X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/e76f531201cd41dfc0ce00be28bcc5c575c7acc5..a5a28b1eabe0533f26d23c4157c97144215b22b7:/youtube_dl/postprocessor/ffmpeg.py diff --git a/youtube_dl/postprocessor/ffmpeg.py b/youtube_dl/postprocessor/ffmpeg.py index 5ed723b..3ea1afc 100644 --- a/youtube_dl/postprocessor/ffmpeg.py +++ b/youtube_dl/postprocessor/ffmpeg.py @@ -4,6 +4,7 @@ import io import os import subprocess import time +import re from .common import AudioConversionError, PostProcessor @@ -22,9 +23,33 @@ from ..utils import ( subtitles_filename, dfxp2srt, ISO639Utils, + replace_extension, ) +EXT_TO_OUT_FORMATS = { + 'aac': 'adts', + 'flac': 'flac', + 'm4a': 'ipod', + 'mka': 'matroska', + 'mkv': 'matroska', + 'mpg': 'mpeg', + 'ogv': 'ogg', + 'ts': 'mpegts', + 'wma': 'asf', + 'wmv': 'asf', +} +ACODECS = { + 'mp3': 'libmp3lame', + 'aac': 'aac', + 'flac': 'flac', + 'm4a': 'aac', + 'opus': 'libopus', + 'vorbis': 'libvorbis', + 'wav': None, +} + + class FFmpegPostProcessorError(PostProcessingError): pass @@ -52,7 +77,7 @@ class FFmpegPostProcessor(PostProcessor): def _determine_executables(self): programs = ['avprobe', 'avconv', 'ffmpeg', 'ffprobe'] - prefer_ffmpeg = self._downloader.params.get('prefer_ffmpeg', False) + prefer_ffmpeg = False self.basename = None self.probe_basename = None @@ -60,6 +85,7 @@ class FFmpegPostProcessor(PostProcessor): self._paths = None self._versions = None if self._downloader: + prefer_ffmpeg = self._downloader.params.get('prefer_ffmpeg', False) location = self._downloader.params.get('ffmpeg_location') if location is not None: if not os.path.exists(location): @@ -125,6 +151,30 @@ class FFmpegPostProcessor(PostProcessor): def probe_executable(self): return self._paths[self.probe_basename] + def get_audio_codec(self, path): + if not self.probe_available: + raise PostProcessingError('ffprobe or avprobe not found. Please install one.') + try: + cmd = [ + encodeFilename(self.probe_executable, True), + encodeArgument('-show_streams'), + encodeFilename(self._ffmpeg_filename_argument(path), True)] + if self._downloader.params.get('verbose', False): + self._downloader.to_screen('[debug] %s command line: %s' % (self.basename, shell_quote(cmd))) + handle = subprocess.Popen(cmd, stderr=compat_subprocess_get_DEVNULL(), stdout=subprocess.PIPE, stdin=subprocess.PIPE) + output = handle.communicate()[0] + if handle.wait() != 0: + return None + except (IOError, OSError): + return None + audio_codec = None + for line in output.decode('ascii', 'ignore').split('\n'): + if line.startswith('codec_name='): + audio_codec = line.split('=')[1].strip() + elif line.strip() == 'codec_type=audio' and audio_codec is not None: + return audio_codec + return None + def run_ffmpeg_multiple_files(self, input_paths, out_path, opts): self.check_version() @@ -161,7 +211,8 @@ class FFmpegPostProcessor(PostProcessor): # Always use 'file:' because the filename may contain ':' (ffmpeg # interprets that as a protocol) or can start with '-' (-- is broken in # ffmpeg, see https://ffmpeg.org/trac/ffmpeg/ticket/2127 for details) - return 'file:' + fn + # Also leave '-' intact in order not to break streaming to stdout. + return 'file:' + fn if fn != '-' else fn class FFmpegExtractAudioPP(FFmpegPostProcessor): @@ -173,31 +224,6 @@ class FFmpegExtractAudioPP(FFmpegPostProcessor): self._preferredquality = preferredquality self._nopostoverwrites = nopostoverwrites - def get_audio_codec(self, path): - - if not self.probe_available: - raise PostProcessingError('ffprobe or avprobe not found. Please install one.') - try: - cmd = [ - encodeFilename(self.probe_executable, True), - encodeArgument('-show_streams'), - encodeFilename(self._ffmpeg_filename_argument(path), True)] - if self._downloader.params.get('verbose', False): - self._downloader.to_screen('[debug] %s command line: %s' % (self.basename, shell_quote(cmd))) - handle = subprocess.Popen(cmd, stderr=compat_subprocess_get_DEVNULL(), stdout=subprocess.PIPE, stdin=subprocess.PIPE) - output = handle.communicate()[0] - if handle.wait() != 0: - return None - except (IOError, OSError): - return None - audio_codec = None - for line in output.decode('ascii', 'ignore').split('\n'): - if line.startswith('codec_name='): - audio_codec = line.split('=')[1].strip() - elif line.strip() == 'codec_type=audio' and audio_codec is not None: - return audio_codec - return None - def run_ffmpeg(self, path, out_path, codec, more_opts): if codec is None: acodec_opts = [] @@ -223,7 +249,7 @@ class FFmpegExtractAudioPP(FFmpegPostProcessor): acodec = 'copy' extension = 'm4a' more_opts = ['-bsf:a', 'aac_adtstoasc'] - elif filecodec in ['aac', 'mp3', 'vorbis', 'opus']: + elif filecodec in ['aac', 'flac', 'mp3', 'vorbis', 'opus']: # Lossless if possible acodec = 'copy' extension = filecodec @@ -242,8 +268,8 @@ class FFmpegExtractAudioPP(FFmpegPostProcessor): else: more_opts += ['-b:a', self._preferredquality + 'k'] else: - # We convert the audio (lossy) - acodec = {'mp3': 'libmp3lame', 'aac': 'aac', 'm4a': 'aac', 'opus': 'opus', 'vorbis': 'libvorbis', 'wav': None}[self._preferredcodec] + # We convert the audio (lossy if codec is lossy) + acodec = ACODECS[self._preferredcodec] extension = self._preferredcodec more_opts = [] if self._preferredquality is not None: @@ -265,6 +291,9 @@ class FFmpegExtractAudioPP(FFmpegPostProcessor): prefix, sep, ext = path.rpartition('.') # not os.path.splitext, since the latter does not work on unicode in all setups new_path = prefix + sep + extension + information['filepath'] = new_path + information['ext'] = extension + # If we download foo.mp3 and convert it to... foo.mp3, then don't delete foo.mp3, silly. if (new_path == path or (self._nopostoverwrites and os.path.exists(encodeFilename(new_path)))): @@ -286,9 +315,6 @@ class FFmpegExtractAudioPP(FFmpegPostProcessor): new_path, time.time(), information['filetime'], errnote='Cannot update utime of audio file') - information['filepath'] = new_path - information['ext'] = extension - return [path], information @@ -317,17 +343,34 @@ class FFmpegVideoConvertorPP(FFmpegPostProcessor): class FFmpegEmbedSubtitlePP(FFmpegPostProcessor): def run(self, information): - if information['ext'] not in ['mp4', 'mkv']: - self._downloader.to_screen('[ffmpeg] Subtitles can only be embedded in mp4 or mkv files') + if information['ext'] not in ('mp4', 'webm', 'mkv'): + self._downloader.to_screen('[ffmpeg] Subtitles can only be embedded in mp4, webm or mkv files') return [], information subtitles = information.get('requested_subtitles') if not subtitles: self._downloader.to_screen('[ffmpeg] There aren\'t any subtitles to embed') return [], information - sub_langs = list(subtitles.keys()) filename = information['filepath'] - sub_filenames = [subtitles_filename(filename, lang, sub_info['ext']) for lang, sub_info in subtitles.items()] + + ext = information['ext'] + sub_langs = [] + sub_filenames = [] + webm_vtt_warn = False + + for lang, sub_info in subtitles.items(): + sub_ext = sub_info['ext'] + if ext != 'webm' or ext == 'webm' and sub_ext == 'vtt': + sub_langs.append(lang) + sub_filenames.append(subtitles_filename(filename, lang, sub_ext)) + else: + if not webm_vtt_warn and ext == 'webm' and sub_ext != 'vtt': + webm_vtt_warn = True + self._downloader.to_screen('[ffmpeg] Only WebVTT subtitles can be embedded in webm files') + + if not sub_langs: + return [], information + input_files = [filename] + sub_filenames opts = [ @@ -357,23 +400,30 @@ class FFmpegEmbedSubtitlePP(FFmpegPostProcessor): class FFmpegMetadataPP(FFmpegPostProcessor): def run(self, info): metadata = {} - if info.get('title') is not None: - metadata['title'] = info['title'] - if info.get('upload_date') is not None: - metadata['date'] = info['upload_date'] - if info.get('artist') is not None: - metadata['artist'] = info['artist'] - elif info.get('uploader') is not None: - metadata['artist'] = info['uploader'] - elif info.get('uploader_id') is not None: - metadata['artist'] = info['uploader_id'] - if info.get('description') is not None: - metadata['description'] = info['description'] - metadata['comment'] = info['description'] - if info.get('webpage_url') is not None: - metadata['purl'] = info['webpage_url'] - if info.get('album') is not None: - metadata['album'] = info['album'] + + def add(meta_list, info_list=None): + if not info_list: + info_list = meta_list + if not isinstance(meta_list, (list, tuple)): + meta_list = (meta_list,) + if not isinstance(info_list, (list, tuple)): + info_list = (info_list,) + for info_f in info_list: + if info.get(info_f) is not None: + for meta_f in meta_list: + metadata[meta_f] = info[info_f] + break + + add('title', ('track', 'title')) + add('date', 'upload_date') + add(('description', 'comment'), 'description') + add('purl', 'webpage_url') + add('track', 'track_number') + add('artist', ('artist', 'creator', 'uploader', 'uploader_id')) + add('genre') + add('album') + add('album_artist') + add('disc', 'disc_number') if not metadata: self._downloader.to_screen('[ffmpeg] There isn\'t any metadata to add') @@ -381,17 +431,40 @@ class FFmpegMetadataPP(FFmpegPostProcessor): filename = info['filepath'] temp_filename = prepend_extension(filename, 'temp') + in_filenames = [filename] + options = [] if info['ext'] == 'm4a': - options = ['-vn', '-acodec', 'copy'] + options.extend(['-vn', '-acodec', 'copy']) else: - options = ['-c', 'copy'] + options.extend(['-c', 'copy']) for (name, value) in metadata.items(): options.extend(['-metadata', '%s=%s' % (name, value)]) + chapters = info.get('chapters', []) + if chapters: + metadata_filename = replace_extension(filename, 'meta') + with io.open(metadata_filename, 'wt', encoding='utf-8') as f: + def ffmpeg_escape(text): + return re.sub(r'(=|;|#|\\|\n)', r'\\\1', text) + + metadata_file_content = ';FFMETADATA1\n' + for chapter in chapters: + metadata_file_content += '[CHAPTER]\nTIMEBASE=1/1000\n' + metadata_file_content += 'START=%d\n' % (chapter['start_time'] * 1000) + metadata_file_content += 'END=%d\n' % (chapter['end_time'] * 1000) + chapter_title = chapter.get('title') + if chapter_title: + metadata_file_content += 'title=%s\n' % ffmpeg_escape(chapter_title) + f.write(metadata_file_content) + in_filenames.append(metadata_filename) + options.extend(['-map_metadata', '1']) + self._downloader.to_screen('[ffmpeg] Adding metadata to \'%s\'' % filename) - self.run_ffmpeg(filename, temp_filename, options) + self.run_ffmpeg_multiple_files(in_filenames, temp_filename, options) + if chapters: + os.remove(metadata_filename) os.remove(encodeFilename(filename)) os.rename(encodeFilename(temp_filename), encodeFilename(filename)) return [], info @@ -462,6 +535,21 @@ class FFmpegFixupM4aPP(FFmpegPostProcessor): return [], info +class FFmpegFixupM3u8PP(FFmpegPostProcessor): + def run(self, info): + filename = info['filepath'] + if self.get_audio_codec(filename) == 'aac': + temp_filename = prepend_extension(filename, 'temp') + + options = ['-c', 'copy', '-f', 'mp4', '-bsf:a', 'aac_adtstoasc'] + self._downloader.to_screen('[ffmpeg] Fixing malformed AAC bitstream in "%s"' % filename) + self.run_ffmpeg(filename, temp_filename, options) + + os.remove(encodeFilename(filename)) + os.rename(encodeFilename(temp_filename), encodeFilename(filename)) + return [], info + + class FFmpegSubtitlesConvertorPP(FFmpegPostProcessor): def __init__(self, downloader=None, format=None): super(FFmpegSubtitlesConvertorPP, self).__init__(downloader) @@ -478,30 +566,32 @@ class FFmpegSubtitlesConvertorPP(FFmpegPostProcessor): self._downloader.to_screen('[ffmpeg] There aren\'t any subtitles to convert') return [], info self._downloader.to_screen('[ffmpeg] Converting subtitles') + sub_filenames = [] for lang, sub in subs.items(): ext = sub['ext'] if ext == new_ext: self._downloader.to_screen( - '[ffmpeg] Subtitle file for %s is already in the requested' - 'format' % new_ext) + '[ffmpeg] Subtitle file for %s is already in the requested format' % new_ext) continue + old_file = subtitles_filename(filename, lang, ext) + sub_filenames.append(old_file) new_file = subtitles_filename(filename, lang, new_ext) - if ext == 'dfxp' or ext == 'ttml': + if ext in ('dfxp', 'ttml', 'tt'): self._downloader.report_warning( 'You have requested to convert dfxp (TTML) subtitles into another format, ' 'which results in style information loss') - dfxp_file = subtitles_filename(filename, lang, ext) + dfxp_file = old_file srt_file = subtitles_filename(filename, lang, 'srt') - with io.open(dfxp_file, 'rt', encoding='utf-8') as f: + with open(dfxp_file, 'rb') as f: srt_data = dfxp2srt(f.read()) with io.open(srt_file, 'wt', encoding='utf-8') as f: f.write(srt_data) + old_file = srt_file - ext = 'srt' subs[lang] = { 'ext': 'srt', 'data': srt_data @@ -509,15 +599,15 @@ class FFmpegSubtitlesConvertorPP(FFmpegPostProcessor): if new_ext == 'srt': continue + else: + sub_filenames.append(srt_file) - self.run_ffmpeg( - subtitles_filename(filename, lang, ext), - new_file, ['-f', new_format]) + self.run_ffmpeg(old_file, new_file, ['-f', new_format]) with io.open(new_file, 'rt', encoding='utf-8') as f: subs[lang] = { - 'ext': ext, + 'ext': new_ext, 'data': f.read(), } - return [], info + return sub_filenames, info