include/vlc_codecs.h

   1 /*****************************************************************************
   2  * codecs.h: codec related structures needed by the demuxers and decoders
   3  *****************************************************************************
   4  * Copyright (C) 1999-2001 the VideoLAN team
   5  * $Id$
   6  *
   7  * Author: Gildas Bazin <gbazin@videolan.org>
   8  *
   9  * This program is free software; you can redistribute it and/or modify
  10  * it under the terms of the GNU General Public License as published by
  11  * the Free Software Foundation; either version 2 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 General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU General Public License
  20  * along with this program; if not, write to the Free Software
  21  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  22  *****************************************************************************/
  23
  24 #ifndef _VLC_CODECS_H
  25 #define _VLC_CODECS_H 1
  26
  27 #ifdef HAVE_ATTRIBUTE_PACKED
  28 #   define ATTR_PACKED __attribute__((__packed__))
  29 #else
  30 #   error FIXME
  31 #endif
  32
  33 /* Structures exported to the demuxers and decoders */
  34
  35 #if !(defined _GUID_DEFINED || defined GUID_DEFINED)
  36 #define GUID_DEFINED
  37 typedef struct _GUID
  38 {
  39     uint32_t Data1;
  40     uint16_t Data2;
  41     uint16_t Data3;
  42     uint8_t  Data4[8];
  43 } GUID, *REFGUID, *LPGUID;
  44 #endif /* GUID_DEFINED */
  45
  46 #ifndef _WAVEFORMATEX_
  47 #define _WAVEFORMATEX_
  48 typedef struct
  49 ATTR_PACKED
  50 _WAVEFORMATEX {
  51     uint16_t   wFormatTag;
  52     uint16_t   nChannels;
  53     uint32_t   nSamplesPerSec;
  54     uint32_t   nAvgBytesPerSec;
  55     uint16_t   nBlockAlign;
  56     uint16_t   wBitsPerSample;
  57     uint16_t   cbSize;
  58 } WAVEFORMATEX, *PWAVEFORMATEX, *NPWAVEFORMATEX, *LPWAVEFORMATEX;
  59 #endif /* _WAVEFORMATEX_ */
  60
  61 #ifndef _WAVEFORMATEXTENSIBLE_
  62 #define _WAVEFORMATEXTENSIBLE_
  63 typedef struct
  64 ATTR_PACKED
  65 _WAVEFORMATEXTENSIBLE {
  66     WAVEFORMATEX Format;
  67     union {
  68         uint16_t wValidBitsPerSample;
  69         uint16_t wSamplesPerBlock;
  70         uint16_t wReserved;
  71     } Samples;
  72     uint32_t     dwChannelMask;
  73     GUID SubFormat;
  74 } WAVEFORMATEXTENSIBLE, *PWAVEFORMATEXTENSIBLE;
  75 #endif /* _WAVEFORMATEXTENSIBLE_ */
  76
  77 #ifndef _WAVEHEADER_
  78 #define _WAVEHEADER_
  79 typedef struct
  80 ATTR_PACKED
  81 _WAVEHEADER {
  82     uint32_t MainChunkID;
  83     uint32_t Length;
  84     uint32_t ChunkTypeID;
  85     uint32_t SubChunkID;
  86     uint32_t SubChunkLength;
  87     uint16_t Format;
  88     uint16_t Modus;
  89     uint32_t SampleFreq;
  90     uint32_t BytesPerSec;
  91     uint16_t BytesPerSample;
  92     uint16_t BitsPerSample;
  93     uint32_t DataChunkID;
  94     uint32_t DataLength;
  95 } WAVEHEADER;
  96 #endif /* _WAVEHEADER_ */
  97
  98 #if !defined(_BITMAPINFOHEADER_) && !defined(WIN32)
  99 #define _BITMAPINFOHEADER_
 100 typedef struct
 101 ATTR_PACKED
 102 {
 103     uint32_t   biSize;
 104     uint32_t   biWidth;
 105     uint32_t   biHeight;
 106     uint16_t   biPlanes;
 107     uint16_t   biBitCount;
 108     uint32_t   biCompression;
 109     uint32_t   biSizeImage;
 110     uint32_t   biXPelsPerMeter;
 111     uint32_t   biYPelsPerMeter;
 112     uint32_t   biClrUsed;
 113     uint32_t   biClrImportant;
 114 } BITMAPINFOHEADER, *PBITMAPINFOHEADER, *LPBITMAPINFOHEADER;
 115
 116 typedef struct
 117 ATTR_PACKED
 118 {
 119     BITMAPINFOHEADER bmiHeader;
 120     int        bmiColors[1];
 121 } BITMAPINFO, *LPBITMAPINFO;
 122 #endif
 123
 124 #ifndef _RECT32_
 125 #define _RECT32_
 126 typedef struct
 127 ATTR_PACKED
 128 {
 129     int left, top, right, bottom;
 130 } RECT32;
 131 #endif
 132
 133 #ifndef _REFERENCE_TIME_
 134 #define _REFERENCE_TIME_
 135 typedef int64_t REFERENCE_TIME;
 136 #endif
 137
 138 #ifndef _VIDEOINFOHEADER_
 139 #define _VIDEOINFOHEADER_
 140 typedef struct
 141 ATTR_PACKED
 142 {
 143     RECT32            rcSource;
 144     RECT32            rcTarget;
 145     uint32_t          dwBitRate;
 146     uint32_t          dwBitErrorRate;
 147     REFERENCE_TIME    AvgTimePerFrame;
 148     BITMAPINFOHEADER  bmiHeader;
 149 } VIDEOINFOHEADER;
 150 #endif
 151
 152 #ifndef _RGBQUAD_
 153 #define _RGBQUAD_
 154 typedef struct
 155 ATTR_PACKED
 156 {
 157     uint8_t rgbBlue;
 158     uint8_t rgbGreen;
 159     uint8_t rgbRed;
 160     uint8_t rgbReserved;
 161 } RGBQUAD1;
 162 #endif
 163
 164 #ifndef _TRUECOLORINFO_
 165 #define _TRUECOLORINFO_
 166 typedef struct
 167 ATTR_PACKED
 168 {
 169     uint32_t dwBitMasks[3];
 170     RGBQUAD1 bmiColors[256];
 171 } TRUECOLORINFO;
 172 #endif
 173
 174 #ifndef _VIDEOINFO_
 175 #define _VIDEOINFO_
 176 typedef struct
 177 ATTR_PACKED
 178 {
 179     RECT32            rcSource;
 180     RECT32            rcTarget;
 181     uint32_t          dwBitRate;
 182     uint32_t          dwBitErrorRate;
 183     REFERENCE_TIME    AvgTimePerFrame;
 184     BITMAPINFOHEADER  bmiHeader;
 185
 186     union
 187     {
 188         RGBQUAD1 bmiColors[256]; /* Colour palette */
 189         uint32_t dwBitMasks[3]; /* True colour masks */
 190         TRUECOLORINFO TrueColorInfo; /* Both of the above */
 191     };
 192
 193 } VIDEOINFO;
 194 #endif
 195
 196 /* WAVE format wFormatTag IDs */
 197 #define WAVE_FORMAT_UNKNOWN             0x0000 /* Microsoft Corporation */
 198 #define WAVE_FORMAT_PCM                 0x0001 /* Microsoft Corporation */
 199 #define WAVE_FORMAT_ADPCM               0x0002 /* Microsoft Corporation */
 200 #define WAVE_FORMAT_IEEE_FLOAT          0x0003 /* Microsoft Corporation */
 201 #define WAVE_FORMAT_ALAW                0x0006 /* Microsoft Corporation */
 202 #define WAVE_FORMAT_MULAW               0x0007 /* Microsoft Corporation */
 203 #define WAVE_FORMAT_DTS_MS              0x0008 /* Microsoft Corporation */
 204 #define WAVE_FORMAT_WMAS                0x000a /* WMA 9 Speech */
 205 #define WAVE_FORMAT_IMA_ADPCM           0x0011 /* Intel Corporation */
 206 #define WAVE_FORMAT_GSM610              0x0031 /* Microsoft Corporation */
 207 #define WAVE_FORMAT_MSNAUDIO            0x0032 /* Microsoft Corporation */
 208 #define WAVE_FORMAT_G726                0x0045 /* ITU-T standard  */
 209 #define WAVE_FORMAT_MPEG                0x0050 /* Microsoft Corporation */
 210 #define WAVE_FORMAT_MPEGLAYER3          0x0055 /* ISO/MPEG Layer3 Format Tag */
 211 #define WAVE_FORMAT_DOLBY_AC3_SPDIF     0x0092 /* Sonic Foundry */
 212
 213 #define WAVE_FORMAT_A52                 0x2000
 214 #define WAVE_FORMAT_DTS                 0x2001
 215 #define WAVE_FORMAT_WMA1                0x0160 /* WMA version 1 */
 216 #define WAVE_FORMAT_WMA2                0x0161 /* WMA (v2) 7, 8, 9 Series */
 217 #define WAVE_FORMAT_WMAP                0x0162 /* WMA 9 Professional */
 218 #define WAVE_FORMAT_WMAL                0x0163 /* WMA 9 Lossless */
 219 #define WAVE_FORMAT_DIVIO_AAC           0x4143
 220 #define WAVE_FORMAT_AAC                 0x00FF
 221 #define WAVE_FORMAT_FFMPEG_AAC          0x706D
 222
 223 /* Need to check these */
 224 #define WAVE_FORMAT_DK3                 0x0061
 225 #define WAVE_FORMAT_DK4                 0x0062
 226
 227 /* At least FFmpeg use that ID: from libavformat/riff.c ('Vo' == 0x566f)
 228  * { CODEC_ID_VORBIS, ('V'<<8)+'o' }, //HACK/FIXME, does vorbis in WAV/AVI have an (in)official id?
 229  */
 230 #define WAVE_FORMAT_VORBIS              0x566f
 231
 232 /* It seems that these IDs are used by braindead & obsolete VorbisACM encoder
 233  * (Windows only)
 234  * A few info is available except VorbisACM source (remember, Windows only)
 235  * (available on http://svn.xiph.org), but it seems that vo3+ at least is
 236  * made of Vorbis data encapsulated in Ogg container...
 237  */
 238 #define WAVE_FORMAT_VORB_1              0x674f
 239 #define WAVE_FORMAT_VORB_2              0x6750
 240 #define WAVE_FORMAT_VORB_3              0x6751
 241 #define WAVE_FORMAT_VORB_1PLUS          0x676f
 242 #define WAVE_FORMAT_VORB_2PLUS          0x6770
 243 #define WAVE_FORMAT_VORB_3PLUS          0x6771
 244
 245 #define WAVE_FORMAT_SPEEX               0xa109 /* Speex audio */
 246
 247
 248 #if !defined(WAVE_FORMAT_EXTENSIBLE)
 249 #define WAVE_FORMAT_EXTENSIBLE          0xFFFE /* Microsoft */
 250 #endif
 251
 252 /* GUID SubFormat IDs */
 253 /* We need both b/c const variables are not compile-time constants in C, giving
 254  * us an error if we use the const GUID in an enum */
 255
 256 #ifndef _KSDATAFORMAT_SUBTYPE_PCM_
 257 #define _KSDATAFORMAT_SUBTYPE_PCM_ {0x00000001, 0x0000, 0x0010, {0x80, 0x00, 0x00, 0xaa, 0x00, 0x38, 0x9b, 0x71}}
 258 static const GUID VLC_KSDATAFORMAT_SUBTYPE_PCM = {0xE923AABF, 0xCB58, 0x4471, {0xA1, 0x19, 0xFF, 0xFA, 0x01, 0xE4, 0xCE, 0x62}};
 259 #define KSDATAFORMAT_SUBTYPE_PCM VLC_KSDATAFORMAT_SUBTYPE_PCM
 260 #endif
 261
 262 #ifndef _KSDATAFORMAT_SUBTYPE_UNKNOWN_
 263 #define _KSDATAFORMAT_SUBTYPE_UNKNOWN_ {0x00000000, 0x0000, 0x0000, {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}}
 264 static const GUID VLC_KSDATAFORMAT_SUBTYPE_UNKNOWN = {0x00000000, 0x0000, 0x0000, {0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00}};
 265 #define KSDATAFORMAT_SUBTYPE_UNKNOWN VLC_KSDATAFORMAT_SUBTYPE_UNKNOWN
 266 #endif
 267
 268 /* Microsoft speaker definitions */
 269 #define WAVE_SPEAKER_FRONT_LEFT             0x1
 270 #define WAVE_SPEAKER_FRONT_RIGHT            0x2
 271 #define WAVE_SPEAKER_FRONT_CENTER           0x4
 272 #define WAVE_SPEAKER_LOW_FREQUENCY          0x8
 273 #define WAVE_SPEAKER_BACK_LEFT              0x10
 274 #define WAVE_SPEAKER_BACK_RIGHT             0x20
 275 #define WAVE_SPEAKER_FRONT_LEFT_OF_CENTER   0x40
 276 #define WAVE_SPEAKER_FRONT_RIGHT_OF_CENTER  0x80
 277 #define WAVE_SPEAKER_BACK_CENTER            0x100
 278 #define WAVE_SPEAKER_SIDE_LEFT              0x200
 279 #define WAVE_SPEAKER_SIDE_RIGHT             0x400
 280 #define WAVE_SPEAKER_TOP_CENTER             0x800
 281 #define WAVE_SPEAKER_TOP_FRONT_LEFT         0x1000
 282 #define WAVE_SPEAKER_TOP_FRONT_CENTER       0x2000
 283 #define WAVE_SPEAKER_TOP_FRONT_RIGHT        0x4000
 284 #define WAVE_SPEAKER_TOP_BACK_LEFT          0x8000
 285 #define WAVE_SPEAKER_TOP_BACK_CENTER        0x10000
 286 #define WAVE_SPEAKER_TOP_BACK_RIGHT         0x20000
 287 #define WAVE_SPEAKER_RESERVED               0x80000000
 288
 289 static struct
 290 {
 291     uint16_t     i_tag;
 292     vlc_fourcc_t i_fourcc;
 293     const char  *psz_name;
 294 }
 295 wave_format_tag_to_fourcc[] =
 296 {
 297     { WAVE_FORMAT_PCM,        VLC_FOURCC( 'a', 'r', 'a', 'w' ), "Raw audio" },
 298     { WAVE_FORMAT_ADPCM,      VLC_FOURCC( 'm', 's', 0x00,0x02), "ADPCM" },
 299     { WAVE_FORMAT_IEEE_FLOAT, VLC_FOURCC( 'a', 'f', 'l', 't' ), "IEEE Float audio" },
 300     { WAVE_FORMAT_ALAW,       VLC_FOURCC( 'a', 'l', 'a', 'w' ), "A-Law" },
 301     { WAVE_FORMAT_MULAW,      VLC_FOURCC( 'm', 'l', 'a', 'w' ), "Mu-Law" },
 302     { WAVE_FORMAT_IMA_ADPCM,  VLC_FOURCC( 'm', 's', 0x00,0x11), "Ima-ADPCM" },
 303     { WAVE_FORMAT_G726,       VLC_FOURCC( 'g', '7', '2', '6' ), "G.726 ADPCM" },
 304     { WAVE_FORMAT_MPEGLAYER3, VLC_FOURCC( 'm', 'p', 'g', 'a' ), "Mpeg Audio" },
 305     { WAVE_FORMAT_MPEG,       VLC_FOURCC( 'm', 'p', 'g', 'a' ), "Mpeg Audio" },
 306     { WAVE_FORMAT_A52,        VLC_FOURCC( 'a', '5', '2', ' ' ), "A/52" },
 307     { WAVE_FORMAT_WMA1,       VLC_FOURCC( 'w', 'm', 'a', '1' ), "Window Media Audio v1" },
 308     { WAVE_FORMAT_WMA2,       VLC_FOURCC( 'w', 'm', 'a', '2' ), "Window Media Audio v2" },
 309     { WAVE_FORMAT_WMA2,       VLC_FOURCC( 'w', 'm', 'a', ' ' ), "Window Media Audio v2" },
 310     { WAVE_FORMAT_WMAP,       VLC_FOURCC( 'w', 'm', 'a', 'p' ), "Window Media Audio 9 Professional" },
 311     { WAVE_FORMAT_WMAL,       VLC_FOURCC( 'w', 'm', 'a', 'l' ), "Window Media Audio 9 Lossless" },
 312     { WAVE_FORMAT_WMAS,       VLC_FOURCC( 'w', 'm', 'a', 's' ), "Window Media Audio 9 Speech" },
 313     { WAVE_FORMAT_DK3,        VLC_FOURCC( 'm', 's', 0x00,0x61), "Duck DK3" },
 314     { WAVE_FORMAT_DK4,        VLC_FOURCC( 'm', 's', 0x00,0x62), "Duck DK4" },
 315     { WAVE_FORMAT_DTS,        VLC_FOURCC( 'd', 't', 's', ' ' ), "DTS Coherent Acoustics" },
 316     { WAVE_FORMAT_DTS_MS,     VLC_FOURCC( 'd', 't', 's', ' ' ), "DTS Coherent Acoustics" },
 317     { WAVE_FORMAT_DIVIO_AAC,  VLC_FOURCC( 'm', 'p', '4', 'a' ), "MPEG-4 Audio (Divio)" },
 318     { WAVE_FORMAT_AAC,        VLC_FOURCC( 'm', 'p', '4', 'a' ), "MPEG-4 Audio" },
 319     { WAVE_FORMAT_FFMPEG_AAC, VLC_FOURCC( 'm', 'p', '4', 'a' ), "MPEG-4 Audio" },
 320     { WAVE_FORMAT_VORBIS,     VLC_FOURCC( 'v', 'o', 'r', 'b' ), "Vorbis Audio" },
 321     { WAVE_FORMAT_VORB_1,     VLC_FOURCC( 'v', 'o', 'r', '1' ), "Vorbis 1 Audio" },
 322     { WAVE_FORMAT_VORB_1PLUS, VLC_FOURCC( 'v', 'o', '1', '+' ), "Vorbis 1+ Audio" },
 323     { WAVE_FORMAT_VORB_2,     VLC_FOURCC( 'v', 'o', 'r', '2' ), "Vorbis 2 Audio" },
 324     { WAVE_FORMAT_VORB_2PLUS, VLC_FOURCC( 'v', 'o', '2', '+' ), "Vorbis 2+ Audio" },
 325     { WAVE_FORMAT_VORB_3,     VLC_FOURCC( 'v', 'o', 'r', '3' ), "Vorbis 3 Audio" },
 326     { WAVE_FORMAT_VORB_3PLUS, VLC_FOURCC( 'v', 'o', '3', '+' ), "Vorbis 3+ Audio" },
 327     { WAVE_FORMAT_SPEEX,      VLC_FOURCC( 's', 'p', 'x', ' ' ), "Speex Audio" },
 328     { WAVE_FORMAT_UNKNOWN,    VLC_FOURCC( 'u', 'n', 'd', 'f' ), "Unknown" }
 329 };
 330
 331 static inline void wf_tag_to_fourcc( uint16_t i_tag, vlc_fourcc_t *fcc,
 332                                      const char **ppsz_name )
 333 {
 334     int i;
 335     for( i = 0; wave_format_tag_to_fourcc[i].i_tag != 0; i++ )
 336     {
 337         if( wave_format_tag_to_fourcc[i].i_tag == i_tag ) break;
 338     }
 339     if( fcc ) *fcc = wave_format_tag_to_fourcc[i].i_fourcc;
 340     if( ppsz_name ) *ppsz_name = wave_format_tag_to_fourcc[i].psz_name;
 341 }
 342
 343 static inline void fourcc_to_wf_tag( vlc_fourcc_t fcc, uint16_t *pi_tag )
 344 {
 345     int i;
 346     for( i = 0; wave_format_tag_to_fourcc[i].i_tag != 0; i++ )
 347     {
 348         if( wave_format_tag_to_fourcc[i].i_fourcc == fcc ) break;
 349     }
 350     if( pi_tag ) *pi_tag = wave_format_tag_to_fourcc[i].i_tag;
 351 }
 352
 353 /* If wFormatTag is WAVEFORMATEXTENSIBLE, we must look at the SubFormat tag
 354  * to determine the actual format.  Microsoft has stopped giving out wFormatTag
 355  * assignments in lieu of letting 3rd parties generate their own GUIDs
 356  */
 357 static struct
 358 {
 359     GUID         guid_tag;
 360     vlc_fourcc_t i_fourcc;
 361     const char  *psz_name;
 362 }
 363 sub_format_tag_to_fourcc[] =
 364 {
 365     { _KSDATAFORMAT_SUBTYPE_PCM_, VLC_FOURCC( 'p', 'c', 'm', ' ' ), "PCM" },
 366     { _KSDATAFORMAT_SUBTYPE_UNKNOWN_, VLC_FOURCC( 'u', 'n', 'd', 'f' ), "Unknown" }
 367 };
 368
 369 /* compares two GUIDs, returns 1 if identical, 0 otherwise */
 370 static inline int guidcmp( const GUID *s1, const GUID *s2 )
 371 {
 372     return( s1->Data1 == s2->Data1 && s1->Data2 == s2->Data2 &&
 373             s1->Data3 == s2->Data3 && !memcmp( s1->Data4, s2->Data4, 8 ) );
 374 }
 375
 376 static inline void sf_tag_to_fourcc( GUID *guid_tag,
 377                                      vlc_fourcc_t *fcc, const char **ppsz_name )
 378 {
 379     int i;
 380
 381     for( i = 0; !guidcmp( &sub_format_tag_to_fourcc[i].guid_tag,
 382                           &KSDATAFORMAT_SUBTYPE_UNKNOWN ); i++ )
 383     {
 384         if( guidcmp( &sub_format_tag_to_fourcc[i].guid_tag, guid_tag ) ) break;
 385     }
 386     if( fcc ) *fcc = sub_format_tag_to_fourcc[i].i_fourcc;
 387     if( ppsz_name ) *ppsz_name = sub_format_tag_to_fourcc[i].psz_name;
 388 }
 389
 390 /**
 391  * Structure to hold information concerning subtitles.
 392  * Used between demuxers and decoders of subtitles.
 393  */
 394 typedef struct es_sys_t
 395 {
 396     char               *psz_header; /* for 'ssa ' and 'subt' */
 397
 398     /* for spudec */
 399     unsigned int        i_orig_height;
 400     unsigned int        i_orig_width;
 401     unsigned int        i_origin_x;
 402     unsigned int        i_origin_y;
 403     unsigned int        i_scale_h;
 404     unsigned int        i_scale_v;
 405     unsigned int        i_alpha;
 406     bool          b_smooth;
 407     mtime_t             i_fade_in;
 408     mtime_t             i_fade_out;
 409     unsigned int        i_align;
 410     mtime_t             i_time_offset;
 411     bool          b_forced_subs;
 412     unsigned int        palette[16];
 413     unsigned int        colors[4];
 414
 415 } subtitle_data_t;
 416
 417 #endif /* "codecs.h" */