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.

1568 lines
46KB

  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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 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 trackID;
  51. AVCodecContext *enc;
  52. int vosLen;
  53. uint8_t *vosData;
  54. MOVIentry** cluster;
  55. } MOVTrack;
  56. typedef struct MOVContext {
  57. int mode;
  58. long time;
  59. int nb_streams;
  60. int mdat_written;
  61. offset_t mdat_pos;
  62. long timescale;
  63. MOVTrack tracks[MAX_STREAMS];
  64. } MOVContext;
  65. static int mov_write_esds_tag(ByteIOContext *pb, MOVTrack* track);
  66. const CodecTag ff_mov_obj_type[] = {
  67. { CODEC_ID_MPEG4 , 32 },
  68. { CODEC_ID_AAC , 64 },
  69. { CODEC_ID_MPEG1VIDEO, 106 },
  70. { CODEC_ID_MPEG2VIDEO, 96 },//mpeg2 profiles
  71. { CODEC_ID_MP2 , 107 },//FIXME mpeg2 mpeg audio -> 105
  72. { CODEC_ID_MP3 , 107 },//FIXME mpeg2 mpeg audio -> 105
  73. { CODEC_ID_H264 , 33 },
  74. { CODEC_ID_H263 , 242 },
  75. { CODEC_ID_H261 , 243 },
  76. { CODEC_ID_MJPEG , 108 },
  77. { CODEC_ID_PCM_S16LE , 224 },
  78. { CODEC_ID_VORBIS , 225 },
  79. { CODEC_ID_AC3 , 226 },
  80. { CODEC_ID_PCM_ALAW , 227 },
  81. { CODEC_ID_PCM_MULAW , 228 },
  82. { CODEC_ID_PCM_S16BE , 230 },
  83. { 0,0 },
  84. };
  85. //FIXME supprt 64bit varaint with wide placeholders
  86. static int updateSize (ByteIOContext *pb, int pos)
  87. {
  88. long curpos = url_ftell(pb);
  89. url_fseek(pb, pos, SEEK_SET);
  90. put_be32(pb, curpos - pos); /* rewrite size */
  91. url_fseek(pb, curpos, SEEK_SET);
  92. return curpos - pos;
  93. }
  94. /* Chunk offset atom */
  95. static int mov_write_stco_tag(ByteIOContext *pb, MOVTrack* track)
  96. {
  97. int i;
  98. int pos = url_ftell(pb);
  99. put_be32(pb, 0); /* size */
  100. put_tag(pb, "stco");
  101. put_be32(pb, 0); /* version & flags */
  102. put_be32(pb, track->entry); /* entry count */
  103. for (i=0; i<track->entry; i++) {
  104. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  105. int id = i % MOV_INDEX_CLUSTER_SIZE;
  106. put_be32(pb, track->cluster[cl][id].pos);
  107. }
  108. return updateSize (pb, pos);
  109. }
  110. /* Sample size atom */
  111. static int mov_write_stsz_tag(ByteIOContext *pb, MOVTrack* track)
  112. {
  113. int equalChunks = 1;
  114. int i, j, entries = 0, tst = -1, oldtst = -1;
  115. int pos = url_ftell(pb);
  116. put_be32(pb, 0); /* size */
  117. put_tag(pb, "stsz");
  118. put_be32(pb, 0); /* version & flags */
  119. for (i=0; i<track->entry; i++) {
  120. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  121. int id = i % MOV_INDEX_CLUSTER_SIZE;
  122. tst = track->cluster[cl][id].size/track->cluster[cl][id].entries;
  123. if(oldtst != -1 && tst != oldtst) {
  124. equalChunks = 0;
  125. }
  126. oldtst = tst;
  127. entries += track->cluster[cl][id].entries;
  128. }
  129. if (equalChunks) {
  130. int sSize = track->cluster[0][0].size/track->cluster[0][0].entries;
  131. put_be32(pb, sSize); // sample size
  132. put_be32(pb, entries); // sample count
  133. }
  134. else {
  135. put_be32(pb, 0); // sample size
  136. put_be32(pb, entries); // sample count
  137. for (i=0; i<track->entry; i++) {
  138. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  139. int id = i % MOV_INDEX_CLUSTER_SIZE;
  140. for ( j=0; j<track->cluster[cl][id].entries; j++) {
  141. put_be32(pb, track->cluster[cl][id].size /
  142. track->cluster[cl][id].entries);
  143. }
  144. }
  145. }
  146. return updateSize (pb, pos);
  147. }
  148. /* Sample to chunk atom */
  149. static int mov_write_stsc_tag(ByteIOContext *pb, MOVTrack* track)
  150. {
  151. int index = 0, oldval = -1, i, entryPos, curpos;
  152. int pos = url_ftell(pb);
  153. put_be32(pb, 0); /* size */
  154. put_tag(pb, "stsc");
  155. put_be32(pb, 0); // version & flags
  156. entryPos = url_ftell(pb);
  157. put_be32(pb, track->entry); // entry count
  158. for (i=0; i<track->entry; i++) {
  159. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  160. int id = i % MOV_INDEX_CLUSTER_SIZE;
  161. if(oldval != track->cluster[cl][id].samplesInChunk)
  162. {
  163. put_be32(pb, i+1); // first chunk
  164. put_be32(pb, track->cluster[cl][id].samplesInChunk); // samples per chunk
  165. put_be32(pb, 0x1); // sample description index
  166. oldval = track->cluster[cl][id].samplesInChunk;
  167. index++;
  168. }
  169. }
  170. curpos = url_ftell(pb);
  171. url_fseek(pb, entryPos, SEEK_SET);
  172. put_be32(pb, index); // rewrite size
  173. url_fseek(pb, curpos, SEEK_SET);
  174. return updateSize (pb, pos);
  175. }
  176. /* Sync sample atom */
  177. static int mov_write_stss_tag(ByteIOContext *pb, MOVTrack* track)
  178. {
  179. long curpos;
  180. int i, index = 0, entryPos;
  181. int pos = url_ftell(pb);
  182. put_be32(pb, 0); // size
  183. put_tag(pb, "stss");
  184. put_be32(pb, 0); // version & flags
  185. entryPos = url_ftell(pb);
  186. put_be32(pb, track->entry); // entry count
  187. for (i=0; i<track->entry; i++) {
  188. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  189. int id = i % MOV_INDEX_CLUSTER_SIZE;
  190. if(track->cluster[cl][id].key_frame == 1) {
  191. put_be32(pb, i+1);
  192. index++;
  193. }
  194. }
  195. curpos = url_ftell(pb);
  196. url_fseek(pb, entryPos, SEEK_SET);
  197. put_be32(pb, index); // rewrite size
  198. url_fseek(pb, curpos, SEEK_SET);
  199. return updateSize (pb, pos);
  200. }
  201. static int mov_write_damr_tag(ByteIOContext *pb)
  202. {
  203. put_be32(pb, 0x11); /* size */
  204. put_tag(pb, "damr");
  205. put_tag(pb, "FFMP");
  206. put_byte(pb, 0);
  207. put_be16(pb, 0x80); /* Mode set (all modes for AMR_NB) */
  208. put_be16(pb, 0xa); /* Mode change period (no restriction) */
  209. //put_be16(pb, 0x81ff); /* Mode set (all modes for AMR_NB) */
  210. //put_be16(pb, 1); /* Mode change period (no restriction) */
  211. return 0x11;
  212. }
  213. static int mov_write_wave_tag(ByteIOContext *pb, MOVTrack* track)
  214. {
  215. int pos = url_ftell(pb);
  216. put_be32(pb, 0); /* size */
  217. put_tag(pb, "wave");
  218. put_be32(pb, 12); /* size */
  219. put_tag(pb, "frma");
  220. put_tag(pb, "mp4a");
  221. put_be32(pb, 12); /* size */
  222. put_tag(pb, "mp4a");
  223. put_be32(pb, 0);
  224. mov_write_esds_tag(pb, track);
  225. put_be32(pb, 12); /* size */
  226. put_tag(pb, "srcq");
  227. put_be32(pb, 0x40);
  228. put_be32(pb, 8); /* size */
  229. put_be32(pb, 0); /* null tag */
  230. return updateSize (pb, pos);
  231. }
  232. static const CodecTag codec_movaudio_tags[] = {
  233. { CODEC_ID_PCM_MULAW, MKTAG('u', 'l', 'a', 'w') },
  234. { CODEC_ID_PCM_ALAW, MKTAG('a', 'l', 'a', 'w') },
  235. { CODEC_ID_ADPCM_IMA_QT, MKTAG('i', 'm', 'a', '4') },
  236. { CODEC_ID_MACE3, MKTAG('M', 'A', 'C', '3') },
  237. { CODEC_ID_MACE6, MKTAG('M', 'A', 'C', '6') },
  238. { CODEC_ID_AAC, MKTAG('m', 'p', '4', 'a') },
  239. { CODEC_ID_AMR_NB, MKTAG('s', 'a', 'm', 'r') },
  240. { CODEC_ID_AMR_WB, MKTAG('s', 'a', 'w', 'b') },
  241. { CODEC_ID_PCM_S16BE, MKTAG('t', 'w', 'o', 's') },
  242. { CODEC_ID_PCM_S16LE, MKTAG('s', 'o', 'w', 't') },
  243. { CODEC_ID_MP3, MKTAG('.', 'm', 'p', '3') },
  244. { 0, 0 },
  245. };
  246. static int mov_write_audio_tag(ByteIOContext *pb, MOVTrack* track)
  247. {
  248. int pos = url_ftell(pb);
  249. int tag;
  250. put_be32(pb, 0); /* size */
  251. tag = track->enc->codec_tag;
  252. if (!tag)
  253. tag = codec_get_tag(codec_movaudio_tags, track->enc->codec_id);
  254. // if no mac fcc found, try with Microsoft tags
  255. if (!tag)
  256. {
  257. int tmp = codec_get_tag(codec_wav_tags, track->enc->codec_id);
  258. tag = MKTAG('m', 's', ((tmp >> 8) & 0xff), (tmp & 0xff));
  259. }
  260. put_le32(pb, tag); // store it byteswapped
  261. put_be32(pb, 0); /* Reserved */
  262. put_be16(pb, 0); /* Reserved */
  263. put_be16(pb, 1); /* Data-reference index, XXX == 1 */
  264. /* SoundDescription */
  265. if(track->mode == MODE_MOV && track->enc->codec_id == CODEC_ID_AAC)
  266. put_be16(pb, 1); /* Version 1 */
  267. else
  268. put_be16(pb, 0); /* Version 0 */
  269. put_be16(pb, 0); /* Revision level */
  270. put_be32(pb, 0); /* Reserved */
  271. put_be16(pb, track->enc->channels); /* Number of channels */
  272. /* TODO: Currently hard-coded to 16-bit, there doesn't seem
  273. to be a good way to get number of bits of audio */
  274. put_be16(pb, 0x10); /* Reserved */
  275. if(track->enc->codec_id == CODEC_ID_AAC ||
  276. track->enc->codec_id == CODEC_ID_MP3)
  277. {
  278. put_be16(pb, 0xfffe); /* compression ID (vbr)*/
  279. }
  280. else
  281. {
  282. put_be16(pb, 0); /* compression ID (= 0) */
  283. }
  284. put_be16(pb, 0); /* packet size (= 0) */
  285. put_be16(pb, track->timescale); /* Time scale */
  286. put_be16(pb, 0); /* Reserved */
  287. if(track->mode == MODE_MOV && track->enc->codec_id == CODEC_ID_AAC)
  288. {
  289. /* SoundDescription V1 extended info */
  290. put_be32(pb, track->enc->frame_size); /* Samples per packet */
  291. put_be32(pb, 1536); /* Bytes per packet */
  292. put_be32(pb, 2); /* Bytes per frame */
  293. put_be32(pb, 2); /* Bytes per sample */
  294. }
  295. if(track->enc->codec_id == CODEC_ID_AAC) {
  296. if( track->mode == MODE_MOV ) mov_write_wave_tag(pb, track);
  297. else mov_write_esds_tag(pb, track);
  298. }
  299. if(track->enc->codec_id == CODEC_ID_AMR_NB)
  300. mov_write_damr_tag(pb);
  301. return updateSize (pb, pos);
  302. }
  303. static int mov_write_d263_tag(ByteIOContext *pb)
  304. {
  305. put_be32(pb, 0xf); /* size */
  306. put_tag(pb, "d263");
  307. put_tag(pb, "FFMP");
  308. put_be16(pb, 0x0a);
  309. put_byte(pb, 0);
  310. return 0xf;
  311. }
  312. /* TODO: No idea about these values */
  313. static int mov_write_svq3_tag(ByteIOContext *pb)
  314. {
  315. put_be32(pb, 0x15);
  316. put_tag(pb, "SMI ");
  317. put_tag(pb, "SEQH");
  318. put_be32(pb, 0x5);
  319. put_be32(pb, 0xe2c0211d);
  320. put_be32(pb, 0xc0000000);
  321. put_byte(pb, 0);
  322. return 0x15;
  323. }
  324. static unsigned int descrLength(unsigned int len)
  325. {
  326. if (len < 0x00000080)
  327. return 2 + len;
  328. else if (len < 0x00004000)
  329. return 3 + len;
  330. else if(len < 0x00200000)
  331. return 4 + len;
  332. else
  333. return 5 + len;
  334. }
  335. static void putDescr(ByteIOContext *pb, int tag, int size)
  336. {
  337. uint32_t len;
  338. uint8_t vals[4];
  339. len = size;
  340. vals[3] = (uint8_t)(len & 0x7f);
  341. len >>= 7;
  342. vals[2] = (uint8_t)((len & 0x7f) | 0x80);
  343. len >>= 7;
  344. vals[1] = (uint8_t)((len & 0x7f) | 0x80);
  345. len >>= 7;
  346. vals[0] = (uint8_t)((len & 0x7f) | 0x80);
  347. put_byte(pb, tag); // DescriptorTag
  348. if (size < 0x00000080)
  349. {
  350. put_byte(pb, vals[3]);
  351. }
  352. else if (size < 0x00004000)
  353. {
  354. put_byte(pb, vals[2]);
  355. put_byte(pb, vals[3]);
  356. }
  357. else if (size < 0x00200000)
  358. {
  359. put_byte(pb, vals[1]);
  360. put_byte(pb, vals[2]);
  361. put_byte(pb, vals[3]);
  362. }
  363. else if (size < 0x10000000)
  364. {
  365. put_byte(pb, vals[0]);
  366. put_byte(pb, vals[1]);
  367. put_byte(pb, vals[2]);
  368. put_byte(pb, vals[3]);
  369. }
  370. }
  371. static int mov_write_esds_tag(ByteIOContext *pb, MOVTrack* track) // Basic
  372. {
  373. int decoderSpecificInfoLen;
  374. int pos = url_ftell(pb);
  375. void *vosDataBackup=track->vosData;
  376. int vosLenBackup=track->vosLen;
  377. // we should be able to have these passed in, via vosData, then we wouldn't need to attack this routine at all
  378. static const char PSPAACData[]={0x13,0x10};
  379. 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 };
  380. if (track->mode == MODE_PSP) // fails on psp if this is not here
  381. {
  382. if (track->enc->codec_id == CODEC_ID_AAC)
  383. {
  384. track->vosLen = 2;
  385. track->vosData = (uint8_t *) PSPAACData;
  386. }
  387. if (track->enc->codec_id == CODEC_ID_MPEG4)
  388. {
  389. track->vosLen = 28;
  390. track->vosData = (uint8_t *) PSPMP4Data;
  391. }
  392. }
  393. decoderSpecificInfoLen = track->vosLen ? descrLength(track->vosLen):0;
  394. put_be32(pb, 0); // size
  395. put_tag(pb, "esds");
  396. put_be32(pb, 0); // Version
  397. // ES descriptor
  398. putDescr(pb, 0x03, 3 + descrLength(13 + decoderSpecificInfoLen) +
  399. descrLength(1));
  400. put_be16(pb, track->trackID);
  401. put_byte(pb, 0x00); // flags (= no flags)
  402. // DecoderConfig descriptor
  403. putDescr(pb, 0x04, 13 + decoderSpecificInfoLen);
  404. // Object type indication
  405. put_byte(pb, codec_get_tag(ff_mov_obj_type, track->enc->codec_id));
  406. // the following fields is made of 6 bits to identify the streamtype (4 for video, 5 for audio)
  407. // plus 1 bit to indicate upstream and 1 bit set to 1 (reserved)
  408. if(track->enc->codec_type == CODEC_TYPE_AUDIO)
  409. put_byte(pb, 0x15); // flags (= Audiostream)
  410. else
  411. put_byte(pb, 0x11); // flags (= Visualstream)
  412. put_byte(pb, track->enc->rc_buffer_size>>(3+16)); // Buffersize DB (24 bits)
  413. put_be16(pb, (track->enc->rc_buffer_size>>3)&0xFFFF); // Buffersize DB
  414. put_be32(pb, FFMAX(track->enc->bit_rate, track->enc->rc_max_rate)); // maxbitrate (FIXME should be max rate in any 1 sec window)
  415. if(track->enc->rc_max_rate != track->enc->rc_min_rate || track->enc->rc_min_rate==0)
  416. put_be32(pb, 0); // vbr
  417. else
  418. put_be32(pb, track->enc->rc_max_rate); // avg bitrate
  419. if (track->vosLen)
  420. {
  421. // DecoderSpecific info descriptor
  422. putDescr(pb, 0x05, track->vosLen);
  423. put_buffer(pb, track->vosData, track->vosLen);
  424. }
  425. track->vosData = vosDataBackup;
  426. track->vosLen = vosLenBackup;
  427. // SL descriptor
  428. putDescr(pb, 0x06, 1);
  429. put_byte(pb, 0x02);
  430. return updateSize (pb, pos);
  431. }
  432. static const CodecTag codec_movvideo_tags[] = {
  433. { CODEC_ID_SVQ1, MKTAG('S', 'V', 'Q', '1') },
  434. { CODEC_ID_SVQ3, MKTAG('S', 'V', 'Q', '3') },
  435. { CODEC_ID_MPEG4, MKTAG('m', 'p', '4', 'v') },
  436. { CODEC_ID_H263, MKTAG('s', '2', '6', '3') },
  437. { CODEC_ID_H264, MKTAG('a', 'v', 'c', '1') },
  438. { CODEC_ID_DVVIDEO, MKTAG('d', 'v', 'c', ' ') },
  439. { 0, 0 },
  440. };
  441. static int mov_write_video_tag(ByteIOContext *pb, MOVTrack* track)
  442. {
  443. int pos = url_ftell(pb);
  444. char compressor_name[32];
  445. int tag;
  446. put_be32(pb, 0); /* size */
  447. tag = track->enc->codec_tag;
  448. if (!tag)
  449. tag = codec_get_tag(codec_movvideo_tags, track->enc->codec_id);
  450. // if no mac fcc found, try with Microsoft tags
  451. if (!tag)
  452. tag = codec_get_tag(codec_bmp_tags, track->enc->codec_id);
  453. put_le32(pb, tag); // store it byteswapped
  454. put_be32(pb, 0); /* Reserved */
  455. put_be16(pb, 0); /* Reserved */
  456. put_be16(pb, 1); /* Data-reference index */
  457. put_be16(pb, 0); /* Codec stream version */
  458. put_be16(pb, 0); /* Codec stream revision (=0) */
  459. put_tag(pb, "FFMP"); /* Vendor */
  460. if(track->enc->codec_id == CODEC_ID_RAWVIDEO) {
  461. put_be32(pb, 0); /* Temporal Quality */
  462. put_be32(pb, 0x400); /* Spatial Quality = lossless*/
  463. } else {
  464. put_be32(pb, 0x200); /* Temporal Quality = normal */
  465. put_be32(pb, 0x200); /* Spatial Quality = normal */
  466. }
  467. put_be16(pb, track->enc->width); /* Video width */
  468. put_be16(pb, track->enc->height); /* Video height */
  469. put_be32(pb, 0x00480000); /* Horizontal resolution 72dpi */
  470. put_be32(pb, 0x00480000); /* Vertical resolution 72dpi */
  471. put_be32(pb, 0); /* Data size (= 0) */
  472. put_be16(pb, 1); /* Frame count (= 1) */
  473. memset(compressor_name,0,32);
  474. if (track->enc->codec && track->enc->codec->name)
  475. strncpy(compressor_name,track->enc->codec->name,31);
  476. put_byte(pb, strlen(compressor_name));
  477. put_buffer(pb, compressor_name, 31);
  478. put_be16(pb, 0x18); /* Reserved */
  479. put_be16(pb, 0xffff); /* Reserved */
  480. if(track->enc->codec_id == CODEC_ID_MPEG4)
  481. mov_write_esds_tag(pb, track);
  482. else if(track->enc->codec_id == CODEC_ID_H263)
  483. mov_write_d263_tag(pb);
  484. else if(track->enc->codec_id == CODEC_ID_SVQ3)
  485. mov_write_svq3_tag(pb);
  486. return updateSize (pb, pos);
  487. }
  488. static int mov_write_stsd_tag(ByteIOContext *pb, MOVTrack* track)
  489. {
  490. int pos = url_ftell(pb);
  491. put_be32(pb, 0); /* size */
  492. put_tag(pb, "stsd");
  493. put_be32(pb, 0); /* version & flags */
  494. put_be32(pb, 1); /* entry count */
  495. if (track->enc->codec_type == CODEC_TYPE_VIDEO)
  496. mov_write_video_tag(pb, track);
  497. else if (track->enc->codec_type == CODEC_TYPE_AUDIO)
  498. mov_write_audio_tag(pb, track);
  499. return updateSize(pb, pos);
  500. }
  501. /* TODO: */
  502. /* Time to sample atom */
  503. static int mov_write_stts_tag(ByteIOContext *pb, MOVTrack* track)
  504. {
  505. put_be32(pb, 0x18); /* size */
  506. put_tag(pb, "stts");
  507. put_be32(pb, 0); /* version & flags */
  508. put_be32(pb, 1); /* entry count */
  509. put_be32(pb, track->sampleCount); /* sample count */
  510. put_be32(pb, track->sampleDuration); /* sample duration */
  511. return 0x18;
  512. }
  513. static int mov_write_dref_tag(ByteIOContext *pb)
  514. {
  515. put_be32(pb, 28); /* size */
  516. put_tag(pb, "dref");
  517. put_be32(pb, 0); /* version & flags */
  518. put_be32(pb, 1); /* entry count */
  519. put_be32(pb, 0xc); /* size */
  520. put_tag(pb, "url ");
  521. put_be32(pb, 1); /* version & flags */
  522. return 28;
  523. }
  524. static int mov_write_stbl_tag(ByteIOContext *pb, MOVTrack* track)
  525. {
  526. int pos = url_ftell(pb);
  527. put_be32(pb, 0); /* size */
  528. put_tag(pb, "stbl");
  529. mov_write_stsd_tag(pb, track);
  530. mov_write_stts_tag(pb, track);
  531. if (track->enc->codec_type == CODEC_TYPE_VIDEO &&
  532. track->hasKeyframes)
  533. mov_write_stss_tag(pb, track);
  534. mov_write_stsc_tag(pb, track);
  535. mov_write_stsz_tag(pb, track);
  536. mov_write_stco_tag(pb, track);
  537. return updateSize(pb, pos);
  538. }
  539. static int mov_write_dinf_tag(ByteIOContext *pb)
  540. {
  541. int pos = url_ftell(pb);
  542. put_be32(pb, 0); /* size */
  543. put_tag(pb, "dinf");
  544. mov_write_dref_tag(pb);
  545. return updateSize(pb, pos);
  546. }
  547. static int mov_write_smhd_tag(ByteIOContext *pb)
  548. {
  549. put_be32(pb, 16); /* size */
  550. put_tag(pb, "smhd");
  551. put_be32(pb, 0); /* version & flags */
  552. put_be16(pb, 0); /* reserved (balance, normally = 0) */
  553. put_be16(pb, 0); /* reserved */
  554. return 16;
  555. }
  556. static int mov_write_vmhd_tag(ByteIOContext *pb)
  557. {
  558. put_be32(pb, 0x14); /* size (always 0x14) */
  559. put_tag(pb, "vmhd");
  560. put_be32(pb, 0x01); /* version & flags */
  561. put_be64(pb, 0); /* reserved (graphics mode = copy) */
  562. return 0x14;
  563. }
  564. static int mov_write_hdlr_tag(ByteIOContext *pb, MOVTrack* track)
  565. {
  566. char *descr, *hdlr, *hdlr_type;
  567. int pos = url_ftell(pb);
  568. if (!track) { /* no media --> data handler */
  569. hdlr = "dhlr";
  570. hdlr_type = "url ";
  571. descr = "DataHandler";
  572. } else {
  573. hdlr = (track->mode == MODE_MOV) ? "mhlr" : "\0\0\0\0";
  574. if (track->enc->codec_type == CODEC_TYPE_VIDEO) {
  575. hdlr_type = "vide";
  576. descr = "VideoHandler";
  577. } else {
  578. hdlr_type = "soun";
  579. descr = "SoundHandler";
  580. }
  581. }
  582. put_be32(pb, 0); /* size */
  583. put_tag(pb, "hdlr");
  584. put_be32(pb, 0); /* Version & flags */
  585. put_buffer(pb, hdlr, 4); /* handler */
  586. put_tag(pb, hdlr_type); /* handler type */
  587. put_be32(pb ,0); /* reserved */
  588. put_be32(pb ,0); /* reserved */
  589. put_be32(pb ,0); /* reserved */
  590. put_byte(pb, strlen(descr)); /* string counter */
  591. put_buffer(pb, descr, strlen(descr)); /* handler description */
  592. return updateSize(pb, pos);
  593. }
  594. static int mov_write_minf_tag(ByteIOContext *pb, MOVTrack* track)
  595. {
  596. int pos = url_ftell(pb);
  597. put_be32(pb, 0); /* size */
  598. put_tag(pb, "minf");
  599. if(track->enc->codec_type == CODEC_TYPE_VIDEO)
  600. mov_write_vmhd_tag(pb);
  601. else
  602. mov_write_smhd_tag(pb);
  603. if (track->mode == MODE_MOV) /* FIXME: Why do it for MODE_MOV only ? */
  604. mov_write_hdlr_tag(pb, NULL);
  605. mov_write_dinf_tag(pb);
  606. mov_write_stbl_tag(pb, track);
  607. return updateSize(pb, pos);
  608. }
  609. static int mov_write_mdhd_tag(ByteIOContext *pb, MOVTrack* track)
  610. {
  611. put_be32(pb, 32); /* size */
  612. put_tag(pb, "mdhd");
  613. put_be32(pb, 0); /* Version & flags */
  614. put_be32(pb, track->time); /* creation time */
  615. put_be32(pb, track->time); /* modification time */
  616. put_be32(pb, track->timescale); /* time scale (sample rate for audio) */
  617. put_be32(pb, track->trackDuration); /* duration */
  618. put_be16(pb, 0); /* language, 0 = english */
  619. put_be16(pb, 0); /* reserved (quality) */
  620. return 32;
  621. }
  622. static int mov_write_mdia_tag(ByteIOContext *pb, MOVTrack* track)
  623. {
  624. int pos = url_ftell(pb);
  625. put_be32(pb, 0); /* size */
  626. put_tag(pb, "mdia");
  627. mov_write_mdhd_tag(pb, track);
  628. mov_write_hdlr_tag(pb, track);
  629. mov_write_minf_tag(pb, track);
  630. return updateSize(pb, pos);
  631. }
  632. static int mov_write_tkhd_tag(ByteIOContext *pb, MOVTrack* track)
  633. {
  634. put_be32(pb, 0x5c); /* size (always 0x5c) */
  635. put_tag(pb, "tkhd");
  636. put_be32(pb, 0xf); /* version & flags (track enabled) */
  637. put_be32(pb, track->time); /* creation time */
  638. put_be32(pb, track->time); /* modification time */
  639. put_be32(pb, track->trackID); /* track-id */
  640. put_be32(pb, 0); /* reserved */
  641. put_be32(pb, av_rescale_rnd(track->trackDuration, globalTimescale, track->timescale, AV_ROUND_UP)); /* duration */
  642. put_be32(pb, 0); /* reserved */
  643. put_be32(pb, 0); /* reserved */
  644. put_be32(pb, 0x0); /* reserved (Layer & Alternate group) */
  645. /* Volume, only for audio */
  646. if(track->enc->codec_type == CODEC_TYPE_AUDIO)
  647. put_be16(pb, 0x0100);
  648. else
  649. put_be16(pb, 0);
  650. put_be16(pb, 0); /* reserved */
  651. /* Matrix structure */
  652. put_be32(pb, 0x00010000); /* reserved */
  653. put_be32(pb, 0x0); /* reserved */
  654. put_be32(pb, 0x0); /* reserved */
  655. put_be32(pb, 0x0); /* reserved */
  656. put_be32(pb, 0x00010000); /* reserved */
  657. put_be32(pb, 0x0); /* reserved */
  658. put_be32(pb, 0x0); /* reserved */
  659. put_be32(pb, 0x0); /* reserved */
  660. put_be32(pb, 0x40000000); /* reserved */
  661. /* Track width and height, for visual only */
  662. if(track->enc->codec_type == CODEC_TYPE_VIDEO) {
  663. double sample_aspect_ratio = av_q2d(track->enc->sample_aspect_ratio);
  664. if( !sample_aspect_ratio ) sample_aspect_ratio = 1;
  665. put_be32(pb, sample_aspect_ratio * track->enc->width*0x10000);
  666. put_be32(pb, track->enc->height*0x10000);
  667. }
  668. else {
  669. put_be32(pb, 0);
  670. put_be32(pb, 0);
  671. }
  672. return 0x5c;
  673. }
  674. // This box seems important for the psp playback ... without it the movie seems to hang
  675. static int mov_write_edts_tag(ByteIOContext *pb, MOVTrack *track)
  676. {
  677. put_be32(pb, 0x24); /* size */
  678. put_tag(pb, "edts");
  679. put_be32(pb, 0x1c); /* size */
  680. put_tag(pb, "elst");
  681. put_be32(pb, 0x0);
  682. put_be32(pb, 0x1);
  683. put_be32(pb, av_rescale_rnd(track->trackDuration, globalTimescale, track->timescale, AV_ROUND_UP)); /* duration ... doesn't seem to effect psp */
  684. put_be32(pb, 0x0);
  685. put_be32(pb, 0x00010000);
  686. return 0x24;
  687. }
  688. // goes at the end of each track! ... Critical for PSP playback ("Incompatible data" without it)
  689. static int mov_write_uuid_tag_psp(ByteIOContext *pb, MOVTrack *mov)
  690. {
  691. put_be32(pb, 0x34); /* size ... reports as 28 in mp4box! */
  692. put_tag(pb, "uuid");
  693. put_tag(pb, "USMT");
  694. put_be32(pb, 0x21d24fce);
  695. put_be32(pb, 0xbb88695c);
  696. put_be32(pb, 0xfac9c740);
  697. put_be32(pb, 0x1c); // another size here!
  698. put_tag(pb, "MTDT");
  699. put_be32(pb, 0x00010012);
  700. put_be32(pb, 0x0a);
  701. put_be32(pb, 0x55c40000);
  702. put_be32(pb, 0x1);
  703. put_be32(pb, 0x0);
  704. return 0x34;
  705. }
  706. static int mov_write_trak_tag(ByteIOContext *pb, MOVTrack* track)
  707. {
  708. int pos = url_ftell(pb);
  709. put_be32(pb, 0); /* size */
  710. put_tag(pb, "trak");
  711. mov_write_tkhd_tag(pb, track);
  712. if (track->mode == MODE_PSP)
  713. mov_write_edts_tag(pb, track); // PSP Movies require edts box
  714. mov_write_mdia_tag(pb, track);
  715. if (track->mode == MODE_PSP)
  716. mov_write_uuid_tag_psp(pb,track); // PSP Movies require this uuid box
  717. return updateSize(pb, pos);
  718. }
  719. #if 0
  720. /* TODO: Not sorted out, but not necessary either */
  721. static int mov_write_iods_tag(ByteIOContext *pb, MOVContext *mov)
  722. {
  723. put_be32(pb, 0x15); /* size */
  724. put_tag(pb, "iods");
  725. put_be32(pb, 0); /* version & flags */
  726. put_be16(pb, 0x1007);
  727. put_byte(pb, 0);
  728. put_be16(pb, 0x4fff);
  729. put_be16(pb, 0xfffe);
  730. put_be16(pb, 0x01ff);
  731. return 0x15;
  732. }
  733. #endif
  734. static int mov_write_mvhd_tag(ByteIOContext *pb, MOVContext *mov)
  735. {
  736. int maxTrackID = 1, maxTrackLen = 0, i;
  737. int64_t maxTrackLenTemp;
  738. put_be32(pb, 0x6c); /* size (always 0x6c) */
  739. put_tag(pb, "mvhd");
  740. put_be32(pb, 0); /* version & flags */
  741. put_be32(pb, mov->time); /* creation time */
  742. put_be32(pb, mov->time); /* modification time */
  743. put_be32(pb, mov->timescale); /* timescale */
  744. for (i=0; i<MAX_STREAMS; i++) {
  745. if(mov->tracks[i].entry > 0) {
  746. maxTrackLenTemp = av_rescale_rnd(mov->tracks[i].trackDuration, globalTimescale, mov->tracks[i].timescale, AV_ROUND_UP);
  747. if(maxTrackLen < maxTrackLenTemp)
  748. maxTrackLen = maxTrackLenTemp;
  749. if(maxTrackID < mov->tracks[i].trackID)
  750. maxTrackID = mov->tracks[i].trackID;
  751. }
  752. }
  753. put_be32(pb, maxTrackLen); /* duration of longest track */
  754. put_be32(pb, 0x00010000); /* reserved (preferred rate) 1.0 = normal */
  755. put_be16(pb, 0x0100); /* reserved (preferred volume) 1.0 = normal */
  756. put_be16(pb, 0); /* reserved */
  757. put_be32(pb, 0); /* reserved */
  758. put_be32(pb, 0); /* reserved */
  759. /* Matrix structure */
  760. put_be32(pb, 0x00010000); /* reserved */
  761. put_be32(pb, 0x0); /* reserved */
  762. put_be32(pb, 0x0); /* reserved */
  763. put_be32(pb, 0x0); /* reserved */
  764. put_be32(pb, 0x00010000); /* reserved */
  765. put_be32(pb, 0x0); /* reserved */
  766. put_be32(pb, 0x0); /* reserved */
  767. put_be32(pb, 0x0); /* reserved */
  768. put_be32(pb, 0x40000000); /* reserved */
  769. put_be32(pb, 0); /* reserved (preview time) */
  770. put_be32(pb, 0); /* reserved (preview duration) */
  771. put_be32(pb, 0); /* reserved (poster time) */
  772. put_be32(pb, 0); /* reserved (selection time) */
  773. put_be32(pb, 0); /* reserved (selection duration) */
  774. put_be32(pb, 0); /* reserved (current time) */
  775. put_be32(pb, maxTrackID+1); /* Next track id */
  776. return 0x6c;
  777. }
  778. static int mov_write_itunes_hdlr_tag(ByteIOContext *pb, MOVContext* mov,
  779. AVFormatContext *s)
  780. {
  781. int pos = url_ftell(pb);
  782. put_be32(pb, 0); /* size */
  783. put_tag(pb, "hdlr");
  784. put_be32(pb, 0);
  785. put_be32(pb, 0);
  786. put_tag(pb, "mdir");
  787. put_tag(pb, "appl");
  788. put_be32(pb, 0);
  789. put_be32(pb, 0);
  790. put_be16(pb, 0);
  791. return updateSize(pb, pos);
  792. }
  793. /* helper function to write a data tag with the specified string as data */
  794. static int mov_write_string_data_tag(ByteIOContext *pb, MOVContext* mov,
  795. AVFormatContext *s, const char *data)
  796. {
  797. int pos = url_ftell(pb);
  798. put_be32(pb, 0); /* size */
  799. put_tag(pb, "data");
  800. put_be32(pb, 1);
  801. put_be32(pb, 0);
  802. put_buffer(pb, data, strlen(data));
  803. return updateSize(pb, pos);
  804. }
  805. /* iTunes name of the song/movie */
  806. static int mov_write_nam_tag(ByteIOContext *pb, MOVContext* mov,
  807. AVFormatContext *s)
  808. {
  809. int size = 0;
  810. if ( s->title[0] ) {
  811. int pos = url_ftell(pb);
  812. put_be32(pb, 0); /* size */
  813. put_tag(pb, "\251nam");
  814. mov_write_string_data_tag(pb, mov, s, s->title);
  815. size = updateSize(pb, pos);
  816. }
  817. return size;
  818. }
  819. /* iTunes name of the artist/performer */
  820. static int mov_write_ART_tag(ByteIOContext *pb, MOVContext* mov,
  821. AVFormatContext *s)
  822. {
  823. int size = 0;
  824. if ( s->author[0] ) {
  825. int pos = url_ftell(pb);
  826. put_be32(pb, 0); /* size */
  827. put_tag(pb, "\251ART");
  828. // we use the author here as this is the only thing that we have...
  829. mov_write_string_data_tag(pb, mov, s, s->author);
  830. size = updateSize(pb, pos);
  831. }
  832. return size;
  833. }
  834. /* iTunes name of the writer */
  835. static int mov_write_wrt_tag(ByteIOContext *pb, MOVContext* mov,
  836. AVFormatContext *s)
  837. {
  838. int size = 0;
  839. if ( s->author[0] ) {
  840. int pos = url_ftell(pb);
  841. put_be32(pb, 0); /* size */
  842. put_tag(pb, "\251wrt");
  843. mov_write_string_data_tag(pb, mov, s, s->author);
  844. size = updateSize(pb, pos);
  845. }
  846. return size;
  847. }
  848. /* iTunes name of the album */
  849. static int mov_write_alb_tag(ByteIOContext *pb, MOVContext* mov,
  850. AVFormatContext *s)
  851. {
  852. int size = 0;
  853. if ( s->album[0] ) {
  854. int pos = url_ftell(pb);
  855. put_be32(pb, 0); /* size */
  856. put_tag(pb, "\251alb");
  857. mov_write_string_data_tag(pb, mov, s, s->album);
  858. size = updateSize(pb, pos);
  859. }
  860. return size;
  861. }
  862. /* iTunes year */
  863. static int mov_write_day_tag(ByteIOContext *pb, MOVContext* mov,
  864. AVFormatContext *s)
  865. {
  866. char year[5];
  867. int size = 0;
  868. if ( s->year ) {
  869. int pos = url_ftell(pb);
  870. put_be32(pb, 0); /* size */
  871. put_tag(pb, "\251day");
  872. snprintf(year, 5, "%04d", s->year);
  873. mov_write_string_data_tag(pb, mov, s, year);
  874. size = updateSize(pb, pos);
  875. }
  876. return size;
  877. }
  878. /* iTunes tool used to create the file */
  879. static int mov_write_too_tag(ByteIOContext *pb, MOVContext* mov,
  880. AVFormatContext *s)
  881. {
  882. int pos = url_ftell(pb);
  883. put_be32(pb, 0); /* size */
  884. put_tag(pb, "\251too");
  885. mov_write_string_data_tag(pb, mov, s, LIBAVFORMAT_IDENT);
  886. return updateSize(pb, pos);
  887. }
  888. /* iTunes comment */
  889. static int mov_write_cmt_tag(ByteIOContext *pb, MOVContext* mov,
  890. AVFormatContext *s)
  891. {
  892. int size = 0;
  893. if ( s->comment[0] ) {
  894. int pos = url_ftell(pb);
  895. put_be32(pb, 0); /* size */
  896. put_tag(pb, "\251cmt");
  897. mov_write_string_data_tag(pb, mov, s, s->comment);
  898. size = updateSize(pb, pos);
  899. }
  900. return size;
  901. }
  902. /* iTunes custom genre */
  903. static int mov_write_gen_tag(ByteIOContext *pb, MOVContext* mov,
  904. AVFormatContext *s)
  905. {
  906. int size = 0;
  907. if ( s->genre[0] ) {
  908. int pos = url_ftell(pb);
  909. put_be32(pb, 0); /* size */
  910. put_tag(pb, "\251gen");
  911. mov_write_string_data_tag(pb, mov, s, s->genre);
  912. size = updateSize(pb, pos);
  913. }
  914. return size;
  915. }
  916. /* iTunes track number */
  917. static int mov_write_trkn_tag(ByteIOContext *pb, MOVContext* mov,
  918. AVFormatContext *s)
  919. {
  920. int size = 0;
  921. if ( s->track ) {
  922. int pos = url_ftell(pb);
  923. put_be32(pb, 0); /* size */
  924. put_tag(pb, "trkn");
  925. {
  926. int pos = url_ftell(pb);
  927. put_be32(pb, 0); /* size */
  928. put_tag(pb, "data");
  929. put_be32(pb, 0); // 8 bytes empty
  930. put_be32(pb, 0);
  931. put_be16(pb, 0); // empty
  932. put_be16(pb, s->track); // track number
  933. put_be16(pb, 0); // total track number
  934. put_be16(pb, 0); // empty
  935. updateSize(pb, pos);
  936. }
  937. size = updateSize(pb, pos);
  938. }
  939. return size;
  940. }
  941. /* iTunes meta data list */
  942. static int mov_write_ilst_tag(ByteIOContext *pb, MOVContext* mov,
  943. AVFormatContext *s)
  944. {
  945. int pos = url_ftell(pb);
  946. put_be32(pb, 0); /* size */
  947. put_tag(pb, "ilst");
  948. mov_write_nam_tag(pb, mov, s);
  949. mov_write_ART_tag(pb, mov, s);
  950. mov_write_wrt_tag(pb, mov, s);
  951. mov_write_alb_tag(pb, mov, s);
  952. mov_write_day_tag(pb, mov, s);
  953. mov_write_too_tag(pb, mov, s);
  954. mov_write_cmt_tag(pb, mov, s);
  955. mov_write_gen_tag(pb, mov, s);
  956. mov_write_trkn_tag(pb, mov, s);
  957. return updateSize(pb, pos);
  958. }
  959. /* iTunes meta data tag */
  960. static int mov_write_meta_tag(ByteIOContext *pb, MOVContext* mov,
  961. AVFormatContext *s)
  962. {
  963. int size = 0;
  964. // only save meta tag if required
  965. if ( s->title[0] || s->author[0] || s->album[0] || s->year ||
  966. s->comment[0] || s->genre[0] || s->track ) {
  967. int pos = url_ftell(pb);
  968. put_be32(pb, 0); /* size */
  969. put_tag(pb, "meta");
  970. put_be32(pb, 0);
  971. mov_write_itunes_hdlr_tag(pb, mov, s);
  972. mov_write_ilst_tag(pb, mov, s);
  973. size = updateSize(pb, pos);
  974. }
  975. return size;
  976. }
  977. static int mov_write_udta_tag(ByteIOContext *pb, MOVContext* mov,
  978. AVFormatContext *s)
  979. {
  980. int pos = url_ftell(pb);
  981. int i;
  982. put_be32(pb, 0); /* size */
  983. put_tag(pb, "udta");
  984. /* iTunes meta data */
  985. mov_write_meta_tag(pb, mov, s);
  986. /* Requirements */
  987. for (i=0; i<MAX_STREAMS; i++) {
  988. if(mov->tracks[i].entry <= 0) continue;
  989. if (mov->tracks[i].enc->codec_id == CODEC_ID_AAC ||
  990. mov->tracks[i].enc->codec_id == CODEC_ID_MPEG4) {
  991. int pos = url_ftell(pb);
  992. put_be32(pb, 0); /* size */
  993. put_tag(pb, "\251req");
  994. put_be16(pb, sizeof("QuickTime 6.0 or greater") - 1);
  995. put_be16(pb, 0);
  996. put_buffer(pb, "QuickTime 6.0 or greater",
  997. sizeof("QuickTime 6.0 or greater") - 1);
  998. updateSize(pb, pos);
  999. break;
  1000. }
  1001. }
  1002. /* Encoder */
  1003. if(mov->tracks[0].enc && !(mov->tracks[0].enc->flags & CODEC_FLAG_BITEXACT))
  1004. {
  1005. int pos = url_ftell(pb);
  1006. put_be32(pb, 0); /* size */
  1007. put_tag(pb, "\251enc");
  1008. put_be16(pb, sizeof(LIBAVFORMAT_IDENT) - 1); /* string length */
  1009. put_be16(pb, 0);
  1010. put_buffer(pb, LIBAVFORMAT_IDENT, sizeof(LIBAVFORMAT_IDENT) - 1);
  1011. updateSize(pb, pos);
  1012. }
  1013. if( s->title[0] )
  1014. {
  1015. int pos = url_ftell(pb);
  1016. put_be32(pb, 0); /* size */
  1017. put_tag(pb, "\251nam");
  1018. put_be16(pb, strlen(s->title)); /* string length */
  1019. put_be16(pb, 0);
  1020. put_buffer(pb, s->title, strlen(s->title));
  1021. updateSize(pb, pos);
  1022. }
  1023. if( s->author[0] )
  1024. {
  1025. int pos = url_ftell(pb);
  1026. put_be32(pb, 0); /* size */
  1027. put_tag(pb, /*"\251aut"*/ "\251day" );
  1028. put_be16(pb, strlen(s->author)); /* string length */
  1029. put_be16(pb, 0);
  1030. put_buffer(pb, s->author, strlen(s->author));
  1031. updateSize(pb, pos);
  1032. }
  1033. if( s->comment[0] )
  1034. {
  1035. int pos = url_ftell(pb);
  1036. put_be32(pb, 0); /* size */
  1037. put_tag(pb, "\251des");
  1038. put_be16(pb, strlen(s->comment)); /* string length */
  1039. put_be16(pb, 0);
  1040. put_buffer(pb, s->comment, strlen(s->comment));
  1041. updateSize(pb, pos);
  1042. }
  1043. return updateSize(pb, pos);
  1044. }
  1045. static int mov_write_moov_tag(ByteIOContext *pb, MOVContext *mov,
  1046. AVFormatContext *s)
  1047. {
  1048. int pos, i;
  1049. pos = url_ftell(pb);
  1050. put_be32(pb, 0); /* size placeholder*/
  1051. put_tag(pb, "moov");
  1052. mov->timescale = globalTimescale;
  1053. for (i=0; i<MAX_STREAMS; i++) {
  1054. if(mov->tracks[i].entry <= 0) continue;
  1055. if(mov->tracks[i].enc->codec_type == CODEC_TYPE_VIDEO) {
  1056. mov->tracks[i].timescale = mov->tracks[i].enc->time_base.den;
  1057. mov->tracks[i].sampleDuration = mov->tracks[i].enc->time_base.num;
  1058. }
  1059. else if(mov->tracks[i].enc->codec_type == CODEC_TYPE_AUDIO) {
  1060. /* If AMR, track timescale = 8000, AMR_WB = 16000 */
  1061. if(mov->tracks[i].enc->codec_id == CODEC_ID_AMR_NB) {
  1062. mov->tracks[i].sampleDuration = 160; // Bytes per chunk
  1063. mov->tracks[i].timescale = 8000;
  1064. }
  1065. else {
  1066. mov->tracks[i].timescale = mov->tracks[i].enc->sample_rate;
  1067. mov->tracks[i].sampleDuration = mov->tracks[i].enc->frame_size;
  1068. }
  1069. }
  1070. mov->tracks[i].trackDuration =
  1071. mov->tracks[i].sampleCount * mov->tracks[i].sampleDuration;
  1072. mov->tracks[i].time = mov->time;
  1073. mov->tracks[i].trackID = i+1;
  1074. }
  1075. mov_write_mvhd_tag(pb, mov);
  1076. //mov_write_iods_tag(pb, mov);
  1077. for (i=0; i<MAX_STREAMS; i++) {
  1078. if(mov->tracks[i].entry > 0) {
  1079. mov_write_trak_tag(pb, &(mov->tracks[i]));
  1080. }
  1081. }
  1082. mov_write_udta_tag(pb, mov, s);
  1083. return updateSize(pb, pos);
  1084. }
  1085. int mov_write_mdat_tag(ByteIOContext *pb, MOVContext* mov)
  1086. {
  1087. mov->mdat_pos = url_ftell(pb);
  1088. put_be32(pb, 0); /* size placeholder*/
  1089. put_tag(pb, "mdat");
  1090. return 0;
  1091. }
  1092. /* TODO: This needs to be more general */
  1093. int mov_write_ftyp_tag(ByteIOContext *pb, AVFormatContext *s)
  1094. {
  1095. MOVContext *mov = s->priv_data;
  1096. put_be32(pb, 0x14 ); /* size */
  1097. put_tag(pb, "ftyp");
  1098. if ( mov->mode == MODE_3GP )
  1099. put_tag(pb, "3gp4");
  1100. else if ( mov->mode == MODE_3G2 )
  1101. put_tag(pb, "3g2a");
  1102. else if ( mov->mode == MODE_PSP )
  1103. put_tag(pb, "MSNV");
  1104. else
  1105. put_tag(pb, "isom");
  1106. put_be32(pb, 0x200 );
  1107. if ( mov->mode == MODE_3GP )
  1108. put_tag(pb, "3gp4");
  1109. else if ( mov->mode == MODE_3G2 )
  1110. put_tag(pb, "3g2a");
  1111. else if ( mov->mode == MODE_PSP )
  1112. put_tag(pb, "MSNV");
  1113. else
  1114. put_tag(pb, "mp41");
  1115. return 0x14;
  1116. }
  1117. static void mov_write_uuidprof_tag(ByteIOContext *pb, AVFormatContext *s)
  1118. {
  1119. int AudioRate = s->streams[1]->codec->sample_rate;
  1120. int FrameRate = ((s->streams[0]->codec->time_base.den) * (0x10000))/ (s->streams[0]->codec->time_base.num);
  1121. //printf("audiorate = %d\n",AudioRate);
  1122. //printf("framerate = %d / %d = 0x%x\n",s->streams[0]->codec->time_base.den,s->streams[0]->codec->time_base.num,FrameRate);
  1123. put_be32(pb, 0x94 ); /* size */
  1124. put_tag(pb, "uuid");
  1125. put_tag(pb, "PROF");
  1126. put_be32(pb, 0x21d24fce ); /* 96 bit UUID */
  1127. put_be32(pb, 0xbb88695c );
  1128. put_be32(pb, 0xfac9c740 );
  1129. put_be32(pb, 0x0 ); /* ? */
  1130. put_be32(pb, 0x3 ); /* 3 sections ? */
  1131. put_be32(pb, 0x14 ); /* size */
  1132. put_tag(pb, "FPRF");
  1133. put_be32(pb, 0x0 ); /* ? */
  1134. put_be32(pb, 0x0 ); /* ? */
  1135. put_be32(pb, 0x0 ); /* ? */
  1136. put_be32(pb, 0x2c ); /* size */
  1137. put_tag(pb, "APRF"); /* audio */
  1138. put_be32(pb, 0x0 );
  1139. put_be32(pb, 0x2 );
  1140. put_tag(pb, "mp4a");
  1141. put_be32(pb, 0x20f );
  1142. put_be32(pb, 0x0 );
  1143. put_be32(pb, 0x40 );
  1144. put_be32(pb, 0x40 );
  1145. put_be32(pb, AudioRate ); //24000 ... audio rate?
  1146. put_be32(pb, 0x2 );
  1147. put_be32(pb, 0x34 ); /* size */
  1148. put_tag(pb, "VPRF"); /* video */
  1149. put_be32(pb, 0x0 );
  1150. put_be32(pb, 0x1 );
  1151. put_tag(pb, "mp4v");
  1152. put_be32(pb, 0x103 );
  1153. put_be32(pb, 0x0 );
  1154. put_be32(pb, 0xc0 );
  1155. put_be32(pb, 0xc0 );
  1156. put_be32(pb, FrameRate); // was 0xefc29
  1157. put_be32(pb, FrameRate ); // was 0xefc29
  1158. put_be16(pb, s->streams[0]->codec->width);
  1159. put_be16(pb, s->streams[0]->codec->height);
  1160. put_be32(pb, 0x010001 );
  1161. }
  1162. static int mov_write_header(AVFormatContext *s)
  1163. {
  1164. ByteIOContext *pb = &s->pb;
  1165. MOVContext *mov = s->priv_data;
  1166. int i;
  1167. for(i=0; i<s->nb_streams; i++){
  1168. AVCodecContext *c= s->streams[i]->codec;
  1169. if (c->codec_type == CODEC_TYPE_VIDEO){
  1170. if (!codec_get_tag(codec_movvideo_tags, c->codec_id)){
  1171. if(!codec_get_tag(codec_bmp_tags, c->codec_id))
  1172. return -1;
  1173. else
  1174. av_log(s, AV_LOG_INFO, "Warning, using MS style video codec tag, the file may be unplayable!\n");
  1175. }
  1176. }else if(c->codec_type == CODEC_TYPE_AUDIO){
  1177. if (!codec_get_tag(codec_movaudio_tags, c->codec_id)){
  1178. if(!codec_get_tag(codec_wav_tags, c->codec_id))
  1179. return -1;
  1180. else
  1181. av_log(s, AV_LOG_INFO, "Warning, using MS style audio codec tag, the file may be unplayable!\n");
  1182. }
  1183. }
  1184. }
  1185. /* Default mode == MP4 */
  1186. mov->mode = MODE_MP4;
  1187. if (s->oformat != NULL) {
  1188. if (!strcmp("3gp", s->oformat->name)) mov->mode = MODE_3GP;
  1189. else if (!strcmp("3g2", s->oformat->name)) mov->mode = MODE_3G2;
  1190. else if (!strcmp("mov", s->oformat->name)) mov->mode = MODE_MOV;
  1191. else if (!strcmp("psp", s->oformat->name)) mov->mode = MODE_PSP;
  1192. if ( mov->mode == MODE_3GP || mov->mode == MODE_3G2 ||
  1193. mov->mode == MODE_MP4 || mov->mode == MODE_PSP )
  1194. mov_write_ftyp_tag(pb,s);
  1195. if ( mov->mode == MODE_PSP ) {
  1196. if ( s->nb_streams != 2 ) {
  1197. av_log(s, AV_LOG_ERROR, "PSP mode need one video and one audio stream\n");
  1198. return -1;
  1199. }
  1200. mov_write_uuidprof_tag(pb,s);
  1201. }
  1202. }
  1203. for (i=0; i<MAX_STREAMS; i++) {
  1204. mov->tracks[i].mode = mov->mode;
  1205. }
  1206. put_flush_packet(pb);
  1207. return 0;
  1208. }
  1209. static int mov_write_packet(AVFormatContext *s, AVPacket *pkt)
  1210. {
  1211. MOVContext *mov = s->priv_data;
  1212. ByteIOContext *pb = &s->pb;
  1213. AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
  1214. MOVTrack* trk = &mov->tracks[pkt->stream_index];
  1215. int cl, id;
  1216. unsigned int samplesInChunk = 0;
  1217. int size= pkt->size;
  1218. if (url_is_streamed(&s->pb)) return 0; /* Can't handle that */
  1219. if (!size) return 0; /* Discard 0 sized packets */
  1220. if (enc->codec_type == CODEC_TYPE_VIDEO ) {
  1221. samplesInChunk = 1;
  1222. }
  1223. else if (enc->codec_type == CODEC_TYPE_AUDIO ) {
  1224. if( enc->codec_id == CODEC_ID_AMR_NB) {
  1225. /* We must find out how many AMR blocks there are in one packet */
  1226. static uint16_t packed_size[16] =
  1227. {13, 14, 16, 18, 20, 21, 27, 32, 6, 0, 0, 0, 0, 0, 0, 0};
  1228. int len = 0;
  1229. while (len < size && samplesInChunk < 100) {
  1230. len += packed_size[(pkt->data[len] >> 3) & 0x0F];
  1231. samplesInChunk++;
  1232. }
  1233. }
  1234. else if(enc->codec_id == CODEC_ID_PCM_ALAW) {
  1235. samplesInChunk = size/enc->channels;
  1236. }
  1237. else if(enc->codec_id == CODEC_ID_PCM_S16BE || enc->codec_id == CODEC_ID_PCM_S16LE) {
  1238. samplesInChunk = size/(2*enc->channels);
  1239. }
  1240. else {
  1241. samplesInChunk = 1;
  1242. }
  1243. }
  1244. if ((enc->codec_id == CODEC_ID_MPEG4 || enc->codec_id == CODEC_ID_AAC)
  1245. && trk->vosLen == 0) {
  1246. // assert(enc->extradata_size);
  1247. trk->vosLen = enc->extradata_size;
  1248. trk->vosData = av_malloc(trk->vosLen);
  1249. memcpy(trk->vosData, enc->extradata, trk->vosLen);
  1250. }
  1251. cl = trk->entry / MOV_INDEX_CLUSTER_SIZE;
  1252. id = trk->entry % MOV_INDEX_CLUSTER_SIZE;
  1253. if (trk->ents_allocated <= trk->entry) {
  1254. trk->cluster = av_realloc(trk->cluster, (cl+1)*sizeof(void*));
  1255. if (!trk->cluster)
  1256. return -1;
  1257. trk->cluster[cl] = av_malloc(MOV_INDEX_CLUSTER_SIZE*sizeof(MOVIentry));
  1258. if (!trk->cluster[cl])
  1259. return -1;
  1260. trk->ents_allocated += MOV_INDEX_CLUSTER_SIZE;
  1261. }
  1262. if (mov->mdat_written == 0) {
  1263. mov_write_mdat_tag(pb, mov);
  1264. mov->mdat_written = 1;
  1265. mov->time = s->timestamp + 0x7C25B080; //1970 based -> 1904 based
  1266. }
  1267. trk->cluster[cl][id].pos = url_ftell(pb);
  1268. trk->cluster[cl][id].samplesInChunk = samplesInChunk;
  1269. trk->cluster[cl][id].size = size;
  1270. trk->cluster[cl][id].entries = samplesInChunk;
  1271. if(enc->codec_type == CODEC_TYPE_VIDEO) {
  1272. trk->cluster[cl][id].key_frame = !!(pkt->flags & PKT_FLAG_KEY);
  1273. if(trk->cluster[cl][id].key_frame)
  1274. trk->hasKeyframes = 1;
  1275. }
  1276. trk->enc = enc;
  1277. trk->entry++;
  1278. trk->sampleCount += samplesInChunk;
  1279. trk->mdat_size += size;
  1280. put_buffer(pb, pkt->data, size);
  1281. put_flush_packet(pb);
  1282. return 0;
  1283. }
  1284. static int mov_write_trailer(AVFormatContext *s)
  1285. {
  1286. MOVContext *mov = s->priv_data;
  1287. ByteIOContext *pb = &s->pb;
  1288. int res = 0;
  1289. int i, j;
  1290. offset_t moov_pos = url_ftell(pb);
  1291. /* Write size of mdat tag */
  1292. for (i=0, j=0; i<MAX_STREAMS; i++) {
  1293. if(mov->tracks[i].ents_allocated > 0) {
  1294. j += mov->tracks[i].mdat_size;
  1295. }
  1296. }
  1297. url_fseek(pb, mov->mdat_pos, SEEK_SET);
  1298. put_be32(pb, j+8);
  1299. url_fseek(pb, moov_pos, SEEK_SET);
  1300. mov_write_moov_tag(pb, mov, s);
  1301. for (i=0; i<MAX_STREAMS; i++) {
  1302. for (j=0; j<mov->tracks[i].ents_allocated/MOV_INDEX_CLUSTER_SIZE; j++) {
  1303. av_free(mov->tracks[i].cluster[j]);
  1304. }
  1305. av_free(mov->tracks[i].cluster);
  1306. if( mov->tracks[i].vosLen ) av_free( mov->tracks[i].vosData );
  1307. mov->tracks[i].cluster = NULL;
  1308. mov->tracks[i].ents_allocated = mov->tracks[i].entry = 0;
  1309. }
  1310. put_flush_packet(pb);
  1311. return res;
  1312. }
  1313. static AVOutputFormat mov_oformat = {
  1314. "mov",
  1315. "mov format",
  1316. NULL,
  1317. "mov",
  1318. sizeof(MOVContext),
  1319. CODEC_ID_AAC,
  1320. CODEC_ID_MPEG4,
  1321. mov_write_header,
  1322. mov_write_packet,
  1323. mov_write_trailer,
  1324. .flags = AVFMT_GLOBALHEADER,
  1325. };
  1326. static AVOutputFormat _3gp_oformat = {
  1327. "3gp",
  1328. "3gp format",
  1329. NULL,
  1330. "3gp",
  1331. sizeof(MOVContext),
  1332. CODEC_ID_AMR_NB,
  1333. CODEC_ID_H263,
  1334. mov_write_header,
  1335. mov_write_packet,
  1336. mov_write_trailer,
  1337. .flags = AVFMT_GLOBALHEADER,
  1338. };
  1339. static AVOutputFormat mp4_oformat = {
  1340. "mp4",
  1341. "mp4 format",
  1342. "application/mp4",
  1343. "mp4,m4a",
  1344. sizeof(MOVContext),
  1345. CODEC_ID_AAC,
  1346. CODEC_ID_MPEG4,
  1347. mov_write_header,
  1348. mov_write_packet,
  1349. mov_write_trailer,
  1350. .flags = AVFMT_GLOBALHEADER,
  1351. };
  1352. static AVOutputFormat psp_oformat = {
  1353. "psp",
  1354. "psp mp4 format",
  1355. NULL,
  1356. "mp4,psp",
  1357. sizeof(MOVContext),
  1358. CODEC_ID_AAC,
  1359. CODEC_ID_MPEG4,
  1360. mov_write_header,
  1361. mov_write_packet,
  1362. mov_write_trailer,
  1363. // .flags = AVFMT_GLOBALHEADER,
  1364. };
  1365. static AVOutputFormat _3g2_oformat = {
  1366. "3g2",
  1367. "3gp2 format",
  1368. NULL,
  1369. "3g2",
  1370. sizeof(MOVContext),
  1371. CODEC_ID_AMR_NB,
  1372. CODEC_ID_H263,
  1373. mov_write_header,
  1374. mov_write_packet,
  1375. mov_write_trailer,
  1376. .flags = AVFMT_GLOBALHEADER,
  1377. };
  1378. int movenc_init(void)
  1379. {
  1380. av_register_output_format(&mov_oformat);
  1381. av_register_output_format(&_3gp_oformat);
  1382. av_register_output_format(&mp4_oformat);
  1383. av_register_output_format(&psp_oformat);
  1384. av_register_output_format(&_3g2_oformat);
  1385. return 0;
  1386. }