FFmpeg  4.0
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 
31 #include "libavutil/spherical.h"
32 #include "libavutil/stereo3d.h"
33 
34 #include "avio.h"
35 #include "internal.h"
36 #include "dv.h"
37 
38 /* isom.c */
39 extern const AVCodecTag ff_mp4_obj_type[];
40 extern const AVCodecTag ff_codec_movvideo_tags[];
41 extern const AVCodecTag ff_codec_movaudio_tags[];
43 
44 int ff_mov_iso639_to_lang(const char lang[4], int mp4);
45 int ff_mov_lang_to_iso639(unsigned code, char to[4]);
46 
47 struct AVAESCTR;
48 
49 /* the QuickTime file format is quite convoluted...
50  * it has lots of index tables, each indexing something in another one...
51  * Here we just use what is needed to read the chunks
52  */
53 
54 typedef struct MOVStts {
55  unsigned int count;
56  int duration;
57 } MOVStts;
58 
59 typedef struct MOVStsc {
60  int first;
61  int count;
62  int id;
63 } MOVStsc;
64 
65 typedef struct MOVElst {
66  int64_t duration;
67  int64_t time;
68  float rate;
69 } MOVElst;
70 
71 typedef struct MOVDref {
72  uint32_t type;
73  char *path;
74  char *dir;
75  char volume[28];
76  char filename[64];
77  int16_t nlvl_to, nlvl_from;
78 } MOVDref;
79 
80 typedef struct MOVAtom {
81  uint32_t type;
82  int64_t size; /* total size (excluding the size and type fields) */
83 } MOVAtom;
84 
85 struct MOVParseTableEntry;
86 
87 typedef struct MOVFragment {
88  unsigned track_id;
89  uint64_t base_data_offset;
90  uint64_t moof_offset;
91  uint64_t implicit_offset;
92  unsigned stsd_id;
93  unsigned duration;
94  unsigned size;
95  unsigned flags;
96 } MOVFragment;
97 
98 typedef struct MOVTrackExt {
99  unsigned track_id;
100  unsigned stsd_id;
101  unsigned duration;
102  unsigned size;
103  unsigned flags;
104 } MOVTrackExt;
105 
106 typedef struct MOVSbgp {
107  unsigned int count;
108  unsigned int index;
109 } MOVSbgp;
110 
111 typedef struct MOVFragmentStreamInfo {
112  int id;
113  int64_t sidx_pts;
114  int64_t first_tfra_pts;
115  int64_t tfdt_dts;
118 
119 typedef struct MOVFragmentIndexItem {
120  int64_t moof_offset;
122  int current;
126 
127 typedef struct MOVFragmentIndex {
129  int complete;
130  int current;
131  int nb_items;
134 
135 typedef struct MOVIndexRange {
136  int64_t start;
137  int64_t end;
138 } MOVIndexRange;
139 
140 typedef struct MOVStreamContext {
143  int ffindex; ///< AVStream index
145  unsigned int chunk_count;
146  int64_t *chunk_offsets;
147  unsigned int stts_count;
149  unsigned int ctts_count;
150  unsigned int ctts_allocated_size;
152  unsigned int stsc_count;
154  unsigned int stsc_index;
156  unsigned int stps_count;
157  unsigned *stps_data; ///< partial sync sample for mpeg-2 open gop
159  unsigned int elst_count;
162  unsigned int sample_size; ///< may contain value calculated from stsd or value from stsz atom
163  unsigned int stsz_sample_size; ///< always contains sample size from stsz atom
164  unsigned int sample_count;
167  unsigned int keyframe_count;
168  int *keyframes;
170  int64_t time_offset; ///< time offset of the edit list entries
171  int64_t min_corrected_pts; ///< minimum Composition time shown by the edits excluding empty edits.
173  int64_t current_index;
176  unsigned int bytes_per_frame;
177  unsigned int samples_per_frame;
179  int pseudo_stream_id; ///< -1 means demux all ids
180  int16_t audio_cid; ///< stsd audio compression id
181  unsigned drefs_count;
183  int dref_id;
185  int width; ///< tkhd width
186  int height; ///< tkhd height
187  int dts_shift; ///< dts shift when ctts is negative
188  uint32_t palette[256];
190  int64_t data_size;
191  uint32_t tmcd_flags; ///< tmcd track flags
192  int64_t track_end; ///< used for dts generation in fragmented movie files
193  int start_pad; ///< amount of samples to skip due to enc-dec delay
194  unsigned int rap_group_count;
196 
199 
200  /** extradata array (and size) for multiple stsd */
205 
212  size_t coll_size;
213 
214  uint32_t format;
215 
216  int has_sidx; // If there is an sidx entry for this stream.
217  struct {
226  struct AVAESCTR* aes_ctr;
227  } cenc;
229 
230 typedef struct MOVContext {
231  const AVClass *class; ///< class for private options
234  int64_t duration; ///< duration of the longest track
235  int found_moov; ///< 'moov' atom has been found
236  int found_mdat; ///< 'mdat' atom has been found
237  int found_hdlr_mdta; ///< 'hdlr' atom with type 'mdta' has been found
238  int trak_index; ///< Index of the current 'trak'
239  char **meta_keys;
240  unsigned meta_keys_count;
243  int isom; ///< 1 if file is ISO Media (mp4/3gp)
244  MOVFragment fragment; ///< current fragment in moof atom
246  unsigned trex_count;
247  int itunes_metadata; ///< metadata are itunes style
250  unsigned int nb_chapter_tracks;
256  int64_t next_root_atom; ///< offset of the next root atom
259  int *bitrates; ///< bitrates read before streams creation
266  unsigned int aax_mode; ///< 'aax' file has been detected
267  uint8_t file_key[20];
268  uint8_t file_iv[20];
277  int32_t movie_display_matrix[3][3]; ///< display matrix from mvhd
278 } MOVContext;
279 
283 void ff_mp4_parse_es_descr(AVIOContext *pb, int *es_id);
284 
285 #define MP4ODescrTag 0x01
286 #define MP4IODescrTag 0x02
287 #define MP4ESDescrTag 0x03
288 #define MP4DecConfigDescrTag 0x04
289 #define MP4DecSpecificDescrTag 0x05
290 #define MP4SLDescrTag 0x06
291 
292 #define MOV_TFHD_BASE_DATA_OFFSET 0x01
293 #define MOV_TFHD_STSD_ID 0x02
294 #define MOV_TFHD_DEFAULT_DURATION 0x08
295 #define MOV_TFHD_DEFAULT_SIZE 0x10
296 #define MOV_TFHD_DEFAULT_FLAGS 0x20
297 #define MOV_TFHD_DURATION_IS_EMPTY 0x010000
298 #define MOV_TFHD_DEFAULT_BASE_IS_MOOF 0x020000
299 
300 #define MOV_TRUN_DATA_OFFSET 0x01
301 #define MOV_TRUN_FIRST_SAMPLE_FLAGS 0x04
302 #define MOV_TRUN_SAMPLE_DURATION 0x100
303 #define MOV_TRUN_SAMPLE_SIZE 0x200
304 #define MOV_TRUN_SAMPLE_FLAGS 0x400
305 #define MOV_TRUN_SAMPLE_CTS 0x800
306 
307 #define MOV_FRAG_SAMPLE_FLAG_DEGRADATION_PRIORITY_MASK 0x0000ffff
308 #define MOV_FRAG_SAMPLE_FLAG_IS_NON_SYNC 0x00010000
309 #define MOV_FRAG_SAMPLE_FLAG_PADDING_MASK 0x000e0000
310 #define MOV_FRAG_SAMPLE_FLAG_REDUNDANCY_MASK 0x00300000
311 #define MOV_FRAG_SAMPLE_FLAG_DEPENDED_MASK 0x00c00000
312 #define MOV_FRAG_SAMPLE_FLAG_DEPENDS_MASK 0x03000000
313 
314 #define MOV_FRAG_SAMPLE_FLAG_DEPENDS_NO 0x02000000
315 #define MOV_FRAG_SAMPLE_FLAG_DEPENDS_YES 0x01000000
316 
317 #define MOV_TKHD_FLAG_ENABLED 0x0001
318 #define MOV_TKHD_FLAG_IN_MOVIE 0x0002
319 #define MOV_TKHD_FLAG_IN_PREVIEW 0x0004
320 #define MOV_TKHD_FLAG_IN_POSTER 0x0008
321 
322 #define MOV_SAMPLE_DEPENDENCY_UNKNOWN 0x0
323 #define MOV_SAMPLE_DEPENDENCY_YES 0x1
324 #define MOV_SAMPLE_DEPENDENCY_NO 0x2
325 
326 
327 #define TAG_IS_AVCI(tag) \
328  ((tag) == MKTAG('a', 'i', '5', 'p') || \
329  (tag) == MKTAG('a', 'i', '5', 'q') || \
330  (tag) == MKTAG('a', 'i', '5', '2') || \
331  (tag) == MKTAG('a', 'i', '5', '3') || \
332  (tag) == MKTAG('a', 'i', '5', '5') || \
333  (tag) == MKTAG('a', 'i', '5', '6') || \
334  (tag) == MKTAG('a', 'i', '1', 'p') || \
335  (tag) == MKTAG('a', 'i', '1', 'q') || \
336  (tag) == MKTAG('a', 'i', '1', '2') || \
337  (tag) == MKTAG('a', 'i', '1', '3') || \
338  (tag) == MKTAG('a', 'i', '1', '5') || \
339  (tag) == MKTAG('a', 'i', '1', '6') || \
340  (tag) == MKTAG('a', 'i', 'v', 'x') || \
341  (tag) == MKTAG('A', 'V', 'i', 'n'))
342 
343 
345 
346 int ff_mov_read_stsd_entries(MOVContext *c, AVIOContext *pb, int entries);
347 void ff_mov_write_chan(AVIOContext *pb, int64_t channel_layout);
348 
349 #define FF_MOV_FLAG_MFRA_AUTO -1
350 #define FF_MOV_FLAG_MFRA_DTS 1
351 #define FF_MOV_FLAG_MFRA_PTS 2
352 
353 /**
354  * Compute codec id for 'lpcm' tag.
355  * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
356  */
357 static inline enum AVCodecID ff_mov_get_lpcm_codec_id(int bps, int flags)
358 {
359  /* lpcm flags:
360  * 0x1 = float
361  * 0x2 = big-endian
362  * 0x4 = signed
363  */
364  return ff_get_pcm_codec_id(bps, flags & 1, flags & 2, flags & 4 ? -1 : 0);
365 }
366 
367 #endif /* AVFORMAT_ISOM_H */
int itunes_metadata
metadata are itunes style
Definition: isom.h:247
int * bitrates
bitrates read before streams creation
Definition: isom.h:259
int64_t current_index
Definition: isom.h:173
int64_t moof_offset
Definition: isom.h:120
Bytestream IO Context.
Definition: avio.h:161
Buffered I/O operations.
uint32_t tmcd_flags
tmcd track flags
Definition: isom.h:191
const AVCodecTag ff_mp4_obj_type[]
Definition: isom.c:34
uint8_t * auxiliary_info_end
Definition: isom.h:220
unsigned int rap_group_count
Definition: isom.h:194
unsigned int elst_count
Definition: isom.h:159
MOVTrackExt * trex_data
Definition: isom.h:245
int export_all
Definition: isom.h:257
int64_t end
Definition: isom.h:137
unsigned track_id
Definition: isom.h:88
uint32_t format
Definition: isom.h:214
unsigned int samples_per_frame
Definition: isom.h:177
int dv_audio_container
Definition: isom.h:178
Definition: isom.h:54
float rate
Definition: isom.h:68
uint64_t base_data_offset
Definition: isom.h:89
int64_t duration
Definition: isom.h:66
size_t spherical_size
Definition: isom.h:209
unsigned int ctts_allocated_size
Definition: isom.h:150
unsigned int stsc_count
Definition: isom.h:152
int has_palette
Definition: isom.h:189
int allocated_size
Definition: isom.h:128
AVMasteringDisplayMetadata * mastering
Definition: isom.h:210
uint8_t * decryption_key
Definition: isom.h:274
int ignore_chapters
Definition: isom.h:254
void * activation_bytes
Definition: isom.h:269
int ff_mp4_read_dec_config_descr(AVFormatContext *fc, AVStream *st, AVIOContext *pb)
Definition: isom.c:498
int16_t audio_cid
stsd audio compression id
Definition: isom.h:180
char ** meta_keys
Definition: isom.h:239
int height
tkhd height
Definition: isom.h:186
uint32_t type
Definition: isom.h:81
MOVElst * elst_data
Definition: isom.h:158
MOVStsc * stsc_data
Definition: isom.h:153
int nb_items
Definition: isom.h:131
int ctts_index
Definition: isom.h:160
int complete
Definition: isom.h:129
unsigned stsd_id
Definition: isom.h:100
AVSphericalMapping * spherical
Definition: isom.h:208
int found_moov
&#39;moov&#39; atom has been found
Definition: isom.h:235
unsigned flags
Definition: isom.h:103
int isom
1 if file is ISO Media (mp4/3gp)
Definition: isom.h:243
int found_mdat
&#39;mdat&#39; atom has been found
Definition: isom.h:236
int width
tkhd width
Definition: isom.h:185
unsigned drefs_count
Definition: isom.h:181
Format I/O context.
Definition: avformat.h:1342
const AVCodecTag ff_codec_movvideo_tags[]
Definition: isom.c:75
MOVDref * drefs
Definition: isom.h:182
const AVCodecTag ff_codec_movaudio_tags[]
Definition: isom.c:312
uint8_t
int first
Definition: isom.h:60
uint8_t * auxiliary_info
Definition: isom.h:219
size_t coll_size
Definition: isom.h:212
int ff_mp4_read_descr_len(AVIOContext *pb)
Definition: isom.c:451
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:164
int count
Definition: isom.h:61
int dts_shift
dts shift when ctts is negative
Definition: isom.h:187
uint8_t * auxiliary_info_pos
Definition: isom.h:221
unsigned int count
Definition: isom.h:107
int16_t nlvl_to
Definition: isom.h:77
int stsd_count
Definition: isom.h:204
int activation_bytes_size
Definition: isom.h:270
static int flags
Definition: log.c:55
uint32_t tag
Definition: movenc.c:1455
int nb_frames_for_fps
Definition: isom.h:197
struct AVAESCTR * aes_ctr
Definition: isom.h:226
const AVCodecTag ff_codec_movsubtitle_tags[]
Definition: isom.c:369
int current_sample
Definition: isom.h:172
unsigned track_id
Definition: isom.h:99
const char * to
Definition: webvttdec.c:34
int64_t time_offset
time offset of the edit list entries
Definition: isom.h:170
unsigned int keyframe_count
Definition: isom.h:167
MOVIndexRange * index_ranges
Definition: isom.h:174
int64_t first_tfra_pts
Definition: isom.h:114
int * extradata_size
Definition: isom.h:202
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:149
int64_t tfdt_dts
Definition: isom.h:115
int stsc_sample
Definition: isom.h:155
int atom_depth
Definition: isom.h:265
MOVIndexRange * current_index_range
Definition: isom.h:175
int * keyframes
Definition: isom.h:168
AVFormatContext * fc
Definition: isom.h:232
int handbrake_version
Definition: isom.h:248
int ctts_sample
Definition: isom.h:161
int keyframe_absent
Definition: isom.h:166
void * audible_fixed_key
Definition: isom.h:271
int64_t min_corrected_pts
minimum Composition time shown by the edits excluding empty edits.
Definition: isom.h:171
int use_subsamples
Definition: isom.h:218
AVContentLightMetadata * coll
Definition: isom.h:211
struct AVAES * aes_decrypt
Definition: isom.h:273
static const uint16_t fc[]
Definition: dcaenc.h:43
uint8_t auxiliary_info_default_size
Definition: isom.h:222
Spherical video.
MOVFragmentStreamInfo * stream_info
Definition: isom.h:124
char * dir
Definition: isom.h:74
int id
Definition: isom.h:62
int decryption_key_len
Definition: isom.h:275
int32_t
int audible_fixed_key_size
Definition: isom.h:272
int advanced_editlist
Definition: isom.h:253
char * path
Definition: isom.h:73
int time_scale
Definition: isom.h:169
Definition: isom.h:106
uint64_t moof_offset
Definition: isom.h:90
MOVStts * ctts_data
Definition: isom.h:151
unsigned size
Definition: isom.h:102
void ff_mov_write_chan(AVIOContext *pb, int64_t channel_layout)
Definition: isom.c:626
int ignore_editlist
Definition: isom.h:252
MOVFragmentIndexItem * item
Definition: isom.h:132
int64_t * chunk_offsets
Definition: isom.h:146
unsigned int index
Definition: isom.h:108
MOVFragmentIndex frag_index
Definition: isom.h:264
int ff_mov_read_esds(AVFormatContext *fc, AVIOContext *pb)
Definition: mov_esds.c:23
size_t auxiliary_info_sizes_count
Definition: isom.h:224
int64_t duration
duration of the longest track
Definition: isom.h:234
Stream structure.
Definition: avformat.h:873
unsigned duration
Definition: isom.h:101
DVDemuxContext * dv_demux
Definition: isom.h:241
int timecode_track
Definition: isom.h:184
int * sample_sizes
Definition: isom.h:165
unsigned duration
Definition: isom.h:93
MOVSbgp * rap_group
Definition: isom.h:195
int duration
Definition: isom.h:56
int64_t auxiliary_info_index
Definition: isom.h:225
int * chapter_tracks
Definition: isom.h:249
unsigned int stsz_sample_size
always contains sample size from stsz atom
Definition: isom.h:163
unsigned trex_count
Definition: isom.h:246
int enable_drefs
Definition: isom.h:276
int64_t data_size
Definition: isom.h:190
int has_looked_for_mfra
Definition: isom.h:263
uint32_t type
Definition: isom.h:72
unsigned int count
Definition: isom.h:55
MOVStts * stts_data
Definition: isom.h:148
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:3120
unsigned int aax_mode
&#39;aax&#39; file has been detected
Definition: isom.h:266
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:244
int64_t track_end
used for dts generation in fragmented movie files
Definition: isom.h:192
Definition: isom.h:80
static enum AVCodecID ff_mov_get_lpcm_codec_id(int bps, int flags)
Compute codec id for &#39;lpcm&#39; tag.
Definition: isom.h:357
int use_mfra_for
Definition: isom.h:262
int64_t time
Definition: isom.h:67
int pb_is_copied
Definition: isom.h:142
int ff_mov_lang_to_iso639(unsigned code, char to[4])
Definition: isom.c:429
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:2444
int next_chunk
Definition: isom.h:144
AVFormatContext * dv_fctx
Definition: isom.h:242
unsigned int stsc_index
Definition: isom.h:154
int ff_mp4_read_descr(AVFormatContext *fc, AVIOContext *pb, int *tag)
Definition: isom.c:464
uint64_t implicit_offset
Definition: isom.h:91
AVStereo3D * stereo3d
Definition: isom.h:207
unsigned * stps_data
partial sync sample for mpeg-2 open gop
Definition: isom.h:157
uint8_t * auxiliary_info_sizes
Definition: isom.h:223
int ff_mov_iso639_to_lang(const char lang[4], int mp4)
Definition: isom.c:402
int64_t sidx_pts
Definition: isom.h:113
int start_pad
amount of samples to skip due to enc-dec delay
Definition: isom.h:193
int trak_index
Index of the current &#39;trak&#39;.
Definition: isom.h:238
int32_t * display_matrix
Definition: isom.h:206
AVIOContext * pb
Definition: isom.h:141
unsigned int bytes_per_frame
Definition: isom.h:176
unsigned flags
Definition: isom.h:95
int bitrates_count
Definition: isom.h:260
int64_t size
Definition: isom.h:82
int export_xmp
Definition: isom.h:258
int seek_individually
Definition: isom.h:255
static double c[64]
unsigned int stps_count
Definition: isom.h:156
unsigned int chunk_count
Definition: isom.h:145
Stereoscopic video.
unsigned bps
Definition: movenc.c:1456
int64_t start
Definition: isom.h:136
unsigned meta_keys_count
Definition: isom.h:240
unsigned stsd_id
Definition: isom.h:92
unsigned int stts_count
Definition: isom.h:147
unsigned int sample_size
may contain value calculated from stsd or value from stsz atom
Definition: isom.h:162
int64_t next_root_atom
offset of the next root atom
Definition: isom.h:256
int last_stsd_index
Definition: isom.h:203
unsigned int nb_chapter_tracks
Definition: isom.h:250
int time_scale
Definition: isom.h:233
void ff_mp4_parse_es_descr(AVIOContext *pb, int *es_id)
Definition: isom.c:473
Definition: isom.h:65
int pseudo_stream_id
-1 means demux all ids
Definition: isom.h:179
int ffindex
AVStream index.
Definition: isom.h:143
uint8_t ** extradata
extradata array (and size) for multiple stsd
Definition: isom.h:201
int use_absolute_path
Definition: isom.h:251
unsigned size
Definition: isom.h:94
int64_t duration_for_fps
Definition: isom.h:198
int moov_retry
Definition: isom.h:261
int found_hdlr_mdta
&#39;hdlr&#39; atom with type &#39;mdta&#39; has been found
Definition: isom.h:237
Definition: isom.h:59
int has_sidx
Definition: isom.h:216
Definition: isom.h:71