mediagoblin/media_types/video/transcoders.py

   1 # GNU MediaGoblin -- federated, autonomous media hosting
   2 # Copyright (C) 2011, 2012 MediaGoblin contributors.  See AUTHORS.
   3 #
   4 # This program is free software: you can redistribute it and/or modify
   5 # it under the terms of the GNU Affero General Public License as published by
   6 # the Free Software Foundation, either version 3 of the License, or
   7 # (at your option) any later version.
   8 #
   9 # This program is distributed in the hope that it will be useful,
  10 # but WITHOUT ANY WARRANTY; without even the implied warranty of
  11 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  12 # GNU Affero General Public License for more details.
  13 #
  14 # You should have received a copy of the GNU Affero General Public License
  15 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
  16
  17 from __future__ import division, print_function
  18
  19 import os
  20 import sys
  21 import logging
  22 import multiprocessing
  23
  24 from mediagoblin.media_types.tools import discover
  25 from mediagoblin.tools.translate import lazy_pass_to_ugettext as _
  26
  27 #os.environ['GST_DEBUG'] = '4,python:4'
  28
  29 old_argv = sys.argv
  30 sys.argv = []
  31
  32 import gi
  33 gi.require_version('Gst', '1.0')
  34 from gi.repository import GObject, Gst, GstPbutils
  35 Gst.init(None)
  36
  37 sys.argv = old_argv
  38 import struct
  39 try:
  40     from PIL import Image
  41 except ImportError:
  42     import Image
  43
  44 _log = logging.getLogger(__name__)
  45
  46 CPU_COUNT = 2
  47
  48 try:
  49     CPU_COUNT = multiprocessing.cpu_count()
  50 except NotImplementedError:
  51     _log.warning('multiprocessing.cpu_count not implemented')
  52
  53 os.putenv('GST_DEBUG_DUMP_DOT_DIR', '/tmp')
  54
  55
  56 def capture_thumb(video_path, dest_path, width=None, height=None, percent=0.5):
  57     def pad_added(element, pad, connect_to):
  58         '''This is a callback to dynamically add element to pipeline'''
  59         caps = pad.query_caps(None)
  60         name = caps.to_string()
  61         _log.debug('on_pad_added: {0}'.format(name))
  62         if name.startswith('video') and not connect_to.is_linked():
  63             pad.link(connect_to)
  64
  65     # construct pipeline: uridecodebin ! videoconvert ! videoscale ! \
  66     # ! CAPS ! appsink
  67     pipeline = Gst.Pipeline()
  68     uridecodebin = Gst.ElementFactory.make('uridecodebin', None)
  69     uridecodebin.set_property('uri', 'file://{0}'.format(video_path))
  70     videoconvert = Gst.ElementFactory.make('videoconvert', None)
  71     uridecodebin.connect('pad-added', pad_added,
  72                          videoconvert.get_static_pad('sink'))
  73     videoscale = Gst.ElementFactory.make('videoscale', None)
  74
  75     # create caps for video scaling
  76     caps_struct = Gst.Structure.new_empty('video/x-raw')
  77     caps_struct.set_value('pixel-aspect-ratio', Gst.Fraction(1, 1))
  78     caps_struct.set_value('format', 'RGB')
  79     if height:
  80         caps_struct.set_value('height', height)
  81     if width:
  82         caps_struct.set_value('width', width)
  83     caps = Gst.Caps.new_empty()
  84     caps.append_structure(caps_struct)
  85
  86     # sink everything to memory
  87     appsink = Gst.ElementFactory.make('appsink', None)
  88     appsink.set_property('caps', caps)
  89
  90     # add everything to pipeline
  91     elements = [uridecodebin, videoconvert, videoscale, appsink]
  92     for e in elements:
  93         pipeline.add(e)
  94     videoconvert.link(videoscale)
  95     videoscale.link(appsink)
  96
  97     # pipeline constructed, starting playing, but first some preparations
  98     # seek to 50% of the file is required
  99     pipeline.set_state(Gst.State.PAUSED)
 100     # timeout of 3 seconds below was set experimentally
 101     state = pipeline.get_state(Gst.SECOND * 3)
 102     if state[0] != Gst.StateChangeReturn.SUCCESS:
 103         _log.warning('state change failed, {0}'.format(state))
 104         return
 105
 106     # get duration
 107     (success, duration) = pipeline.query_duration(Gst.Format.TIME)
 108     if not success:
 109         _log.warning('query_duration failed')
 110         return
 111
 112     seek_to = int(duration * int(percent * 100) / 100)
 113     _log.debug('Seeking to {0} of {1}'.format(
 114             float(seek_to) / Gst.SECOND, float(duration) / Gst.SECOND))
 115     seek = pipeline.seek_simple(Gst.Format.TIME, Gst.SeekFlags.FLUSH, seek_to)
 116     if not seek:
 117         _log.warning('seek failed')
 118         return
 119
 120     # get sample, retrieve it's format and save
 121     sample = appsink.emit("pull-preroll")
 122     if not sample:
 123         _log.warning('could not get sample')
 124         return
 125     caps = sample.get_caps()
 126     if not caps:
 127         _log.warning('could not get snapshot format')
 128         return
 129     structure = caps.get_structure(0)
 130     (success, width) = structure.get_int('width')
 131     (success, height) = structure.get_int('height')
 132     buffer = sample.get_buffer()
 133
 134     # get the image from the buffer and save it to disk
 135     im = Image.frombytes('RGB', (width, height),
 136                          buffer.extract_dup(0, buffer.get_size()))
 137     im.save(dest_path)
 138     _log.info('thumbnail saved to {0}'.format(dest_path))
 139
 140     # cleanup
 141     pipeline.set_state(Gst.State.NULL)
 142
 143
 144 class VideoTranscoder(object):
 145     '''
 146     Video transcoder
 147
 148     Transcodes the SRC video file to a VP8 WebM video file at DST
 149
 150      - Produces a WebM vp8 and vorbis video file.
 151     '''
 152     def __init__(self):
 153         _log.info('Initializing VideoTranscoder...')
 154         self.progress_percentage = None
 155         self.loop = GObject.MainLoop()
 156
 157     def transcode(self, src, dst, **kwargs):
 158         '''
 159         Transcode a video file into a 'medium'-sized version.
 160         '''
 161         self.source_path = src
 162         self.destination_path = dst
 163
 164         # vp8enc options
 165         self.destination_dimensions = kwargs.get('dimensions', (640, 640))
 166         self.vp8_quality = kwargs.get('vp8_quality', 8)
 167         # Number of threads used by vp8enc:
 168         # number of real cores - 1 as per recommendation on
 169         # <http://www.webmproject.org/tools/encoder-parameters/#6-multi-threaded-encode-and-decode>
 170         self.vp8_threads = kwargs.get('vp8_threads', CPU_COUNT - 1)
 171
 172         # 0 means auto-detect, but dict.get() only falls back to CPU_COUNT
 173         # if value is None, this will correct our incompatibility with
 174         # dict.get()
 175         # This will also correct cases where there's only 1 CPU core, see
 176         # original self.vp8_threads assignment above.
 177         if self.vp8_threads == 0:
 178             self.vp8_threads = CPU_COUNT
 179
 180         # vorbisenc options
 181         self.vorbis_quality = kwargs.get('vorbis_quality', 0.3)
 182
 183         self._progress_callback = kwargs.get('progress_callback') or None
 184
 185         if not type(self.destination_dimensions) == tuple:
 186             raise Exception('dimensions must be tuple: (width, height)')
 187
 188         self._setup_pipeline()
 189         self.data = discover(self.source_path)
 190         self._link_elements()
 191         self.__setup_videoscale_capsfilter()
 192         self.pipeline.set_state(Gst.State.PLAYING)
 193         _log.info('Transcoding...')
 194         _log.debug('Initializing MainLoop()')
 195         self.loop.run()
 196
 197
 198     def _setup_pipeline(self):
 199         _log.debug('Setting up transcoding pipeline')
 200         # Create the pipeline bin.
 201         self.pipeline = Gst.Pipeline.new('VideoTranscoderPipeline')
 202
 203         # Create all GStreamer elements, starting with
 204         # filesrc & decoder
 205         self.filesrc = Gst.ElementFactory.make('filesrc', 'filesrc')
 206         self.filesrc.set_property('location', self.source_path)
 207         self.pipeline.add(self.filesrc)
 208
 209         self.decoder = Gst.ElementFactory.make('decodebin', 'decoder')
 210         self.decoder.connect('pad-added', self._on_dynamic_pad)
 211         self.pipeline.add(self.decoder)
 212
 213         # Video elements
 214         self.videoqueue = Gst.ElementFactory.make('queue', 'videoqueue')
 215         self.pipeline.add(self.videoqueue)
 216
 217         self.videorate = Gst.ElementFactory.make('videorate', 'videorate')
 218         self.pipeline.add(self.videorate)
 219
 220         self.videoconvert = Gst.ElementFactory.make('videoconvert',
 221                                                     'videoconvert')
 222         self.pipeline.add(self.videoconvert)
 223
 224         self.videoscale = Gst.ElementFactory.make('videoscale', 'videoscale')
 225         self.pipeline.add(self.videoscale)
 226
 227         self.capsfilter = Gst.ElementFactory.make('capsfilter', 'capsfilter')
 228         self.pipeline.add(self.capsfilter)
 229
 230         self.vp8enc = Gst.ElementFactory.make('vp8enc', 'vp8enc')
 231         self.vp8enc.set_property('threads', self.vp8_threads)
 232         self.pipeline.add(self.vp8enc)
 233
 234         # Audio elements
 235         self.audioqueue = Gst.ElementFactory.make('queue', 'audioqueue')
 236         self.pipeline.add(self.audioqueue)
 237
 238         self.audiorate = Gst.ElementFactory.make('audiorate', 'audiorate')
 239         self.audiorate.set_property('tolerance', 80000000)
 240         self.pipeline.add(self.audiorate)
 241
 242         self.audioconvert = Gst.ElementFactory.make('audioconvert', 'audioconvert')
 243         self.pipeline.add(self.audioconvert)
 244         self.audiocapsfilter = Gst.ElementFactory.make('capsfilter',
 245                                                        'audiocapsfilter')
 246         audiocaps = Gst.Caps.new_empty()
 247         audiocaps_struct = Gst.Structure.new_empty('audio/x-raw')
 248         audiocaps.append_structure(audiocaps_struct)
 249         self.audiocapsfilter.set_property('caps', audiocaps)
 250         self.pipeline.add(self.audiocapsfilter)
 251
 252         self.vorbisenc = Gst.ElementFactory.make('vorbisenc', 'vorbisenc')
 253         self.vorbisenc.set_property('quality', self.vorbis_quality)
 254         self.pipeline.add(self.vorbisenc)
 255
 256         # WebMmux & filesink
 257         self.webmmux = Gst.ElementFactory.make('webmmux', 'webmmux')
 258         self.pipeline.add(self.webmmux)
 259
 260         self.filesink = Gst.ElementFactory.make('filesink', 'filesink')
 261         self.filesink.set_property('location', self.destination_path)
 262         self.pipeline.add(self.filesink)
 263
 264         # Progressreport
 265         self.progressreport = Gst.ElementFactory.make(
 266             'progressreport', 'progressreport')
 267         # Update every second
 268         self.progressreport.set_property('update-freq', 1)
 269         self.progressreport.set_property('silent', True)
 270         self.pipeline.add(self.progressreport)
 271
 272     def _link_elements(self):
 273         '''
 274         Link all the elements
 275
 276         This code depends on data from the discoverer and is called
 277         from __discovered
 278         '''
 279         _log.debug('linking elements')
 280         # Link the filesrc element to the decoder. The decoder then emits
 281         # 'new-decoded-pad' which links decoded src pads to either a video
 282         # or audio sink
 283         self.filesrc.link(self.decoder)
 284         # link the rest
 285         self.videoqueue.link(self.videorate)
 286         self.videorate.link(self.videoconvert)
 287         self.videoconvert.link(self.videoscale)
 288         self.videoscale.link(self.capsfilter)
 289         self.capsfilter.link(self.vp8enc)
 290         self.vp8enc.link(self.webmmux)
 291
 292         if self.data.get_audio_streams():
 293             self.audioqueue.link(self.audiorate)
 294             self.audiorate.link(self.audioconvert)
 295             self.audioconvert.link(self.audiocapsfilter)
 296             self.audiocapsfilter.link(self.vorbisenc)
 297             self.vorbisenc.link(self.webmmux)
 298         self.webmmux.link(self.progressreport)
 299         self.progressreport.link(self.filesink)
 300
 301         # Setup the message bus and connect _on_message to the pipeline
 302         self._setup_bus()
 303
 304     def _on_dynamic_pad(self, dbin, pad):
 305         '''
 306         Callback called when ``decodebin`` has a pad that we can connect to
 307         '''
 308         # Intersect the capabilities of the video sink and the pad src
 309         # Then check if they have no common capabilities.
 310         if (self.videorate.get_static_pad('sink').get_pad_template()
 311                 .get_caps().intersect(pad.query_caps()).is_empty()):
 312             # It is NOT a video src pad.
 313             _log.debug('linking audio to the pad dynamically')
 314             pad.link(self.audioqueue.get_static_pad('sink'))
 315         else:
 316             # It IS a video src pad.
 317             _log.debug('linking video to the pad dynamically')
 318             pad.link(self.videoqueue.get_static_pad('sink'))
 319
 320     def _setup_bus(self):
 321         self.bus = self.pipeline.get_bus()
 322         self.bus.add_signal_watch()
 323         self.bus.connect('message', self._on_message)
 324
 325     def __setup_videoscale_capsfilter(self):
 326         '''
 327         Sets up the output format (width, height) for the video
 328         '''
 329         caps_struct = Gst.Structure.new_empty('video/x-raw')
 330         caps_struct.set_value('pixel-aspect-ratio', Gst.Fraction(1, 1))
 331         caps_struct.set_value('framerate', Gst.Fraction(30, 1))
 332         video_info = self.data.get_video_streams()[0]
 333         if video_info.get_height() > video_info.get_width():
 334             # portrait
 335             caps_struct.set_value('height', self.destination_dimensions[1])
 336         else:
 337             # landscape
 338             caps_struct.set_value('width', self.destination_dimensions[0])
 339         caps = Gst.Caps.new_empty()
 340         caps.append_structure(caps_struct)
 341         self.capsfilter.set_property('caps', caps)
 342
 343     def _on_message(self, bus, message):
 344         _log.debug((bus, message, message.type))
 345         if message.type == Gst.MessageType.EOS:
 346             self.dst_data = discover(self.destination_path)
 347             self.__stop()
 348             _log.info('Done')
 349         elif message.type == Gst.MessageType.ELEMENT:
 350             if message.has_name('progress'):
 351                 structure = message.get_structure()
 352                 # Update progress state if it has changed
 353                 (success, percent) = structure.get_int('percent')
 354                 if self.progress_percentage != percent and success:
 355                     self.progress_percentage = percent
 356                     if self._progress_callback:
 357                         self._progress_callback(percent)
 358                     _log.info('{percent}% done...'.format(percent=percent))
 359         elif message.type == Gst.MessageType.ERROR:
 360             _log.error('Got error: {0}'.format(message.parse_error()))
 361             self.dst_data = None
 362             self.__stop()
 363
 364     def __stop(self):
 365         _log.debug(self.loop)
 366
 367         if hasattr(self, 'pipeline'):
 368             # Stop executing the pipeline
 369             self.pipeline.set_state(Gst.State.NULL)
 370
 371         # This kills the loop, mercifully
 372         GObject.idle_add(self.__stop_mainloop)
 373
 374     def __stop_mainloop(self):
 375         '''
 376         Wrapper for GObject.MainLoop.quit()
 377
 378         This wrapper makes us able to see if self.loop.quit has been called
 379         '''
 380         _log.info('Terminating MainLoop')
 381
 382         self.loop.quit()
 383
 384
 385 if __name__ == '__main__':
 386     os.nice(19)
 387     from optparse import OptionParser
 388
 389     parser = OptionParser(
 390         usage='%prog [-v] -a [ video | thumbnail | discover ] SRC [ DEST ]')
 391
 392     parser.add_option('-a', '--action',
 393                       dest='action',
 394                       help='One of "video", "discover" or "thumbnail"')
 395
 396     parser.add_option('-v',
 397                       dest='verbose',
 398                       action='store_true',
 399                       help='Output debug information')
 400
 401     parser.add_option('-q',
 402                       dest='quiet',
 403                       action='store_true',
 404                       help='Dear program, please be quiet unless *error*')
 405
 406     parser.add_option('-w', '--width',
 407                       type=int,
 408                       default=180)
 409
 410     (options, args) = parser.parse_args()
 411
 412     if options.verbose:
 413         _log.setLevel(logging.DEBUG)
 414     else:
 415         _log.setLevel(logging.INFO)
 416
 417     if options.quiet:
 418         _log.setLevel(logging.ERROR)
 419
 420     _log.debug(args)
 421
 422     if not len(args) == 2 and not options.action == 'discover':
 423         parser.print_help()
 424         sys.exit()
 425
 426     transcoder = VideoTranscoder()
 427
 428     if options.action == 'thumbnail':
 429         args.append(options.width)
 430         VideoThumbnailerMarkII(*args)
 431     elif options.action == 'video':
 432         def cb(data):
 433             print('I\'m a callback!')
 434         transcoder.transcode(*args, progress_callback=cb)
 435     elif options.action == 'discover':
 436         print(transcoder.discover(*args))