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