] + sub_filenames
 
         opts = [
-            '-c', 'copy', '-map', '0', '-dn',
+            *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',
-            # Don't copy Apple TV chapters track, bin_data (see #19042, #19024,
-            # https://trac.ffmpeg.org/ticket/6016)
-            '-map', '-0:d',
         ]
-        if information['ext'] == 'mp4':
-            opts += ['-c:s', 'mov_text']
         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
@@ -644,7 +654,7 @@ def run(self, information):
         os.replace(temp_filename, filename)
 
         files_to_delete = [] if self._already_have_subtitle else sub_filenames
-        return files_to_delete, information
+        return files_to_delete, info
 
 
 class FFmpegMetadataPP(FFmpegPostProcessor):
@@ -657,11 +667,10 @@ def __init__(self, downloader, add_metadata=True, add_chapters=True, add_infojso
 
     @staticmethod
     def _options(target_ext):
-        yield from ('-map', '0', '-dn')
-        if target_ext == 'm4a':
+        audio_only = target_ext == 'm4a'
+        yield from FFmpegPostProcessor.stream_copy_opts(not audio_only)
+        if audio_only:
             yield from ('-vn', '-acodec', 'copy')
-        else:
-            yield from ('-c', 'copy')
 
     @PostProcessor._restrict_to(images=False)
     def run(self, info):
@@ -692,14 +701,13 @@ def run(self, info):
         self.run_ffmpeg_multiple_files(
             (filename, metadata_filename), temp_filename,
             itertools.chain(self._options(info['ext']), *options))
-        for file in filter(None, files_to_delete):
-            os.remove(file)  # Don't obey --keep-files
+        self._delete_downloaded_files(*files_to_delete)
         os.replace(temp_filename, filename)
         return [], info
 
     @staticmethod
     def _get_chapter_opts(chapters, metadata_filename):
-        with io.open(metadata_filename, 'wt', encoding='utf-8') as f:
+        with open(metadata_filename, 'wt', encoding='utf-8') as f:
             def ffmpeg_escape(text):
                 return re.sub(r'([\\=;#\n])', r'\\\1', text)
 
@@ -715,15 +723,16 @@ def ffmpeg_escape(text):
         yield ('-map_metadata', '1')
 
     def _get_metadata_opts(self, info):
-        metadata = {}
-        meta_prefix = 'meta_'
+        meta_prefix = 'meta'
+        metadata = collections.defaultdict(dict)
 
         def add(meta_list, info_list=None):
             value = next((
-                str(info[key]) for key in [meta_prefix] + list(variadic(info_list or meta_list))
+                str(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):
-                metadata.update({meta_f: value for meta_f in variadic(meta_list)})
+                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)})
 
         # See [1-4] for some info on media metadata/metadata supported
         # by ffmpeg.
@@ -747,22 +756,29 @@ def add(meta_list, info_list=None):
         add('episode_sort', 'episode_number')
         if 'embed-metadata' in self.get_param('compat_opts', []):
             add('comment', 'description')
-            metadata.pop('synopsis', None)
+            metadata['common'].pop('synopsis', None)
 
+        meta_regex = rf'{re.escape(meta_prefix)}(?P<i>\d+)?_(?P<key>.+)'
         for key, value in info.items():
-            if value is not None and key != meta_prefix and key.startswith(meta_prefix):
-                metadata[key[len(meta_prefix):]] = value
+            mobj = re.fullmatch(meta_regex, key)
+            if value is not None and mobj:
+                metadata[mobj.group('i') or 'common'][mobj.group('key')] = value
+
+        # Write id3v1 metadata also since Windows Explorer can't handle id3v2 tags
+        yield ('-write_id3v1', '1')
 
-        for name, value in metadata.items():
+        for name, value in metadata['common'].items():
             yield ('-metadata', f'{name}={value}')
 
         stream_idx = 0
         for fmt in info.get('requested_formats') or []:
             stream_count = 2 if 'none' not in (fmt.get('vcodec'), fmt.get('acodec')) else 1
-            if fmt.get('language'):
-                lang = ISO639Utils.short2long(fmt['language']) or fmt['language']
-                for i in range(stream_count):
-                    yield ('-metadata:s:%d' % (stream_idx + i), 'language=%s' % lang)
+            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):
@@ -783,8 +799,11 @@ def _get_infojson_opts(self, info, infofn):
             yield ('-map', '-0:%d' % old_stream)
             new_stream -= 1
 
-        yield ('-attach', infofn,
-               '-metadata:s:%d' % new_stream, 'mimetype=application/json')
+        yield (
+            '-attach', infofn,
+            f'-metadata:s:{new_stream}', 'mimetype=application/json',
+            f'-metadata:s:{new_stream}', 'filename=info.json',
+        )
 
 
 class FFmpegMergerPP(FFmpegPostProcessor):
@@ -841,7 +860,7 @@ def run(self, info):
         stretched_ratio = info.get('stretched_ratio')
         if stretched_ratio not in (None, 1):
             self._fixup('Fixing aspect ratio', info['filepath'], [
-                '-c', 'copy', '-map', '0', '-dn', '-aspect', '%f' % stretched_ratio])
+                *self.stream_copy_opts(), '-aspect', '%f' % stretched_ratio])
         return [], info
 
 
@@ -849,17 +868,27 @@ 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'], [
-                '-c', 'copy', '-map', '0', '-dn', '-f', 'mp4'])
+            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 self.get_audio_codec(info['filepath']) == 'aac':
-            self._fixup('Fixing malformed AAC bitstream', info['filepath'], [
-                '-c', 'copy', '-map', '0', '-dn', '-f', 'mp4', '-bsf:a', 'aac_adtstoasc'])
+        if all(self._needs_fixup(info)):
+            self._fixup('Fixing MPEG-TS in MP4 container', info['filepath'], [
+                *self.stream_copy_opts(), '-f', 'mp4', '-bsf:a', 'aac_adtstoasc'])
         return [], info
 
 
@@ -867,7 +896,7 @@ class FFmpegFixupTimestampPP(FFmpegFixupPostProcessor):
 
     def __init__(self, downloader=None, trim=0.001):
         # "trim" should be used when the video contains unintended packets
-        super(FFmpegFixupTimestampPP, self).__init__(downloader)
+        super().__init__(downloader)
         assert isinstance(trim, (int, float))
         self.trim = str(trim)
 
@@ -880,22 +909,32 @@ def run(self, info):
             opts = ['-vf', 'setpts=PTS-STARTPTS']
         else:
             opts = ['-c', 'copy', '-bsf', 'setts=ts=TS-STARTPTS']
-        self._fixup('Fixing frame timestamp', info['filepath'], opts + ['-map', '0', '-dn', '-ss', self.trim])
+        self._fixup('Fixing frame timestamp', info['filepath'], opts + [*self.stream_copy_opts(False), '-ss', self.trim])
         return [], info
 
 
-class FFmpegFixupDurationPP(FFmpegFixupPostProcessor):
+class FFmpegCopyStreamPP(FFmpegFixupPostProcessor):
+    MESSAGE = 'Copying stream'
+
     @PostProcessor._restrict_to(images=False)
     def run(self, info):
-        self._fixup('Fixing video duration', info['filepath'], ['-c', 'copy', '-map', '0', '-dn'])
+        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 = ('srt', 'vtt', 'ass', 'lrc')
 
     def __init__(self, downloader=None, format=None):
-        super(FFmpegSubtitlesConvertorPP, self).__init__(downloader)
+        super().__init__(downloader)
         self.format = format
 
     def run(self, info):
@@ -937,7 +976,7 @@ def run(self, info):
                 with open(dfxp_file, 'rb') as f:
                     srt_data = dfxp2srt(f.read())
 
-                with io.open(srt_file, 'wt', encoding='utf-8') as f:
+                with open(srt_file, 'wt', encoding='utf-8') as f:
                     f.write(srt_data)
                 old_file = srt_file
 
@@ -954,7 +993,7 @@ def run(self, info):
 
             self.run_ffmpeg(old_file, new_file, ['-f', new_format])
 
-            with io.open(new_file, 'rt', encoding='utf-8') as f:
+            with open(new_file, encoding='utf-8') as f:
                 subs[lang] = {
                     'ext': new_ext,
                     'data': f.read(),
@@ -1007,17 +1046,17 @@ def run(self, info):
         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, ['-c', 'copy'])])
+            self.real_run_ffmpeg([(in_file, opts)], [(destination, self.stream_copy_opts())])
         if in_file != info['filepath']:
-            os.remove(in_file)
+            self._delete_downloaded_files(in_file, msg=None)
         return [], info
 
 
 class FFmpegThumbnailsConvertorPP(FFmpegPostProcessor):
-    SUPPORTED_EXTS = ('jpg', 'png')
+    SUPPORTED_EXTS = ('jpg', 'png', 'webp')
 
     def __init__(self, downloader=None, format=None):
-        super(FFmpegThumbnailsConvertorPP, self).__init__(downloader)
+        super().__init__(downloader)
         self.format = format
 
     @staticmethod
@@ -1048,7 +1087,7 @@ def _options(target_ext):
     def convert_thumbnail(self, thumbnail_filename, target_ext):
         thumbnail_conv_filename = replace_extension(thumbnail_filename, target_ext)
 
-        self.to_screen('Converting thumbnail "%s" to %s' % (thumbnail_filename, target_ext))
+        self.to_screen(f'Converting thumbnail "{thumbnail_filename}" to {target_ext}')
         self.real_run_ffmpeg(
             [(thumbnail_filename, ['-f', 'image2', '-pattern_type', 'none'])],
             [(thumbnail_conv_filename.replace('%', '%%'), self._options(target_ext))])
@@ -1058,12 +1097,12 @@ def run(self, info):
         files_to_delete = []
         has_thumbnail = False
 
-        for idx, thumbnail_dict in enumerate(info['thumbnails']):
-            if 'filepath' not in thumbnail_dict:
+        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']
             _, thumbnail_ext = os.path.splitext(original_thumbnail)
             if thumbnail_ext:
                 thumbnail_ext = thumbnail_ext[1:].lower()
@@ -1080,3 +1119,57 @@ def run(self, info):
         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')
+
+        ie_copy = self._downloader._playlist_infodict(info)
+        exts = traverse_obj(entries, (..., 'requested_downloads', 0, 'ext'), (..., 'ext'))
+        ie_copy['ext'] = exts[0] if len(set(exts)) == 1 else 'mkv'
+        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