FFmpeg  4.1.4
isom.h
Go to the documentation of this file.
1 /*
2  * ISO Media common code
3  * copyright (c) 2001 Fabrice Bellard
4  * copyright (c) 2002 Francois Revol <revol@free.fr>
5  * copyright (c) 2006 Baptiste Coudurier <baptiste.coudurier@free.fr>
6  *
7  * This file is part of FFmpeg.
8  *
9  * FFmpeg is free software; you can redistribute it and/or
10  * modify it under the terms of the GNU Lesser General Public
11  * License as published by the Free Software Foundation; either
12  * version 2.1 of the License, or (at your option) any later version.
13  *
14  * FFmpeg 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 GNU
17  * Lesser General Public License for more details.
18  *
19  * You should have received a copy of the GNU Lesser General Public
20  * License along with FFmpeg; if not, write to the Free Software
21  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22  */
23 
24 #ifndef AVFORMAT_ISOM_H
25 #define AVFORMAT_ISOM_H
26 
27 #include <stddef.h>
28 #include <stdint.h>
29 
32 #include "libavutil/spherical.h"
33 #include "libavutil/stereo3d.h"
34 
35 #include "avio.h"
36 #include "internal.h"
37 #include "dv.h"
38 
39 /* isom.c */
40 extern const AVCodecTag ff_mp4_obj_type[];
41 extern const AVCodecTag ff_codec_movvideo_tags[];
42 extern const AVCodecTag ff_codec_movaudio_tags[];
44 extern const AVCodecTag ff_codec_movdata_tags[];
45 
46 int ff_mov_iso639_to_lang(const char lang[4], int mp4);
47 int ff_mov_lang_to_iso639(unsigned code, char to[4]);
48 
49 struct AVAESCTR;
50 
51 /* the QuickTime file format is quite convoluted...
52  * it has lots of index tables, each indexing something in another one...
53  * Here we just use what is needed to read the chunks
54  */
55 
56 typedef struct MOVStts {
57  unsigned int count;
58  int duration;
59 } MOVStts;
60 
61 typedef struct MOVStsc {
62  int first;
63  int count;
64  int id;
65 } MOVStsc;
66 
67 typedef struct MOVElst {
68  int64_t duration;
69  int64_t time;
70  float rate;
71 } MOVElst;
72 
73 typedef struct MOVDref {
74  uint32_t type;
75  char *path;
76  char *dir;
77  char volume[28];
78  char filename[64];
79  int16_t nlvl_to, nlvl_from;
80 } MOVDref;
81 
82 typedef struct MOVAtom {
83  uint32_t type;
84  int64_t size; /* total size (excluding the size and type fields) */
85 } MOVAtom;
86 
87 struct MOVParseTableEntry;
88 
89 typedef struct MOVFragment {
91  unsigned track_id;
92  uint64_t base_data_offset;
93  uint64_t moof_offset;
94  uint64_t implicit_offset;
95  unsigned stsd_id;
96  unsigned duration;
97  unsigned size;
98  unsigned flags;
99 } MOVFragment;
100 
101 typedef struct MOVTrackExt {
102  unsigned track_id;
103  unsigned stsd_id;
104  unsigned duration;
105  unsigned size;
106  unsigned flags;
107 } MOVTrackExt;
108 
109 typedef struct MOVSbgp {
110  unsigned int count;
111  unsigned int index;
112 } MOVSbgp;
113 
114 typedef struct MOVEncryptionIndex {
115  // Individual encrypted samples. If there are no elements, then the default
116  // settings will be used.
117  unsigned int nb_encrypted_samples;
119 
123  uint64_t* auxiliary_offsets; ///< Absolute seek position
126 
127 typedef struct MOVFragmentStreamInfo {
128  int id;
129  int64_t sidx_pts;
130  int64_t first_tfra_pts;
131  int64_t tfdt_dts;
135 
136 typedef struct MOVFragmentIndexItem {
137  int64_t moof_offset;
139  int current;
143 
144 typedef struct MOVFragmentIndex {
146  int complete;
147  int current;
148  int nb_items;
151 
152 typedef struct MOVIndexRange {
153  int64_t start;
154  int64_t end;
155 } MOVIndexRange;
156 
157 typedef struct MOVStreamContext {
160  int ffindex; ///< AVStream index
162  unsigned int chunk_count;
163  int64_t *chunk_offsets;
164  unsigned int stts_count;
166  unsigned int ctts_count;
167  unsigned int ctts_allocated_size;
169  unsigned int stsc_count;
171  unsigned int stsc_index;
173  unsigned int stps_count;
174  unsigned *stps_data; ///< partial sync sample for mpeg-2 open gop
176  unsigned int elst_count;
179  unsigned int sample_size; ///< may contain value calculated from stsd or value from stsz atom
180  unsigned int stsz_sample_size; ///< always contains sample size from stsz atom
181  unsigned int sample_count;
184  unsigned int keyframe_count;
185  int *keyframes;
187  int64_t time_offset; ///< time offset of the edit list entries
188  int64_t min_corrected_pts; ///< minimum Composition time shown by the edits excluding empty edits.
190  int64_t current_index;
193  unsigned int bytes_per_frame;
194  unsigned int samples_per_frame;
196  int pseudo_stream_id; ///< -1 means demux all ids
197  int16_t audio_cid; ///< stsd audio compression id
198  unsigned drefs_count;
200  int dref_id;
202  int width; ///< tkhd width
203  int height; ///< tkhd height
204  int dts_shift; ///< dts shift when ctts is negative
205  uint32_t palette[256];
207  int64_t data_size;
208  uint32_t tmcd_flags; ///< tmcd track flags
209  int64_t track_end; ///< used for dts generation in fragmented movie files
210  int start_pad; ///< amount of samples to skip due to enc-dec delay
211  unsigned int rap_group_count;
213 
216 
217  /** extradata array (and size) for multiple stsd */
223 
230  size_t coll_size;
231 
232  uint32_t format;
233 
234  int has_sidx; // If there is an sidx entry for this stream.
235  struct {
236  struct AVAESCTR* aes_ctr;
237  unsigned int per_sample_iv_size; // Either 0, 8, or 16.
240  } cenc;
242 
243 typedef struct MOVContext {
244  const AVClass *class; ///< class for private options
247  int64_t duration; ///< duration of the longest track
248  int found_moov; ///< 'moov' atom has been found
249  int found_mdat; ///< 'mdat' atom has been found
250  int found_hdlr_mdta; ///< 'hdlr' atom with type 'mdta' has been found
251  int trak_index; ///< Index of the current 'trak'
252  char **meta_keys;
253  unsigned meta_keys_count;
256  int isom; ///< 1 if file is ISO Media (mp4/3gp)
257  MOVFragment fragment; ///< current fragment in moof atom
259  unsigned trex_count;
260  int itunes_metadata; ///< metadata are itunes style
263  unsigned int nb_chapter_tracks;
269  int64_t next_root_atom; ///< offset of the next root atom
272  int *bitrates; ///< bitrates read before streams creation
279  unsigned int aax_mode; ///< 'aax' file has been detected
280  uint8_t file_key[20];
281  uint8_t file_iv[20];
290  int32_t movie_display_matrix[3][3]; ///< display matrix from mvhd
291 } MOVContext;
292 
296 void ff_mp4_parse_es_descr(AVIOContext *pb, int *es_id);
297 
298 #define MP4ODescrTag 0x01
299 #define MP4IODescrTag 0x02
300 #define MP4ESDescrTag 0x03
301 #define MP4DecConfigDescrTag 0x04
302 #define MP4DecSpecificDescrTag 0x05
303 #define MP4SLDescrTag 0x06
304 
305 #define MOV_TFHD_BASE_DATA_OFFSET 0x01
306 #define MOV_TFHD_STSD_ID 0x02
307 #define MOV_TFHD_DEFAULT_DURATION 0x08
308 #define MOV_TFHD_DEFAULT_SIZE 0x10
309 #define MOV_TFHD_DEFAULT_FLAGS 0x20
310 #define MOV_TFHD_DURATION_IS_EMPTY 0x010000
311 #define MOV_TFHD_DEFAULT_BASE_IS_MOOF 0x020000
312 
313 #define MOV_TRUN_DATA_OFFSET 0x01
314 #define MOV_TRUN_FIRST_SAMPLE_FLAGS 0x04
315 #define MOV_TRUN_SAMPLE_DURATION 0x100
316 #define MOV_TRUN_SAMPLE_SIZE 0x200
317 #define MOV_TRUN_SAMPLE_FLAGS 0x400
318 #define MOV_TRUN_SAMPLE_CTS 0x800
319 
320 #define MOV_FRAG_SAMPLE_FLAG_DEGRADATION_PRIORITY_MASK 0x0000ffff
321 #define MOV_FRAG_SAMPLE_FLAG_IS_NON_SYNC 0x00010000
322 #define MOV_FRAG_SAMPLE_FLAG_PADDING_MASK 0x000e0000
323 #define MOV_FRAG_SAMPLE_FLAG_REDUNDANCY_MASK 0x00300000
324 #define MOV_FRAG_SAMPLE_FLAG_DEPENDED_MASK 0x00c00000
325 #define MOV_FRAG_SAMPLE_FLAG_DEPENDS_MASK 0x03000000
326 
327 #define MOV_FRAG_SAMPLE_FLAG_DEPENDS_NO 0x02000000
328 #define MOV_FRAG_SAMPLE_FLAG_DEPENDS_YES 0x01000000
329 
330 #define MOV_TKHD_FLAG_ENABLED 0x0001
331 #define MOV_TKHD_FLAG_IN_MOVIE 0x0002
332 #define MOV_TKHD_FLAG_IN_PREVIEW 0x0004
333 #define MOV_TKHD_FLAG_IN_POSTER 0x0008
334 
335 #define MOV_SAMPLE_DEPENDENCY_UNKNOWN 0x0
336 #define MOV_SAMPLE_DEPENDENCY_YES 0x1
337 #define MOV_SAMPLE_DEPENDENCY_NO 0x2
338 
339 
340 #define TAG_IS_AVCI(tag) \
341  ((tag) == MKTAG('a', 'i', '5', 'p') || \
342  (tag) == MKTAG('a', 'i', '5', 'q') || \
343  (tag) == MKTAG('a', 'i', '5', '2') || \
344  (tag) == MKTAG('a', 'i', '5', '3') || \
345  (tag) == MKTAG('a', 'i', '5', '5') || \
346  (tag) == MKTAG('a', 'i', '5', '6') || \
347  (tag) == MKTAG('a', 'i', '1', 'p') || \
348  (tag) == MKTAG('a', 'i', '1', 'q') || \
349  (tag) == MKTAG('a', 'i', '1', '2') || \
350  (tag) == MKTAG('a', 'i', '1', '3') || \
351  (tag) == MKTAG('a', 'i', '1', '5') || \
352  (tag) == MKTAG('a', 'i', '1', '6') || \
353  (tag) == MKTAG('a', 'i', 'v', 'x') || \
354  (tag) == MKTAG('A', 'V', 'i', 'n'))
355 
356 
358 
359 int ff_mov_read_stsd_entries(MOVContext *c, AVIOContext *pb, int entries);
360 void ff_mov_write_chan(AVIOContext *pb, int64_t channel_layout);
361 
362 #define FF_MOV_FLAG_MFRA_AUTO -1
363 #define FF_MOV_FLAG_MFRA_DTS 1
364 #define FF_MOV_FLAG_MFRA_PTS 2
365 
366 /**
367  * Compute codec id for 'lpcm' tag.
368  * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
369  */
370 static inline enum AVCodecID ff_mov_get_lpcm_codec_id(int bps, int flags)
371 {
372  /* lpcm flags:
373  * 0x1 = float
374  * 0x2 = big-endian
375  * 0x4 = signed
376  */
377  return ff_get_pcm_codec_id(bps, flags & 1, flags & 2, flags & 4 ? -1 : 0);
378 }
379 
380 #endif /* AVFORMAT_ISOM_H */
int itunes_metadata
metadata are itunes style
Definition: isom.h:260
int * bitrates
bitrates read before streams creation
Definition: isom.h:272
int64_t current_index
Definition: isom.h:190
int64_t moof_offset
Definition: isom.h:137
size_t auxiliary_offsets_count
Definition: isom.h:124
Bytestream IO Context.
Definition: avio.h:161
Buffered I/O operations.
uint32_t tmcd_flags
tmcd track flags
Definition: isom.h:208
const AVCodecTag ff_mp4_obj_type[]
Definition: isom.c:34
unsigned int rap_group_count
Definition: isom.h:211
int found_tfhd
Definition: isom.h:90
unsigned int elst_count
Definition: isom.h:176
MOVTrackExt * trex_data
Definition: isom.h:258
int export_all
Definition: isom.h:270
int64_t end
Definition: isom.h:154
unsigned track_id
Definition: isom.h:91
uint32_t format
Definition: isom.h:232
unsigned int samples_per_frame
Definition: isom.h:194
int dv_audio_container
Definition: isom.h:195
Definition: isom.h:56
float rate
Definition: isom.h:70
uint64_t base_data_offset
Definition: isom.h:92
int64_t duration
Definition: isom.h:68
size_t spherical_size
Definition: isom.h:227
unsigned int ctts_allocated_size
Definition: isom.h:167
unsigned int stsc_count
Definition: isom.h:169
int has_palette
Definition: isom.h:206
AVEncryptionInfo * default_encrypted_sample
Definition: isom.h:238
int allocated_size
Definition: isom.h:145
AVMasteringDisplayMetadata * mastering
Definition: isom.h:228
uint8_t * decryption_key
Definition: isom.h:287
This describes encryption info for a packet.
int ignore_chapters
Definition: isom.h:267
void * activation_bytes
Definition: isom.h:282
AVEncryptionInfo ** encrypted_samples
Definition: isom.h:118
uint64_t * auxiliary_offsets
Absolute seek position.
Definition: isom.h:123
int ff_mp4_read_dec_config_descr(AVFormatContext *fc, AVStream *st, AVIOContext *pb)
Definition: isom.c:504
int16_t audio_cid
stsd audio compression id
Definition: isom.h:197
char ** meta_keys
Definition: isom.h:252
int height
tkhd height
Definition: isom.h:203
uint32_t type
Definition: isom.h:83
MOVElst * elst_data
Definition: isom.h:175
MOVStsc * stsc_data
Definition: isom.h:170
int nb_items
Definition: isom.h:148
int ctts_index
Definition: isom.h:177
int complete
Definition: isom.h:146
unsigned stsd_id
Definition: isom.h:103
AVSphericalMapping * spherical
Definition: isom.h:226
int found_moov
&#39;moov&#39; atom has been found
Definition: isom.h:248
unsigned flags
Definition: isom.h:106
int isom
1 if file is ISO Media (mp4/3gp)
Definition: isom.h:256
int found_mdat
&#39;mdat&#39; atom has been found
Definition: isom.h:249
int width
tkhd width
Definition: isom.h:202
unsigned drefs_count
Definition: isom.h:198
Format I/O context.
Definition: avformat.h:1351
const AVCodecTag ff_codec_movvideo_tags[]
Definition: isom.c:75
MOVDref * drefs
Definition: isom.h:199
const AVCodecTag ff_codec_movaudio_tags[]
Definition: isom.c:313
uint8_t
int first
Definition: isom.h:62
size_t coll_size
Definition: isom.h:230
int ff_mp4_read_descr_len(AVIOContext *pb)
Definition: isom.c:457
Stereo 3D type: this structure describes how two videos are packed within a single video surface...
Definition: stereo3d.h:176
unsigned int sample_count
Definition: isom.h:181
int count
Definition: isom.h:63
int dts_shift
dts shift when ctts is negative
Definition: isom.h:204
int stsd_version
Definition: isom.h:222
unsigned int count
Definition: isom.h:110
int16_t nlvl_to
Definition: isom.h:79
int stsd_count
Definition: isom.h:221
int activation_bytes_size
Definition: isom.h:283
uint32_t tag
Definition: movenc.c:1478
int nb_frames_for_fps
Definition: isom.h:214
struct AVAESCTR * aes_ctr
Definition: isom.h:236
const AVCodecTag ff_codec_movsubtitle_tags[]
Definition: isom.c:370
int current_sample
Definition: isom.h:189
unsigned track_id
Definition: isom.h:102
const char * to
Definition: webvttdec.c:34
int64_t time_offset
time offset of the edit list entries
Definition: isom.h:187
unsigned int keyframe_count
Definition: isom.h:184
MOVIndexRange * index_ranges
Definition: isom.h:191
int64_t first_tfra_pts
Definition: isom.h:130
uint8_t * auxiliary_info_sizes
Definition: isom.h:120
int * extradata_size
Definition: isom.h:219
#define fc(width, name, range_min, range_max)
Definition: cbs_av1.c:596
Content light level needed by to transmit HDR over HDMI (CTA-861.3).
AVCodecID
Identify the syntax and semantics of the bitstream.
Definition: avcodec.h:215
unsigned int ctts_count
Definition: isom.h:166
int64_t tfdt_dts
Definition: isom.h:131
int stsc_sample
Definition: isom.h:172
int atom_depth
Definition: isom.h:278
MOVIndexRange * current_index_range
Definition: isom.h:192
int * keyframes
Definition: isom.h:185
AVFormatContext * fc
Definition: isom.h:245
int handbrake_version
Definition: isom.h:261
int ctts_sample
Definition: isom.h:178
int keyframe_absent
Definition: isom.h:183
void * audible_fixed_key
Definition: isom.h:284
int64_t min_corrected_pts
minimum Composition time shown by the edits excluding empty edits.
Definition: isom.h:188
AVContentLightMetadata * coll
Definition: isom.h:229
struct AVAES * aes_decrypt
Definition: isom.h:286
Spherical video.
MOVFragmentStreamInfo * stream_info
Definition: isom.h:141
char * dir
Definition: isom.h:76
int id
Definition: isom.h:64
uint8_t auxiliary_info_default_size
Definition: isom.h:122
int decryption_key_len
Definition: isom.h:288
int32_t
int audible_fixed_key_size
Definition: isom.h:285
int advanced_editlist
Definition: isom.h:266
char * path
Definition: isom.h:75
int time_scale
Definition: isom.h:186
Definition: isom.h:109
uint64_t moof_offset
Definition: isom.h:93
MOVStts * ctts_data
Definition: isom.h:168
unsigned size
Definition: isom.h:105
void ff_mov_write_chan(AVIOContext *pb, int64_t channel_layout)
Definition: isom.c:632
int ignore_editlist
Definition: isom.h:265
MOVFragmentIndexItem * item
Definition: isom.h:149
int64_t * chunk_offsets
Definition: isom.h:163
unsigned int index
Definition: isom.h:111
MOVFragmentIndex frag_index
Definition: isom.h:277
const AVCodecTag ff_codec_movdata_tags[]
Definition: isom.c:377
int ff_mov_read_esds(AVFormatContext *fc, AVIOContext *pb)
Definition: mov_esds.c:23
int64_t duration
duration of the longest track
Definition: isom.h:247
Stream structure.
Definition: avformat.h:874
unsigned duration
Definition: isom.h:104
DVDemuxContext * dv_demux
Definition: isom.h:254
int timecode_track
Definition: isom.h:201
int * sample_sizes
Definition: isom.h:182
unsigned duration
Definition: isom.h:96
MOVSbgp * rap_group
Definition: isom.h:212
int duration
Definition: isom.h:58
int * chapter_tracks
Definition: isom.h:262
unsigned int stsz_sample_size
always contains sample size from stsz atom
Definition: isom.h:180
unsigned trex_count
Definition: isom.h:259
int enable_drefs
Definition: isom.h:289
int64_t data_size
Definition: isom.h:207
int has_looked_for_mfra
Definition: isom.h:276
uint32_t type
Definition: isom.h:74
unsigned int count
Definition: isom.h:57
MOVStts * stts_data
Definition: isom.h:165
MOVEncryptionIndex * encryption_index
Definition: isom.h:133
Describe the class of an AVClass context structure.
Definition: log.h:67
enum AVCodecID ff_get_pcm_codec_id(int bps, int flt, int be, int sflags)
Select a PCM codec based on the given parameters.
Definition: utils.c:3137
unsigned int aax_mode
&#39;aax&#39; file has been detected
Definition: isom.h:279
Mastering display metadata capable of representing the color volume of the display used to master the...
MOVFragment fragment
current fragment in moof atom
Definition: isom.h:257
int64_t track_end
used for dts generation in fragmented movie files
Definition: isom.h:209
Definition: isom.h:82
static enum AVCodecID ff_mov_get_lpcm_codec_id(int bps, int flags)
Compute codec id for &#39;lpcm&#39; tag.
Definition: isom.h:370
int use_mfra_for
Definition: isom.h:275
int64_t time
Definition: isom.h:69
int pb_is_copied
Definition: isom.h:159
int ff_mov_lang_to_iso639(unsigned code, char to[4])
Definition: isom.c:435
This structure describes how to handle spherical videos, outlining information about projection...
Definition: spherical.h:82
int ff_mov_read_stsd_entries(MOVContext *c, AVIOContext *pb, int entries)
Definition: mov.c:2460
int next_chunk
Definition: isom.h:161
AVFormatContext * dv_fctx
Definition: isom.h:255
unsigned int stsc_index
Definition: isom.h:171
int ff_mp4_read_descr(AVFormatContext *fc, AVIOContext *pb, int *tag)
Definition: isom.c:470
uint64_t implicit_offset
Definition: isom.h:94
#define flags(name, subs,...)
Definition: cbs_av1.c:606
AVStereo3D * stereo3d
Definition: isom.h:225
unsigned int per_sample_iv_size
Definition: isom.h:237
unsigned * stps_data
partial sync sample for mpeg-2 open gop
Definition: isom.h:174
int ff_mov_iso639_to_lang(const char lang[4], int mp4)
Definition: isom.c:408
int64_t sidx_pts
Definition: isom.h:129
int start_pad
amount of samples to skip due to enc-dec delay
Definition: isom.h:210
int trak_index
Index of the current &#39;trak&#39;.
Definition: isom.h:251
int32_t * display_matrix
Definition: isom.h:224
AVIOContext * pb
Definition: isom.h:158
unsigned int bytes_per_frame
Definition: isom.h:193
unsigned flags
Definition: isom.h:98
int bitrates_count
Definition: isom.h:273
int64_t size
Definition: isom.h:84
int export_xmp
Definition: isom.h:271
int seek_individually
Definition: isom.h:268
static double c[64]
unsigned int stps_count
Definition: isom.h:173
unsigned int chunk_count
Definition: isom.h:162
Stereoscopic video.
unsigned bps
Definition: movenc.c:1479
int64_t start
Definition: isom.h:153
unsigned meta_keys_count
Definition: isom.h:253
unsigned stsd_id
Definition: isom.h:95
MOVEncryptionIndex * encryption_index
Definition: isom.h:239
unsigned int stts_count
Definition: isom.h:164
unsigned int sample_size
may contain value calculated from stsd or value from stsz atom
Definition: isom.h:179
int64_t next_root_atom
offset of the next root atom
Definition: isom.h:269
int last_stsd_index
Definition: isom.h:220
unsigned int nb_encrypted_samples
Definition: isom.h:117
unsigned int nb_chapter_tracks
Definition: isom.h:263
int time_scale
Definition: isom.h:246
void ff_mp4_parse_es_descr(AVIOContext *pb, int *es_id)
Definition: isom.c:479
Definition: isom.h:67
int pseudo_stream_id
-1 means demux all ids
Definition: isom.h:196
int ffindex
AVStream index.
Definition: isom.h:160
uint8_t ** extradata
extradata array (and size) for multiple stsd
Definition: isom.h:218
int use_absolute_path
Definition: isom.h:264
unsigned size
Definition: isom.h:97
int64_t duration_for_fps
Definition: isom.h:215
int moov_retry
Definition: isom.h:274
size_t auxiliary_info_sample_count
Definition: isom.h:121
int found_hdlr_mdta
&#39;hdlr&#39; atom with type &#39;mdta&#39; has been found
Definition: isom.h:250
Definition: isom.h:61
int has_sidx
Definition: isom.h:234
Definition: isom.h:73