modules/codec/speex.c

   1 /*****************************************************************************
   2  * speex.c: speex decoder/packetizer/encoder module making use of libspeex.
   3  *****************************************************************************
   4  * Copyright (C) 2003-2009 VLC authors and VideoLAN
   5  * $Id$
   6  *
   7  * Authors: Gildas Bazin <gbazin@videolan.org>
   8  *
   9  * This program is free software; you can redistribute it and/or modify it
  10  * under the terms of the GNU Lesser General Public License as published by
  11  * the Free Software Foundation; either version 2.1 of the License, or
  12  * (at your option) any later version.
  13  *
  14  * This program is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  17  * GNU Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public License
  20  * along with this program; if not, write to the Free Software Foundation,
  21  * Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  22  *****************************************************************************/
  23
  24 /*****************************************************************************
  25  * Preamble
  26  *****************************************************************************/
  27 #ifdef HAVE_CONFIG_H
  28 # include "config.h"
  29 #endif
  30
  31 #include <vlc_common.h>
  32 #include <vlc_plugin.h>
  33 #include <vlc_input.h>
  34 #include <vlc_codec.h>
  35 #include "../demux/xiph.h"
  36
  37 #include <ogg/ogg.h>
  38 #include <speex/speex.h>
  39 #include <speex/speex_header.h>
  40 #include <speex/speex_stereo.h>
  41 #include <speex/speex_callbacks.h>
  42
  43 #include <assert.h>
  44
  45 /*****************************************************************************
  46  * Module descriptor
  47  *****************************************************************************/
  48 static int  OpenDecoder   ( vlc_object_t * );
  49 static int  OpenPacketizer( vlc_object_t * );
  50 static void CloseDecoder  ( vlc_object_t * );
  51
  52 #ifdef ENABLE_SOUT
  53 static int OpenEncoder   ( vlc_object_t * );
  54 static void CloseEncoder ( vlc_object_t * );
  55 #endif
  56
  57 #define ENC_CFG_PREFIX "sout-speex-"
  58
  59 #define ENC_MODE_TEXT N_("Mode" )
  60 #define ENC_MODE_LONGTEXT N_( \
  61     "Enforce the mode of the encoder." )
  62
  63 #define ENC_QUALITY_TEXT N_("Encoding quality")
  64 #define ENC_QUALITY_LONGTEXT N_( \
  65     "Enforce a quality between 0 (low) and 10 (high)." )
  66
  67 #define ENC_COMPLEXITY_TEXT N_("Encoding complexity" )
  68 #define ENC_COMPLEXITY_LONGTEXT N_( \
  69     "Enforce the complexity of the encoder." )
  70
  71 #define ENC_MAXBITRATE_TEXT N_( "Maximal bitrate" )
  72 #define ENC_MAXBITRATE_LONGTEXT N_( \
  73     "Enforce the maximal VBR bitrate" )
  74
  75 #define ENC_CBR_TEXT N_( "CBR encoding" )
  76 #define ENC_CBR_LONGTEXT N_( \
  77     "Enforce a constant bitrate encoding (CBR) instead of default " \
  78     "variable bitrate encoding (VBR)." )
  79
  80 #define ENC_VAD_TEXT N_( "Voice activity detection" )
  81 #define ENC_VAD_LONGTEXT N_( \
  82     "Enable voice activity detection (VAD). It is automatically " \
  83     "activated in VBR mode." )
  84
  85 #define ENC_DTX_TEXT N_( "Discontinuous Transmission" )
  86 #define ENC_DTX_LONGTEXT N_( \
  87     "Enable discontinuous transmission (DTX)." )
  88
  89 static const int pi_enc_mode_values[] = { 0, 1, 2 };
  90 static const char * const ppsz_enc_mode_descriptions[] = {
  91     N_("Narrow-band (8kHz)"), N_("Wide-band (16kHz)"), N_("Ultra-wideband (32kHz)"), NULL
  92 };
  93
  94 vlc_module_begin ()
  95     set_category( CAT_INPUT )
  96     set_subcategory( SUBCAT_INPUT_ACODEC )
  97
  98     set_description( N_("Speex audio decoder") )
  99     set_capability( "audio decoder", 100 )
 100     set_shortname( N_("Speex") )
 101     set_callbacks( OpenDecoder, CloseDecoder )
 102
 103     add_submodule ()
 104     set_description( N_("Speex audio packetizer") )
 105     set_capability( "packetizer", 100 )
 106     set_callbacks( OpenPacketizer, CloseDecoder )
 107
 108 #ifdef ENABLE_SOUT
 109     add_submodule ()
 110     set_description( N_("Speex audio encoder") )
 111     set_capability( "encoder", 100 )
 112     set_callbacks( OpenEncoder, CloseEncoder )
 113
 114     add_integer( ENC_CFG_PREFIX "mode", 0, ENC_MODE_TEXT,
 115                  ENC_MODE_LONGTEXT, false )
 116         change_integer_list( pi_enc_mode_values, ppsz_enc_mode_descriptions )
 117
 118     add_integer( ENC_CFG_PREFIX "complexity", 3, ENC_COMPLEXITY_TEXT,
 119                  ENC_COMPLEXITY_LONGTEXT, false )
 120         change_integer_range( 1, 10 )
 121
 122     add_bool( ENC_CFG_PREFIX "cbr", false, ENC_CBR_TEXT,
 123                  ENC_CBR_LONGTEXT, false )
 124
 125     add_float( ENC_CFG_PREFIX "quality", 8.0, ENC_QUALITY_TEXT,
 126                ENC_QUALITY_LONGTEXT, false )
 127         change_float_range( 0.0, 10.0 )
 128
 129     add_integer( ENC_CFG_PREFIX "max-bitrate", 0, ENC_MAXBITRATE_TEXT,
 130                  ENC_MAXBITRATE_LONGTEXT, false )
 131
 132     add_bool( ENC_CFG_PREFIX "vad", true, ENC_VAD_TEXT,
 133                  ENC_VAD_LONGTEXT, false )
 134
 135     add_bool( ENC_CFG_PREFIX "dtx", false, ENC_DTX_TEXT,
 136                  ENC_DTX_LONGTEXT, false )
 137
 138     /* TODO agc, noise suppression, */
 139 #endif
 140
 141 vlc_module_end ()
 142
 143 static const char *const ppsz_enc_options[] = {
 144     "mode", "complexity", "cbr", "quality", "max-bitrate", "vad", "dtx", NULL
 145 };
 146
 147 /*****************************************************************************
 148  * decoder_sys_t : speex decoder descriptor
 149  *****************************************************************************/
 150 typedef struct
 151 {
 152     /* Module mode */
 153     bool b_packetizer;
 154
 155     /*
 156      * Input properties
 157      */
 158     bool b_has_headers;
 159     int i_frame_in_packet;
 160
 161     /*
 162      * Speex properties
 163      */
 164     SpeexBits bits;
 165     SpeexHeader *p_header;
 166     SpeexStereoState stereo;
 167     void *p_state;
 168     unsigned int rtp_rate;
 169
 170     /*
 171      * Common properties
 172      */
 173     date_t end_date;
 174
 175 } decoder_sys_t;
 176
 177 /****************************************************************************
 178  * Local prototypes
 179  ****************************************************************************/
 180
 181 static block_t *Packetize  ( decoder_t *, block_t ** );
 182 static int      DecodeAudio  ( decoder_t *, block_t * );
 183 static int      DecodeRtpSpeexPacket( decoder_t *, block_t *);
 184 static int      ProcessHeaders( decoder_t * );
 185 static int      ProcessInitialHeader ( decoder_t *, ogg_packet * );
 186 static block_t *ProcessPacket( decoder_t *, ogg_packet *, block_t ** );
 187 static void Flush( decoder_t * );
 188
 189 static block_t *DecodePacket( decoder_t *, ogg_packet * );
 190 static block_t *SendPacket( decoder_t *, block_t * );
 191
 192 static void ParseSpeexComments( decoder_t *, ogg_packet * );
 193
 194 static int OpenCommon( vlc_object_t *p_this, bool b_packetizer )
 195 {
 196     decoder_t *p_dec = (decoder_t*)p_this;
 197     decoder_sys_t *p_sys;
 198
 199     if( p_dec->fmt_in.i_codec != VLC_CODEC_SPEEX )
 200         return VLC_EGENERIC;
 201
 202     /* Allocate the memory needed to store the decoder's structure */
 203     if( ( p_dec->p_sys = p_sys = malloc(sizeof(decoder_sys_t)) ) == NULL )
 204         return VLC_ENOMEM;
 205     p_sys->bits.buf_size = 0;
 206     p_sys->b_packetizer = b_packetizer;
 207     p_sys->rtp_rate = p_dec->fmt_in.audio.i_rate;
 208     p_sys->b_has_headers = false;
 209
 210     date_Set( &p_sys->end_date, VLC_TICK_INVALID );
 211
 212     if( b_packetizer )
 213     {
 214         p_dec->fmt_out.i_codec = VLC_CODEC_SPEEX;
 215         p_dec->pf_packetize    = Packetize;
 216     }
 217     else
 218     {
 219         /* Set output properties */
 220         p_dec->fmt_out.i_codec = VLC_CODEC_S16N;
 221
 222         /*
 223           Set callbacks
 224           If the codec is spxr then this decoder is
 225           being invoked on a Speex stream arriving via RTP.
 226           A special decoder callback is used.
 227         */
 228         if (p_dec->fmt_in.i_original_fourcc == VLC_FOURCC('s', 'p', 'x', 'r'))
 229         {
 230             msg_Dbg( p_dec, "Using RTP version of Speex decoder @ rate %d.",
 231             p_dec->fmt_in.audio.i_rate );
 232             p_dec->pf_decode = DecodeRtpSpeexPacket;
 233         }
 234         else
 235         {
 236             p_dec->pf_decode = DecodeAudio;
 237         }
 238     }
 239     p_dec->pf_flush        = Flush;
 240
 241     p_sys->p_state = NULL;
 242     p_sys->p_header = NULL;
 243     p_sys->i_frame_in_packet = 0;
 244
 245     return VLC_SUCCESS;
 246 }
 247
 248 /*****************************************************************************
 249  * OpenDecoder: probe the decoder and return score
 250  *****************************************************************************/
 251 static int OpenDecoder( vlc_object_t *p_this )
 252 {
 253     return OpenCommon( p_this, false );
 254 }
 255
 256 static int OpenPacketizer( vlc_object_t *p_this )
 257 {
 258     return OpenCommon( p_this, true );
 259 }
 260
 261 static int CreateDefaultHeader( decoder_t *p_dec )
 262 {
 263     ogg_packet oggpacket;
 264     SpeexHeader *p_header = malloc( sizeof(SpeexHeader) );
 265     if( !p_header )
 266         return VLC_ENOMEM;
 267
 268     const int rate = p_dec->fmt_in.audio.i_rate;
 269     const unsigned i_mode = (rate / 8000) >> 1;
 270
 271     const SpeexMode *mode;
 272     int ret = VLC_SUCCESS;
 273     oggpacket.packet = NULL;
 274
 275     switch( rate )
 276     {
 277         case 8000:
 278         case 16000:
 279         case 32000:
 280             mode = speex_lib_get_mode( i_mode );
 281             break;
 282         default:
 283             msg_Err( p_dec, "Unexpected rate %d", rate );
 284             ret = VLC_EGENERIC;
 285             goto cleanup;
 286     }
 287
 288     speex_init_header( p_header, rate, p_dec->fmt_in.audio.i_channels, mode );
 289     p_header->frames_per_packet = 160 << i_mode;
 290
 291     oggpacket.packet = (unsigned char *) speex_header_to_packet( p_header,
 292             (int *) &oggpacket.bytes );
 293     if( !oggpacket.packet )
 294     {
 295         ret = VLC_ENOMEM;
 296         goto cleanup;
 297     }
 298
 299     oggpacket.b_o_s = 1;
 300     oggpacket.e_o_s = 0;
 301     oggpacket.granulepos = -1;
 302     oggpacket.packetno = 0;
 303
 304     ret = ProcessInitialHeader( p_dec, &oggpacket );
 305
 306     if( ret != VLC_SUCCESS )
 307     {
 308         msg_Err( p_dec, "default Speex header is corrupted" );
 309     }
 310
 311 cleanup:
 312     free( oggpacket.packet );
 313     free( p_header );
 314
 315     return ret;
 316 }
 317
 318
 319 /****************************************************************************
 320  * DecodeBlock: the whole thing
 321  ****************************************************************************
 322  * This function must be fed with ogg packets.
 323  ****************************************************************************/
 324 static block_t *DecodeBlock( decoder_t *p_dec, block_t **pp_block )
 325 {
 326     decoder_sys_t *p_sys = p_dec->p_sys;
 327     ogg_packet oggpacket;
 328
 329     block_t *block = *pp_block;
 330
 331     if( block != NULL )
 332     {
 333         if( block->i_flags & (BLOCK_FLAG_CORRUPTED|BLOCK_FLAG_DISCONTINUITY) )
 334         {
 335             Flush( p_dec );
 336             if( block->i_flags & BLOCK_FLAG_CORRUPTED )
 337             {
 338                 block_Release( block );
 339                 *pp_block = NULL;
 340                 return NULL;
 341             }
 342         }
 343         /* Block to Ogg packet */
 344         oggpacket.packet = block->p_buffer;
 345         oggpacket.bytes = block->i_buffer;
 346     }
 347     else
 348     {
 349         if( p_sys->b_packetizer ) return NULL;
 350
 351         /* Block to Ogg packet */
 352         oggpacket.packet = NULL;
 353         oggpacket.bytes = 0;
 354     }
 355
 356     oggpacket.granulepos = -1;
 357     oggpacket.b_o_s = 0;
 358     oggpacket.e_o_s = 0;
 359     oggpacket.packetno = 0;
 360
 361     /* Check for headers */
 362     if( !p_sys->b_has_headers )
 363     {
 364         if( !p_dec->fmt_in.p_extra )
 365         {
 366             msg_Warn( p_dec, "Header missing, using default settings" );
 367
 368             if( CreateDefaultHeader( p_dec ) )
 369             {
 370                 if( block != NULL )
 371                     block_Release( block );
 372                 return NULL;
 373             }
 374         }
 375         else if( ProcessHeaders( p_dec ) )
 376         {
 377             if( block != NULL )
 378                 block_Release( block );
 379             return NULL;
 380         }
 381         p_sys->b_has_headers = true;
 382     }
 383
 384     return ProcessPacket( p_dec, &oggpacket, pp_block );
 385 }
 386
 387 static int DecodeAudio( decoder_t *p_dec, block_t *p_block )
 388 {
 389     if( p_block == NULL ) /* No Drain */
 390         return VLCDEC_SUCCESS;
 391
 392     block_t **pp_block = &p_block, *p_out;
 393     while( ( p_out = DecodeBlock( p_dec, pp_block ) ) != NULL )
 394         decoder_QueueAudio( p_dec, p_out );
 395     return VLCDEC_SUCCESS;
 396 }
 397
 398 static block_t *Packetize( decoder_t *p_dec, block_t **pp_block )
 399 {
 400     if( pp_block == NULL ) /* No Drain */
 401         return NULL;
 402     return DecodeBlock( p_dec, pp_block );
 403 }
 404
 405 /*****************************************************************************
 406  * ProcessHeaders: process Speex headers.
 407  *****************************************************************************/
 408 static int ProcessHeaders( decoder_t *p_dec )
 409 {
 410     decoder_sys_t *p_sys = p_dec->p_sys;
 411     ogg_packet oggpacket;
 412
 413     unsigned pi_size[XIPH_MAX_HEADER_COUNT];
 414     void     *pp_data[XIPH_MAX_HEADER_COUNT];
 415     unsigned i_count;
 416     if( xiph_SplitHeaders( pi_size, pp_data, &i_count,
 417                            p_dec->fmt_in.i_extra, p_dec->fmt_in.p_extra) )
 418         return VLC_EGENERIC;
 419     if( i_count < 2 )
 420         return VLC_EGENERIC;;
 421
 422     oggpacket.granulepos = -1;
 423     oggpacket.e_o_s = 0;
 424     oggpacket.packetno = 0;
 425
 426     /* Take care of the initial Vorbis header */
 427     oggpacket.b_o_s = 1; /* yes this actually is a b_o_s packet :) */
 428     oggpacket.bytes  = pi_size[0];
 429     oggpacket.packet = pp_data[0];
 430     if( ProcessInitialHeader( p_dec, &oggpacket ) != VLC_SUCCESS )
 431     {
 432         msg_Err( p_dec, "initial Speex header is corrupted" );
 433         return VLC_EGENERIC;;
 434     }
 435
 436     /* The next packet in order is the comments header */
 437     oggpacket.b_o_s = 0;
 438     oggpacket.bytes  = pi_size[1];
 439     oggpacket.packet = pp_data[1];
 440     ParseSpeexComments( p_dec, &oggpacket );
 441
 442     if( p_sys->b_packetizer )
 443     {
 444         void* p_extra = realloc( p_dec->fmt_out.p_extra,
 445                                  p_dec->fmt_in.i_extra );
 446         if( unlikely( p_extra == NULL ) )
 447         {
 448             return VLC_ENOMEM;
 449         }
 450         p_dec->fmt_out.p_extra = p_extra;
 451         p_dec->fmt_out.i_extra = p_dec->fmt_in.i_extra;
 452         memcpy( p_dec->fmt_out.p_extra,
 453                 p_dec->fmt_in.p_extra, p_dec->fmt_out.i_extra );
 454     }
 455
 456     return VLC_SUCCESS;
 457 }
 458
 459 /*****************************************************************************
 460  * ProcessInitialHeader: processes the inital Speex header packet.
 461  *****************************************************************************/
 462 static int ProcessInitialHeader( decoder_t *p_dec, ogg_packet *p_oggpacket )
 463 {
 464     decoder_sys_t *p_sys = p_dec->p_sys;
 465
 466     void *p_state;
 467     SpeexHeader *p_header;
 468     const SpeexMode *p_mode;
 469     SpeexCallback callback;
 470
 471     p_sys->p_header = p_header =
 472         speex_packet_to_header( (char *)p_oggpacket->packet,
 473                                 p_oggpacket->bytes );
 474     if( !p_header )
 475     {
 476         msg_Err( p_dec, "cannot read Speex header" );
 477         return VLC_EGENERIC;
 478     }
 479     if( p_header->mode >= SPEEX_NB_MODES || p_header->mode < 0 )
 480     {
 481         msg_Err( p_dec, "mode number %d does not (yet/any longer) exist in "
 482                  "this version of libspeex.", p_header->mode );
 483         return VLC_EGENERIC;
 484     }
 485
 486     p_mode = speex_mode_list[p_header->mode];
 487     if( p_mode == NULL )
 488         return VLC_EGENERIC;
 489
 490     if( p_header->speex_version_id > 1 )
 491     {
 492         msg_Err( p_dec, "this file was encoded with Speex bit-stream "
 493                  "version %d which is not supported by this decoder.",
 494                  p_header->speex_version_id );
 495         return VLC_EGENERIC;
 496     }
 497
 498     if( p_mode->bitstream_version < p_header->mode_bitstream_version )
 499     {
 500         msg_Err( p_dec, "file encoded with a newer version of Speex." );
 501         return VLC_EGENERIC;
 502     }
 503     if( p_mode->bitstream_version > p_header->mode_bitstream_version )
 504     {
 505         msg_Err( p_dec, "file encoded with an older version of Speex." );
 506         return VLC_EGENERIC;
 507     }
 508
 509     msg_Dbg( p_dec, "Speex %d Hz audio using %s mode %s%s",
 510              p_header->rate, p_mode->modeName,
 511              ( p_header->nb_channels == 1 ) ? " (mono" : " (stereo",
 512              p_header->vbr ? ", VBR)" : ")" );
 513
 514     /* Take care of speex decoder init */
 515     speex_bits_init( &p_sys->bits );
 516     p_sys->p_state = p_state = speex_decoder_init( p_mode );
 517     if( !p_state )
 518     {
 519         msg_Err( p_dec, "decoder initialization failed" );
 520         return VLC_EGENERIC;
 521     }
 522
 523     if( p_header->nb_channels == 2 )
 524     {
 525         SpeexStereoState stereo = SPEEX_STEREO_STATE_INIT;
 526         p_sys->stereo = stereo;
 527         callback.callback_id = SPEEX_INBAND_STEREO;
 528         callback.func = speex_std_stereo_request_handler;
 529         callback.data = &p_sys->stereo;
 530         speex_decoder_ctl( p_state, SPEEX_SET_HANDLER, &callback );
 531     }
 532     if( p_header->nb_channels <= 0 ||
 533         p_header->nb_channels > 5 )
 534     {
 535         msg_Err( p_dec, "invalid number of channels (not between 1 and 5): %i",
 536                  p_header->nb_channels );
 537         return VLC_EGENERIC;
 538     }
 539
 540     /* Setup the format */
 541     p_dec->fmt_out.audio.i_physical_channels =
 542         vlc_chan_maps[p_header->nb_channels];
 543     p_dec->fmt_out.audio.i_channels = p_header->nb_channels;
 544     p_dec->fmt_out.audio.i_rate = p_header->rate;
 545
 546     date_Init( &p_sys->end_date, p_header->rate, 1 );
 547
 548     return VLC_SUCCESS;
 549 }
 550
 551 /*****************************************************************************
 552  * Flush:
 553  *****************************************************************************/
 554 static void Flush( decoder_t *p_dec )
 555 {
 556     decoder_sys_t *p_sys = p_dec->p_sys;
 557
 558     date_Set( &p_sys->end_date, VLC_TICK_INVALID );
 559 }
 560
 561 /*****************************************************************************
 562  * ProcessPacket: processes a Speex packet.
 563  *****************************************************************************/
 564 static block_t *ProcessPacket( decoder_t *p_dec, ogg_packet *p_oggpacket,
 565                                block_t **pp_block )
 566 {
 567     decoder_sys_t *p_sys = p_dec->p_sys;
 568     block_t *p_block = *pp_block;
 569
 570     /* Date management */
 571     if( p_block && p_block->i_pts != VLC_TICK_INVALID &&
 572         p_block->i_pts != date_Get( &p_sys->end_date ) )
 573     {
 574         date_Set( &p_sys->end_date, p_block->i_pts );
 575     }
 576
 577     if( date_Get( &p_sys->end_date ) == VLC_TICK_INVALID )
 578     {
 579         /* We've just started the stream, wait for the first PTS. */
 580         if( p_block ) block_Release( p_block );
 581         return NULL;
 582     }
 583
 584     *pp_block = NULL; /* To avoid being fed the same packet again */
 585
 586     if( p_sys->b_packetizer )
 587     {
 588         if ( p_sys->p_header->frames_per_packet > 1 )
 589         {
 590             short *p_frame_holder = NULL;
 591             int i_bits_before = 0, i_bits_after = 0, i_bytes_in_speex_frame = 0,
 592                 i_pcm_output_size = 0, i_bits_in_speex_frame = 0;
 593             block_t *p_new_block = NULL;
 594
 595             i_pcm_output_size = p_sys->p_header->frame_size;
 596             p_frame_holder = (short*)xmalloc( sizeof(short)*i_pcm_output_size );
 597
 598             speex_bits_read_from( &p_sys->bits, (char*)p_oggpacket->packet,
 599                 p_oggpacket->bytes);
 600             i_bits_before = speex_bits_remaining( &p_sys->bits );
 601             speex_decode_int(p_sys->p_state, &p_sys->bits, p_frame_holder);
 602             i_bits_after = speex_bits_remaining( &p_sys->bits );
 603
 604             i_bits_in_speex_frame = i_bits_before - i_bits_after;
 605             i_bytes_in_speex_frame = ( i_bits_in_speex_frame +
 606                 (8 - (i_bits_in_speex_frame % 8)) )
 607                 / 8;
 608
 609             p_new_block = block_Alloc( i_bytes_in_speex_frame );
 610             memset( p_new_block->p_buffer, 0xff, i_bytes_in_speex_frame );
 611
 612             /*
 613              * Copy the first frame in this packet to a new packet.
 614              */
 615             speex_bits_rewind( &p_sys->bits );
 616             speex_bits_write( &p_sys->bits,
 617                 (char*)p_new_block->p_buffer,
 618                 (int)i_bytes_in_speex_frame );
 619
 620             /*
 621              * Move the remaining part of the original packet (subsequent
 622              * frames, if there are any) into the beginning
 623              * of the original packet so
 624              * they are preserved following the realloc.
 625              * Note: Any bits that
 626              * remain in the initial packet
 627              * are "filler" if they do not constitute
 628              * an entire byte.
 629              */
 630             if ( i_bits_after > 7 )
 631             {
 632                 /* round-down since we rounded-up earlier (to include
 633              * the speex terminator code.
 634              */
 635                 i_bytes_in_speex_frame--;
 636                 speex_bits_write( &p_sys->bits,
 637                     (char*)p_block->p_buffer,
 638                     p_block->i_buffer - i_bytes_in_speex_frame );
 639                 p_block = block_Realloc( p_block,
 640                     0,
 641                     p_block->i_buffer-i_bytes_in_speex_frame );
 642                 *pp_block = p_block;
 643             }
 644             else
 645             {
 646                 speex_bits_reset( &p_sys->bits );
 647             }
 648
 649             free( p_frame_holder );
 650             return SendPacket( p_dec, p_new_block);
 651         }
 652         else
 653         {
 654                 return SendPacket( p_dec, p_block );
 655         }
 656     }
 657     else
 658     {
 659         block_t *p_aout_buffer = DecodePacket( p_dec, p_oggpacket );
 660
 661         if( p_block )
 662             block_Release( p_block );
 663         return p_aout_buffer;
 664     }
 665 }
 666
 667 static int DecodeRtpSpeexPacket( decoder_t *p_dec, block_t *p_speex_bit_block )
 668 {
 669     decoder_sys_t *p_sys = p_dec->p_sys;
 670     block_t *p_aout_buffer;
 671     int i_decode_ret;
 672     unsigned int i_speex_frame_size;
 673
 674     if ( !p_speex_bit_block || p_speex_bit_block->i_pts == VLC_TICK_INVALID )
 675         return VLCDEC_SUCCESS;
 676
 677     /*
 678       If the SpeexBits buffer size is 0 (a default value),
 679       we know that a proper initialization has not yet been done.
 680     */
 681     if ( p_sys->bits.buf_size==0 )
 682     {
 683         p_sys->p_header = malloc(sizeof(SpeexHeader));
 684         if ( !p_sys->p_header )
 685         {
 686             msg_Err( p_dec, "Could not allocate a Speex header.");
 687             return VLCDEC_SUCCESS;
 688         }
 689
 690         const SpeexMode *mode = speex_lib_get_mode((p_sys->rtp_rate / 8000) >> 1);
 691
 692         speex_init_header( p_sys->p_header,p_sys->rtp_rate, 1, mode );
 693         speex_bits_init( &p_sys->bits );
 694         p_sys->p_state = speex_decoder_init( mode );
 695         if ( !p_sys->p_state )
 696         {
 697             msg_Err( p_dec, "Could not allocate a Speex decoder." );
 698             free( p_sys->p_header );
 699             return VLCDEC_SUCCESS;
 700         }
 701
 702         /*
 703           Assume that variable bit rate is enabled. Also assume
 704           that there is only one frame per packet.
 705         */
 706         p_sys->p_header->vbr = 1;
 707         p_sys->p_header->frames_per_packet = 1;
 708
 709         p_dec->fmt_out.audio.i_channels = p_sys->p_header->nb_channels;
 710         p_dec->fmt_out.audio.i_physical_channels =
 711             vlc_chan_maps[p_sys->p_header->nb_channels];
 712         p_dec->fmt_out.audio.i_rate = p_sys->p_header->rate;
 713
 714         if ( speex_mode_query( &speex_nb_mode,
 715                                SPEEX_MODE_FRAME_SIZE,
 716                                &i_speex_frame_size ) )
 717         {
 718             msg_Err( p_dec, "Could not determine the frame size." );
 719             speex_decoder_destroy( p_sys->p_state );
 720             free( p_sys->p_header );
 721             return VLCDEC_SUCCESS;
 722         }
 723         p_dec->fmt_out.audio.i_bytes_per_frame = i_speex_frame_size;
 724
 725         date_Init(&p_sys->end_date, p_sys->p_header->rate, 1);
 726     }
 727
 728     /*
 729       If the SpeexBits are initialized but there is
 730       still no header, an error must be thrown.
 731     */
 732     if ( !p_sys->p_header )
 733     {
 734         msg_Err( p_dec, "There is no valid Speex header found." );
 735         return VLCDEC_SUCCESS;
 736     }
 737
 738     if ( date_Get( &p_sys->end_date ) == VLC_TICK_INVALID )
 739         date_Set( &p_sys->end_date, p_speex_bit_block->i_dts );
 740
 741     /*
 742       Ask for a new audio output buffer and make sure
 743       we get one.
 744     */
 745     if( decoder_UpdateAudioFormat( p_dec ) )
 746         p_aout_buffer = NULL;
 747     else
 748         p_aout_buffer = decoder_NewAudioBuffer( p_dec,
 749             p_sys->p_header->frame_size );
 750     if ( !p_aout_buffer || p_aout_buffer->i_buffer == 0 )
 751     {
 752         msg_Err(p_dec, "Oops: No new buffer was returned!");
 753         return VLCDEC_SUCCESS;
 754     }
 755
 756     /*
 757       Read the Speex payload into the SpeexBits buffer.
 758     */
 759     speex_bits_read_from( &p_sys->bits,
 760         (char*)p_speex_bit_block->p_buffer,
 761         p_speex_bit_block->i_buffer );
 762
 763     /*
 764       Decode the input and ensure that no errors
 765       were encountered.
 766     */
 767     i_decode_ret = speex_decode_int( p_sys->p_state, &p_sys->bits,
 768             (int16_t*)p_aout_buffer->p_buffer );
 769     if ( i_decode_ret < 0 )
 770     {
 771         msg_Err( p_dec, "Decoding failed. Perhaps we have a bad stream?" );
 772         return VLCDEC_SUCCESS;
 773     }
 774
 775     /*
 776       Handle date management on the audio output buffer.
 777     */
 778     p_aout_buffer->i_pts = date_Get( &p_sys->end_date );
 779     p_aout_buffer->i_length = date_Increment( &p_sys->end_date,
 780         p_sys->p_header->frame_size ) - p_aout_buffer->i_pts;
 781
 782
 783     p_sys->i_frame_in_packet++;
 784     block_Release( p_speex_bit_block );
 785     decoder_QueueAudio( p_dec, p_aout_buffer );
 786     return VLCDEC_SUCCESS;
 787 }
 788
 789 /*****************************************************************************
 790  * DecodePacket: decodes a Speex packet.
 791  *****************************************************************************/
 792 static block_t *DecodePacket( decoder_t *p_dec, ogg_packet *p_oggpacket )
 793 {
 794     decoder_sys_t *p_sys = p_dec->p_sys;
 795
 796     if( p_oggpacket->bytes )
 797     {
 798         /* Copy Ogg packet to Speex bitstream */
 799         speex_bits_read_from( &p_sys->bits, (char *)p_oggpacket->packet,
 800                               p_oggpacket->bytes );
 801         p_sys->i_frame_in_packet = 0;
 802     }
 803
 804     /* Decode one frame at a time */
 805     if( p_sys->i_frame_in_packet < p_sys->p_header->frames_per_packet )
 806     {
 807         block_t *p_aout_buffer;
 808         if( p_sys->p_header->frame_size == 0 )
 809             return NULL;
 810
 811         if( decoder_UpdateAudioFormat( p_dec ) )
 812             return NULL;
 813         p_aout_buffer =
 814             decoder_NewAudioBuffer( p_dec, p_sys->p_header->frame_size );
 815         if( !p_aout_buffer )
 816         {
 817             return NULL;
 818         }
 819
 820         switch( speex_decode_int( p_sys->p_state, &p_sys->bits,
 821                                   (int16_t *)p_aout_buffer->p_buffer ) )
 822         {
 823             case -2:
 824                 msg_Err( p_dec, "decoding error: corrupted stream?" );
 825             case -1: /* End of stream */
 826                 return NULL;
 827         }
 828
 829         if( speex_bits_remaining( &p_sys->bits ) < 0 )
 830         {
 831             msg_Err( p_dec, "decoding overflow: corrupted stream?" );
 832         }
 833
 834         if( p_sys->p_header->nb_channels == 2 )
 835             speex_decode_stereo_int( (int16_t *)p_aout_buffer->p_buffer,
 836                                      p_sys->p_header->frame_size,
 837                                      &p_sys->stereo );
 838
 839         /* Date management */
 840         p_aout_buffer->i_pts = date_Get( &p_sys->end_date );
 841         p_aout_buffer->i_length =
 842             date_Increment( &p_sys->end_date, p_sys->p_header->frame_size )
 843             - p_aout_buffer->i_pts;
 844
 845         p_sys->i_frame_in_packet++;
 846
 847         return p_aout_buffer;
 848     }
 849     else
 850     {
 851         return NULL;
 852     }
 853 }
 854
 855 /*****************************************************************************
 856  * SendPacket: send an ogg packet to the stream output.
 857  *****************************************************************************/
 858 static block_t *SendPacket( decoder_t *p_dec, block_t *p_block )
 859 {
 860     decoder_sys_t *p_sys = p_dec->p_sys;
 861
 862     /* Date management */
 863     p_block->i_dts = p_block->i_pts = date_Get( &p_sys->end_date );
 864
 865     p_block->i_length =
 866         date_Increment( &p_sys->end_date,
 867                             p_sys->p_header->frame_size ) -
 868         p_block->i_pts;
 869
 870     return p_block;
 871 }
 872
 873 /*****************************************************************************
 874  * ParseSpeexComments:
 875  *****************************************************************************/
 876
 877 static void ParseSpeexComments( decoder_t *p_dec, ogg_packet *p_oggpacket )
 878 {
 879     decoder_sys_t *p_sys = p_dec->p_sys;
 880     const SpeexMode *p_mode;
 881
 882     assert( p_sys->p_header->mode < SPEEX_NB_MODES );
 883
 884     p_mode = speex_mode_list[p_sys->p_header->mode];
 885     assert( p_mode != NULL );
 886
 887     if( !p_dec->p_description )
 888     {
 889         p_dec->p_description = vlc_meta_New();
 890         if( !p_dec->p_description )
 891             return;
 892     }
 893
 894     /* */
 895     char *psz_mode;
 896     if( asprintf( &psz_mode, "%s%s", p_mode->modeName, p_sys->p_header->vbr ? " VBR" : "" ) >= 0 )
 897     {
 898         vlc_meta_AddExtra( p_dec->p_description, _("Mode"), psz_mode );
 899         free( psz_mode );
 900     }
 901
 902     /* TODO: finish comments parsing */
 903     VLC_UNUSED( p_oggpacket );
 904 }
 905
 906 /*****************************************************************************
 907  * CloseDecoder: speex decoder destruction
 908  *****************************************************************************/
 909 static void CloseDecoder( vlc_object_t *p_this )
 910 {
 911     decoder_t * p_dec = (decoder_t *)p_this;
 912     decoder_sys_t *p_sys = p_dec->p_sys;
 913
 914     if( p_sys->p_state )
 915     {
 916         speex_decoder_destroy( p_sys->p_state );
 917         speex_bits_destroy( &p_sys->bits );
 918     }
 919
 920     free( p_sys->p_header );
 921     free( p_sys );
 922 }
 923
 924 #ifdef ENABLE_SOUT
 925 /*****************************************************************************
 926  * encoder_sys_t: encoder descriptor
 927  *****************************************************************************/
 928 #define MAX_FRAME_BYTES 2000
 929
 930 typedef struct
 931 {
 932     /*
 933      * Input properties
 934      */
 935     char *p_buffer;
 936     char p_buffer_out[MAX_FRAME_BYTES];
 937
 938     /*
 939      * Speex properties
 940      */
 941     SpeexBits bits;
 942     SpeexHeader header;
 943     SpeexStereoState stereo;
 944     void *p_state;
 945
 946     int i_frames_per_packet;
 947     int i_frames_in_packet;
 948
 949     int i_frame_length;
 950     int i_samples_delay;
 951     int i_frame_size;
 952 } encoder_sys_t;
 953
 954 static block_t *Encode   ( encoder_t *, block_t * );
 955
 956 /*****************************************************************************
 957  * OpenEncoder: probe the encoder and return score
 958  *****************************************************************************/
 959 static int OpenEncoder( vlc_object_t *p_this )
 960 {
 961     encoder_t *p_enc = (encoder_t *)p_this;
 962     encoder_sys_t *p_sys;
 963     const SpeexMode *p_speex_mode = &speex_nb_mode;
 964     int i_tmp, i;
 965     const char *pp_header[2];
 966     int pi_header[2];
 967     uint8_t *p_extra;
 968
 969     if( p_enc->fmt_out.i_codec != VLC_CODEC_SPEEX &&
 970         !p_enc->obj.force )
 971     {
 972         return VLC_EGENERIC;
 973     }
 974
 975     config_ChainParse( p_enc, ENC_CFG_PREFIX, ppsz_enc_options, p_enc->p_cfg );
 976     switch( var_GetInteger( p_enc, ENC_CFG_PREFIX "mode" ) )
 977     {
 978     case 1:
 979         msg_Dbg( p_enc, "Using wideband" );
 980         p_speex_mode = &speex_wb_mode;
 981         break;
 982     case 2:
 983         msg_Dbg( p_enc, "Using ultra-wideband" );
 984         p_speex_mode = &speex_uwb_mode;
 985         break;
 986     default:
 987         msg_Dbg( p_enc, "Using narrowband" );
 988         p_speex_mode = &speex_nb_mode;
 989         break;
 990     }
 991
 992     /* Allocate the memory needed to store the decoder's structure */
 993     if( ( p_sys = (encoder_sys_t *)malloc(sizeof(encoder_sys_t)) ) == NULL )
 994         return VLC_ENOMEM;
 995     p_enc->p_sys = p_sys;
 996     p_enc->pf_encode_audio = Encode;
 997     p_enc->fmt_in.i_codec = VLC_CODEC_S16N;
 998     p_enc->fmt_out.i_codec = VLC_CODEC_SPEEX;
 999
1000     speex_init_header( &p_sys->header, p_enc->fmt_in.audio.i_rate,
1001                        1, p_speex_mode );
1002
1003     p_sys->header.frames_per_packet = 1;
1004     p_sys->header.vbr = var_GetBool( p_enc, ENC_CFG_PREFIX "cbr" ) ? 0 : 1;
1005     p_sys->header.nb_channels = p_enc->fmt_in.audio.i_channels;
1006
1007     /* Create a new encoder state in narrowband mode */
1008     p_sys->p_state = speex_encoder_init( p_speex_mode );
1009
1010     /* Parameters */
1011     i_tmp = var_GetInteger( p_enc, ENC_CFG_PREFIX "complexity" );
1012     speex_encoder_ctl( p_sys->p_state, SPEEX_SET_COMPLEXITY, &i_tmp );
1013
1014     i_tmp = var_GetBool( p_enc, ENC_CFG_PREFIX "cbr" ) ? 0 : 1;
1015     speex_encoder_ctl( p_sys->p_state, SPEEX_SET_VBR, &i_tmp );
1016
1017     if( i_tmp == 0 ) /* CBR */
1018     {
1019         i_tmp = var_GetFloat( p_enc, ENC_CFG_PREFIX "quality" );
1020         speex_encoder_ctl( p_sys->p_state, SPEEX_SET_QUALITY, &i_tmp );
1021
1022         i_tmp = var_GetBool( p_enc, ENC_CFG_PREFIX "vad" ) ? 1 : 0;
1023         speex_encoder_ctl( p_sys->p_state, SPEEX_SET_VAD, &i_tmp );
1024     }
1025     else
1026     {
1027         float f_tmp;
1028
1029         f_tmp = var_GetFloat( p_enc, ENC_CFG_PREFIX "quality" );
1030         speex_encoder_ctl( p_sys->p_state, SPEEX_SET_VBR_QUALITY, &f_tmp );
1031
1032         i_tmp = var_GetInteger( p_enc, ENC_CFG_PREFIX "max-bitrate" );
1033         if( i_tmp > 0 )
1034 #ifdef SPEEX_SET_VBR_MAX_BITRATE
1035             speex_encoder_ctl( p_sys->p_state, SPEEX_SET_VBR_MAX_BITRATE, &i_tmp );
1036 #else
1037             msg_Dbg( p_enc, "max-bitrate cannot be set in this version of libspeex");
1038 #endif
1039     }
1040
1041     i_tmp = var_GetBool( p_enc, ENC_CFG_PREFIX "dtx" ) ? 1 : 0;
1042     speex_encoder_ctl( p_sys->p_state, SPEEX_SET_DTX, &i_tmp );
1043
1044
1045     /*Initialization of the structure that holds the bits*/
1046     speex_bits_init( &p_sys->bits );
1047
1048     p_sys->i_frames_in_packet = 0;
1049     p_sys->i_samples_delay = 0;
1050
1051     speex_encoder_ctl( p_sys->p_state, SPEEX_GET_FRAME_SIZE,
1052                        &p_sys->i_frame_length );
1053
1054     p_sys->i_frame_size = p_sys->i_frame_length *
1055         sizeof(int16_t) * p_enc->fmt_in.audio.i_channels;
1056     p_sys->p_buffer = xmalloc( p_sys->i_frame_size );
1057
1058     /* Create and store headers */
1059     pp_header[0] = speex_header_to_packet( &p_sys->header, &pi_header[0] );
1060     pp_header[1] = "ENCODER=VLC media player";
1061     pi_header[1] = sizeof("ENCODER=VLC media player");
1062
1063     p_enc->fmt_out.i_extra = 3 * 2 + pi_header[0] + pi_header[1];
1064     p_extra = p_enc->fmt_out.p_extra = xmalloc( p_enc->fmt_out.i_extra );
1065     for( i = 0; i < 2; i++ )
1066     {
1067         *(p_extra++) = pi_header[i] >> 8;
1068         *(p_extra++) = pi_header[i] & 0xFF;
1069         memcpy( p_extra, pp_header[i], pi_header[i] );
1070         p_extra += pi_header[i];
1071     }
1072
1073     msg_Dbg( p_enc, "encoding: frame size:%d, channels:%d, samplerate:%d",
1074              p_sys->i_frame_size, p_enc->fmt_in.audio.i_channels,
1075              p_enc->fmt_in.audio.i_rate );
1076
1077     return VLC_SUCCESS;
1078 }
1079
1080 /****************************************************************************
1081  * Encode: the whole thing
1082  ****************************************************************************
1083  * This function spits out ogg packets.
1084  ****************************************************************************/
1085 static block_t *Encode( encoder_t *p_enc, block_t *p_aout_buf )
1086 {
1087     encoder_sys_t *p_sys = p_enc->p_sys;
1088     block_t *p_block, *p_chain = NULL;
1089
1090     /* Encoder gets NULL when it's time to flush */
1091     if( unlikely( !p_aout_buf ) ) return NULL;
1092
1093     unsigned char *p_buffer = p_aout_buf->p_buffer;
1094     unsigned i_samples = p_aout_buf->i_nb_samples;
1095     int i_samples_delay = p_sys->i_samples_delay;
1096
1097     vlc_tick_t i_pts = p_aout_buf->i_pts -
1098                 vlc_tick_from_samples( p_sys->i_samples_delay,
1099                             p_enc->fmt_in.audio.i_rate );
1100
1101     p_sys->i_samples_delay += i_samples;
1102
1103     while( p_sys->i_samples_delay >= p_sys->i_frame_length )
1104     {
1105         int16_t *p_samples;
1106         int i_out;
1107
1108         if( i_samples_delay )
1109         {
1110             /* Take care of the left-over from last time */
1111             int i_delay_size = i_samples_delay * 2 *
1112                                  p_enc->fmt_in.audio.i_channels;
1113             int i_size = p_sys->i_frame_size - i_delay_size;
1114
1115             p_samples = (int16_t *)p_sys->p_buffer;
1116             memcpy( p_sys->p_buffer + i_delay_size, p_buffer, i_size );
1117             p_buffer -= i_delay_size;
1118             i_samples += i_samples_delay;
1119             i_samples_delay = 0;
1120         }
1121         else
1122         {
1123             p_samples = (int16_t *)p_buffer;
1124         }
1125
1126         /* Encode current frame */
1127         if( p_enc->fmt_in.audio.i_channels == 2 )
1128             speex_encode_stereo_int( p_samples, p_sys->i_frame_length,
1129                                      &p_sys->bits );
1130
1131 #if 0
1132         if( p_sys->preprocess )
1133             speex_preprocess( p_sys->preprocess, p_samples, NULL );
1134 #endif
1135
1136         speex_encode_int( p_sys->p_state, p_samples, &p_sys->bits );
1137
1138         p_buffer += p_sys->i_frame_size;
1139         p_sys->i_samples_delay -= p_sys->i_frame_length;
1140         i_samples -= p_sys->i_frame_length;
1141
1142         p_sys->i_frames_in_packet++;
1143
1144         if( p_sys->i_frames_in_packet < p_sys->header.frames_per_packet )
1145             continue;
1146
1147         p_sys->i_frames_in_packet = 0;
1148
1149         speex_bits_insert_terminator( &p_sys->bits );
1150         i_out = speex_bits_write( &p_sys->bits, p_sys->p_buffer_out,
1151                                   MAX_FRAME_BYTES );
1152         speex_bits_reset( &p_sys->bits );
1153
1154         p_block = block_Alloc( i_out );
1155         memcpy( p_block->p_buffer, p_sys->p_buffer_out, i_out );
1156
1157         p_block->i_length = vlc_tick_from_samples(
1158             p_sys->i_frame_length * p_sys->header.frames_per_packet,
1159             p_enc->fmt_in.audio.i_rate );
1160
1161         p_block->i_dts = p_block->i_pts = i_pts;
1162
1163         /* Update pts */
1164         i_pts += p_block->i_length;
1165         block_ChainAppend( &p_chain, p_block );
1166
1167     }
1168
1169     /* Backup the remaining raw samples */
1170     if( i_samples )
1171     {
1172         memcpy( p_sys->p_buffer + i_samples_delay * 2 *
1173                 p_enc->fmt_in.audio.i_channels, p_buffer,
1174                 i_samples * 2 * p_enc->fmt_in.audio.i_channels );
1175     }
1176
1177     return p_chain;
1178 }
1179
1180 /*****************************************************************************
1181  * CloseEncoder: encoder destruction
1182  *****************************************************************************/
1183 static void CloseEncoder( vlc_object_t *p_this )
1184 {
1185     encoder_t *p_enc = (encoder_t *)p_this;
1186     encoder_sys_t *p_sys = p_enc->p_sys;
1187
1188     speex_encoder_destroy( p_sys->p_state );
1189     speex_bits_destroy( &p_sys->bits );
1190
1191     free( p_sys->p_buffer );
1192     free( p_sys );
1193 }
1194 #endif