You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1692 lines
50KB

  1. /*
  2. * MOV, 3GP, MP4 encoder.
  3. * Copyright (c) 2003 Thomas Raivio.
  4. * Copyright (c) 2004 Gildas Bazin <gbazin at videolan dot org>.
  5. *
  6. * This library is free software; you can redistribute it and/or
  7. * modify it under the terms of the GNU Lesser General Public
  8. * License as published by the Free Software Foundation; either
  9. * version 2 of the License, or (at your option) any later version.
  10. *
  11. * This library is distributed in the hope that it will be useful,
  12. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  14. * Lesser General Public License for more details.
  15. *
  16. * You should have received a copy of the GNU Lesser General Public
  17. * License along with this library; if not, write to the Free Software
  18. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19. */
  20. #include "avformat.h"
  21. #include "avi.h"
  22. #include "avio.h"
  23. #include "mov.h"
  24. #undef NDEBUG
  25. #include <assert.h>
  26. #define MOV_INDEX_CLUSTER_SIZE 16384
  27. #define globalTimescale 1000
  28. #define MODE_MP4 0
  29. #define MODE_MOV 1
  30. #define MODE_3GP 2
  31. #define MODE_PSP 3 // example working PSP command line:
  32. // ffmpeg -i testinput.avi -f psp -r 14.985 -s 320x240 -b 768 -ar 24000 -ab 32 M4V00001.MP4
  33. #define MODE_3G2 4
  34. typedef struct MOVIentry {
  35. unsigned int flags, size;
  36. uint64_t pos;
  37. unsigned int samplesInChunk;
  38. char key_frame;
  39. unsigned int entries;
  40. int64_t cts;
  41. } MOVIentry;
  42. typedef struct MOVIndex {
  43. int mode;
  44. int entry;
  45. uint64_t mdat_size;
  46. int ents_allocated;
  47. long timescale;
  48. long time;
  49. long trackDuration;
  50. long sampleCount;
  51. long sampleDuration;
  52. int hasKeyframes;
  53. int hasBframes;
  54. int language;
  55. int trackID;
  56. AVCodecContext *enc;
  57. int vosLen;
  58. uint8_t *vosData;
  59. MOVIentry** cluster;
  60. } MOVTrack;
  61. typedef struct MOVContext {
  62. int mode;
  63. long time;
  64. int nb_streams;
  65. int mdat_written;
  66. offset_t mdat_pos;
  67. long timescale;
  68. MOVTrack tracks[MAX_STREAMS];
  69. } MOVContext;
  70. static int mov_write_esds_tag(ByteIOContext *pb, MOVTrack* track);
  71. /* output language code from iso639 language name */
  72. extern int ff_mov_iso639_to_lang(const char *lang, int mp4);
  73. //FIXME supprt 64bit varaint with wide placeholders
  74. static offset_t updateSize (ByteIOContext *pb, offset_t pos)
  75. {
  76. offset_t curpos = url_ftell(pb);
  77. url_fseek(pb, pos, SEEK_SET);
  78. put_be32(pb, curpos - pos); /* rewrite size */
  79. url_fseek(pb, curpos, SEEK_SET);
  80. return curpos - pos;
  81. }
  82. /* Chunk offset atom */
  83. static int mov_write_stco_tag(ByteIOContext *pb, MOVTrack* track)
  84. {
  85. int i;
  86. int mode64 = 0; // use 32 bit size variant if possible
  87. offset_t pos = url_ftell(pb);
  88. put_be32(pb, 0); /* size */
  89. if (pos > UINT32_MAX) {
  90. mode64 = 1;
  91. put_tag(pb, "co64");
  92. } else
  93. put_tag(pb, "stco");
  94. put_be32(pb, 0); /* version & flags */
  95. put_be32(pb, track->entry); /* entry count */
  96. for (i=0; i<track->entry; i++) {
  97. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  98. int id = i % MOV_INDEX_CLUSTER_SIZE;
  99. if(mode64 == 1)
  100. put_be64(pb, track->cluster[cl][id].pos);
  101. else
  102. put_be32(pb, track->cluster[cl][id].pos);
  103. }
  104. return updateSize (pb, pos);
  105. }
  106. /* Sample size atom */
  107. static int mov_write_stsz_tag(ByteIOContext *pb, MOVTrack* track)
  108. {
  109. int equalChunks = 1;
  110. int i, j, entries = 0, tst = -1, oldtst = -1;
  111. offset_t pos = url_ftell(pb);
  112. put_be32(pb, 0); /* size */
  113. put_tag(pb, "stsz");
  114. put_be32(pb, 0); /* version & flags */
  115. for (i=0; i<track->entry; i++) {
  116. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  117. int id = i % MOV_INDEX_CLUSTER_SIZE;
  118. tst = track->cluster[cl][id].size/track->cluster[cl][id].entries;
  119. if(oldtst != -1 && tst != oldtst) {
  120. equalChunks = 0;
  121. }
  122. oldtst = tst;
  123. entries += track->cluster[cl][id].entries;
  124. }
  125. if (equalChunks) {
  126. int sSize = track->cluster[0][0].size/track->cluster[0][0].entries;
  127. put_be32(pb, sSize); // sample size
  128. put_be32(pb, entries); // sample count
  129. }
  130. else {
  131. put_be32(pb, 0); // sample size
  132. put_be32(pb, entries); // sample count
  133. for (i=0; i<track->entry; i++) {
  134. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  135. int id = i % MOV_INDEX_CLUSTER_SIZE;
  136. for ( j=0; j<track->cluster[cl][id].entries; j++) {
  137. put_be32(pb, track->cluster[cl][id].size /
  138. track->cluster[cl][id].entries);
  139. }
  140. }
  141. }
  142. return updateSize (pb, pos);
  143. }
  144. /* Sample to chunk atom */
  145. static int mov_write_stsc_tag(ByteIOContext *pb, MOVTrack* track)
  146. {
  147. int index = 0, oldval = -1, i;
  148. offset_t entryPos, curpos;
  149. offset_t pos = url_ftell(pb);
  150. put_be32(pb, 0); /* size */
  151. put_tag(pb, "stsc");
  152. put_be32(pb, 0); // version & flags
  153. entryPos = url_ftell(pb);
  154. put_be32(pb, track->entry); // entry count
  155. for (i=0; i<track->entry; i++) {
  156. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  157. int id = i % MOV_INDEX_CLUSTER_SIZE;
  158. if(oldval != track->cluster[cl][id].samplesInChunk)
  159. {
  160. put_be32(pb, i+1); // first chunk
  161. put_be32(pb, track->cluster[cl][id].samplesInChunk); // samples per chunk
  162. put_be32(pb, 0x1); // sample description index
  163. oldval = track->cluster[cl][id].samplesInChunk;
  164. index++;
  165. }
  166. }
  167. curpos = url_ftell(pb);
  168. url_fseek(pb, entryPos, SEEK_SET);
  169. put_be32(pb, index); // rewrite size
  170. url_fseek(pb, curpos, SEEK_SET);
  171. return updateSize (pb, pos);
  172. }
  173. /* Sync sample atom */
  174. static int mov_write_stss_tag(ByteIOContext *pb, MOVTrack* track)
  175. {
  176. offset_t curpos, entryPos;
  177. int i, index = 0;
  178. offset_t pos = url_ftell(pb);
  179. put_be32(pb, 0); // size
  180. put_tag(pb, "stss");
  181. put_be32(pb, 0); // version & flags
  182. entryPos = url_ftell(pb);
  183. put_be32(pb, track->entry); // entry count
  184. for (i=0; i<track->entry; i++) {
  185. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  186. int id = i % MOV_INDEX_CLUSTER_SIZE;
  187. if(track->cluster[cl][id].key_frame == 1) {
  188. put_be32(pb, i+1);
  189. index++;
  190. }
  191. }
  192. curpos = url_ftell(pb);
  193. url_fseek(pb, entryPos, SEEK_SET);
  194. put_be32(pb, index); // rewrite size
  195. url_fseek(pb, curpos, SEEK_SET);
  196. return updateSize (pb, pos);
  197. }
  198. static int mov_write_damr_tag(ByteIOContext *pb)
  199. {
  200. put_be32(pb, 0x11); /* size */
  201. put_tag(pb, "damr");
  202. put_tag(pb, "FFMP");
  203. put_byte(pb, 0);
  204. put_be16(pb, 0x80); /* Mode set (all modes for AMR_NB) */
  205. put_be16(pb, 0xa); /* Mode change period (no restriction) */
  206. //put_be16(pb, 0x81ff); /* Mode set (all modes for AMR_NB) */
  207. //put_be16(pb, 1); /* Mode change period (no restriction) */
  208. return 0x11;
  209. }
  210. static int mov_write_wave_tag(ByteIOContext *pb, MOVTrack* track)
  211. {
  212. offset_t pos = url_ftell(pb);
  213. put_be32(pb, 0); /* size */
  214. put_tag(pb, "wave");
  215. put_be32(pb, 12); /* size */
  216. put_tag(pb, "frma");
  217. put_tag(pb, "mp4a");
  218. put_be32(pb, 12); /* size */
  219. put_tag(pb, "mp4a");
  220. put_be32(pb, 0);
  221. mov_write_esds_tag(pb, track);
  222. put_be32(pb, 12); /* size */
  223. put_tag(pb, "srcq");
  224. put_be32(pb, 0x40);
  225. put_be32(pb, 8); /* size */
  226. put_be32(pb, 0); /* null tag */
  227. return updateSize (pb, pos);
  228. }
  229. static const CodecTag codec_movaudio_tags[] = {
  230. { CODEC_ID_PCM_MULAW, MKTAG('u', 'l', 'a', 'w') },
  231. { CODEC_ID_PCM_ALAW, MKTAG('a', 'l', 'a', 'w') },
  232. { CODEC_ID_ADPCM_IMA_QT, MKTAG('i', 'm', 'a', '4') },
  233. { CODEC_ID_MACE3, MKTAG('M', 'A', 'C', '3') },
  234. { CODEC_ID_MACE6, MKTAG('M', 'A', 'C', '6') },
  235. { CODEC_ID_AAC, MKTAG('m', 'p', '4', 'a') },
  236. { CODEC_ID_AMR_NB, MKTAG('s', 'a', 'm', 'r') },
  237. { CODEC_ID_AMR_WB, MKTAG('s', 'a', 'w', 'b') },
  238. { CODEC_ID_PCM_S16BE, MKTAG('t', 'w', 'o', 's') },
  239. { CODEC_ID_PCM_S16LE, MKTAG('s', 'o', 'w', 't') },
  240. { CODEC_ID_MP3, MKTAG('.', 'm', 'p', '3') },
  241. { CODEC_ID_NONE, 0 },
  242. };
  243. static int mov_write_audio_tag(ByteIOContext *pb, MOVTrack* track)
  244. {
  245. offset_t pos = url_ftell(pb);
  246. int tag;
  247. put_be32(pb, 0); /* size */
  248. tag = track->enc->codec_tag;
  249. if (!tag)
  250. tag = codec_get_tag(codec_movaudio_tags, track->enc->codec_id);
  251. // if no mac fcc found, try with Microsoft tags
  252. if (!tag)
  253. {
  254. int tmp = codec_get_tag(codec_wav_tags, track->enc->codec_id);
  255. tag = MKTAG('m', 's', ((tmp >> 8) & 0xff), (tmp & 0xff));
  256. }
  257. put_le32(pb, tag); // store it byteswapped
  258. put_be32(pb, 0); /* Reserved */
  259. put_be16(pb, 0); /* Reserved */
  260. put_be16(pb, 1); /* Data-reference index, XXX == 1 */
  261. /* SoundDescription */
  262. if(track->mode == MODE_MOV && track->enc->codec_id == CODEC_ID_AAC)
  263. put_be16(pb, 1); /* Version 1 */
  264. else
  265. put_be16(pb, 0); /* Version 0 */
  266. put_be16(pb, 0); /* Revision level */
  267. put_be32(pb, 0); /* Reserved */
  268. put_be16(pb, track->enc->channels); /* Number of channels */
  269. /* TODO: Currently hard-coded to 16-bit, there doesn't seem
  270. to be a good way to get number of bits of audio */
  271. put_be16(pb, 0x10); /* Reserved */
  272. if(track->enc->codec_id == CODEC_ID_AAC ||
  273. track->enc->codec_id == CODEC_ID_MP3)
  274. {
  275. put_be16(pb, 0xfffe); /* compression ID (vbr)*/
  276. }
  277. else
  278. {
  279. put_be16(pb, 0); /* compression ID (= 0) */
  280. }
  281. put_be16(pb, 0); /* packet size (= 0) */
  282. put_be16(pb, track->timescale); /* Time scale */
  283. put_be16(pb, 0); /* Reserved */
  284. if(track->mode == MODE_MOV && track->enc->codec_id == CODEC_ID_AAC)
  285. {
  286. /* SoundDescription V1 extended info */
  287. put_be32(pb, track->enc->frame_size); /* Samples per packet */
  288. put_be32(pb, 1536); /* Bytes per packet */
  289. put_be32(pb, 2); /* Bytes per frame */
  290. put_be32(pb, 2); /* Bytes per sample */
  291. }
  292. if(track->enc->codec_id == CODEC_ID_AAC) {
  293. if( track->mode == MODE_MOV ) mov_write_wave_tag(pb, track);
  294. else mov_write_esds_tag(pb, track);
  295. }
  296. if(track->enc->codec_id == CODEC_ID_AMR_NB)
  297. mov_write_damr_tag(pb);
  298. return updateSize (pb, pos);
  299. }
  300. static int mov_write_d263_tag(ByteIOContext *pb)
  301. {
  302. put_be32(pb, 0xf); /* size */
  303. put_tag(pb, "d263");
  304. put_tag(pb, "FFMP");
  305. put_be16(pb, 0x0a);
  306. put_byte(pb, 0);
  307. return 0xf;
  308. }
  309. /* TODO: No idea about these values */
  310. static int mov_write_svq3_tag(ByteIOContext *pb)
  311. {
  312. put_be32(pb, 0x15);
  313. put_tag(pb, "SMI ");
  314. put_tag(pb, "SEQH");
  315. put_be32(pb, 0x5);
  316. put_be32(pb, 0xe2c0211d);
  317. put_be32(pb, 0xc0000000);
  318. put_byte(pb, 0);
  319. return 0x15;
  320. }
  321. static unsigned int descrLength(unsigned int len)
  322. {
  323. if (len < 0x00000080)
  324. return 2 + len;
  325. else if (len < 0x00004000)
  326. return 3 + len;
  327. else if(len < 0x00200000)
  328. return 4 + len;
  329. else
  330. return 5 + len;
  331. }
  332. static void putDescr(ByteIOContext *pb, int tag, int size)
  333. {
  334. uint32_t len;
  335. uint8_t vals[4];
  336. len = size;
  337. vals[3] = (uint8_t)(len & 0x7f);
  338. len >>= 7;
  339. vals[2] = (uint8_t)((len & 0x7f) | 0x80);
  340. len >>= 7;
  341. vals[1] = (uint8_t)((len & 0x7f) | 0x80);
  342. len >>= 7;
  343. vals[0] = (uint8_t)((len & 0x7f) | 0x80);
  344. put_byte(pb, tag); // DescriptorTag
  345. if (size < 0x00000080)
  346. {
  347. put_byte(pb, vals[3]);
  348. }
  349. else if (size < 0x00004000)
  350. {
  351. put_byte(pb, vals[2]);
  352. put_byte(pb, vals[3]);
  353. }
  354. else if (size < 0x00200000)
  355. {
  356. put_byte(pb, vals[1]);
  357. put_byte(pb, vals[2]);
  358. put_byte(pb, vals[3]);
  359. }
  360. else if (size < 0x10000000)
  361. {
  362. put_byte(pb, vals[0]);
  363. put_byte(pb, vals[1]);
  364. put_byte(pb, vals[2]);
  365. put_byte(pb, vals[3]);
  366. }
  367. }
  368. static int mov_write_esds_tag(ByteIOContext *pb, MOVTrack* track) // Basic
  369. {
  370. int decoderSpecificInfoLen;
  371. offset_t pos = url_ftell(pb);
  372. decoderSpecificInfoLen = track->vosLen ? descrLength(track->vosLen):0;
  373. put_be32(pb, 0); // size
  374. put_tag(pb, "esds");
  375. put_be32(pb, 0); // Version
  376. // ES descriptor
  377. putDescr(pb, 0x03, 3 + descrLength(13 + decoderSpecificInfoLen) +
  378. descrLength(1));
  379. put_be16(pb, track->trackID);
  380. put_byte(pb, 0x00); // flags (= no flags)
  381. // DecoderConfig descriptor
  382. putDescr(pb, 0x04, 13 + decoderSpecificInfoLen);
  383. // Object type indication
  384. put_byte(pb, codec_get_tag(ff_mov_obj_type, track->enc->codec_id));
  385. // the following fields is made of 6 bits to identify the streamtype (4 for video, 5 for audio)
  386. // plus 1 bit to indicate upstream and 1 bit set to 1 (reserved)
  387. if(track->enc->codec_type == CODEC_TYPE_AUDIO)
  388. put_byte(pb, 0x15); // flags (= Audiostream)
  389. else
  390. put_byte(pb, 0x11); // flags (= Visualstream)
  391. put_byte(pb, track->enc->rc_buffer_size>>(3+16)); // Buffersize DB (24 bits)
  392. put_be16(pb, (track->enc->rc_buffer_size>>3)&0xFFFF); // Buffersize DB
  393. put_be32(pb, FFMAX(track->enc->bit_rate, track->enc->rc_max_rate)); // maxbitrate (FIXME should be max rate in any 1 sec window)
  394. if(track->enc->rc_max_rate != track->enc->rc_min_rate || track->enc->rc_min_rate==0)
  395. put_be32(pb, 0); // vbr
  396. else
  397. put_be32(pb, track->enc->rc_max_rate); // avg bitrate
  398. if (track->vosLen)
  399. {
  400. // DecoderSpecific info descriptor
  401. putDescr(pb, 0x05, track->vosLen);
  402. put_buffer(pb, track->vosData, track->vosLen);
  403. }
  404. // SL descriptor
  405. putDescr(pb, 0x06, 1);
  406. put_byte(pb, 0x02);
  407. return updateSize (pb, pos);
  408. }
  409. static const CodecTag codec_movvideo_tags[] = {
  410. { CODEC_ID_SVQ1, MKTAG('S', 'V', 'Q', '1') },
  411. { CODEC_ID_SVQ3, MKTAG('S', 'V', 'Q', '3') },
  412. { CODEC_ID_MPEG4, MKTAG('m', 'p', '4', 'v') },
  413. { CODEC_ID_H263, MKTAG('s', '2', '6', '3') },
  414. { CODEC_ID_H264, MKTAG('a', 'v', 'c', '1') },
  415. /* special handling in mov_find_video_codec_tag */
  416. { CODEC_ID_DVVIDEO, MKTAG('d', 'v', 'c', ' ') }, /* DV NTSC */
  417. { CODEC_ID_DVVIDEO, MKTAG('d', 'v', 'c', 'p') }, /* DV PAL */
  418. { CODEC_ID_DVVIDEO, MKTAG('d', 'v', 'p', 'p') }, /* DVCPRO PAL */
  419. { CODEC_ID_DVVIDEO, MKTAG('d', 'v', '5', 'n') }, /* DVCPRO50 NTSC */
  420. { CODEC_ID_DVVIDEO, MKTAG('d', 'v', '5', 'p') }, /* DVCPRO50 PAL */
  421. { CODEC_ID_NONE, 0 },
  422. };
  423. static int mov_find_video_codec_tag(MOVTrack* track)
  424. {
  425. int tag;
  426. tag = track->enc->codec_tag;
  427. if (!tag) {
  428. if (track->enc->codec_id == CODEC_ID_DVVIDEO) {
  429. if (track->enc->height == 480) { /* NTSC */
  430. if (track->enc->pix_fmt == PIX_FMT_YUV422P)
  431. tag = MKTAG('d', 'v', '5', 'n');
  432. else
  433. tag = MKTAG('d', 'v', 'c', ' ');
  434. } else { /* assume PAL */
  435. if (track->enc->pix_fmt == PIX_FMT_YUV422P)
  436. tag = MKTAG('d', 'v', '5', 'p');
  437. else if (track->enc->pix_fmt == PIX_FMT_YUV420P)
  438. tag = MKTAG('d', 'v', 'p', 'p');
  439. else
  440. tag = MKTAG('d', 'v', 'c', 'p');
  441. }
  442. } else {
  443. tag = codec_get_tag(codec_movvideo_tags, track->enc->codec_id);
  444. }
  445. }
  446. // if no mac fcc found, try with Microsoft tags
  447. if (!tag)
  448. tag = codec_get_tag(codec_bmp_tags, track->enc->codec_id);
  449. assert(tag);
  450. return tag;
  451. }
  452. static int mov_write_video_tag(ByteIOContext *pb, MOVTrack* track)
  453. {
  454. offset_t pos = url_ftell(pb);
  455. char compressor_name[32];
  456. int tag;
  457. put_be32(pb, 0); /* size */
  458. tag = mov_find_video_codec_tag(track);
  459. put_le32(pb, tag); // store it byteswapped
  460. put_be32(pb, 0); /* Reserved */
  461. put_be16(pb, 0); /* Reserved */
  462. put_be16(pb, 1); /* Data-reference index */
  463. put_be16(pb, 0); /* Codec stream version */
  464. put_be16(pb, 0); /* Codec stream revision (=0) */
  465. put_tag(pb, "FFMP"); /* Vendor */
  466. if(track->enc->codec_id == CODEC_ID_RAWVIDEO) {
  467. put_be32(pb, 0); /* Temporal Quality */
  468. put_be32(pb, 0x400); /* Spatial Quality = lossless*/
  469. } else {
  470. put_be32(pb, 0x200); /* Temporal Quality = normal */
  471. put_be32(pb, 0x200); /* Spatial Quality = normal */
  472. }
  473. put_be16(pb, track->enc->width); /* Video width */
  474. put_be16(pb, track->enc->height); /* Video height */
  475. put_be32(pb, 0x00480000); /* Horizontal resolution 72dpi */
  476. put_be32(pb, 0x00480000); /* Vertical resolution 72dpi */
  477. put_be32(pb, 0); /* Data size (= 0) */
  478. put_be16(pb, 1); /* Frame count (= 1) */
  479. memset(compressor_name,0,32);
  480. if (track->enc->codec && track->enc->codec->name)
  481. strncpy(compressor_name,track->enc->codec->name,31);
  482. put_byte(pb, strlen(compressor_name));
  483. put_buffer(pb, compressor_name, 31);
  484. put_be16(pb, 0x18); /* Reserved */
  485. put_be16(pb, 0xffff); /* Reserved */
  486. if(track->enc->codec_id == CODEC_ID_MPEG4)
  487. mov_write_esds_tag(pb, track);
  488. else if(track->enc->codec_id == CODEC_ID_H263)
  489. mov_write_d263_tag(pb);
  490. else if(track->enc->codec_id == CODEC_ID_SVQ3)
  491. mov_write_svq3_tag(pb);
  492. return updateSize (pb, pos);
  493. }
  494. static int mov_write_stsd_tag(ByteIOContext *pb, MOVTrack* track)
  495. {
  496. offset_t pos = url_ftell(pb);
  497. put_be32(pb, 0); /* size */
  498. put_tag(pb, "stsd");
  499. put_be32(pb, 0); /* version & flags */
  500. put_be32(pb, 1); /* entry count */
  501. if (track->enc->codec_type == CODEC_TYPE_VIDEO)
  502. mov_write_video_tag(pb, track);
  503. else if (track->enc->codec_type == CODEC_TYPE_AUDIO)
  504. mov_write_audio_tag(pb, track);
  505. return updateSize(pb, pos);
  506. }
  507. static int mov_write_ctts_tag(ByteIOContext *pb, MOVTrack* track)
  508. {
  509. Time2Sample *ctts_entries;
  510. uint32_t entries = 0;
  511. uint32_t atom_size;
  512. int i;
  513. ctts_entries = av_malloc((track->entry + 1) * sizeof(*ctts_entries)); /* worst case */
  514. ctts_entries[0].count = 1;
  515. ctts_entries[0].duration = track->cluster[0][0].cts;
  516. for (i=1; i<track->entry; i++) {
  517. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  518. int id = i % MOV_INDEX_CLUSTER_SIZE;
  519. if (track->cluster[cl][id].cts == ctts_entries[entries].duration) {
  520. ctts_entries[entries].count++; /* compress */
  521. } else {
  522. entries++;
  523. ctts_entries[entries].duration = track->cluster[cl][id].cts;
  524. ctts_entries[entries].count = 1;
  525. }
  526. }
  527. entries++; /* last one */
  528. atom_size = 16 + (entries * 8);
  529. put_be32(pb, atom_size); /* size */
  530. put_tag(pb, "ctts");
  531. put_be32(pb, 0); /* version & flags */
  532. put_be32(pb, entries); /* entry count */
  533. for (i=0; i<entries; i++) {
  534. put_be32(pb, ctts_entries[i].count);
  535. put_be32(pb, ctts_entries[i].duration);
  536. }
  537. av_free(ctts_entries);
  538. return atom_size;
  539. }
  540. /* TODO: */
  541. /* Time to sample atom */
  542. static int mov_write_stts_tag(ByteIOContext *pb, MOVTrack* track)
  543. {
  544. put_be32(pb, 0x18); /* size */
  545. put_tag(pb, "stts");
  546. put_be32(pb, 0); /* version & flags */
  547. put_be32(pb, 1); /* entry count */
  548. put_be32(pb, track->sampleCount); /* sample count */
  549. put_be32(pb, track->sampleDuration); /* sample duration */
  550. return 0x18;
  551. }
  552. static int mov_write_dref_tag(ByteIOContext *pb)
  553. {
  554. put_be32(pb, 28); /* size */
  555. put_tag(pb, "dref");
  556. put_be32(pb, 0); /* version & flags */
  557. put_be32(pb, 1); /* entry count */
  558. put_be32(pb, 0xc); /* size */
  559. put_tag(pb, "url ");
  560. put_be32(pb, 1); /* version & flags */
  561. return 28;
  562. }
  563. static int mov_write_stbl_tag(ByteIOContext *pb, MOVTrack* track)
  564. {
  565. offset_t pos = url_ftell(pb);
  566. put_be32(pb, 0); /* size */
  567. put_tag(pb, "stbl");
  568. mov_write_stsd_tag(pb, track);
  569. mov_write_stts_tag(pb, track);
  570. if (track->enc->codec_type == CODEC_TYPE_VIDEO &&
  571. track->hasKeyframes)
  572. mov_write_stss_tag(pb, track);
  573. if (track->enc->codec_type == CODEC_TYPE_VIDEO &&
  574. track->hasBframes)
  575. mov_write_ctts_tag(pb, track);
  576. mov_write_stsc_tag(pb, track);
  577. mov_write_stsz_tag(pb, track);
  578. mov_write_stco_tag(pb, track);
  579. return updateSize(pb, pos);
  580. }
  581. static int mov_write_dinf_tag(ByteIOContext *pb)
  582. {
  583. offset_t pos = url_ftell(pb);
  584. put_be32(pb, 0); /* size */
  585. put_tag(pb, "dinf");
  586. mov_write_dref_tag(pb);
  587. return updateSize(pb, pos);
  588. }
  589. static int mov_write_smhd_tag(ByteIOContext *pb)
  590. {
  591. put_be32(pb, 16); /* size */
  592. put_tag(pb, "smhd");
  593. put_be32(pb, 0); /* version & flags */
  594. put_be16(pb, 0); /* reserved (balance, normally = 0) */
  595. put_be16(pb, 0); /* reserved */
  596. return 16;
  597. }
  598. static int mov_write_vmhd_tag(ByteIOContext *pb)
  599. {
  600. put_be32(pb, 0x14); /* size (always 0x14) */
  601. put_tag(pb, "vmhd");
  602. put_be32(pb, 0x01); /* version & flags */
  603. put_be64(pb, 0); /* reserved (graphics mode = copy) */
  604. return 0x14;
  605. }
  606. static int mov_write_hdlr_tag(ByteIOContext *pb, MOVTrack* track)
  607. {
  608. char *descr, *hdlr, *hdlr_type;
  609. offset_t pos = url_ftell(pb);
  610. if (!track) { /* no media --> data handler */
  611. hdlr = "dhlr";
  612. hdlr_type = "url ";
  613. descr = "DataHandler";
  614. } else {
  615. hdlr = (track->mode == MODE_MOV) ? "mhlr" : "\0\0\0\0";
  616. if (track->enc->codec_type == CODEC_TYPE_VIDEO) {
  617. hdlr_type = "vide";
  618. descr = "VideoHandler";
  619. } else {
  620. hdlr_type = "soun";
  621. descr = "SoundHandler";
  622. }
  623. }
  624. put_be32(pb, 0); /* size */
  625. put_tag(pb, "hdlr");
  626. put_be32(pb, 0); /* Version & flags */
  627. put_buffer(pb, hdlr, 4); /* handler */
  628. put_tag(pb, hdlr_type); /* handler type */
  629. put_be32(pb ,0); /* reserved */
  630. put_be32(pb ,0); /* reserved */
  631. put_be32(pb ,0); /* reserved */
  632. put_byte(pb, strlen(descr)); /* string counter */
  633. put_buffer(pb, descr, strlen(descr)); /* handler description */
  634. return updateSize(pb, pos);
  635. }
  636. static int mov_write_minf_tag(ByteIOContext *pb, MOVTrack* track)
  637. {
  638. offset_t pos = url_ftell(pb);
  639. put_be32(pb, 0); /* size */
  640. put_tag(pb, "minf");
  641. if(track->enc->codec_type == CODEC_TYPE_VIDEO)
  642. mov_write_vmhd_tag(pb);
  643. else
  644. mov_write_smhd_tag(pb);
  645. if (track->mode == MODE_MOV) /* FIXME: Why do it for MODE_MOV only ? */
  646. mov_write_hdlr_tag(pb, NULL);
  647. mov_write_dinf_tag(pb);
  648. mov_write_stbl_tag(pb, track);
  649. return updateSize(pb, pos);
  650. }
  651. static int mov_write_mdhd_tag(ByteIOContext *pb, MOVTrack* track)
  652. {
  653. put_be32(pb, 32); /* size */
  654. put_tag(pb, "mdhd");
  655. put_be32(pb, 0); /* Version & flags */
  656. put_be32(pb, track->time); /* creation time */
  657. put_be32(pb, track->time); /* modification time */
  658. put_be32(pb, track->timescale); /* time scale (sample rate for audio) */
  659. put_be32(pb, track->trackDuration); /* duration */
  660. put_be16(pb, track->language); /* language */
  661. put_be16(pb, 0); /* reserved (quality) */
  662. return 32;
  663. }
  664. static int mov_write_mdia_tag(ByteIOContext *pb, MOVTrack* track)
  665. {
  666. offset_t pos = url_ftell(pb);
  667. put_be32(pb, 0); /* size */
  668. put_tag(pb, "mdia");
  669. mov_write_mdhd_tag(pb, track);
  670. mov_write_hdlr_tag(pb, track);
  671. mov_write_minf_tag(pb, track);
  672. return updateSize(pb, pos);
  673. }
  674. static int mov_write_tkhd_tag(ByteIOContext *pb, MOVTrack* track)
  675. {
  676. put_be32(pb, 0x5c); /* size (always 0x5c) */
  677. put_tag(pb, "tkhd");
  678. put_be32(pb, 0xf); /* version & flags (track enabled) */
  679. put_be32(pb, track->time); /* creation time */
  680. put_be32(pb, track->time); /* modification time */
  681. put_be32(pb, track->trackID); /* track-id */
  682. put_be32(pb, 0); /* reserved */
  683. put_be32(pb, av_rescale_rnd(track->trackDuration, globalTimescale, track->timescale, AV_ROUND_UP)); /* duration */
  684. put_be32(pb, 0); /* reserved */
  685. put_be32(pb, 0); /* reserved */
  686. put_be32(pb, 0x0); /* reserved (Layer & Alternate group) */
  687. /* Volume, only for audio */
  688. if(track->enc->codec_type == CODEC_TYPE_AUDIO)
  689. put_be16(pb, 0x0100);
  690. else
  691. put_be16(pb, 0);
  692. put_be16(pb, 0); /* reserved */
  693. /* Matrix structure */
  694. put_be32(pb, 0x00010000); /* reserved */
  695. put_be32(pb, 0x0); /* reserved */
  696. put_be32(pb, 0x0); /* reserved */
  697. put_be32(pb, 0x0); /* reserved */
  698. put_be32(pb, 0x00010000); /* reserved */
  699. put_be32(pb, 0x0); /* reserved */
  700. put_be32(pb, 0x0); /* reserved */
  701. put_be32(pb, 0x0); /* reserved */
  702. put_be32(pb, 0x40000000); /* reserved */
  703. /* Track width and height, for visual only */
  704. if(track->enc->codec_type == CODEC_TYPE_VIDEO) {
  705. double sample_aspect_ratio = av_q2d(track->enc->sample_aspect_ratio);
  706. if( !sample_aspect_ratio ) sample_aspect_ratio = 1;
  707. put_be32(pb, sample_aspect_ratio * track->enc->width*0x10000);
  708. put_be32(pb, track->enc->height*0x10000);
  709. }
  710. else {
  711. put_be32(pb, 0);
  712. put_be32(pb, 0);
  713. }
  714. return 0x5c;
  715. }
  716. // This box seems important for the psp playback ... without it the movie seems to hang
  717. static int mov_write_edts_tag(ByteIOContext *pb, MOVTrack *track)
  718. {
  719. put_be32(pb, 0x24); /* size */
  720. put_tag(pb, "edts");
  721. put_be32(pb, 0x1c); /* size */
  722. put_tag(pb, "elst");
  723. put_be32(pb, 0x0);
  724. put_be32(pb, 0x1);
  725. put_be32(pb, av_rescale_rnd(track->trackDuration, globalTimescale, track->timescale, AV_ROUND_UP)); /* duration ... doesn't seem to effect psp */
  726. put_be32(pb, 0x0);
  727. put_be32(pb, 0x00010000);
  728. return 0x24;
  729. }
  730. // goes at the end of each track! ... Critical for PSP playback ("Incompatible data" without it)
  731. static int mov_write_uuid_tag_psp(ByteIOContext *pb, MOVTrack *mov)
  732. {
  733. put_be32(pb, 0x34); /* size ... reports as 28 in mp4box! */
  734. put_tag(pb, "uuid");
  735. put_tag(pb, "USMT");
  736. put_be32(pb, 0x21d24fce);
  737. put_be32(pb, 0xbb88695c);
  738. put_be32(pb, 0xfac9c740);
  739. put_be32(pb, 0x1c); // another size here!
  740. put_tag(pb, "MTDT");
  741. put_be32(pb, 0x00010012);
  742. put_be32(pb, 0x0a);
  743. put_be32(pb, 0x55c40000);
  744. put_be32(pb, 0x1);
  745. put_be32(pb, 0x0);
  746. return 0x34;
  747. }
  748. static int mov_write_trak_tag(ByteIOContext *pb, MOVTrack* track)
  749. {
  750. offset_t pos = url_ftell(pb);
  751. put_be32(pb, 0); /* size */
  752. put_tag(pb, "trak");
  753. mov_write_tkhd_tag(pb, track);
  754. if (track->mode == MODE_PSP)
  755. mov_write_edts_tag(pb, track); // PSP Movies require edts box
  756. mov_write_mdia_tag(pb, track);
  757. if (track->mode == MODE_PSP)
  758. mov_write_uuid_tag_psp(pb,track); // PSP Movies require this uuid box
  759. return updateSize(pb, pos);
  760. }
  761. #if 0
  762. /* TODO: Not sorted out, but not necessary either */
  763. static int mov_write_iods_tag(ByteIOContext *pb, MOVContext *mov)
  764. {
  765. put_be32(pb, 0x15); /* size */
  766. put_tag(pb, "iods");
  767. put_be32(pb, 0); /* version & flags */
  768. put_be16(pb, 0x1007);
  769. put_byte(pb, 0);
  770. put_be16(pb, 0x4fff);
  771. put_be16(pb, 0xfffe);
  772. put_be16(pb, 0x01ff);
  773. return 0x15;
  774. }
  775. #endif
  776. static int mov_write_mvhd_tag(ByteIOContext *pb, MOVContext *mov)
  777. {
  778. int maxTrackID = 1, i;
  779. int64_t maxTrackLenTemp, maxTrackLen = 0;
  780. put_be32(pb, 0x6c); /* size (always 0x6c) */
  781. put_tag(pb, "mvhd");
  782. put_be32(pb, 0); /* version & flags */
  783. put_be32(pb, mov->time); /* creation time */
  784. put_be32(pb, mov->time); /* modification time */
  785. put_be32(pb, mov->timescale); /* timescale */
  786. for (i=0; i<MAX_STREAMS; i++) {
  787. if(mov->tracks[i].entry > 0) {
  788. maxTrackLenTemp = av_rescale_rnd(mov->tracks[i].trackDuration, globalTimescale, mov->tracks[i].timescale, AV_ROUND_UP);
  789. if(maxTrackLen < maxTrackLenTemp)
  790. maxTrackLen = maxTrackLenTemp;
  791. if(maxTrackID < mov->tracks[i].trackID)
  792. maxTrackID = mov->tracks[i].trackID;
  793. }
  794. }
  795. put_be32(pb, maxTrackLen); /* duration of longest track */
  796. put_be32(pb, 0x00010000); /* reserved (preferred rate) 1.0 = normal */
  797. put_be16(pb, 0x0100); /* reserved (preferred volume) 1.0 = normal */
  798. put_be16(pb, 0); /* reserved */
  799. put_be32(pb, 0); /* reserved */
  800. put_be32(pb, 0); /* reserved */
  801. /* Matrix structure */
  802. put_be32(pb, 0x00010000); /* reserved */
  803. put_be32(pb, 0x0); /* reserved */
  804. put_be32(pb, 0x0); /* reserved */
  805. put_be32(pb, 0x0); /* reserved */
  806. put_be32(pb, 0x00010000); /* reserved */
  807. put_be32(pb, 0x0); /* reserved */
  808. put_be32(pb, 0x0); /* reserved */
  809. put_be32(pb, 0x0); /* reserved */
  810. put_be32(pb, 0x40000000); /* reserved */
  811. put_be32(pb, 0); /* reserved (preview time) */
  812. put_be32(pb, 0); /* reserved (preview duration) */
  813. put_be32(pb, 0); /* reserved (poster time) */
  814. put_be32(pb, 0); /* reserved (selection time) */
  815. put_be32(pb, 0); /* reserved (selection duration) */
  816. put_be32(pb, 0); /* reserved (current time) */
  817. put_be32(pb, maxTrackID+1); /* Next track id */
  818. return 0x6c;
  819. }
  820. static int mov_write_itunes_hdlr_tag(ByteIOContext *pb, MOVContext* mov,
  821. AVFormatContext *s)
  822. {
  823. offset_t pos = url_ftell(pb);
  824. put_be32(pb, 0); /* size */
  825. put_tag(pb, "hdlr");
  826. put_be32(pb, 0);
  827. put_be32(pb, 0);
  828. put_tag(pb, "mdir");
  829. put_tag(pb, "appl");
  830. put_be32(pb, 0);
  831. put_be32(pb, 0);
  832. put_be16(pb, 0);
  833. return updateSize(pb, pos);
  834. }
  835. /* helper function to write a data tag with the specified string as data */
  836. static int mov_write_string_data_tag(ByteIOContext *pb, MOVContext* mov,
  837. AVFormatContext *s, const char *data)
  838. {
  839. offset_t pos = url_ftell(pb);
  840. put_be32(pb, 0); /* size */
  841. put_tag(pb, "data");
  842. put_be32(pb, 1);
  843. put_be32(pb, 0);
  844. put_buffer(pb, data, strlen(data));
  845. return updateSize(pb, pos);
  846. }
  847. /* iTunes name of the song/movie */
  848. static int mov_write_nam_tag(ByteIOContext *pb, MOVContext* mov,
  849. AVFormatContext *s)
  850. {
  851. int size = 0;
  852. if ( s->title[0] ) {
  853. offset_t pos = url_ftell(pb);
  854. put_be32(pb, 0); /* size */
  855. put_tag(pb, "\251nam");
  856. mov_write_string_data_tag(pb, mov, s, s->title);
  857. size = updateSize(pb, pos);
  858. }
  859. return size;
  860. }
  861. /* iTunes name of the artist/performer */
  862. static int mov_write_ART_tag(ByteIOContext *pb, MOVContext* mov,
  863. AVFormatContext *s)
  864. {
  865. int size = 0;
  866. if ( s->author[0] ) {
  867. offset_t pos = url_ftell(pb);
  868. put_be32(pb, 0); /* size */
  869. put_tag(pb, "\251ART");
  870. // we use the author here as this is the only thing that we have...
  871. mov_write_string_data_tag(pb, mov, s, s->author);
  872. size = updateSize(pb, pos);
  873. }
  874. return size;
  875. }
  876. /* iTunes name of the writer */
  877. static int mov_write_wrt_tag(ByteIOContext *pb, MOVContext* mov,
  878. AVFormatContext *s)
  879. {
  880. int size = 0;
  881. if ( s->author[0] ) {
  882. offset_t pos = url_ftell(pb);
  883. put_be32(pb, 0); /* size */
  884. put_tag(pb, "\251wrt");
  885. mov_write_string_data_tag(pb, mov, s, s->author);
  886. size = updateSize(pb, pos);
  887. }
  888. return size;
  889. }
  890. /* iTunes name of the album */
  891. static int mov_write_alb_tag(ByteIOContext *pb, MOVContext* mov,
  892. AVFormatContext *s)
  893. {
  894. int size = 0;
  895. if ( s->album[0] ) {
  896. offset_t pos = url_ftell(pb);
  897. put_be32(pb, 0); /* size */
  898. put_tag(pb, "\251alb");
  899. mov_write_string_data_tag(pb, mov, s, s->album);
  900. size = updateSize(pb, pos);
  901. }
  902. return size;
  903. }
  904. /* iTunes year */
  905. static int mov_write_day_tag(ByteIOContext *pb, MOVContext* mov,
  906. AVFormatContext *s)
  907. {
  908. char year[5];
  909. int size = 0;
  910. if ( s->year ) {
  911. offset_t pos = url_ftell(pb);
  912. put_be32(pb, 0); /* size */
  913. put_tag(pb, "\251day");
  914. snprintf(year, 5, "%04d", s->year);
  915. mov_write_string_data_tag(pb, mov, s, year);
  916. size = updateSize(pb, pos);
  917. }
  918. return size;
  919. }
  920. /* iTunes tool used to create the file */
  921. static int mov_write_too_tag(ByteIOContext *pb, MOVContext* mov,
  922. AVFormatContext *s)
  923. {
  924. offset_t pos = url_ftell(pb);
  925. put_be32(pb, 0); /* size */
  926. put_tag(pb, "\251too");
  927. mov_write_string_data_tag(pb, mov, s, LIBAVFORMAT_IDENT);
  928. return updateSize(pb, pos);
  929. }
  930. /* iTunes comment */
  931. static int mov_write_cmt_tag(ByteIOContext *pb, MOVContext* mov,
  932. AVFormatContext *s)
  933. {
  934. int size = 0;
  935. if ( s->comment[0] ) {
  936. offset_t pos = url_ftell(pb);
  937. put_be32(pb, 0); /* size */
  938. put_tag(pb, "\251cmt");
  939. mov_write_string_data_tag(pb, mov, s, s->comment);
  940. size = updateSize(pb, pos);
  941. }
  942. return size;
  943. }
  944. /* iTunes custom genre */
  945. static int mov_write_gen_tag(ByteIOContext *pb, MOVContext* mov,
  946. AVFormatContext *s)
  947. {
  948. int size = 0;
  949. if ( s->genre[0] ) {
  950. offset_t pos = url_ftell(pb);
  951. put_be32(pb, 0); /* size */
  952. put_tag(pb, "\251gen");
  953. mov_write_string_data_tag(pb, mov, s, s->genre);
  954. size = updateSize(pb, pos);
  955. }
  956. return size;
  957. }
  958. /* iTunes track number */
  959. static int mov_write_trkn_tag(ByteIOContext *pb, MOVContext* mov,
  960. AVFormatContext *s)
  961. {
  962. int size = 0;
  963. if ( s->track ) {
  964. offset_t pos = url_ftell(pb);
  965. put_be32(pb, 0); /* size */
  966. put_tag(pb, "trkn");
  967. {
  968. offset_t pos = url_ftell(pb);
  969. put_be32(pb, 0); /* size */
  970. put_tag(pb, "data");
  971. put_be32(pb, 0); // 8 bytes empty
  972. put_be32(pb, 0);
  973. put_be16(pb, 0); // empty
  974. put_be16(pb, s->track); // track number
  975. put_be16(pb, 0); // total track number
  976. put_be16(pb, 0); // empty
  977. updateSize(pb, pos);
  978. }
  979. size = updateSize(pb, pos);
  980. }
  981. return size;
  982. }
  983. /* iTunes meta data list */
  984. static int mov_write_ilst_tag(ByteIOContext *pb, MOVContext* mov,
  985. AVFormatContext *s)
  986. {
  987. offset_t pos = url_ftell(pb);
  988. put_be32(pb, 0); /* size */
  989. put_tag(pb, "ilst");
  990. mov_write_nam_tag(pb, mov, s);
  991. mov_write_ART_tag(pb, mov, s);
  992. mov_write_wrt_tag(pb, mov, s);
  993. mov_write_alb_tag(pb, mov, s);
  994. mov_write_day_tag(pb, mov, s);
  995. mov_write_too_tag(pb, mov, s);
  996. mov_write_cmt_tag(pb, mov, s);
  997. mov_write_gen_tag(pb, mov, s);
  998. mov_write_trkn_tag(pb, mov, s);
  999. return updateSize(pb, pos);
  1000. }
  1001. /* iTunes meta data tag */
  1002. static int mov_write_meta_tag(ByteIOContext *pb, MOVContext* mov,
  1003. AVFormatContext *s)
  1004. {
  1005. int size = 0;
  1006. // only save meta tag if required
  1007. if ( s->title[0] || s->author[0] || s->album[0] || s->year ||
  1008. s->comment[0] || s->genre[0] || s->track ) {
  1009. offset_t pos = url_ftell(pb);
  1010. put_be32(pb, 0); /* size */
  1011. put_tag(pb, "meta");
  1012. put_be32(pb, 0);
  1013. mov_write_itunes_hdlr_tag(pb, mov, s);
  1014. mov_write_ilst_tag(pb, mov, s);
  1015. size = updateSize(pb, pos);
  1016. }
  1017. return size;
  1018. }
  1019. static int mov_write_udta_tag(ByteIOContext *pb, MOVContext* mov,
  1020. AVFormatContext *s)
  1021. {
  1022. offset_t pos = url_ftell(pb);
  1023. int i;
  1024. put_be32(pb, 0); /* size */
  1025. put_tag(pb, "udta");
  1026. /* iTunes meta data */
  1027. mov_write_meta_tag(pb, mov, s);
  1028. /* Requirements */
  1029. for (i=0; i<MAX_STREAMS; i++) {
  1030. if(mov->tracks[i].entry <= 0) continue;
  1031. if (mov->tracks[i].enc->codec_id == CODEC_ID_AAC ||
  1032. mov->tracks[i].enc->codec_id == CODEC_ID_MPEG4) {
  1033. offset_t pos = url_ftell(pb);
  1034. put_be32(pb, 0); /* size */
  1035. put_tag(pb, "\251req");
  1036. put_be16(pb, sizeof("QuickTime 6.0 or greater") - 1);
  1037. put_be16(pb, 0);
  1038. put_buffer(pb, "QuickTime 6.0 or greater",
  1039. sizeof("QuickTime 6.0 or greater") - 1);
  1040. updateSize(pb, pos);
  1041. break;
  1042. }
  1043. }
  1044. /* Encoder */
  1045. if(mov->tracks[0].enc && !(mov->tracks[0].enc->flags & CODEC_FLAG_BITEXACT))
  1046. {
  1047. offset_t pos = url_ftell(pb);
  1048. put_be32(pb, 0); /* size */
  1049. put_tag(pb, "\251enc");
  1050. put_be16(pb, sizeof(LIBAVFORMAT_IDENT) - 1); /* string length */
  1051. put_be16(pb, 0);
  1052. put_buffer(pb, LIBAVFORMAT_IDENT, sizeof(LIBAVFORMAT_IDENT) - 1);
  1053. updateSize(pb, pos);
  1054. }
  1055. if( s->title[0] )
  1056. {
  1057. offset_t pos = url_ftell(pb);
  1058. put_be32(pb, 0); /* size */
  1059. put_tag(pb, "\251nam");
  1060. put_be16(pb, strlen(s->title)); /* string length */
  1061. put_be16(pb, 0);
  1062. put_buffer(pb, s->title, strlen(s->title));
  1063. updateSize(pb, pos);
  1064. }
  1065. if( s->author[0] )
  1066. {
  1067. offset_t pos = url_ftell(pb);
  1068. put_be32(pb, 0); /* size */
  1069. put_tag(pb, /*"\251aut"*/ "\251day" );
  1070. put_be16(pb, strlen(s->author)); /* string length */
  1071. put_be16(pb, 0);
  1072. put_buffer(pb, s->author, strlen(s->author));
  1073. updateSize(pb, pos);
  1074. }
  1075. if( s->comment[0] )
  1076. {
  1077. offset_t pos = url_ftell(pb);
  1078. put_be32(pb, 0); /* size */
  1079. put_tag(pb, "\251des");
  1080. put_be16(pb, strlen(s->comment)); /* string length */
  1081. put_be16(pb, 0);
  1082. put_buffer(pb, s->comment, strlen(s->comment));
  1083. updateSize(pb, pos);
  1084. }
  1085. return updateSize(pb, pos);
  1086. }
  1087. static size_t ascii_to_wc (ByteIOContext *pb, char *b, size_t n)
  1088. {
  1089. size_t i;
  1090. unsigned char c;
  1091. for (i = 0; i < n - 1; i++) {
  1092. c = b[i];
  1093. if (! (0x20 <= c && c <= 0x7f ))
  1094. c = 0x3f; /* '?' */
  1095. put_be16(pb, c);
  1096. }
  1097. put_be16(pb, 0x00);
  1098. return 2*n;
  1099. }
  1100. static uint16_t language_code (char *str)
  1101. {
  1102. return ((((str[0]-'a') & 0x1F)<<10) + (((str[1]-'a') & 0x1F)<<5) + ((str[2]-'a') & 0x1F));
  1103. }
  1104. static int mov_write_uuidusmt_tag (ByteIOContext *pb, AVFormatContext *s)
  1105. {
  1106. size_t len, size;
  1107. offset_t pos, curpos;
  1108. size = 0;
  1109. if (s->title[0]) {
  1110. pos = url_ftell(pb);
  1111. put_be32(pb, 0); /* size placeholder*/
  1112. put_tag(pb, "uuid");
  1113. put_tag(pb, "USMT");
  1114. put_be32(pb, 0x21d24fce ); /* 96 bit UUID */
  1115. put_be32(pb, 0xbb88695c );
  1116. put_be32(pb, 0xfac9c740 );
  1117. size += 24;
  1118. put_be32(pb, 0); /* size placeholder*/
  1119. put_tag(pb, "MTDT");
  1120. put_be16(pb, 1);
  1121. size += 10;
  1122. // Title
  1123. len = strlen(s->title)+1;
  1124. put_be16(pb, len*2+10); /* size */
  1125. put_be32(pb, 0x01); /* type */
  1126. put_be16(pb, language_code("und")); /* language */
  1127. put_be16(pb, 0x01); /* ? */
  1128. ascii_to_wc (pb, s->title, len);
  1129. size += len*2+10;
  1130. // size
  1131. curpos = url_ftell(pb);
  1132. url_fseek(pb, pos, SEEK_SET);
  1133. put_be32(pb, size);
  1134. url_fseek(pb, pos+24, SEEK_SET);
  1135. put_be32(pb, size-24);
  1136. url_fseek(pb, curpos, SEEK_SET);
  1137. }
  1138. return size;
  1139. }
  1140. static int mov_write_moov_tag(ByteIOContext *pb, MOVContext *mov,
  1141. AVFormatContext *s)
  1142. {
  1143. int i;
  1144. offset_t pos = url_ftell(pb);
  1145. put_be32(pb, 0); /* size placeholder*/
  1146. put_tag(pb, "moov");
  1147. mov->timescale = globalTimescale;
  1148. for (i=0; i<MAX_STREAMS; i++) {
  1149. if(mov->tracks[i].entry <= 0) continue;
  1150. if(mov->tracks[i].enc->codec_type == CODEC_TYPE_VIDEO) {
  1151. mov->tracks[i].timescale = mov->tracks[i].enc->time_base.den;
  1152. mov->tracks[i].sampleDuration = mov->tracks[i].enc->time_base.num;
  1153. }
  1154. else if(mov->tracks[i].enc->codec_type == CODEC_TYPE_AUDIO) {
  1155. /* If AMR, track timescale = 8000, AMR_WB = 16000 */
  1156. if(mov->tracks[i].enc->codec_id == CODEC_ID_AMR_NB) {
  1157. mov->tracks[i].sampleDuration = 160; // Bytes per chunk
  1158. mov->tracks[i].timescale = 8000;
  1159. }
  1160. else {
  1161. mov->tracks[i].timescale = mov->tracks[i].enc->sample_rate;
  1162. mov->tracks[i].sampleDuration = mov->tracks[i].enc->frame_size;
  1163. }
  1164. }
  1165. mov->tracks[i].trackDuration =
  1166. mov->tracks[i].sampleCount * mov->tracks[i].sampleDuration;
  1167. mov->tracks[i].time = mov->time;
  1168. mov->tracks[i].trackID = i+1;
  1169. }
  1170. mov_write_mvhd_tag(pb, mov);
  1171. //mov_write_iods_tag(pb, mov);
  1172. for (i=0; i<MAX_STREAMS; i++) {
  1173. if(mov->tracks[i].entry > 0) {
  1174. mov_write_trak_tag(pb, &(mov->tracks[i]));
  1175. }
  1176. }
  1177. if (mov->mode == MODE_PSP)
  1178. mov_write_uuidusmt_tag(pb, s);
  1179. else
  1180. mov_write_udta_tag(pb, mov, s);
  1181. return updateSize(pb, pos);
  1182. }
  1183. int mov_write_mdat_tag(ByteIOContext *pb, MOVContext* mov)
  1184. {
  1185. put_be32(pb, 8); // placeholder for extended size field (64 bit)
  1186. put_tag(pb, "wide");
  1187. mov->mdat_pos = url_ftell(pb);
  1188. put_be32(pb, 0); /* size placeholder*/
  1189. put_tag(pb, "mdat");
  1190. return 0;
  1191. }
  1192. /* TODO: This needs to be more general */
  1193. static void mov_write_ftyp_tag (ByteIOContext *pb, AVFormatContext *s)
  1194. {
  1195. MOVContext *mov = s->priv_data;
  1196. put_be32(pb, 0x14 ); /* size */
  1197. put_tag(pb, "ftyp");
  1198. if ( mov->mode == MODE_3GP )
  1199. put_tag(pb, "3gp4");
  1200. else if ( mov->mode == MODE_3G2 )
  1201. put_tag(pb, "3g2a");
  1202. else if ( mov->mode == MODE_PSP )
  1203. put_tag(pb, "MSNV");
  1204. else if ( mov->mode == MODE_MP4 )
  1205. put_tag(pb, "isom");
  1206. else
  1207. put_tag(pb, "qt ");
  1208. put_be32(pb, 0x200 );
  1209. if ( mov->mode == MODE_3GP )
  1210. put_tag(pb, "3gp4");
  1211. else if ( mov->mode == MODE_3G2 )
  1212. put_tag(pb, "3g2a");
  1213. else if ( mov->mode == MODE_PSP )
  1214. put_tag(pb, "MSNV");
  1215. else if ( mov->mode == MODE_MP4 )
  1216. put_tag(pb, "mp41");
  1217. else
  1218. put_tag(pb, "qt ");
  1219. }
  1220. static void mov_write_uuidprof_tag(ByteIOContext *pb, AVFormatContext *s)
  1221. {
  1222. AVCodecContext *VideoCodec = s->streams[0]->codec;
  1223. AVCodecContext *AudioCodec = s->streams[1]->codec;
  1224. int AudioRate = AudioCodec->sample_rate;
  1225. int FrameRate = ((VideoCodec->time_base.den) * (0x10000))/ (VideoCodec->time_base.num);
  1226. put_be32(pb, 0x94 ); /* size */
  1227. put_tag(pb, "uuid");
  1228. put_tag(pb, "PROF");
  1229. put_be32(pb, 0x21d24fce ); /* 96 bit UUID */
  1230. put_be32(pb, 0xbb88695c );
  1231. put_be32(pb, 0xfac9c740 );
  1232. put_be32(pb, 0x0 ); /* ? */
  1233. put_be32(pb, 0x3 ); /* 3 sections ? */
  1234. put_be32(pb, 0x14 ); /* size */
  1235. put_tag(pb, "FPRF");
  1236. put_be32(pb, 0x0 ); /* ? */
  1237. put_be32(pb, 0x0 ); /* ? */
  1238. put_be32(pb, 0x0 ); /* ? */
  1239. put_be32(pb, 0x2c ); /* size */
  1240. put_tag(pb, "APRF"); /* audio */
  1241. put_be32(pb, 0x0 );
  1242. put_be32(pb, 0x2 ); /* TrackID */
  1243. put_tag(pb, "mp4a");
  1244. put_be32(pb, 0x20f );
  1245. put_be32(pb, 0x0 );
  1246. put_be32(pb, AudioCodec->bit_rate / 1000);
  1247. put_be32(pb, AudioCodec->bit_rate / 1000);
  1248. put_be32(pb, AudioRate );
  1249. put_be32(pb, AudioCodec->channels );
  1250. put_be32(pb, 0x34 ); /* size */
  1251. put_tag(pb, "VPRF"); /* video */
  1252. put_be32(pb, 0x0 );
  1253. put_be32(pb, 0x1 ); /* TrackID */
  1254. put_tag(pb, "mp4v");
  1255. put_be32(pb, 0x103 );
  1256. put_be32(pb, 0x0 );
  1257. put_be32(pb, VideoCodec->bit_rate / 1000);
  1258. put_be32(pb, VideoCodec->bit_rate / 1000);
  1259. put_be32(pb, FrameRate);
  1260. put_be32(pb, FrameRate);
  1261. put_be16(pb, VideoCodec->width);
  1262. put_be16(pb, VideoCodec->height);
  1263. put_be32(pb, 0x010001); /* ? */
  1264. }
  1265. static int mov_write_header(AVFormatContext *s)
  1266. {
  1267. ByteIOContext *pb = &s->pb;
  1268. MOVContext *mov = s->priv_data;
  1269. int i;
  1270. /* Default mode == MP4 */
  1271. mov->mode = MODE_MP4;
  1272. if (s->oformat != NULL) {
  1273. if (!strcmp("3gp", s->oformat->name)) mov->mode = MODE_3GP;
  1274. else if (!strcmp("3g2", s->oformat->name)) mov->mode = MODE_3G2;
  1275. else if (!strcmp("mov", s->oformat->name)) mov->mode = MODE_MOV;
  1276. else if (!strcmp("psp", s->oformat->name)) mov->mode = MODE_PSP;
  1277. mov_write_ftyp_tag(pb,s);
  1278. if ( mov->mode == MODE_PSP ) {
  1279. if ( s->nb_streams != 2 ) {
  1280. av_log(s, AV_LOG_ERROR, "PSP mode need one video and one audio stream\n");
  1281. return -1;
  1282. }
  1283. mov_write_uuidprof_tag(pb,s);
  1284. }
  1285. }
  1286. for(i=0; i<s->nb_streams; i++){
  1287. AVCodecContext *c= s->streams[i]->codec;
  1288. if(c->codec_type == CODEC_TYPE_VIDEO){
  1289. av_set_pts_info(s->streams[i], 64, 1, c->time_base.den);
  1290. if (!codec_get_tag(codec_movvideo_tags, c->codec_id)){
  1291. if(!codec_get_tag(codec_bmp_tags, c->codec_id))
  1292. return -1;
  1293. else
  1294. av_log(s, AV_LOG_INFO, "Warning, using MS style video codec tag, the file may be unplayable!\n");
  1295. }
  1296. }else if(c->codec_type == CODEC_TYPE_AUDIO){
  1297. av_set_pts_info(s->streams[i], 64, 1, c->sample_rate);
  1298. if (!codec_get_tag(codec_movaudio_tags, c->codec_id)){
  1299. if(!codec_get_tag(codec_wav_tags, c->codec_id))
  1300. return -1;
  1301. else
  1302. av_log(s, AV_LOG_INFO, "Warning, using MS style audio codec tag, the file may be unplayable!\n");
  1303. }
  1304. }
  1305. mov->tracks[i].language = ff_mov_iso639_to_lang(s->streams[i]->language, mov->mode != MODE_MOV);
  1306. }
  1307. for (i=0; i<MAX_STREAMS; i++) {
  1308. mov->tracks[i].mode = mov->mode;
  1309. }
  1310. put_flush_packet(pb);
  1311. return 0;
  1312. }
  1313. static int mov_write_packet(AVFormatContext *s, AVPacket *pkt)
  1314. {
  1315. MOVContext *mov = s->priv_data;
  1316. ByteIOContext *pb = &s->pb;
  1317. AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
  1318. MOVTrack* trk = &mov->tracks[pkt->stream_index];
  1319. int cl, id;
  1320. unsigned int samplesInChunk = 0;
  1321. int size= pkt->size;
  1322. if (url_is_streamed(&s->pb)) return 0; /* Can't handle that */
  1323. if (!size) return 0; /* Discard 0 sized packets */
  1324. if (enc->codec_type == CODEC_TYPE_VIDEO ) {
  1325. samplesInChunk = 1;
  1326. }
  1327. else if (enc->codec_type == CODEC_TYPE_AUDIO ) {
  1328. if( enc->codec_id == CODEC_ID_AMR_NB) {
  1329. /* We must find out how many AMR blocks there are in one packet */
  1330. static uint16_t packed_size[16] =
  1331. {13, 14, 16, 18, 20, 21, 27, 32, 6, 0, 0, 0, 0, 0, 0, 0};
  1332. int len = 0;
  1333. while (len < size && samplesInChunk < 100) {
  1334. len += packed_size[(pkt->data[len] >> 3) & 0x0F];
  1335. samplesInChunk++;
  1336. }
  1337. }
  1338. else if(enc->codec_id == CODEC_ID_PCM_ALAW) {
  1339. samplesInChunk = size/enc->channels;
  1340. }
  1341. else if(enc->codec_id == CODEC_ID_PCM_S16BE || enc->codec_id == CODEC_ID_PCM_S16LE) {
  1342. samplesInChunk = size/(2*enc->channels);
  1343. }
  1344. else {
  1345. samplesInChunk = 1;
  1346. }
  1347. }
  1348. if ((enc->codec_id == CODEC_ID_MPEG4 || enc->codec_id == CODEC_ID_AAC)
  1349. && trk->vosLen == 0) {
  1350. // assert(enc->extradata_size);
  1351. trk->vosLen = enc->extradata_size;
  1352. trk->vosData = av_malloc(trk->vosLen);
  1353. memcpy(trk->vosData, enc->extradata, trk->vosLen);
  1354. }
  1355. cl = trk->entry / MOV_INDEX_CLUSTER_SIZE;
  1356. id = trk->entry % MOV_INDEX_CLUSTER_SIZE;
  1357. if (trk->ents_allocated <= trk->entry) {
  1358. trk->cluster = av_realloc(trk->cluster, (cl+1)*sizeof(void*));
  1359. if (!trk->cluster)
  1360. return -1;
  1361. trk->cluster[cl] = av_malloc(MOV_INDEX_CLUSTER_SIZE*sizeof(MOVIentry));
  1362. if (!trk->cluster[cl])
  1363. return -1;
  1364. trk->ents_allocated += MOV_INDEX_CLUSTER_SIZE;
  1365. }
  1366. if (mov->mdat_written == 0) {
  1367. mov_write_mdat_tag(pb, mov);
  1368. mov->mdat_written = 1;
  1369. mov->time = s->timestamp + 0x7C25B080; //1970 based -> 1904 based
  1370. }
  1371. trk->cluster[cl][id].pos = url_ftell(pb);
  1372. trk->cluster[cl][id].samplesInChunk = samplesInChunk;
  1373. trk->cluster[cl][id].size = size;
  1374. trk->cluster[cl][id].entries = samplesInChunk;
  1375. if(enc->codec_type == CODEC_TYPE_VIDEO) {
  1376. if (pkt->dts != pkt->pts)
  1377. trk->hasBframes = 1;
  1378. trk->cluster[cl][id].cts = pkt->pts - pkt->dts;
  1379. trk->cluster[cl][id].key_frame = !!(pkt->flags & PKT_FLAG_KEY);
  1380. if(trk->cluster[cl][id].key_frame)
  1381. trk->hasKeyframes = 1;
  1382. }
  1383. trk->enc = enc;
  1384. trk->entry++;
  1385. trk->sampleCount += samplesInChunk;
  1386. trk->mdat_size += size;
  1387. put_buffer(pb, pkt->data, size);
  1388. put_flush_packet(pb);
  1389. return 0;
  1390. }
  1391. static int mov_write_trailer(AVFormatContext *s)
  1392. {
  1393. MOVContext *mov = s->priv_data;
  1394. ByteIOContext *pb = &s->pb;
  1395. int res = 0;
  1396. int i;
  1397. uint64_t j;
  1398. offset_t moov_pos = url_ftell(pb);
  1399. /* Write size of mdat tag */
  1400. for (i=0, j=0; i<MAX_STREAMS; i++) {
  1401. if(mov->tracks[i].ents_allocated > 0) {
  1402. j += mov->tracks[i].mdat_size;
  1403. }
  1404. }
  1405. if (j+8 <= UINT32_MAX) {
  1406. url_fseek(pb, mov->mdat_pos, SEEK_SET);
  1407. put_be32(pb, j+8);
  1408. } else {
  1409. /* overwrite 'wide' placeholder atom */
  1410. url_fseek(pb, mov->mdat_pos - 8, SEEK_SET);
  1411. put_be32(pb, 1); /* special value: real atom size will be 64 bit value after tag field */
  1412. put_tag(pb, "mdat");
  1413. put_be64(pb, j+16);
  1414. }
  1415. url_fseek(pb, moov_pos, SEEK_SET);
  1416. mov_write_moov_tag(pb, mov, s);
  1417. for (i=0; i<MAX_STREAMS; i++) {
  1418. for (j=0; j<mov->tracks[i].ents_allocated/MOV_INDEX_CLUSTER_SIZE; j++) {
  1419. av_free(mov->tracks[i].cluster[j]);
  1420. }
  1421. av_free(mov->tracks[i].cluster);
  1422. if( mov->tracks[i].vosLen ) av_free( mov->tracks[i].vosData );
  1423. mov->tracks[i].cluster = NULL;
  1424. mov->tracks[i].ents_allocated = mov->tracks[i].entry = 0;
  1425. }
  1426. put_flush_packet(pb);
  1427. return res;
  1428. }
  1429. static AVOutputFormat mov_oformat = {
  1430. "mov",
  1431. "mov format",
  1432. NULL,
  1433. "mov",
  1434. sizeof(MOVContext),
  1435. CODEC_ID_AAC,
  1436. CODEC_ID_MPEG4,
  1437. mov_write_header,
  1438. mov_write_packet,
  1439. mov_write_trailer,
  1440. .flags = AVFMT_GLOBALHEADER,
  1441. };
  1442. static AVOutputFormat _3gp_oformat = {
  1443. "3gp",
  1444. "3gp format",
  1445. NULL,
  1446. "3gp",
  1447. sizeof(MOVContext),
  1448. CODEC_ID_AMR_NB,
  1449. CODEC_ID_H263,
  1450. mov_write_header,
  1451. mov_write_packet,
  1452. mov_write_trailer,
  1453. .flags = AVFMT_GLOBALHEADER,
  1454. };
  1455. static AVOutputFormat mp4_oformat = {
  1456. "mp4",
  1457. "mp4 format",
  1458. "application/mp4",
  1459. "mp4,m4a",
  1460. sizeof(MOVContext),
  1461. CODEC_ID_AAC,
  1462. CODEC_ID_MPEG4,
  1463. mov_write_header,
  1464. mov_write_packet,
  1465. mov_write_trailer,
  1466. .flags = AVFMT_GLOBALHEADER,
  1467. };
  1468. static AVOutputFormat psp_oformat = {
  1469. "psp",
  1470. "psp mp4 format",
  1471. NULL,
  1472. "mp4,psp",
  1473. sizeof(MOVContext),
  1474. CODEC_ID_AAC,
  1475. CODEC_ID_MPEG4,
  1476. mov_write_header,
  1477. mov_write_packet,
  1478. mov_write_trailer,
  1479. .flags = AVFMT_GLOBALHEADER,
  1480. };
  1481. static AVOutputFormat _3g2_oformat = {
  1482. "3g2",
  1483. "3gp2 format",
  1484. NULL,
  1485. "3g2",
  1486. sizeof(MOVContext),
  1487. CODEC_ID_AMR_NB,
  1488. CODEC_ID_H263,
  1489. mov_write_header,
  1490. mov_write_packet,
  1491. mov_write_trailer,
  1492. .flags = AVFMT_GLOBALHEADER,
  1493. };
  1494. int movenc_init(void)
  1495. {
  1496. av_register_output_format(&mov_oformat);
  1497. av_register_output_format(&_3gp_oformat);
  1498. av_register_output_format(&mp4_oformat);
  1499. av_register_output_format(&psp_oformat);
  1500. av_register_output_format(&_3g2_oformat);
  1501. return 0;
  1502. }