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.

1574 lines
47KB

  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. #undef NDEBUG
  24. #include <assert.h>
  25. #define MOV_INDEX_CLUSTER_SIZE 16384
  26. #define globalTimescale 1000
  27. #define MODE_MP4 0
  28. #define MODE_MOV 1
  29. #define MODE_3GP 2
  30. #define MODE_PSP 3 // example working PSP command line:
  31. // ffmpeg -i testinput.avi -f psp -r 14.985 -s 320x240 -b 768 -ar 24000 -ab 32 M4V00001.MP4
  32. #define MODE_3G2 4
  33. typedef struct MOVIentry {
  34. unsigned int flags, pos, size;
  35. unsigned int samplesInChunk;
  36. char key_frame;
  37. unsigned int entries;
  38. } MOVIentry;
  39. typedef struct MOVIndex {
  40. int mode;
  41. int entry;
  42. int mdat_size;
  43. int ents_allocated;
  44. long timescale;
  45. long time;
  46. long trackDuration;
  47. long sampleCount;
  48. long sampleDuration;
  49. int hasKeyframes;
  50. int language;
  51. int trackID;
  52. AVCodecContext *enc;
  53. int vosLen;
  54. uint8_t *vosData;
  55. MOVIentry** cluster;
  56. } MOVTrack;
  57. typedef struct MOVContext {
  58. int mode;
  59. long time;
  60. int nb_streams;
  61. int mdat_written;
  62. offset_t mdat_pos;
  63. long timescale;
  64. MOVTrack tracks[MAX_STREAMS];
  65. } MOVContext;
  66. static int mov_write_esds_tag(ByteIOContext *pb, MOVTrack* track);
  67. /* output language code from iso639 language name */
  68. extern int ff_mov_iso639_to_lang(const char *lang, int mp4);
  69. const CodecTag ff_mov_obj_type[] = {
  70. { CODEC_ID_MPEG4 , 32 },
  71. { CODEC_ID_AAC , 64 },
  72. { CODEC_ID_MPEG1VIDEO, 106 },
  73. { CODEC_ID_MPEG2VIDEO, 96 },//mpeg2 profiles
  74. { CODEC_ID_MP2 , 107 },//FIXME mpeg2 mpeg audio -> 105
  75. { CODEC_ID_MP3 , 107 },//FIXME mpeg2 mpeg audio -> 105
  76. { CODEC_ID_H264 , 33 },
  77. { CODEC_ID_H263 , 242 },
  78. { CODEC_ID_H261 , 243 },
  79. { CODEC_ID_MJPEG , 108 },
  80. { CODEC_ID_PCM_S16LE , 224 },
  81. { CODEC_ID_VORBIS , 225 },
  82. { CODEC_ID_AC3 , 226 },
  83. { CODEC_ID_PCM_ALAW , 227 },
  84. { CODEC_ID_PCM_MULAW , 228 },
  85. { CODEC_ID_PCM_S16BE , 230 },
  86. { 0,0 },
  87. };
  88. //FIXME supprt 64bit varaint with wide placeholders
  89. static int updateSize (ByteIOContext *pb, int pos)
  90. {
  91. long curpos = url_ftell(pb);
  92. url_fseek(pb, pos, SEEK_SET);
  93. put_be32(pb, curpos - pos); /* rewrite size */
  94. url_fseek(pb, curpos, SEEK_SET);
  95. return curpos - pos;
  96. }
  97. /* Chunk offset atom */
  98. static int mov_write_stco_tag(ByteIOContext *pb, MOVTrack* track)
  99. {
  100. int i;
  101. int pos = url_ftell(pb);
  102. put_be32(pb, 0); /* size */
  103. put_tag(pb, "stco");
  104. put_be32(pb, 0); /* version & flags */
  105. put_be32(pb, track->entry); /* entry count */
  106. for (i=0; i<track->entry; i++) {
  107. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  108. int id = i % MOV_INDEX_CLUSTER_SIZE;
  109. put_be32(pb, track->cluster[cl][id].pos);
  110. }
  111. return updateSize (pb, pos);
  112. }
  113. /* Sample size atom */
  114. static int mov_write_stsz_tag(ByteIOContext *pb, MOVTrack* track)
  115. {
  116. int equalChunks = 1;
  117. int i, j, entries = 0, tst = -1, oldtst = -1;
  118. int pos = url_ftell(pb);
  119. put_be32(pb, 0); /* size */
  120. put_tag(pb, "stsz");
  121. put_be32(pb, 0); /* version & flags */
  122. for (i=0; i<track->entry; i++) {
  123. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  124. int id = i % MOV_INDEX_CLUSTER_SIZE;
  125. tst = track->cluster[cl][id].size/track->cluster[cl][id].entries;
  126. if(oldtst != -1 && tst != oldtst) {
  127. equalChunks = 0;
  128. }
  129. oldtst = tst;
  130. entries += track->cluster[cl][id].entries;
  131. }
  132. if (equalChunks) {
  133. int sSize = track->cluster[0][0].size/track->cluster[0][0].entries;
  134. put_be32(pb, sSize); // sample size
  135. put_be32(pb, entries); // sample count
  136. }
  137. else {
  138. put_be32(pb, 0); // sample size
  139. put_be32(pb, entries); // sample count
  140. for (i=0; i<track->entry; i++) {
  141. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  142. int id = i % MOV_INDEX_CLUSTER_SIZE;
  143. for ( j=0; j<track->cluster[cl][id].entries; j++) {
  144. put_be32(pb, track->cluster[cl][id].size /
  145. track->cluster[cl][id].entries);
  146. }
  147. }
  148. }
  149. return updateSize (pb, pos);
  150. }
  151. /* Sample to chunk atom */
  152. static int mov_write_stsc_tag(ByteIOContext *pb, MOVTrack* track)
  153. {
  154. int index = 0, oldval = -1, i, entryPos, curpos;
  155. int pos = url_ftell(pb);
  156. put_be32(pb, 0); /* size */
  157. put_tag(pb, "stsc");
  158. put_be32(pb, 0); // version & flags
  159. entryPos = url_ftell(pb);
  160. put_be32(pb, track->entry); // entry count
  161. for (i=0; i<track->entry; i++) {
  162. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  163. int id = i % MOV_INDEX_CLUSTER_SIZE;
  164. if(oldval != track->cluster[cl][id].samplesInChunk)
  165. {
  166. put_be32(pb, i+1); // first chunk
  167. put_be32(pb, track->cluster[cl][id].samplesInChunk); // samples per chunk
  168. put_be32(pb, 0x1); // sample description index
  169. oldval = track->cluster[cl][id].samplesInChunk;
  170. index++;
  171. }
  172. }
  173. curpos = url_ftell(pb);
  174. url_fseek(pb, entryPos, SEEK_SET);
  175. put_be32(pb, index); // rewrite size
  176. url_fseek(pb, curpos, SEEK_SET);
  177. return updateSize (pb, pos);
  178. }
  179. /* Sync sample atom */
  180. static int mov_write_stss_tag(ByteIOContext *pb, MOVTrack* track)
  181. {
  182. long curpos;
  183. int i, index = 0, entryPos;
  184. int pos = url_ftell(pb);
  185. put_be32(pb, 0); // size
  186. put_tag(pb, "stss");
  187. put_be32(pb, 0); // version & flags
  188. entryPos = url_ftell(pb);
  189. put_be32(pb, track->entry); // entry count
  190. for (i=0; i<track->entry; i++) {
  191. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  192. int id = i % MOV_INDEX_CLUSTER_SIZE;
  193. if(track->cluster[cl][id].key_frame == 1) {
  194. put_be32(pb, i+1);
  195. index++;
  196. }
  197. }
  198. curpos = url_ftell(pb);
  199. url_fseek(pb, entryPos, SEEK_SET);
  200. put_be32(pb, index); // rewrite size
  201. url_fseek(pb, curpos, SEEK_SET);
  202. return updateSize (pb, pos);
  203. }
  204. static int mov_write_damr_tag(ByteIOContext *pb)
  205. {
  206. put_be32(pb, 0x11); /* size */
  207. put_tag(pb, "damr");
  208. put_tag(pb, "FFMP");
  209. put_byte(pb, 0);
  210. put_be16(pb, 0x80); /* Mode set (all modes for AMR_NB) */
  211. put_be16(pb, 0xa); /* Mode change period (no restriction) */
  212. //put_be16(pb, 0x81ff); /* Mode set (all modes for AMR_NB) */
  213. //put_be16(pb, 1); /* Mode change period (no restriction) */
  214. return 0x11;
  215. }
  216. static int mov_write_wave_tag(ByteIOContext *pb, MOVTrack* track)
  217. {
  218. int pos = url_ftell(pb);
  219. put_be32(pb, 0); /* size */
  220. put_tag(pb, "wave");
  221. put_be32(pb, 12); /* size */
  222. put_tag(pb, "frma");
  223. put_tag(pb, "mp4a");
  224. put_be32(pb, 12); /* size */
  225. put_tag(pb, "mp4a");
  226. put_be32(pb, 0);
  227. mov_write_esds_tag(pb, track);
  228. put_be32(pb, 12); /* size */
  229. put_tag(pb, "srcq");
  230. put_be32(pb, 0x40);
  231. put_be32(pb, 8); /* size */
  232. put_be32(pb, 0); /* null tag */
  233. return updateSize (pb, pos);
  234. }
  235. static const CodecTag codec_movaudio_tags[] = {
  236. { CODEC_ID_PCM_MULAW, MKTAG('u', 'l', 'a', 'w') },
  237. { CODEC_ID_PCM_ALAW, MKTAG('a', 'l', 'a', 'w') },
  238. { CODEC_ID_ADPCM_IMA_QT, MKTAG('i', 'm', 'a', '4') },
  239. { CODEC_ID_MACE3, MKTAG('M', 'A', 'C', '3') },
  240. { CODEC_ID_MACE6, MKTAG('M', 'A', 'C', '6') },
  241. { CODEC_ID_AAC, MKTAG('m', 'p', '4', 'a') },
  242. { CODEC_ID_AMR_NB, MKTAG('s', 'a', 'm', 'r') },
  243. { CODEC_ID_AMR_WB, MKTAG('s', 'a', 'w', 'b') },
  244. { CODEC_ID_PCM_S16BE, MKTAG('t', 'w', 'o', 's') },
  245. { CODEC_ID_PCM_S16LE, MKTAG('s', 'o', 'w', 't') },
  246. { CODEC_ID_MP3, MKTAG('.', 'm', 'p', '3') },
  247. { 0, 0 },
  248. };
  249. static int mov_write_audio_tag(ByteIOContext *pb, MOVTrack* track)
  250. {
  251. int pos = url_ftell(pb);
  252. int tag;
  253. put_be32(pb, 0); /* size */
  254. tag = track->enc->codec_tag;
  255. if (!tag)
  256. tag = codec_get_tag(codec_movaudio_tags, track->enc->codec_id);
  257. // if no mac fcc found, try with Microsoft tags
  258. if (!tag)
  259. {
  260. int tmp = codec_get_tag(codec_wav_tags, track->enc->codec_id);
  261. tag = MKTAG('m', 's', ((tmp >> 8) & 0xff), (tmp & 0xff));
  262. }
  263. put_le32(pb, tag); // store it byteswapped
  264. put_be32(pb, 0); /* Reserved */
  265. put_be16(pb, 0); /* Reserved */
  266. put_be16(pb, 1); /* Data-reference index, XXX == 1 */
  267. /* SoundDescription */
  268. if(track->mode == MODE_MOV && track->enc->codec_id == CODEC_ID_AAC)
  269. put_be16(pb, 1); /* Version 1 */
  270. else
  271. put_be16(pb, 0); /* Version 0 */
  272. put_be16(pb, 0); /* Revision level */
  273. put_be32(pb, 0); /* Reserved */
  274. put_be16(pb, track->enc->channels); /* Number of channels */
  275. /* TODO: Currently hard-coded to 16-bit, there doesn't seem
  276. to be a good way to get number of bits of audio */
  277. put_be16(pb, 0x10); /* Reserved */
  278. if(track->enc->codec_id == CODEC_ID_AAC ||
  279. track->enc->codec_id == CODEC_ID_MP3)
  280. {
  281. put_be16(pb, 0xfffe); /* compression ID (vbr)*/
  282. }
  283. else
  284. {
  285. put_be16(pb, 0); /* compression ID (= 0) */
  286. }
  287. put_be16(pb, 0); /* packet size (= 0) */
  288. put_be16(pb, track->timescale); /* Time scale */
  289. put_be16(pb, 0); /* Reserved */
  290. if(track->mode == MODE_MOV && track->enc->codec_id == CODEC_ID_AAC)
  291. {
  292. /* SoundDescription V1 extended info */
  293. put_be32(pb, track->enc->frame_size); /* Samples per packet */
  294. put_be32(pb, 1536); /* Bytes per packet */
  295. put_be32(pb, 2); /* Bytes per frame */
  296. put_be32(pb, 2); /* Bytes per sample */
  297. }
  298. if(track->enc->codec_id == CODEC_ID_AAC) {
  299. if( track->mode == MODE_MOV ) mov_write_wave_tag(pb, track);
  300. else mov_write_esds_tag(pb, track);
  301. }
  302. if(track->enc->codec_id == CODEC_ID_AMR_NB)
  303. mov_write_damr_tag(pb);
  304. return updateSize (pb, pos);
  305. }
  306. static int mov_write_d263_tag(ByteIOContext *pb)
  307. {
  308. put_be32(pb, 0xf); /* size */
  309. put_tag(pb, "d263");
  310. put_tag(pb, "FFMP");
  311. put_be16(pb, 0x0a);
  312. put_byte(pb, 0);
  313. return 0xf;
  314. }
  315. /* TODO: No idea about these values */
  316. static int mov_write_svq3_tag(ByteIOContext *pb)
  317. {
  318. put_be32(pb, 0x15);
  319. put_tag(pb, "SMI ");
  320. put_tag(pb, "SEQH");
  321. put_be32(pb, 0x5);
  322. put_be32(pb, 0xe2c0211d);
  323. put_be32(pb, 0xc0000000);
  324. put_byte(pb, 0);
  325. return 0x15;
  326. }
  327. static unsigned int descrLength(unsigned int len)
  328. {
  329. if (len < 0x00000080)
  330. return 2 + len;
  331. else if (len < 0x00004000)
  332. return 3 + len;
  333. else if(len < 0x00200000)
  334. return 4 + len;
  335. else
  336. return 5 + len;
  337. }
  338. static void putDescr(ByteIOContext *pb, int tag, int size)
  339. {
  340. uint32_t len;
  341. uint8_t vals[4];
  342. len = size;
  343. vals[3] = (uint8_t)(len & 0x7f);
  344. len >>= 7;
  345. vals[2] = (uint8_t)((len & 0x7f) | 0x80);
  346. len >>= 7;
  347. vals[1] = (uint8_t)((len & 0x7f) | 0x80);
  348. len >>= 7;
  349. vals[0] = (uint8_t)((len & 0x7f) | 0x80);
  350. put_byte(pb, tag); // DescriptorTag
  351. if (size < 0x00000080)
  352. {
  353. put_byte(pb, vals[3]);
  354. }
  355. else if (size < 0x00004000)
  356. {
  357. put_byte(pb, vals[2]);
  358. put_byte(pb, vals[3]);
  359. }
  360. else if (size < 0x00200000)
  361. {
  362. put_byte(pb, vals[1]);
  363. put_byte(pb, vals[2]);
  364. put_byte(pb, vals[3]);
  365. }
  366. else if (size < 0x10000000)
  367. {
  368. put_byte(pb, vals[0]);
  369. put_byte(pb, vals[1]);
  370. put_byte(pb, vals[2]);
  371. put_byte(pb, vals[3]);
  372. }
  373. }
  374. static int mov_write_esds_tag(ByteIOContext *pb, MOVTrack* track) // Basic
  375. {
  376. int decoderSpecificInfoLen;
  377. int pos = url_ftell(pb);
  378. void *vosDataBackup=track->vosData;
  379. int vosLenBackup=track->vosLen;
  380. // we should be able to have these passed in, via vosData, then we wouldn't need to attack this routine at all
  381. static const char PSPAACData[]={0x13,0x10};
  382. static const char PSPMP4Data[]={0x00,0x00,0x01,0xB0,0x03,0x00,0x00,0x01,0xB5,0x09,0x00,0x00,0x01,0x00,0x00,0x00,0x01,0x20,0x00,0x84,0x5D,0x4C,0x28,0x50,0x20,0xF0,0xA3,0x1F };
  383. if (track->mode == MODE_PSP) // fails on psp if this is not here
  384. {
  385. if (track->enc->codec_id == CODEC_ID_AAC)
  386. {
  387. track->vosLen = 2;
  388. track->vosData = (uint8_t *) PSPAACData;
  389. }
  390. if (track->enc->codec_id == CODEC_ID_MPEG4)
  391. {
  392. track->vosLen = 28;
  393. track->vosData = (uint8_t *) PSPMP4Data;
  394. }
  395. }
  396. decoderSpecificInfoLen = track->vosLen ? descrLength(track->vosLen):0;
  397. put_be32(pb, 0); // size
  398. put_tag(pb, "esds");
  399. put_be32(pb, 0); // Version
  400. // ES descriptor
  401. putDescr(pb, 0x03, 3 + descrLength(13 + decoderSpecificInfoLen) +
  402. descrLength(1));
  403. put_be16(pb, track->trackID);
  404. put_byte(pb, 0x00); // flags (= no flags)
  405. // DecoderConfig descriptor
  406. putDescr(pb, 0x04, 13 + decoderSpecificInfoLen);
  407. // Object type indication
  408. put_byte(pb, codec_get_tag(ff_mov_obj_type, track->enc->codec_id));
  409. // the following fields is made of 6 bits to identify the streamtype (4 for video, 5 for audio)
  410. // plus 1 bit to indicate upstream and 1 bit set to 1 (reserved)
  411. if(track->enc->codec_type == CODEC_TYPE_AUDIO)
  412. put_byte(pb, 0x15); // flags (= Audiostream)
  413. else
  414. put_byte(pb, 0x11); // flags (= Visualstream)
  415. put_byte(pb, track->enc->rc_buffer_size>>(3+16)); // Buffersize DB (24 bits)
  416. put_be16(pb, (track->enc->rc_buffer_size>>3)&0xFFFF); // Buffersize DB
  417. put_be32(pb, FFMAX(track->enc->bit_rate, track->enc->rc_max_rate)); // maxbitrate (FIXME should be max rate in any 1 sec window)
  418. if(track->enc->rc_max_rate != track->enc->rc_min_rate || track->enc->rc_min_rate==0)
  419. put_be32(pb, 0); // vbr
  420. else
  421. put_be32(pb, track->enc->rc_max_rate); // avg bitrate
  422. if (track->vosLen)
  423. {
  424. // DecoderSpecific info descriptor
  425. putDescr(pb, 0x05, track->vosLen);
  426. put_buffer(pb, track->vosData, track->vosLen);
  427. }
  428. track->vosData = vosDataBackup;
  429. track->vosLen = vosLenBackup;
  430. // SL descriptor
  431. putDescr(pb, 0x06, 1);
  432. put_byte(pb, 0x02);
  433. return updateSize (pb, pos);
  434. }
  435. static const CodecTag codec_movvideo_tags[] = {
  436. { CODEC_ID_SVQ1, MKTAG('S', 'V', 'Q', '1') },
  437. { CODEC_ID_SVQ3, MKTAG('S', 'V', 'Q', '3') },
  438. { CODEC_ID_MPEG4, MKTAG('m', 'p', '4', 'v') },
  439. { CODEC_ID_H263, MKTAG('s', '2', '6', '3') },
  440. { CODEC_ID_H264, MKTAG('a', 'v', 'c', '1') },
  441. { CODEC_ID_DVVIDEO, MKTAG('d', 'v', 'c', ' ') },
  442. { 0, 0 },
  443. };
  444. static int mov_write_video_tag(ByteIOContext *pb, MOVTrack* track)
  445. {
  446. int pos = url_ftell(pb);
  447. char compressor_name[32];
  448. int tag;
  449. put_be32(pb, 0); /* size */
  450. tag = track->enc->codec_tag;
  451. if (!tag)
  452. tag = codec_get_tag(codec_movvideo_tags, track->enc->codec_id);
  453. // if no mac fcc found, try with Microsoft tags
  454. if (!tag)
  455. tag = codec_get_tag(codec_bmp_tags, track->enc->codec_id);
  456. put_le32(pb, tag); // store it byteswapped
  457. put_be32(pb, 0); /* Reserved */
  458. put_be16(pb, 0); /* Reserved */
  459. put_be16(pb, 1); /* Data-reference index */
  460. put_be16(pb, 0); /* Codec stream version */
  461. put_be16(pb, 0); /* Codec stream revision (=0) */
  462. put_tag(pb, "FFMP"); /* Vendor */
  463. if(track->enc->codec_id == CODEC_ID_RAWVIDEO) {
  464. put_be32(pb, 0); /* Temporal Quality */
  465. put_be32(pb, 0x400); /* Spatial Quality = lossless*/
  466. } else {
  467. put_be32(pb, 0x200); /* Temporal Quality = normal */
  468. put_be32(pb, 0x200); /* Spatial Quality = normal */
  469. }
  470. put_be16(pb, track->enc->width); /* Video width */
  471. put_be16(pb, track->enc->height); /* Video height */
  472. put_be32(pb, 0x00480000); /* Horizontal resolution 72dpi */
  473. put_be32(pb, 0x00480000); /* Vertical resolution 72dpi */
  474. put_be32(pb, 0); /* Data size (= 0) */
  475. put_be16(pb, 1); /* Frame count (= 1) */
  476. memset(compressor_name,0,32);
  477. if (track->enc->codec && track->enc->codec->name)
  478. strncpy(compressor_name,track->enc->codec->name,31);
  479. put_byte(pb, strlen(compressor_name));
  480. put_buffer(pb, compressor_name, 31);
  481. put_be16(pb, 0x18); /* Reserved */
  482. put_be16(pb, 0xffff); /* Reserved */
  483. if(track->enc->codec_id == CODEC_ID_MPEG4)
  484. mov_write_esds_tag(pb, track);
  485. else if(track->enc->codec_id == CODEC_ID_H263)
  486. mov_write_d263_tag(pb);
  487. else if(track->enc->codec_id == CODEC_ID_SVQ3)
  488. mov_write_svq3_tag(pb);
  489. return updateSize (pb, pos);
  490. }
  491. static int mov_write_stsd_tag(ByteIOContext *pb, MOVTrack* track)
  492. {
  493. int pos = url_ftell(pb);
  494. put_be32(pb, 0); /* size */
  495. put_tag(pb, "stsd");
  496. put_be32(pb, 0); /* version & flags */
  497. put_be32(pb, 1); /* entry count */
  498. if (track->enc->codec_type == CODEC_TYPE_VIDEO)
  499. mov_write_video_tag(pb, track);
  500. else if (track->enc->codec_type == CODEC_TYPE_AUDIO)
  501. mov_write_audio_tag(pb, track);
  502. return updateSize(pb, pos);
  503. }
  504. /* TODO: */
  505. /* Time to sample atom */
  506. static int mov_write_stts_tag(ByteIOContext *pb, MOVTrack* track)
  507. {
  508. put_be32(pb, 0x18); /* size */
  509. put_tag(pb, "stts");
  510. put_be32(pb, 0); /* version & flags */
  511. put_be32(pb, 1); /* entry count */
  512. put_be32(pb, track->sampleCount); /* sample count */
  513. put_be32(pb, track->sampleDuration); /* sample duration */
  514. return 0x18;
  515. }
  516. static int mov_write_dref_tag(ByteIOContext *pb)
  517. {
  518. put_be32(pb, 28); /* size */
  519. put_tag(pb, "dref");
  520. put_be32(pb, 0); /* version & flags */
  521. put_be32(pb, 1); /* entry count */
  522. put_be32(pb, 0xc); /* size */
  523. put_tag(pb, "url ");
  524. put_be32(pb, 1); /* version & flags */
  525. return 28;
  526. }
  527. static int mov_write_stbl_tag(ByteIOContext *pb, MOVTrack* track)
  528. {
  529. int pos = url_ftell(pb);
  530. put_be32(pb, 0); /* size */
  531. put_tag(pb, "stbl");
  532. mov_write_stsd_tag(pb, track);
  533. mov_write_stts_tag(pb, track);
  534. if (track->enc->codec_type == CODEC_TYPE_VIDEO &&
  535. track->hasKeyframes)
  536. mov_write_stss_tag(pb, track);
  537. mov_write_stsc_tag(pb, track);
  538. mov_write_stsz_tag(pb, track);
  539. mov_write_stco_tag(pb, track);
  540. return updateSize(pb, pos);
  541. }
  542. static int mov_write_dinf_tag(ByteIOContext *pb)
  543. {
  544. int pos = url_ftell(pb);
  545. put_be32(pb, 0); /* size */
  546. put_tag(pb, "dinf");
  547. mov_write_dref_tag(pb);
  548. return updateSize(pb, pos);
  549. }
  550. static int mov_write_smhd_tag(ByteIOContext *pb)
  551. {
  552. put_be32(pb, 16); /* size */
  553. put_tag(pb, "smhd");
  554. put_be32(pb, 0); /* version & flags */
  555. put_be16(pb, 0); /* reserved (balance, normally = 0) */
  556. put_be16(pb, 0); /* reserved */
  557. return 16;
  558. }
  559. static int mov_write_vmhd_tag(ByteIOContext *pb)
  560. {
  561. put_be32(pb, 0x14); /* size (always 0x14) */
  562. put_tag(pb, "vmhd");
  563. put_be32(pb, 0x01); /* version & flags */
  564. put_be64(pb, 0); /* reserved (graphics mode = copy) */
  565. return 0x14;
  566. }
  567. static int mov_write_hdlr_tag(ByteIOContext *pb, MOVTrack* track)
  568. {
  569. char *descr, *hdlr, *hdlr_type;
  570. int pos = url_ftell(pb);
  571. if (!track) { /* no media --> data handler */
  572. hdlr = "dhlr";
  573. hdlr_type = "url ";
  574. descr = "DataHandler";
  575. } else {
  576. hdlr = (track->mode == MODE_MOV) ? "mhlr" : "\0\0\0\0";
  577. if (track->enc->codec_type == CODEC_TYPE_VIDEO) {
  578. hdlr_type = "vide";
  579. descr = "VideoHandler";
  580. } else {
  581. hdlr_type = "soun";
  582. descr = "SoundHandler";
  583. }
  584. }
  585. put_be32(pb, 0); /* size */
  586. put_tag(pb, "hdlr");
  587. put_be32(pb, 0); /* Version & flags */
  588. put_buffer(pb, hdlr, 4); /* handler */
  589. put_tag(pb, hdlr_type); /* handler type */
  590. put_be32(pb ,0); /* reserved */
  591. put_be32(pb ,0); /* reserved */
  592. put_be32(pb ,0); /* reserved */
  593. put_byte(pb, strlen(descr)); /* string counter */
  594. put_buffer(pb, descr, strlen(descr)); /* handler description */
  595. return updateSize(pb, pos);
  596. }
  597. static int mov_write_minf_tag(ByteIOContext *pb, MOVTrack* track)
  598. {
  599. int pos = url_ftell(pb);
  600. put_be32(pb, 0); /* size */
  601. put_tag(pb, "minf");
  602. if(track->enc->codec_type == CODEC_TYPE_VIDEO)
  603. mov_write_vmhd_tag(pb);
  604. else
  605. mov_write_smhd_tag(pb);
  606. if (track->mode == MODE_MOV) /* FIXME: Why do it for MODE_MOV only ? */
  607. mov_write_hdlr_tag(pb, NULL);
  608. mov_write_dinf_tag(pb);
  609. mov_write_stbl_tag(pb, track);
  610. return updateSize(pb, pos);
  611. }
  612. static int mov_write_mdhd_tag(ByteIOContext *pb, MOVTrack* track)
  613. {
  614. put_be32(pb, 32); /* size */
  615. put_tag(pb, "mdhd");
  616. put_be32(pb, 0); /* Version & flags */
  617. put_be32(pb, track->time); /* creation time */
  618. put_be32(pb, track->time); /* modification time */
  619. put_be32(pb, track->timescale); /* time scale (sample rate for audio) */
  620. put_be32(pb, track->trackDuration); /* duration */
  621. put_be16(pb, track->language); /* language */
  622. put_be16(pb, 0); /* reserved (quality) */
  623. return 32;
  624. }
  625. static int mov_write_mdia_tag(ByteIOContext *pb, MOVTrack* track)
  626. {
  627. int pos = url_ftell(pb);
  628. put_be32(pb, 0); /* size */
  629. put_tag(pb, "mdia");
  630. mov_write_mdhd_tag(pb, track);
  631. mov_write_hdlr_tag(pb, track);
  632. mov_write_minf_tag(pb, track);
  633. return updateSize(pb, pos);
  634. }
  635. static int mov_write_tkhd_tag(ByteIOContext *pb, MOVTrack* track)
  636. {
  637. put_be32(pb, 0x5c); /* size (always 0x5c) */
  638. put_tag(pb, "tkhd");
  639. put_be32(pb, 0xf); /* version & flags (track enabled) */
  640. put_be32(pb, track->time); /* creation time */
  641. put_be32(pb, track->time); /* modification time */
  642. put_be32(pb, track->trackID); /* track-id */
  643. put_be32(pb, 0); /* reserved */
  644. put_be32(pb, av_rescale_rnd(track->trackDuration, globalTimescale, track->timescale, AV_ROUND_UP)); /* duration */
  645. put_be32(pb, 0); /* reserved */
  646. put_be32(pb, 0); /* reserved */
  647. put_be32(pb, 0x0); /* reserved (Layer & Alternate group) */
  648. /* Volume, only for audio */
  649. if(track->enc->codec_type == CODEC_TYPE_AUDIO)
  650. put_be16(pb, 0x0100);
  651. else
  652. put_be16(pb, 0);
  653. put_be16(pb, 0); /* reserved */
  654. /* Matrix structure */
  655. put_be32(pb, 0x00010000); /* reserved */
  656. put_be32(pb, 0x0); /* reserved */
  657. put_be32(pb, 0x0); /* reserved */
  658. put_be32(pb, 0x0); /* reserved */
  659. put_be32(pb, 0x00010000); /* reserved */
  660. put_be32(pb, 0x0); /* reserved */
  661. put_be32(pb, 0x0); /* reserved */
  662. put_be32(pb, 0x0); /* reserved */
  663. put_be32(pb, 0x40000000); /* reserved */
  664. /* Track width and height, for visual only */
  665. if(track->enc->codec_type == CODEC_TYPE_VIDEO) {
  666. double sample_aspect_ratio = av_q2d(track->enc->sample_aspect_ratio);
  667. if( !sample_aspect_ratio ) sample_aspect_ratio = 1;
  668. put_be32(pb, sample_aspect_ratio * track->enc->width*0x10000);
  669. put_be32(pb, track->enc->height*0x10000);
  670. }
  671. else {
  672. put_be32(pb, 0);
  673. put_be32(pb, 0);
  674. }
  675. return 0x5c;
  676. }
  677. // This box seems important for the psp playback ... without it the movie seems to hang
  678. static int mov_write_edts_tag(ByteIOContext *pb, MOVTrack *track)
  679. {
  680. put_be32(pb, 0x24); /* size */
  681. put_tag(pb, "edts");
  682. put_be32(pb, 0x1c); /* size */
  683. put_tag(pb, "elst");
  684. put_be32(pb, 0x0);
  685. put_be32(pb, 0x1);
  686. put_be32(pb, av_rescale_rnd(track->trackDuration, globalTimescale, track->timescale, AV_ROUND_UP)); /* duration ... doesn't seem to effect psp */
  687. put_be32(pb, 0x0);
  688. put_be32(pb, 0x00010000);
  689. return 0x24;
  690. }
  691. // goes at the end of each track! ... Critical for PSP playback ("Incompatible data" without it)
  692. static int mov_write_uuid_tag_psp(ByteIOContext *pb, MOVTrack *mov)
  693. {
  694. put_be32(pb, 0x34); /* size ... reports as 28 in mp4box! */
  695. put_tag(pb, "uuid");
  696. put_tag(pb, "USMT");
  697. put_be32(pb, 0x21d24fce);
  698. put_be32(pb, 0xbb88695c);
  699. put_be32(pb, 0xfac9c740);
  700. put_be32(pb, 0x1c); // another size here!
  701. put_tag(pb, "MTDT");
  702. put_be32(pb, 0x00010012);
  703. put_be32(pb, 0x0a);
  704. put_be32(pb, 0x55c40000);
  705. put_be32(pb, 0x1);
  706. put_be32(pb, 0x0);
  707. return 0x34;
  708. }
  709. static int mov_write_trak_tag(ByteIOContext *pb, MOVTrack* track)
  710. {
  711. int pos = url_ftell(pb);
  712. put_be32(pb, 0); /* size */
  713. put_tag(pb, "trak");
  714. mov_write_tkhd_tag(pb, track);
  715. if (track->mode == MODE_PSP)
  716. mov_write_edts_tag(pb, track); // PSP Movies require edts box
  717. mov_write_mdia_tag(pb, track);
  718. if (track->mode == MODE_PSP)
  719. mov_write_uuid_tag_psp(pb,track); // PSP Movies require this uuid box
  720. return updateSize(pb, pos);
  721. }
  722. #if 0
  723. /* TODO: Not sorted out, but not necessary either */
  724. static int mov_write_iods_tag(ByteIOContext *pb, MOVContext *mov)
  725. {
  726. put_be32(pb, 0x15); /* size */
  727. put_tag(pb, "iods");
  728. put_be32(pb, 0); /* version & flags */
  729. put_be16(pb, 0x1007);
  730. put_byte(pb, 0);
  731. put_be16(pb, 0x4fff);
  732. put_be16(pb, 0xfffe);
  733. put_be16(pb, 0x01ff);
  734. return 0x15;
  735. }
  736. #endif
  737. static int mov_write_mvhd_tag(ByteIOContext *pb, MOVContext *mov)
  738. {
  739. int maxTrackID = 1, maxTrackLen = 0, i;
  740. int64_t maxTrackLenTemp;
  741. put_be32(pb, 0x6c); /* size (always 0x6c) */
  742. put_tag(pb, "mvhd");
  743. put_be32(pb, 0); /* version & flags */
  744. put_be32(pb, mov->time); /* creation time */
  745. put_be32(pb, mov->time); /* modification time */
  746. put_be32(pb, mov->timescale); /* timescale */
  747. for (i=0; i<MAX_STREAMS; i++) {
  748. if(mov->tracks[i].entry > 0) {
  749. maxTrackLenTemp = av_rescale_rnd(mov->tracks[i].trackDuration, globalTimescale, mov->tracks[i].timescale, AV_ROUND_UP);
  750. if(maxTrackLen < maxTrackLenTemp)
  751. maxTrackLen = maxTrackLenTemp;
  752. if(maxTrackID < mov->tracks[i].trackID)
  753. maxTrackID = mov->tracks[i].trackID;
  754. }
  755. }
  756. put_be32(pb, maxTrackLen); /* duration of longest track */
  757. put_be32(pb, 0x00010000); /* reserved (preferred rate) 1.0 = normal */
  758. put_be16(pb, 0x0100); /* reserved (preferred volume) 1.0 = normal */
  759. put_be16(pb, 0); /* reserved */
  760. put_be32(pb, 0); /* reserved */
  761. put_be32(pb, 0); /* reserved */
  762. /* Matrix structure */
  763. put_be32(pb, 0x00010000); /* reserved */
  764. put_be32(pb, 0x0); /* reserved */
  765. put_be32(pb, 0x0); /* reserved */
  766. put_be32(pb, 0x0); /* reserved */
  767. put_be32(pb, 0x00010000); /* reserved */
  768. put_be32(pb, 0x0); /* reserved */
  769. put_be32(pb, 0x0); /* reserved */
  770. put_be32(pb, 0x0); /* reserved */
  771. put_be32(pb, 0x40000000); /* reserved */
  772. put_be32(pb, 0); /* reserved (preview time) */
  773. put_be32(pb, 0); /* reserved (preview duration) */
  774. put_be32(pb, 0); /* reserved (poster time) */
  775. put_be32(pb, 0); /* reserved (selection time) */
  776. put_be32(pb, 0); /* reserved (selection duration) */
  777. put_be32(pb, 0); /* reserved (current time) */
  778. put_be32(pb, maxTrackID+1); /* Next track id */
  779. return 0x6c;
  780. }
  781. static int mov_write_itunes_hdlr_tag(ByteIOContext *pb, MOVContext* mov,
  782. AVFormatContext *s)
  783. {
  784. int pos = url_ftell(pb);
  785. put_be32(pb, 0); /* size */
  786. put_tag(pb, "hdlr");
  787. put_be32(pb, 0);
  788. put_be32(pb, 0);
  789. put_tag(pb, "mdir");
  790. put_tag(pb, "appl");
  791. put_be32(pb, 0);
  792. put_be32(pb, 0);
  793. put_be16(pb, 0);
  794. return updateSize(pb, pos);
  795. }
  796. /* helper function to write a data tag with the specified string as data */
  797. static int mov_write_string_data_tag(ByteIOContext *pb, MOVContext* mov,
  798. AVFormatContext *s, const char *data)
  799. {
  800. int pos = url_ftell(pb);
  801. put_be32(pb, 0); /* size */
  802. put_tag(pb, "data");
  803. put_be32(pb, 1);
  804. put_be32(pb, 0);
  805. put_buffer(pb, data, strlen(data));
  806. return updateSize(pb, pos);
  807. }
  808. /* iTunes name of the song/movie */
  809. static int mov_write_nam_tag(ByteIOContext *pb, MOVContext* mov,
  810. AVFormatContext *s)
  811. {
  812. int size = 0;
  813. if ( s->title[0] ) {
  814. int pos = url_ftell(pb);
  815. put_be32(pb, 0); /* size */
  816. put_tag(pb, "\251nam");
  817. mov_write_string_data_tag(pb, mov, s, s->title);
  818. size = updateSize(pb, pos);
  819. }
  820. return size;
  821. }
  822. /* iTunes name of the artist/performer */
  823. static int mov_write_ART_tag(ByteIOContext *pb, MOVContext* mov,
  824. AVFormatContext *s)
  825. {
  826. int size = 0;
  827. if ( s->author[0] ) {
  828. int pos = url_ftell(pb);
  829. put_be32(pb, 0); /* size */
  830. put_tag(pb, "\251ART");
  831. // we use the author here as this is the only thing that we have...
  832. mov_write_string_data_tag(pb, mov, s, s->author);
  833. size = updateSize(pb, pos);
  834. }
  835. return size;
  836. }
  837. /* iTunes name of the writer */
  838. static int mov_write_wrt_tag(ByteIOContext *pb, MOVContext* mov,
  839. AVFormatContext *s)
  840. {
  841. int size = 0;
  842. if ( s->author[0] ) {
  843. int pos = url_ftell(pb);
  844. put_be32(pb, 0); /* size */
  845. put_tag(pb, "\251wrt");
  846. mov_write_string_data_tag(pb, mov, s, s->author);
  847. size = updateSize(pb, pos);
  848. }
  849. return size;
  850. }
  851. /* iTunes name of the album */
  852. static int mov_write_alb_tag(ByteIOContext *pb, MOVContext* mov,
  853. AVFormatContext *s)
  854. {
  855. int size = 0;
  856. if ( s->album[0] ) {
  857. int pos = url_ftell(pb);
  858. put_be32(pb, 0); /* size */
  859. put_tag(pb, "\251alb");
  860. mov_write_string_data_tag(pb, mov, s, s->album);
  861. size = updateSize(pb, pos);
  862. }
  863. return size;
  864. }
  865. /* iTunes year */
  866. static int mov_write_day_tag(ByteIOContext *pb, MOVContext* mov,
  867. AVFormatContext *s)
  868. {
  869. char year[5];
  870. int size = 0;
  871. if ( s->year ) {
  872. int pos = url_ftell(pb);
  873. put_be32(pb, 0); /* size */
  874. put_tag(pb, "\251day");
  875. snprintf(year, 5, "%04d", s->year);
  876. mov_write_string_data_tag(pb, mov, s, year);
  877. size = updateSize(pb, pos);
  878. }
  879. return size;
  880. }
  881. /* iTunes tool used to create the file */
  882. static int mov_write_too_tag(ByteIOContext *pb, MOVContext* mov,
  883. AVFormatContext *s)
  884. {
  885. int pos = url_ftell(pb);
  886. put_be32(pb, 0); /* size */
  887. put_tag(pb, "\251too");
  888. mov_write_string_data_tag(pb, mov, s, LIBAVFORMAT_IDENT);
  889. return updateSize(pb, pos);
  890. }
  891. /* iTunes comment */
  892. static int mov_write_cmt_tag(ByteIOContext *pb, MOVContext* mov,
  893. AVFormatContext *s)
  894. {
  895. int size = 0;
  896. if ( s->comment[0] ) {
  897. int pos = url_ftell(pb);
  898. put_be32(pb, 0); /* size */
  899. put_tag(pb, "\251cmt");
  900. mov_write_string_data_tag(pb, mov, s, s->comment);
  901. size = updateSize(pb, pos);
  902. }
  903. return size;
  904. }
  905. /* iTunes custom genre */
  906. static int mov_write_gen_tag(ByteIOContext *pb, MOVContext* mov,
  907. AVFormatContext *s)
  908. {
  909. int size = 0;
  910. if ( s->genre[0] ) {
  911. int pos = url_ftell(pb);
  912. put_be32(pb, 0); /* size */
  913. put_tag(pb, "\251gen");
  914. mov_write_string_data_tag(pb, mov, s, s->genre);
  915. size = updateSize(pb, pos);
  916. }
  917. return size;
  918. }
  919. /* iTunes track number */
  920. static int mov_write_trkn_tag(ByteIOContext *pb, MOVContext* mov,
  921. AVFormatContext *s)
  922. {
  923. int size = 0;
  924. if ( s->track ) {
  925. int pos = url_ftell(pb);
  926. put_be32(pb, 0); /* size */
  927. put_tag(pb, "trkn");
  928. {
  929. int pos = url_ftell(pb);
  930. put_be32(pb, 0); /* size */
  931. put_tag(pb, "data");
  932. put_be32(pb, 0); // 8 bytes empty
  933. put_be32(pb, 0);
  934. put_be16(pb, 0); // empty
  935. put_be16(pb, s->track); // track number
  936. put_be16(pb, 0); // total track number
  937. put_be16(pb, 0); // empty
  938. updateSize(pb, pos);
  939. }
  940. size = updateSize(pb, pos);
  941. }
  942. return size;
  943. }
  944. /* iTunes meta data list */
  945. static int mov_write_ilst_tag(ByteIOContext *pb, MOVContext* mov,
  946. AVFormatContext *s)
  947. {
  948. int pos = url_ftell(pb);
  949. put_be32(pb, 0); /* size */
  950. put_tag(pb, "ilst");
  951. mov_write_nam_tag(pb, mov, s);
  952. mov_write_ART_tag(pb, mov, s);
  953. mov_write_wrt_tag(pb, mov, s);
  954. mov_write_alb_tag(pb, mov, s);
  955. mov_write_day_tag(pb, mov, s);
  956. mov_write_too_tag(pb, mov, s);
  957. mov_write_cmt_tag(pb, mov, s);
  958. mov_write_gen_tag(pb, mov, s);
  959. mov_write_trkn_tag(pb, mov, s);
  960. return updateSize(pb, pos);
  961. }
  962. /* iTunes meta data tag */
  963. static int mov_write_meta_tag(ByteIOContext *pb, MOVContext* mov,
  964. AVFormatContext *s)
  965. {
  966. int size = 0;
  967. // only save meta tag if required
  968. if ( s->title[0] || s->author[0] || s->album[0] || s->year ||
  969. s->comment[0] || s->genre[0] || s->track ) {
  970. int pos = url_ftell(pb);
  971. put_be32(pb, 0); /* size */
  972. put_tag(pb, "meta");
  973. put_be32(pb, 0);
  974. mov_write_itunes_hdlr_tag(pb, mov, s);
  975. mov_write_ilst_tag(pb, mov, s);
  976. size = updateSize(pb, pos);
  977. }
  978. return size;
  979. }
  980. static int mov_write_udta_tag(ByteIOContext *pb, MOVContext* mov,
  981. AVFormatContext *s)
  982. {
  983. int pos = url_ftell(pb);
  984. int i;
  985. put_be32(pb, 0); /* size */
  986. put_tag(pb, "udta");
  987. /* iTunes meta data */
  988. mov_write_meta_tag(pb, mov, s);
  989. /* Requirements */
  990. for (i=0; i<MAX_STREAMS; i++) {
  991. if(mov->tracks[i].entry <= 0) continue;
  992. if (mov->tracks[i].enc->codec_id == CODEC_ID_AAC ||
  993. mov->tracks[i].enc->codec_id == CODEC_ID_MPEG4) {
  994. int pos = url_ftell(pb);
  995. put_be32(pb, 0); /* size */
  996. put_tag(pb, "\251req");
  997. put_be16(pb, sizeof("QuickTime 6.0 or greater") - 1);
  998. put_be16(pb, 0);
  999. put_buffer(pb, "QuickTime 6.0 or greater",
  1000. sizeof("QuickTime 6.0 or greater") - 1);
  1001. updateSize(pb, pos);
  1002. break;
  1003. }
  1004. }
  1005. /* Encoder */
  1006. if(mov->tracks[0].enc && !(mov->tracks[0].enc->flags & CODEC_FLAG_BITEXACT))
  1007. {
  1008. int pos = url_ftell(pb);
  1009. put_be32(pb, 0); /* size */
  1010. put_tag(pb, "\251enc");
  1011. put_be16(pb, sizeof(LIBAVFORMAT_IDENT) - 1); /* string length */
  1012. put_be16(pb, 0);
  1013. put_buffer(pb, LIBAVFORMAT_IDENT, sizeof(LIBAVFORMAT_IDENT) - 1);
  1014. updateSize(pb, pos);
  1015. }
  1016. if( s->title[0] )
  1017. {
  1018. int pos = url_ftell(pb);
  1019. put_be32(pb, 0); /* size */
  1020. put_tag(pb, "\251nam");
  1021. put_be16(pb, strlen(s->title)); /* string length */
  1022. put_be16(pb, 0);
  1023. put_buffer(pb, s->title, strlen(s->title));
  1024. updateSize(pb, pos);
  1025. }
  1026. if( s->author[0] )
  1027. {
  1028. int pos = url_ftell(pb);
  1029. put_be32(pb, 0); /* size */
  1030. put_tag(pb, /*"\251aut"*/ "\251day" );
  1031. put_be16(pb, strlen(s->author)); /* string length */
  1032. put_be16(pb, 0);
  1033. put_buffer(pb, s->author, strlen(s->author));
  1034. updateSize(pb, pos);
  1035. }
  1036. if( s->comment[0] )
  1037. {
  1038. int pos = url_ftell(pb);
  1039. put_be32(pb, 0); /* size */
  1040. put_tag(pb, "\251des");
  1041. put_be16(pb, strlen(s->comment)); /* string length */
  1042. put_be16(pb, 0);
  1043. put_buffer(pb, s->comment, strlen(s->comment));
  1044. updateSize(pb, pos);
  1045. }
  1046. return updateSize(pb, pos);
  1047. }
  1048. static int mov_write_moov_tag(ByteIOContext *pb, MOVContext *mov,
  1049. AVFormatContext *s)
  1050. {
  1051. int pos, i;
  1052. pos = url_ftell(pb);
  1053. put_be32(pb, 0); /* size placeholder*/
  1054. put_tag(pb, "moov");
  1055. mov->timescale = globalTimescale;
  1056. for (i=0; i<MAX_STREAMS; i++) {
  1057. if(mov->tracks[i].entry <= 0) continue;
  1058. if(mov->tracks[i].enc->codec_type == CODEC_TYPE_VIDEO) {
  1059. mov->tracks[i].timescale = mov->tracks[i].enc->time_base.den;
  1060. mov->tracks[i].sampleDuration = mov->tracks[i].enc->time_base.num;
  1061. }
  1062. else if(mov->tracks[i].enc->codec_type == CODEC_TYPE_AUDIO) {
  1063. /* If AMR, track timescale = 8000, AMR_WB = 16000 */
  1064. if(mov->tracks[i].enc->codec_id == CODEC_ID_AMR_NB) {
  1065. mov->tracks[i].sampleDuration = 160; // Bytes per chunk
  1066. mov->tracks[i].timescale = 8000;
  1067. }
  1068. else {
  1069. mov->tracks[i].timescale = mov->tracks[i].enc->sample_rate;
  1070. mov->tracks[i].sampleDuration = mov->tracks[i].enc->frame_size;
  1071. }
  1072. }
  1073. mov->tracks[i].trackDuration =
  1074. mov->tracks[i].sampleCount * mov->tracks[i].sampleDuration;
  1075. mov->tracks[i].time = mov->time;
  1076. mov->tracks[i].trackID = i+1;
  1077. }
  1078. mov_write_mvhd_tag(pb, mov);
  1079. //mov_write_iods_tag(pb, mov);
  1080. for (i=0; i<MAX_STREAMS; i++) {
  1081. if(mov->tracks[i].entry > 0) {
  1082. mov_write_trak_tag(pb, &(mov->tracks[i]));
  1083. }
  1084. }
  1085. mov_write_udta_tag(pb, mov, s);
  1086. return updateSize(pb, pos);
  1087. }
  1088. int mov_write_mdat_tag(ByteIOContext *pb, MOVContext* mov)
  1089. {
  1090. mov->mdat_pos = url_ftell(pb);
  1091. put_be32(pb, 0); /* size placeholder*/
  1092. put_tag(pb, "mdat");
  1093. return 0;
  1094. }
  1095. /* TODO: This needs to be more general */
  1096. int mov_write_ftyp_tag(ByteIOContext *pb, AVFormatContext *s)
  1097. {
  1098. MOVContext *mov = s->priv_data;
  1099. put_be32(pb, 0x14 ); /* size */
  1100. put_tag(pb, "ftyp");
  1101. if ( mov->mode == MODE_3GP )
  1102. put_tag(pb, "3gp4");
  1103. else if ( mov->mode == MODE_3G2 )
  1104. put_tag(pb, "3g2a");
  1105. else if ( mov->mode == MODE_PSP )
  1106. put_tag(pb, "MSNV");
  1107. else
  1108. put_tag(pb, "isom");
  1109. put_be32(pb, 0x200 );
  1110. if ( mov->mode == MODE_3GP )
  1111. put_tag(pb, "3gp4");
  1112. else if ( mov->mode == MODE_3G2 )
  1113. put_tag(pb, "3g2a");
  1114. else if ( mov->mode == MODE_PSP )
  1115. put_tag(pb, "MSNV");
  1116. else
  1117. put_tag(pb, "mp41");
  1118. return 0x14;
  1119. }
  1120. static void mov_write_uuidprof_tag(ByteIOContext *pb, AVFormatContext *s)
  1121. {
  1122. int AudioRate = s->streams[1]->codec->sample_rate;
  1123. int FrameRate = ((s->streams[0]->codec->time_base.den) * (0x10000))/ (s->streams[0]->codec->time_base.num);
  1124. //printf("audiorate = %d\n",AudioRate);
  1125. //printf("framerate = %d / %d = 0x%x\n",s->streams[0]->codec->time_base.den,s->streams[0]->codec->time_base.num,FrameRate);
  1126. put_be32(pb, 0x94 ); /* size */
  1127. put_tag(pb, "uuid");
  1128. put_tag(pb, "PROF");
  1129. put_be32(pb, 0x21d24fce ); /* 96 bit UUID */
  1130. put_be32(pb, 0xbb88695c );
  1131. put_be32(pb, 0xfac9c740 );
  1132. put_be32(pb, 0x0 ); /* ? */
  1133. put_be32(pb, 0x3 ); /* 3 sections ? */
  1134. put_be32(pb, 0x14 ); /* size */
  1135. put_tag(pb, "FPRF");
  1136. put_be32(pb, 0x0 ); /* ? */
  1137. put_be32(pb, 0x0 ); /* ? */
  1138. put_be32(pb, 0x0 ); /* ? */
  1139. put_be32(pb, 0x2c ); /* size */
  1140. put_tag(pb, "APRF"); /* audio */
  1141. put_be32(pb, 0x0 );
  1142. put_be32(pb, 0x2 );
  1143. put_tag(pb, "mp4a");
  1144. put_be32(pb, 0x20f );
  1145. put_be32(pb, 0x0 );
  1146. put_be32(pb, 0x40 );
  1147. put_be32(pb, 0x40 );
  1148. put_be32(pb, AudioRate ); //24000 ... audio rate?
  1149. put_be32(pb, 0x2 );
  1150. put_be32(pb, 0x34 ); /* size */
  1151. put_tag(pb, "VPRF"); /* video */
  1152. put_be32(pb, 0x0 );
  1153. put_be32(pb, 0x1 );
  1154. put_tag(pb, "mp4v");
  1155. put_be32(pb, 0x103 );
  1156. put_be32(pb, 0x0 );
  1157. put_be32(pb, 0xc0 );
  1158. put_be32(pb, 0xc0 );
  1159. put_be32(pb, FrameRate); // was 0xefc29
  1160. put_be32(pb, FrameRate ); // was 0xefc29
  1161. put_be16(pb, s->streams[0]->codec->width);
  1162. put_be16(pb, s->streams[0]->codec->height);
  1163. put_be32(pb, 0x010001 );
  1164. }
  1165. static int mov_write_header(AVFormatContext *s)
  1166. {
  1167. ByteIOContext *pb = &s->pb;
  1168. MOVContext *mov = s->priv_data;
  1169. int i;
  1170. for(i=0; i<s->nb_streams; i++){
  1171. AVCodecContext *c= s->streams[i]->codec;
  1172. if (c->codec_type == CODEC_TYPE_VIDEO){
  1173. if (!codec_get_tag(codec_movvideo_tags, c->codec_id)){
  1174. if(!codec_get_tag(codec_bmp_tags, c->codec_id))
  1175. return -1;
  1176. else
  1177. av_log(s, AV_LOG_INFO, "Warning, using MS style video codec tag, the file may be unplayable!\n");
  1178. }
  1179. }else if(c->codec_type == CODEC_TYPE_AUDIO){
  1180. if (!codec_get_tag(codec_movaudio_tags, c->codec_id)){
  1181. if(!codec_get_tag(codec_wav_tags, c->codec_id))
  1182. return -1;
  1183. else
  1184. av_log(s, AV_LOG_INFO, "Warning, using MS style audio codec tag, the file may be unplayable!\n");
  1185. }
  1186. }
  1187. /* don't know yet if mp4 or not */
  1188. mov->tracks[i].language = ff_mov_iso639_to_lang(s->streams[i]->language, 1);
  1189. }
  1190. /* Default mode == MP4 */
  1191. mov->mode = MODE_MP4;
  1192. if (s->oformat != NULL) {
  1193. if (!strcmp("3gp", s->oformat->name)) mov->mode = MODE_3GP;
  1194. else if (!strcmp("3g2", s->oformat->name)) mov->mode = MODE_3G2;
  1195. else if (!strcmp("mov", s->oformat->name)) mov->mode = MODE_MOV;
  1196. else if (!strcmp("psp", s->oformat->name)) mov->mode = MODE_PSP;
  1197. if ( mov->mode == MODE_3GP || mov->mode == MODE_3G2 ||
  1198. mov->mode == MODE_MP4 || mov->mode == MODE_PSP )
  1199. mov_write_ftyp_tag(pb,s);
  1200. if ( mov->mode == MODE_PSP ) {
  1201. if ( s->nb_streams != 2 ) {
  1202. av_log(s, AV_LOG_ERROR, "PSP mode need one video and one audio stream\n");
  1203. return -1;
  1204. }
  1205. mov_write_uuidprof_tag(pb,s);
  1206. }
  1207. }
  1208. for (i=0; i<MAX_STREAMS; i++) {
  1209. mov->tracks[i].mode = mov->mode;
  1210. }
  1211. put_flush_packet(pb);
  1212. return 0;
  1213. }
  1214. static int mov_write_packet(AVFormatContext *s, AVPacket *pkt)
  1215. {
  1216. MOVContext *mov = s->priv_data;
  1217. ByteIOContext *pb = &s->pb;
  1218. AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
  1219. MOVTrack* trk = &mov->tracks[pkt->stream_index];
  1220. int cl, id;
  1221. unsigned int samplesInChunk = 0;
  1222. int size= pkt->size;
  1223. if (url_is_streamed(&s->pb)) return 0; /* Can't handle that */
  1224. if (!size) return 0; /* Discard 0 sized packets */
  1225. if (enc->codec_type == CODEC_TYPE_VIDEO ) {
  1226. samplesInChunk = 1;
  1227. }
  1228. else if (enc->codec_type == CODEC_TYPE_AUDIO ) {
  1229. if( enc->codec_id == CODEC_ID_AMR_NB) {
  1230. /* We must find out how many AMR blocks there are in one packet */
  1231. static uint16_t packed_size[16] =
  1232. {13, 14, 16, 18, 20, 21, 27, 32, 6, 0, 0, 0, 0, 0, 0, 0};
  1233. int len = 0;
  1234. while (len < size && samplesInChunk < 100) {
  1235. len += packed_size[(pkt->data[len] >> 3) & 0x0F];
  1236. samplesInChunk++;
  1237. }
  1238. }
  1239. else if(enc->codec_id == CODEC_ID_PCM_ALAW) {
  1240. samplesInChunk = size/enc->channels;
  1241. }
  1242. else if(enc->codec_id == CODEC_ID_PCM_S16BE || enc->codec_id == CODEC_ID_PCM_S16LE) {
  1243. samplesInChunk = size/(2*enc->channels);
  1244. }
  1245. else {
  1246. samplesInChunk = 1;
  1247. }
  1248. }
  1249. if ((enc->codec_id == CODEC_ID_MPEG4 || enc->codec_id == CODEC_ID_AAC)
  1250. && trk->vosLen == 0) {
  1251. // assert(enc->extradata_size);
  1252. trk->vosLen = enc->extradata_size;
  1253. trk->vosData = av_malloc(trk->vosLen);
  1254. memcpy(trk->vosData, enc->extradata, trk->vosLen);
  1255. }
  1256. cl = trk->entry / MOV_INDEX_CLUSTER_SIZE;
  1257. id = trk->entry % MOV_INDEX_CLUSTER_SIZE;
  1258. if (trk->ents_allocated <= trk->entry) {
  1259. trk->cluster = av_realloc(trk->cluster, (cl+1)*sizeof(void*));
  1260. if (!trk->cluster)
  1261. return -1;
  1262. trk->cluster[cl] = av_malloc(MOV_INDEX_CLUSTER_SIZE*sizeof(MOVIentry));
  1263. if (!trk->cluster[cl])
  1264. return -1;
  1265. trk->ents_allocated += MOV_INDEX_CLUSTER_SIZE;
  1266. }
  1267. if (mov->mdat_written == 0) {
  1268. mov_write_mdat_tag(pb, mov);
  1269. mov->mdat_written = 1;
  1270. mov->time = s->timestamp + 0x7C25B080; //1970 based -> 1904 based
  1271. }
  1272. trk->cluster[cl][id].pos = url_ftell(pb);
  1273. trk->cluster[cl][id].samplesInChunk = samplesInChunk;
  1274. trk->cluster[cl][id].size = size;
  1275. trk->cluster[cl][id].entries = samplesInChunk;
  1276. if(enc->codec_type == CODEC_TYPE_VIDEO) {
  1277. trk->cluster[cl][id].key_frame = !!(pkt->flags & PKT_FLAG_KEY);
  1278. if(trk->cluster[cl][id].key_frame)
  1279. trk->hasKeyframes = 1;
  1280. }
  1281. trk->enc = enc;
  1282. trk->entry++;
  1283. trk->sampleCount += samplesInChunk;
  1284. trk->mdat_size += size;
  1285. put_buffer(pb, pkt->data, size);
  1286. put_flush_packet(pb);
  1287. return 0;
  1288. }
  1289. static int mov_write_trailer(AVFormatContext *s)
  1290. {
  1291. MOVContext *mov = s->priv_data;
  1292. ByteIOContext *pb = &s->pb;
  1293. int res = 0;
  1294. int i, j;
  1295. offset_t moov_pos = url_ftell(pb);
  1296. /* Write size of mdat tag */
  1297. for (i=0, j=0; i<MAX_STREAMS; i++) {
  1298. if(mov->tracks[i].ents_allocated > 0) {
  1299. j += mov->tracks[i].mdat_size;
  1300. }
  1301. }
  1302. url_fseek(pb, mov->mdat_pos, SEEK_SET);
  1303. put_be32(pb, j+8);
  1304. url_fseek(pb, moov_pos, SEEK_SET);
  1305. mov_write_moov_tag(pb, mov, s);
  1306. for (i=0; i<MAX_STREAMS; i++) {
  1307. for (j=0; j<mov->tracks[i].ents_allocated/MOV_INDEX_CLUSTER_SIZE; j++) {
  1308. av_free(mov->tracks[i].cluster[j]);
  1309. }
  1310. av_free(mov->tracks[i].cluster);
  1311. if( mov->tracks[i].vosLen ) av_free( mov->tracks[i].vosData );
  1312. mov->tracks[i].cluster = NULL;
  1313. mov->tracks[i].ents_allocated = mov->tracks[i].entry = 0;
  1314. }
  1315. put_flush_packet(pb);
  1316. return res;
  1317. }
  1318. static AVOutputFormat mov_oformat = {
  1319. "mov",
  1320. "mov format",
  1321. NULL,
  1322. "mov",
  1323. sizeof(MOVContext),
  1324. CODEC_ID_AAC,
  1325. CODEC_ID_MPEG4,
  1326. mov_write_header,
  1327. mov_write_packet,
  1328. mov_write_trailer,
  1329. .flags = AVFMT_GLOBALHEADER,
  1330. };
  1331. static AVOutputFormat _3gp_oformat = {
  1332. "3gp",
  1333. "3gp format",
  1334. NULL,
  1335. "3gp",
  1336. sizeof(MOVContext),
  1337. CODEC_ID_AMR_NB,
  1338. CODEC_ID_H263,
  1339. mov_write_header,
  1340. mov_write_packet,
  1341. mov_write_trailer,
  1342. .flags = AVFMT_GLOBALHEADER,
  1343. };
  1344. static AVOutputFormat mp4_oformat = {
  1345. "mp4",
  1346. "mp4 format",
  1347. "application/mp4",
  1348. "mp4,m4a",
  1349. sizeof(MOVContext),
  1350. CODEC_ID_AAC,
  1351. CODEC_ID_MPEG4,
  1352. mov_write_header,
  1353. mov_write_packet,
  1354. mov_write_trailer,
  1355. .flags = AVFMT_GLOBALHEADER,
  1356. };
  1357. static AVOutputFormat psp_oformat = {
  1358. "psp",
  1359. "psp mp4 format",
  1360. NULL,
  1361. "mp4,psp",
  1362. sizeof(MOVContext),
  1363. CODEC_ID_AAC,
  1364. CODEC_ID_MPEG4,
  1365. mov_write_header,
  1366. mov_write_packet,
  1367. mov_write_trailer,
  1368. // .flags = AVFMT_GLOBALHEADER,
  1369. };
  1370. static AVOutputFormat _3g2_oformat = {
  1371. "3g2",
  1372. "3gp2 format",
  1373. NULL,
  1374. "3g2",
  1375. sizeof(MOVContext),
  1376. CODEC_ID_AMR_NB,
  1377. CODEC_ID_H263,
  1378. mov_write_header,
  1379. mov_write_packet,
  1380. mov_write_trailer,
  1381. .flags = AVFMT_GLOBALHEADER,
  1382. };
  1383. int movenc_init(void)
  1384. {
  1385. av_register_output_format(&mov_oformat);
  1386. av_register_output_format(&_3gp_oformat);
  1387. av_register_output_format(&mp4_oformat);
  1388. av_register_output_format(&psp_oformat);
  1389. av_register_output_format(&_3g2_oformat);
  1390. return 0;
  1391. }