libhb/muxmp4.c

   1 /* $Id: muxmp4.c,v 1.24 2005/11/04 13:09:41 titer Exp $
   2
   3    This file is part of the HandBrake source code.
   4    Homepage: <http://handbrake.m0k.org/>.
   5    It may be used under the terms of the GNU General Public License. */
   6
   7 /* libmp4v2 header */
   8 #include "mp4.h"
   9
  10 #include "hb.h"
  11
  12 void AddIPodUUID(MP4FileHandle, MP4TrackId);
  13
  14 struct hb_mux_object_s
  15 {
  16     HB_MUX_COMMON;
  17
  18     hb_job_t * job;
  19
  20     /* libmp4v2 handle */
  21     MP4FileHandle file;
  22
  23     /* Cumulated durations so far, in timescale units (see MP4Mux) */
  24     uint64_t sum_dur;
  25
  26     /* Chapter state information for muxing */
  27     MP4TrackId chapter_track;
  28     int current_chapter;
  29     uint64_t chapter_duration;
  30 };
  31
  32 struct hb_mux_data_s
  33 {
  34     MP4TrackId track;
  35 };
  36
  37 struct hb_text_sample_s
  38 {
  39     uint8_t     sample[1280];
  40     uint32_t    length;
  41     MP4Duration duration;
  42 };
  43
  44 /**********************************************************************
  45  * MP4CreateTextSample
  46  **********************************************************************
  47  * Creates a buffer for a text track sample
  48  *********************************************************************/
  49 static struct hb_text_sample_s *MP4CreateTextSample( char *textString, uint64_t duration )
  50 {
  51     struct hb_text_sample_s *sample = NULL;
  52     int stringLength = strlen(textString);
  53     int x;
  54
  55     if( stringLength < 1024 )
  56     {
  57         sample = malloc( sizeof( struct hb_text_sample_s ) );
  58
  59         //textLength = (stringLength; // Account for BOM
  60         sample->length = stringLength + 2 + 12; // Account for text length code and other marker
  61         sample->duration = (MP4Duration)duration;
  62
  63         // 2-byte length marker
  64         sample->sample[0] = (stringLength >> 8) & 0xff;
  65         sample->sample[1] = stringLength & 0xff;
  66
  67         strncpy( (char *)&(sample->sample[2]), textString, stringLength );
  68
  69         x = 2 + stringLength;
  70
  71         // Modifier Length Marker
  72         sample->sample[x] = 0x00;
  73         sample->sample[x+1] = 0x00;
  74         sample->sample[x+2] = 0x00;
  75         sample->sample[x+3] = 0x0C;
  76
  77         // Modifier Type Code
  78         sample->sample[x+4] = 'e';
  79         sample->sample[x+5] = 'n';
  80         sample->sample[x+6] = 'c';
  81         sample->sample[x+7] = 'd';
  82
  83         // Modifier Value
  84         sample->sample[x+8] = 0x00;
  85         sample->sample[x+9] = 0x00;
  86         sample->sample[x+10] = (256 >> 8) & 0xff;
  87         sample->sample[x+11] = 256 & 0xff;
  88     }
  89
  90     return sample;
  91 }
  92
  93 /**********************************************************************
  94  * MP4GenerateChapterSample
  95  **********************************************************************
  96  * Creates a buffer for a text track sample
  97  *********************************************************************/
  98 static struct hb_text_sample_s *MP4GenerateChapterSample( hb_mux_object_t * m, uint64_t duration )
  99 {
 100     int chapter = m->current_chapter;
 101     hb_chapter_t *chapter_data = hb_list_item( m->job->title->list_chapter, chapter - 1 );
 102     char tmp_buffer[1024];
 103     char *string = tmp_buffer;
 104
 105     tmp_buffer[0] = '\0';
 106
 107     if( chapter_data != NULL )
 108     {
 109         string = chapter_data->title;
 110     }
 111
 112     if( strlen(string) == 0 || strlen(string) >= 1024 )
 113     {
 114         snprintf( tmp_buffer, 1023, "Chapter %03i", chapter );
 115         string = tmp_buffer;
 116     }
 117
 118     return MP4CreateTextSample( string, duration );
 119 }
 120
 121
 122 /**********************************************************************
 123  * MP4Init
 124  **********************************************************************
 125  * Allocates hb_mux_data_t structures, create file and write headers
 126  *********************************************************************/
 127 static int MP4Init( hb_mux_object_t * m )
 128 {
 129     hb_job_t   * job   = m->job;
 130     hb_title_t * title = job->title;
 131
 132     hb_audio_t    * audio;
 133     hb_mux_data_t * mux_data;
 134     int i;
 135     u_int16_t language_code;
 136
 137     /* Flags for enabling/disabling tracks in an MP4. */
 138     typedef enum { TRACK_DISABLED = 0x0, TRACK_ENABLED = 0x1, TRACK_IN_MOVIE = 0x2, TRACK_IN_PREVIEW = 0x4, TRACK_IN_POSTER = 0x8}  track_header_flags;
 139
 140
 141     /* Create an empty mp4 file */
 142     if (job->largeFileSize)
 143     /* Use 64-bit MP4 file */
 144     {
 145         m->file = MP4Create( job->file, MP4_DETAILS_ERROR, MP4_CREATE_64BIT_DATA );
 146         hb_log("Using 64-bit MP4 formatting.");
 147     }
 148     else
 149     /* Limit MP4s to less than 4 GB */
 150     {
 151         m->file = MP4Create( job->file, MP4_DETAILS_ERROR, 0 );
 152     }
 153
 154     if (m->file == MP4_INVALID_FILE_HANDLE)
 155     {
 156         hb_error("muxmp4.c: MP4Create failed!");
 157         *job->die = 1;
 158         return 0;
 159     }
 160
 161     /* Video track */
 162     mux_data      = malloc( sizeof( hb_mux_data_t ) );
 163     job->mux_data = mux_data;
 164
 165     /* When using the standard 90000 timescale, QuickTime tends to have
 166        synchronization issues (audio not playing at the correct speed).
 167        To workaround this, we use the audio samplerate as the
 168        timescale */
 169     if (!(MP4SetTimeScale( m->file, job->arate )))
 170     {
 171         hb_error("muxmp4.c: MP4SetTimeScale failed!");
 172         *job->die = 1;
 173         return 0;
 174     }
 175
 176     if( job->vcodec == HB_VCODEC_X264 )
 177     {
 178         /* Stolen from mp4creator */
 179         if(!(MP4SetVideoProfileLevel( m->file, 0x7F )))
 180         {
 181             hb_error("muxmp4.c: MP4SetVideoProfileLevel failed!");
 182             *job->die = 1;
 183             return 0;
 184         }
 185
 186                 mux_data->track = MP4AddH264VideoTrack( m->file, job->arate,
 187                         MP4_INVALID_DURATION, job->width, job->height,
 188                         job->config.h264.sps[1], /* AVCProfileIndication */
 189                         job->config.h264.sps[2], /* profile_compat */
 190                         job->config.h264.sps[3], /* AVCLevelIndication */
 191                         3 );      /* 4 bytes length before each NAL unit */
 192
 193
 194         MP4AddH264SequenceParameterSet( m->file, mux_data->track,
 195                 job->config.h264.sps, job->config.h264.sps_length );
 196         MP4AddH264PictureParameterSet( m->file, mux_data->track,
 197                 job->config.h264.pps, job->config.h264.pps_length );
 198
 199                 if( job->h264_level == 30 || job->ipod_atom)
 200                 {
 201                         hb_log("About to add iPod atom");
 202                         AddIPodUUID(m->file, mux_data->track);
 203                 }
 204
 205     }
 206     else /* FFmpeg or XviD */
 207     {
 208         if(!(MP4SetVideoProfileLevel( m->file, MPEG4_SP_L3 )))
 209         {
 210             hb_error("muxmp4.c: MP4SetVideoProfileLevel failed!");
 211             *job->die = 1;
 212             return 0;
 213         }
 214         mux_data->track = MP4AddVideoTrack( m->file, job->arate,
 215                 MP4_INVALID_DURATION, job->width, job->height,
 216                 MP4_MPEG4_VIDEO_TYPE );
 217         if (mux_data->track == MP4_INVALID_TRACK_ID)
 218         {
 219             hb_error("muxmp4.c: MP4AddVideoTrack failed!");
 220             *job->die = 1;
 221             return 0;
 222         }
 223
 224
 225         /* VOL from FFmpeg or XviD */
 226         if (!(MP4SetTrackESConfiguration( m->file, mux_data->track,
 227                 job->config.mpeg4.bytes, job->config.mpeg4.length )))
 228         {
 229             hb_error("muxmp4.c: MP4SetTrackESConfiguration failed!");
 230             *job->die = 1;
 231             return 0;
 232         }
 233     }
 234
 235         /* apply the anamorphic transformation matrix if needed */
 236
 237         if( job->pixel_ratio ) {
 238
 239                 uint8_t* val;
 240                 uint8_t nval[38];
 241                 uint32_t *ptr32 = (uint32_t*) (nval + 2);
 242                 uint32_t size;
 243
 244                 MP4GetBytesProperty(m->file, "moov.trak.tkhd.reserved3", &val, &size);
 245
 246                 if (size == 38) {
 247
 248                         memcpy(nval, val, size);
 249
 250                         float width, height;
 251                         width = job->pixel_aspect_width;
 252                         height = job->pixel_aspect_height;
 253
 254            uint32_t pixelRatioInt;
 255            if (width >= height)
 256            {
 257                pixelRatioInt = (uint32_t)((width / height) * 0x10000);
 258
 259 #ifdef WORDS_BIGENDIAN
 260                ptr32[0] = pixelRatioInt;
 261 #else
 262                /* we need to switch the endianness, as the file format expects big endian */
 263                ptr32[0] = ((pixelRatioInt & 0x000000FF) << 24) + ((pixelRatioInt & 0x0000FF00) << 8) + ((pixelRatioInt & 0x00FF0000) >> 8) + ((pixelRatioInt & 0xFF000000) >> 24);
 264 #endif
 265
 266            }
 267            else
 268            {
 269                pixelRatioInt = (uint32_t)((height / width) * 0x10000);
 270 #ifdef WORDS_BIGENDIAN
 271                ptr32[4] = pixelRatioInt;
 272 #else
 273                 /* we need to switch the endianness, as the file format expects big endian */
 274                 ptr32[4] = ((pixelRatioInt & 0x000000FF) << 24) + ((pixelRatioInt & 0x0000FF00) << 8) + ((pixelRatioInt & 0x00FF0000) >> 8) + ((pixelRatioInt & 0xFF000000) >> 24);
 275 #endif
 276             }
 277
 278
 279                         if(!MP4SetBytesProperty(m->file, "moov.trak.tkhd.reserved3", nval, size)) {
 280                                 hb_log("Problem setting transform matrix");
 281                         }
 282
 283                 }
 284
 285         }
 286
 287         /* end of transformation matrix */
 288
 289         /* firstAudioTrack will be used to reference the first audio track when we add a chapter track */
 290         MP4TrackId firstAudioTrack = 0;
 291
 292         /* add the audio tracks */
 293     for( i = 0; i < hb_list_count( title->list_audio ); i++ )
 294     {
 295         static u_int8_t reserved2[16] = {
 296                 0x00, 0x00, 0x00, 0x00,
 297                 0x00, 0x00, 0x00, 0x00,
 298                 0x00, 0x02, 0x00, 0x10,
 299                 0x00, 0x00, 0x00, 0x00,
 300             };
 301
 302         audio = hb_list_item( title->list_audio, i );
 303         mux_data = malloc( sizeof( hb_mux_data_t ) );
 304         audio->mux_data = mux_data;
 305
 306         mux_data->track = MP4AddAudioTrack( m->file,
 307                 job->arate, 1024, MP4_MPEG4_AUDIO_TYPE );
 308         MP4SetAudioProfileLevel( m->file, 0x0F );
 309         MP4SetTrackESConfiguration( m->file, mux_data->track,
 310                 audio->config.aac.bytes, audio->config.aac.length );
 311
 312         /* Set the language for this track */
 313         /* The language is stored as 5-bit text - 0x60 */
 314         language_code = audio->iso639_2[0] - 0x60;   language_code <<= 5;
 315         language_code |= audio->iso639_2[1] - 0x60;  language_code <<= 5;
 316         language_code |= audio->iso639_2[2] - 0x60;
 317         MP4SetTrackIntegerProperty(m->file, mux_data->track, "mdia.mdhd.language", language_code);
 318
 319         /* Set the correct number of channels for this track */
 320         reserved2[9] = (u_int8_t)HB_AMIXDOWN_GET_DISCRETE_CHANNEL_COUNT(audio->amixdown);
 321         MP4SetTrackBytesProperty(m->file, mux_data->track, "mdia.minf.stbl.stsd.mp4a.reserved2", reserved2, sizeof(reserved2));
 322
 323         /* Set the audio track alternate group */
 324         MP4SetTrackIntegerProperty(m->file, mux_data->track, "tkhd.alternate_group", 1);
 325
 326         /* If we ever upgrade mpeg4ip, the line above should be replaced with the line below.*/
 327 //        MP4SetTrackIntegerProperty(m->file, mux_data->track, "mdia.minf.stbl.stsd.mp4a.channels",  (u_int16_t)HB_AMIXDOWN_GET_DISCRETE_CHANNEL_COUNT(audio->amixdown));
 328
 329         /* store a reference to the first audio track,
 330         so we can use it to feed the chapter text track's sample rate */
 331         if (i == 0) {
 332             firstAudioTrack = mux_data->track;
 333
 334             /* Enable the first audio track */
 335             MP4SetTrackIntegerProperty(m->file, mux_data->track, "tkhd.flags", (TRACK_ENABLED | TRACK_IN_MOVIE));
 336         }
 337
 338         else
 339             /* Disable the other audio tracks so QuickTime doesn't play
 340                them all at once. */
 341         {
 342             MP4SetTrackIntegerProperty(m->file, mux_data->track, "tkhd.flags", (TRACK_DISABLED | TRACK_IN_MOVIE));
 343             hb_log("Disabled extra audio track %i", mux_data->track-1);
 344         }
 345
 346     }
 347
 348         if (job->chapter_markers)
 349     {
 350                 /* add a text track for the chapters */
 351                 MP4TrackId textTrack;
 352                 textTrack = MP4AddChapterTextTrack(m->file, firstAudioTrack);
 353
 354         m->chapter_track = textTrack;
 355         m->chapter_duration = 0;
 356         m->current_chapter = job->chapter_start;
 357         }
 358
 359     /* Add encoded-by metadata listing version and build date */
 360     char *tool_string;
 361     tool_string = (char *)malloc(80);
 362     snprintf( tool_string, 80, "HandBrake %s %i", HB_VERSION, HB_BUILD);
 363     MP4SetMetadataTool(m->file, tool_string);
 364     free(tool_string);
 365
 366     return 0;
 367 }
 368
 369 static int MP4Mux( hb_mux_object_t * m, hb_mux_data_t * mux_data,
 370                    hb_buffer_t * buf )
 371 {
 372     hb_job_t * job = m->job;
 373
 374     uint64_t duration;
 375
 376     if( mux_data == job->mux_data )
 377     {
 378         /* Add the sample before the new frame.
 379            It is important that this be calculated prior to the duration
 380            of the new video sample, as we want to sync to right after it.
 381            (This is because of how durations for text tracks work in QT) */
 382         if( job->chapter_markers && buf->new_chap )
 383         {
 384             struct hb_text_sample_s *sample;
 385
 386             /* If this is an x264 encode with bframes the IDR frame we're
 387                trying to mark will be displayed offset by its renderOffset
 388                so we need to offset the chapter by the same amount.
 389                MP4 render offsets don't seem to work for text tracks so
 390                we have to fudge the duration instead. */
 391             duration = m->sum_dur - m->chapter_duration;
 392
 393             if ( job->areBframes )
 394             {
 395                 duration += buf->renderOffset * job->arate / 90000;
 396             }
 397
 398             sample = MP4GenerateChapterSample( m, duration );
 399
 400             if( !MP4WriteSample(m->file,
 401                                 m->chapter_track,
 402                                 sample->sample,
 403                                 sample->length,
 404                                 sample->duration,
 405                                 0, true) )
 406             {
 407                 hb_error("Failed to write to output file, disk full?");
 408                 *job->die = 1;
 409             }
 410             free(sample);
 411             m->current_chapter++;
 412             m->chapter_duration += duration;
 413         }
 414
 415         /* Video */
 416         /* Because we use the audio samplerate as the timescale,
 417            we have to use potentially variable durations so the video
 418            doesn't go out of sync */
 419         int64_t bias = ( buf->start * job->arate / 90000 ) - m->sum_dur;
 420         duration = ( buf->stop - buf->start ) * job->arate / 90000 + bias;
 421         m->sum_dur += duration;
 422     }
 423     else
 424     {
 425         /* Audio */
 426         duration = MP4_INVALID_DURATION;
 427     }
 428
 429     /* Here's where the sample actually gets muxed.
 430        If it's an audio sample, don't offset the sample's playback.
 431        If it's a video sample and there are no b-frames, ditto.
 432        If there are b-frames, offset by the initDelay plus the
 433        difference between the presentation time stamp x264 gives
 434        and the decoding time stamp from the buffer data. */
 435     if( !MP4WriteSample( m->file,
 436                          mux_data->track,
 437                          buf->data,
 438                          buf->size,
 439                          duration,
 440                          ((mux_data->track != 1) ||
 441                           (job->areBframes==0) ||
 442                           (job->vcodec != HB_VCODEC_X264)) ? 0 : (  buf->renderOffset * job->arate / 90000),
 443                          ((buf->frametype & HB_FRAME_KEY) != 0) ) )
 444     {
 445         hb_error("Failed to write to output file, disk full?");
 446         *job->die = 1;
 447     }
 448
 449     return 0;
 450 }
 451
 452 static int MP4End( hb_mux_object_t * m )
 453 {
 454     hb_job_t   * job   = m->job;
 455
 456     /* Write our final chapter marker */
 457     if( m->job->chapter_markers )
 458     {
 459         struct hb_text_sample_s *sample = MP4GenerateChapterSample( m, (m->sum_dur - m->chapter_duration) );
 460
 461         if( !MP4WriteSample(m->file,
 462                             m->chapter_track,
 463                             sample->sample,
 464                             sample->length,
 465                             sample->duration,
 466                             0, true) )
 467         {
 468             hb_error("Failed to write to output file, disk full?");
 469             *job->die = 1;
 470         }
 471         free(sample);
 472     }
 473
 474     if (job->areBframes)
 475     {
 476            // Insert track edit to get A/V back in sync.  The edit amount is
 477            // the rendering offset of the first sample.
 478            MP4AddTrackEdit(m->file, 1, MP4_INVALID_EDIT_ID, MP4GetSampleRenderingOffset(m->file,1,1),
 479                MP4GetTrackDuration(m->file, 1), 0);
 480             if ( m->job->chapter_markers )
 481             {
 482                 // apply same edit to chapter track to keep it in sync with video
 483                 MP4AddTrackEdit(m->file, m->chapter_track, MP4_INVALID_EDIT_ID,
 484                                 MP4GetSampleRenderingOffset(m->file,1,1),
 485                                 MP4GetTrackDuration(m->file, m->chapter_track), 0);
 486             }
 487      }
 488
 489     MP4Close( m->file );
 490
 491     if ( job->mp4_optimize )
 492     {
 493         hb_log( "muxmp4: optimizing file" );
 494         char filename[1024]; memset( filename, 0, 1024 );
 495         snprintf( filename, 1024, "%s.tmp", job->file );
 496         MP4Optimize( job->file, filename, MP4_DETAILS_ERROR );
 497         remove( job->file );
 498         rename( filename, job->file );
 499     }
 500
 501     return 0;
 502 }
 503
 504 hb_mux_object_t * hb_mux_mp4_init( hb_job_t * job )
 505 {
 506     hb_mux_object_t * m = calloc( sizeof( hb_mux_object_t ), 1 );
 507     m->init      = MP4Init;
 508     m->mux       = MP4Mux;
 509     m->end       = MP4End;
 510     m->job       = job;
 511     return m;
 512 }
 513