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.

1634 lines
48KB

  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, size;
  35. uint64_t pos;
  36. unsigned int samplesInChunk;
  37. char key_frame;
  38. unsigned int entries;
  39. } MOVIentry;
  40. typedef struct MOVIndex {
  41. int mode;
  42. int entry;
  43. uint64_t mdat_size;
  44. int ents_allocated;
  45. long timescale;
  46. long time;
  47. long trackDuration;
  48. long sampleCount;
  49. long sampleDuration;
  50. int hasKeyframes;
  51. int language;
  52. int trackID;
  53. AVCodecContext *enc;
  54. int vosLen;
  55. uint8_t *vosData;
  56. MOVIentry** cluster;
  57. } MOVTrack;
  58. typedef struct MOVContext {
  59. int mode;
  60. long time;
  61. int nb_streams;
  62. int mdat_written;
  63. offset_t mdat_pos;
  64. long timescale;
  65. MOVTrack tracks[MAX_STREAMS];
  66. } MOVContext;
  67. static int mov_write_esds_tag(ByteIOContext *pb, MOVTrack* track);
  68. /* output language code from iso639 language name */
  69. extern int ff_mov_iso639_to_lang(const char *lang, int mp4);
  70. const CodecTag ff_mov_obj_type[] = {
  71. { CODEC_ID_MPEG4 , 32 },
  72. { CODEC_ID_AAC , 64 },
  73. { CODEC_ID_MPEG1VIDEO, 106 },
  74. { CODEC_ID_MPEG2VIDEO, 96 },//mpeg2 profiles
  75. { CODEC_ID_MP2 , 107 },//FIXME mpeg2 mpeg audio -> 105
  76. { CODEC_ID_MP3 , 107 },//FIXME mpeg2 mpeg audio -> 105
  77. { CODEC_ID_H264 , 33 },
  78. { CODEC_ID_H263 , 242 },
  79. { CODEC_ID_H261 , 243 },
  80. { CODEC_ID_MJPEG , 108 },
  81. { CODEC_ID_PCM_S16LE , 224 },
  82. { CODEC_ID_VORBIS , 225 },
  83. { CODEC_ID_AC3 , 226 },
  84. { CODEC_ID_PCM_ALAW , 227 },
  85. { CODEC_ID_PCM_MULAW , 228 },
  86. { CODEC_ID_PCM_S16BE , 230 },
  87. { 0,0 },
  88. };
  89. //FIXME supprt 64bit varaint with wide placeholders
  90. static offset_t updateSize (ByteIOContext *pb, offset_t pos)
  91. {
  92. offset_t curpos = url_ftell(pb);
  93. url_fseek(pb, pos, SEEK_SET);
  94. put_be32(pb, curpos - pos); /* rewrite size */
  95. url_fseek(pb, curpos, SEEK_SET);
  96. return curpos - pos;
  97. }
  98. /* Chunk offset atom */
  99. static int mov_write_stco_tag(ByteIOContext *pb, MOVTrack* track)
  100. {
  101. int i;
  102. int mode64 = 0; // use 32 bit size variant if possible
  103. offset_t pos = url_ftell(pb);
  104. put_be32(pb, 0); /* size */
  105. if (pos > UINT32_MAX) {
  106. mode64 = 1;
  107. put_tag(pb, "co64");
  108. } else
  109. put_tag(pb, "stco");
  110. put_be32(pb, 0); /* version & flags */
  111. put_be32(pb, track->entry); /* entry count */
  112. for (i=0; i<track->entry; i++) {
  113. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  114. int id = i % MOV_INDEX_CLUSTER_SIZE;
  115. if(mode64 == 1)
  116. put_be64(pb, track->cluster[cl][id].pos);
  117. else
  118. put_be32(pb, track->cluster[cl][id].pos);
  119. }
  120. return updateSize (pb, pos);
  121. }
  122. /* Sample size atom */
  123. static int mov_write_stsz_tag(ByteIOContext *pb, MOVTrack* track)
  124. {
  125. int equalChunks = 1;
  126. int i, j, entries = 0, tst = -1, oldtst = -1;
  127. offset_t pos = url_ftell(pb);
  128. put_be32(pb, 0); /* size */
  129. put_tag(pb, "stsz");
  130. put_be32(pb, 0); /* version & flags */
  131. for (i=0; i<track->entry; i++) {
  132. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  133. int id = i % MOV_INDEX_CLUSTER_SIZE;
  134. tst = track->cluster[cl][id].size/track->cluster[cl][id].entries;
  135. if(oldtst != -1 && tst != oldtst) {
  136. equalChunks = 0;
  137. }
  138. oldtst = tst;
  139. entries += track->cluster[cl][id].entries;
  140. }
  141. if (equalChunks) {
  142. int sSize = track->cluster[0][0].size/track->cluster[0][0].entries;
  143. put_be32(pb, sSize); // sample size
  144. put_be32(pb, entries); // sample count
  145. }
  146. else {
  147. put_be32(pb, 0); // sample size
  148. put_be32(pb, entries); // sample count
  149. for (i=0; i<track->entry; i++) {
  150. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  151. int id = i % MOV_INDEX_CLUSTER_SIZE;
  152. for ( j=0; j<track->cluster[cl][id].entries; j++) {
  153. put_be32(pb, track->cluster[cl][id].size /
  154. track->cluster[cl][id].entries);
  155. }
  156. }
  157. }
  158. return updateSize (pb, pos);
  159. }
  160. /* Sample to chunk atom */
  161. static int mov_write_stsc_tag(ByteIOContext *pb, MOVTrack* track)
  162. {
  163. int index = 0, oldval = -1, i;
  164. offset_t entryPos, curpos;
  165. offset_t pos = url_ftell(pb);
  166. put_be32(pb, 0); /* size */
  167. put_tag(pb, "stsc");
  168. put_be32(pb, 0); // version & flags
  169. entryPos = url_ftell(pb);
  170. put_be32(pb, track->entry); // entry count
  171. for (i=0; i<track->entry; i++) {
  172. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  173. int id = i % MOV_INDEX_CLUSTER_SIZE;
  174. if(oldval != track->cluster[cl][id].samplesInChunk)
  175. {
  176. put_be32(pb, i+1); // first chunk
  177. put_be32(pb, track->cluster[cl][id].samplesInChunk); // samples per chunk
  178. put_be32(pb, 0x1); // sample description index
  179. oldval = track->cluster[cl][id].samplesInChunk;
  180. index++;
  181. }
  182. }
  183. curpos = url_ftell(pb);
  184. url_fseek(pb, entryPos, SEEK_SET);
  185. put_be32(pb, index); // rewrite size
  186. url_fseek(pb, curpos, SEEK_SET);
  187. return updateSize (pb, pos);
  188. }
  189. /* Sync sample atom */
  190. static int mov_write_stss_tag(ByteIOContext *pb, MOVTrack* track)
  191. {
  192. offset_t curpos, entryPos;
  193. int i, index = 0;
  194. offset_t pos = url_ftell(pb);
  195. put_be32(pb, 0); // size
  196. put_tag(pb, "stss");
  197. put_be32(pb, 0); // version & flags
  198. entryPos = url_ftell(pb);
  199. put_be32(pb, track->entry); // entry count
  200. for (i=0; i<track->entry; i++) {
  201. int cl = i / MOV_INDEX_CLUSTER_SIZE;
  202. int id = i % MOV_INDEX_CLUSTER_SIZE;
  203. if(track->cluster[cl][id].key_frame == 1) {
  204. put_be32(pb, i+1);
  205. index++;
  206. }
  207. }
  208. curpos = url_ftell(pb);
  209. url_fseek(pb, entryPos, SEEK_SET);
  210. put_be32(pb, index); // rewrite size
  211. url_fseek(pb, curpos, SEEK_SET);
  212. return updateSize (pb, pos);
  213. }
  214. static int mov_write_damr_tag(ByteIOContext *pb)
  215. {
  216. put_be32(pb, 0x11); /* size */
  217. put_tag(pb, "damr");
  218. put_tag(pb, "FFMP");
  219. put_byte(pb, 0);
  220. put_be16(pb, 0x80); /* Mode set (all modes for AMR_NB) */
  221. put_be16(pb, 0xa); /* Mode change period (no restriction) */
  222. //put_be16(pb, 0x81ff); /* Mode set (all modes for AMR_NB) */
  223. //put_be16(pb, 1); /* Mode change period (no restriction) */
  224. return 0x11;
  225. }
  226. static int mov_write_wave_tag(ByteIOContext *pb, MOVTrack* track)
  227. {
  228. offset_t pos = url_ftell(pb);
  229. put_be32(pb, 0); /* size */
  230. put_tag(pb, "wave");
  231. put_be32(pb, 12); /* size */
  232. put_tag(pb, "frma");
  233. put_tag(pb, "mp4a");
  234. put_be32(pb, 12); /* size */
  235. put_tag(pb, "mp4a");
  236. put_be32(pb, 0);
  237. mov_write_esds_tag(pb, track);
  238. put_be32(pb, 12); /* size */
  239. put_tag(pb, "srcq");
  240. put_be32(pb, 0x40);
  241. put_be32(pb, 8); /* size */
  242. put_be32(pb, 0); /* null tag */
  243. return updateSize (pb, pos);
  244. }
  245. static const CodecTag codec_movaudio_tags[] = {
  246. { CODEC_ID_PCM_MULAW, MKTAG('u', 'l', 'a', 'w') },
  247. { CODEC_ID_PCM_ALAW, MKTAG('a', 'l', 'a', 'w') },
  248. { CODEC_ID_ADPCM_IMA_QT, MKTAG('i', 'm', 'a', '4') },
  249. { CODEC_ID_MACE3, MKTAG('M', 'A', 'C', '3') },
  250. { CODEC_ID_MACE6, MKTAG('M', 'A', 'C', '6') },
  251. { CODEC_ID_AAC, MKTAG('m', 'p', '4', 'a') },
  252. { CODEC_ID_AMR_NB, MKTAG('s', 'a', 'm', 'r') },
  253. { CODEC_ID_AMR_WB, MKTAG('s', 'a', 'w', 'b') },
  254. { CODEC_ID_PCM_S16BE, MKTAG('t', 'w', 'o', 's') },
  255. { CODEC_ID_PCM_S16LE, MKTAG('s', 'o', 'w', 't') },
  256. { CODEC_ID_MP3, MKTAG('.', 'm', 'p', '3') },
  257. { 0, 0 },
  258. };
  259. static int mov_write_audio_tag(ByteIOContext *pb, MOVTrack* track)
  260. {
  261. offset_t pos = url_ftell(pb);
  262. int tag;
  263. put_be32(pb, 0); /* size */
  264. tag = track->enc->codec_tag;
  265. if (!tag)
  266. tag = codec_get_tag(codec_movaudio_tags, track->enc->codec_id);
  267. // if no mac fcc found, try with Microsoft tags
  268. if (!tag)
  269. {
  270. int tmp = codec_get_tag(codec_wav_tags, track->enc->codec_id);
  271. tag = MKTAG('m', 's', ((tmp >> 8) & 0xff), (tmp & 0xff));
  272. }
  273. put_le32(pb, tag); // store it byteswapped
  274. put_be32(pb, 0); /* Reserved */
  275. put_be16(pb, 0); /* Reserved */
  276. put_be16(pb, 1); /* Data-reference index, XXX == 1 */
  277. /* SoundDescription */
  278. if(track->mode == MODE_MOV && track->enc->codec_id == CODEC_ID_AAC)
  279. put_be16(pb, 1); /* Version 1 */
  280. else
  281. put_be16(pb, 0); /* Version 0 */
  282. put_be16(pb, 0); /* Revision level */
  283. put_be32(pb, 0); /* Reserved */
  284. put_be16(pb, track->enc->channels); /* Number of channels */
  285. /* TODO: Currently hard-coded to 16-bit, there doesn't seem
  286. to be a good way to get number of bits of audio */
  287. put_be16(pb, 0x10); /* Reserved */
  288. if(track->enc->codec_id == CODEC_ID_AAC ||
  289. track->enc->codec_id == CODEC_ID_MP3)
  290. {
  291. put_be16(pb, 0xfffe); /* compression ID (vbr)*/
  292. }
  293. else
  294. {
  295. put_be16(pb, 0); /* compression ID (= 0) */
  296. }
  297. put_be16(pb, 0); /* packet size (= 0) */
  298. put_be16(pb, track->timescale); /* Time scale */
  299. put_be16(pb, 0); /* Reserved */
  300. if(track->mode == MODE_MOV && track->enc->codec_id == CODEC_ID_AAC)
  301. {
  302. /* SoundDescription V1 extended info */
  303. put_be32(pb, track->enc->frame_size); /* Samples per packet */
  304. put_be32(pb, 1536); /* Bytes per packet */
  305. put_be32(pb, 2); /* Bytes per frame */
  306. put_be32(pb, 2); /* Bytes per sample */
  307. }
  308. if(track->enc->codec_id == CODEC_ID_AAC) {
  309. if( track->mode == MODE_MOV ) mov_write_wave_tag(pb, track);
  310. else mov_write_esds_tag(pb, track);
  311. }
  312. if(track->enc->codec_id == CODEC_ID_AMR_NB)
  313. mov_write_damr_tag(pb);
  314. return updateSize (pb, pos);
  315. }
  316. static int mov_write_d263_tag(ByteIOContext *pb)
  317. {
  318. put_be32(pb, 0xf); /* size */
  319. put_tag(pb, "d263");
  320. put_tag(pb, "FFMP");
  321. put_be16(pb, 0x0a);
  322. put_byte(pb, 0);
  323. return 0xf;
  324. }
  325. /* TODO: No idea about these values */
  326. static int mov_write_svq3_tag(ByteIOContext *pb)
  327. {
  328. put_be32(pb, 0x15);
  329. put_tag(pb, "SMI ");
  330. put_tag(pb, "SEQH");
  331. put_be32(pb, 0x5);
  332. put_be32(pb, 0xe2c0211d);
  333. put_be32(pb, 0xc0000000);
  334. put_byte(pb, 0);
  335. return 0x15;
  336. }
  337. static unsigned int descrLength(unsigned int len)
  338. {
  339. if (len < 0x00000080)
  340. return 2 + len;
  341. else if (len < 0x00004000)
  342. return 3 + len;
  343. else if(len < 0x00200000)
  344. return 4 + len;
  345. else
  346. return 5 + len;
  347. }
  348. static void putDescr(ByteIOContext *pb, int tag, int size)
  349. {
  350. uint32_t len;
  351. uint8_t vals[4];
  352. len = size;
  353. vals[3] = (uint8_t)(len & 0x7f);
  354. len >>= 7;
  355. vals[2] = (uint8_t)((len & 0x7f) | 0x80);
  356. len >>= 7;
  357. vals[1] = (uint8_t)((len & 0x7f) | 0x80);
  358. len >>= 7;
  359. vals[0] = (uint8_t)((len & 0x7f) | 0x80);
  360. put_byte(pb, tag); // DescriptorTag
  361. if (size < 0x00000080)
  362. {
  363. put_byte(pb, vals[3]);
  364. }
  365. else if (size < 0x00004000)
  366. {
  367. put_byte(pb, vals[2]);
  368. put_byte(pb, vals[3]);
  369. }
  370. else if (size < 0x00200000)
  371. {
  372. put_byte(pb, vals[1]);
  373. put_byte(pb, vals[2]);
  374. put_byte(pb, vals[3]);
  375. }
  376. else if (size < 0x10000000)
  377. {
  378. put_byte(pb, vals[0]);
  379. put_byte(pb, vals[1]);
  380. put_byte(pb, vals[2]);
  381. put_byte(pb, vals[3]);
  382. }
  383. }
  384. static int mov_write_esds_tag(ByteIOContext *pb, MOVTrack* track) // Basic
  385. {
  386. int decoderSpecificInfoLen;
  387. offset_t pos = url_ftell(pb);
  388. decoderSpecificInfoLen = track->vosLen ? descrLength(track->vosLen):0;
  389. put_be32(pb, 0); // size
  390. put_tag(pb, "esds");
  391. put_be32(pb, 0); // Version
  392. // ES descriptor
  393. putDescr(pb, 0x03, 3 + descrLength(13 + decoderSpecificInfoLen) +
  394. descrLength(1));
  395. put_be16(pb, track->trackID);
  396. put_byte(pb, 0x00); // flags (= no flags)
  397. // DecoderConfig descriptor
  398. putDescr(pb, 0x04, 13 + decoderSpecificInfoLen);
  399. // Object type indication
  400. put_byte(pb, codec_get_tag(ff_mov_obj_type, track->enc->codec_id));
  401. // the following fields is made of 6 bits to identify the streamtype (4 for video, 5 for audio)
  402. // plus 1 bit to indicate upstream and 1 bit set to 1 (reserved)
  403. if(track->enc->codec_type == CODEC_TYPE_AUDIO)
  404. put_byte(pb, 0x15); // flags (= Audiostream)
  405. else
  406. put_byte(pb, 0x11); // flags (= Visualstream)
  407. put_byte(pb, track->enc->rc_buffer_size>>(3+16)); // Buffersize DB (24 bits)
  408. put_be16(pb, (track->enc->rc_buffer_size>>3)&0xFFFF); // Buffersize DB
  409. put_be32(pb, FFMAX(track->enc->bit_rate, track->enc->rc_max_rate)); // maxbitrate (FIXME should be max rate in any 1 sec window)
  410. if(track->enc->rc_max_rate != track->enc->rc_min_rate || track->enc->rc_min_rate==0)
  411. put_be32(pb, 0); // vbr
  412. else
  413. put_be32(pb, track->enc->rc_max_rate); // avg bitrate
  414. if (track->vosLen)
  415. {
  416. // DecoderSpecific info descriptor
  417. putDescr(pb, 0x05, track->vosLen);
  418. put_buffer(pb, track->vosData, track->vosLen);
  419. }
  420. // SL descriptor
  421. putDescr(pb, 0x06, 1);
  422. put_byte(pb, 0x02);
  423. return updateSize (pb, pos);
  424. }
  425. static const CodecTag codec_movvideo_tags[] = {
  426. { CODEC_ID_SVQ1, MKTAG('S', 'V', 'Q', '1') },
  427. { CODEC_ID_SVQ3, MKTAG('S', 'V', 'Q', '3') },
  428. { CODEC_ID_MPEG4, MKTAG('m', 'p', '4', 'v') },
  429. { CODEC_ID_H263, MKTAG('s', '2', '6', '3') },
  430. { CODEC_ID_H264, MKTAG('a', 'v', 'c', '1') },
  431. { CODEC_ID_DVVIDEO, MKTAG('d', 'v', 'c', ' ') },
  432. { 0, 0 },
  433. };
  434. static int mov_write_video_tag(ByteIOContext *pb, MOVTrack* track)
  435. {
  436. offset_t pos = url_ftell(pb);
  437. char compressor_name[32];
  438. int tag;
  439. put_be32(pb, 0); /* size */
  440. tag = track->enc->codec_tag;
  441. if (!tag)
  442. tag = codec_get_tag(codec_movvideo_tags, track->enc->codec_id);
  443. // if no mac fcc found, try with Microsoft tags
  444. if (!tag)
  445. tag = codec_get_tag(codec_bmp_tags, track->enc->codec_id);
  446. put_le32(pb, tag); // store it byteswapped
  447. put_be32(pb, 0); /* Reserved */
  448. put_be16(pb, 0); /* Reserved */
  449. put_be16(pb, 1); /* Data-reference index */
  450. put_be16(pb, 0); /* Codec stream version */
  451. put_be16(pb, 0); /* Codec stream revision (=0) */
  452. put_tag(pb, "FFMP"); /* Vendor */
  453. if(track->enc->codec_id == CODEC_ID_RAWVIDEO) {
  454. put_be32(pb, 0); /* Temporal Quality */
  455. put_be32(pb, 0x400); /* Spatial Quality = lossless*/
  456. } else {
  457. put_be32(pb, 0x200); /* Temporal Quality = normal */
  458. put_be32(pb, 0x200); /* Spatial Quality = normal */
  459. }
  460. put_be16(pb, track->enc->width); /* Video width */
  461. put_be16(pb, track->enc->height); /* Video height */
  462. put_be32(pb, 0x00480000); /* Horizontal resolution 72dpi */
  463. put_be32(pb, 0x00480000); /* Vertical resolution 72dpi */
  464. put_be32(pb, 0); /* Data size (= 0) */
  465. put_be16(pb, 1); /* Frame count (= 1) */
  466. memset(compressor_name,0,32);
  467. if (track->enc->codec && track->enc->codec->name)
  468. strncpy(compressor_name,track->enc->codec->name,31);
  469. put_byte(pb, strlen(compressor_name));
  470. put_buffer(pb, compressor_name, 31);
  471. put_be16(pb, 0x18); /* Reserved */
  472. put_be16(pb, 0xffff); /* Reserved */
  473. if(track->enc->codec_id == CODEC_ID_MPEG4)
  474. mov_write_esds_tag(pb, track);
  475. else if(track->enc->codec_id == CODEC_ID_H263)
  476. mov_write_d263_tag(pb);
  477. else if(track->enc->codec_id == CODEC_ID_SVQ3)
  478. mov_write_svq3_tag(pb);
  479. return updateSize (pb, pos);
  480. }
  481. static int mov_write_stsd_tag(ByteIOContext *pb, MOVTrack* track)
  482. {
  483. offset_t pos = url_ftell(pb);
  484. put_be32(pb, 0); /* size */
  485. put_tag(pb, "stsd");
  486. put_be32(pb, 0); /* version & flags */
  487. put_be32(pb, 1); /* entry count */
  488. if (track->enc->codec_type == CODEC_TYPE_VIDEO)
  489. mov_write_video_tag(pb, track);
  490. else if (track->enc->codec_type == CODEC_TYPE_AUDIO)
  491. mov_write_audio_tag(pb, track);
  492. return updateSize(pb, pos);
  493. }
  494. /* TODO: */
  495. /* Time to sample atom */
  496. static int mov_write_stts_tag(ByteIOContext *pb, MOVTrack* track)
  497. {
  498. put_be32(pb, 0x18); /* size */
  499. put_tag(pb, "stts");
  500. put_be32(pb, 0); /* version & flags */
  501. put_be32(pb, 1); /* entry count */
  502. put_be32(pb, track->sampleCount); /* sample count */
  503. put_be32(pb, track->sampleDuration); /* sample duration */
  504. return 0x18;
  505. }
  506. static int mov_write_dref_tag(ByteIOContext *pb)
  507. {
  508. put_be32(pb, 28); /* size */
  509. put_tag(pb, "dref");
  510. put_be32(pb, 0); /* version & flags */
  511. put_be32(pb, 1); /* entry count */
  512. put_be32(pb, 0xc); /* size */
  513. put_tag(pb, "url ");
  514. put_be32(pb, 1); /* version & flags */
  515. return 28;
  516. }
  517. static int mov_write_stbl_tag(ByteIOContext *pb, MOVTrack* track)
  518. {
  519. offset_t pos = url_ftell(pb);
  520. put_be32(pb, 0); /* size */
  521. put_tag(pb, "stbl");
  522. mov_write_stsd_tag(pb, track);
  523. mov_write_stts_tag(pb, track);
  524. if (track->enc->codec_type == CODEC_TYPE_VIDEO &&
  525. track->hasKeyframes)
  526. mov_write_stss_tag(pb, track);
  527. mov_write_stsc_tag(pb, track);
  528. mov_write_stsz_tag(pb, track);
  529. mov_write_stco_tag(pb, track);
  530. return updateSize(pb, pos);
  531. }
  532. static int mov_write_dinf_tag(ByteIOContext *pb)
  533. {
  534. offset_t pos = url_ftell(pb);
  535. put_be32(pb, 0); /* size */
  536. put_tag(pb, "dinf");
  537. mov_write_dref_tag(pb);
  538. return updateSize(pb, pos);
  539. }
  540. static int mov_write_smhd_tag(ByteIOContext *pb)
  541. {
  542. put_be32(pb, 16); /* size */
  543. put_tag(pb, "smhd");
  544. put_be32(pb, 0); /* version & flags */
  545. put_be16(pb, 0); /* reserved (balance, normally = 0) */
  546. put_be16(pb, 0); /* reserved */
  547. return 16;
  548. }
  549. static int mov_write_vmhd_tag(ByteIOContext *pb)
  550. {
  551. put_be32(pb, 0x14); /* size (always 0x14) */
  552. put_tag(pb, "vmhd");
  553. put_be32(pb, 0x01); /* version & flags */
  554. put_be64(pb, 0); /* reserved (graphics mode = copy) */
  555. return 0x14;
  556. }
  557. static int mov_write_hdlr_tag(ByteIOContext *pb, MOVTrack* track)
  558. {
  559. char *descr, *hdlr, *hdlr_type;
  560. offset_t pos = url_ftell(pb);
  561. if (!track) { /* no media --> data handler */
  562. hdlr = "dhlr";
  563. hdlr_type = "url ";
  564. descr = "DataHandler";
  565. } else {
  566. hdlr = (track->mode == MODE_MOV) ? "mhlr" : "\0\0\0\0";
  567. if (track->enc->codec_type == CODEC_TYPE_VIDEO) {
  568. hdlr_type = "vide";
  569. descr = "VideoHandler";
  570. } else {
  571. hdlr_type = "soun";
  572. descr = "SoundHandler";
  573. }
  574. }
  575. put_be32(pb, 0); /* size */
  576. put_tag(pb, "hdlr");
  577. put_be32(pb, 0); /* Version & flags */
  578. put_buffer(pb, hdlr, 4); /* handler */
  579. put_tag(pb, hdlr_type); /* handler type */
  580. put_be32(pb ,0); /* reserved */
  581. put_be32(pb ,0); /* reserved */
  582. put_be32(pb ,0); /* reserved */
  583. put_byte(pb, strlen(descr)); /* string counter */
  584. put_buffer(pb, descr, strlen(descr)); /* handler description */
  585. return updateSize(pb, pos);
  586. }
  587. static int mov_write_minf_tag(ByteIOContext *pb, MOVTrack* track)
  588. {
  589. offset_t pos = url_ftell(pb);
  590. put_be32(pb, 0); /* size */
  591. put_tag(pb, "minf");
  592. if(track->enc->codec_type == CODEC_TYPE_VIDEO)
  593. mov_write_vmhd_tag(pb);
  594. else
  595. mov_write_smhd_tag(pb);
  596. if (track->mode == MODE_MOV) /* FIXME: Why do it for MODE_MOV only ? */
  597. mov_write_hdlr_tag(pb, NULL);
  598. mov_write_dinf_tag(pb);
  599. mov_write_stbl_tag(pb, track);
  600. return updateSize(pb, pos);
  601. }
  602. static int mov_write_mdhd_tag(ByteIOContext *pb, MOVTrack* track)
  603. {
  604. put_be32(pb, 32); /* size */
  605. put_tag(pb, "mdhd");
  606. put_be32(pb, 0); /* Version & flags */
  607. put_be32(pb, track->time); /* creation time */
  608. put_be32(pb, track->time); /* modification time */
  609. put_be32(pb, track->timescale); /* time scale (sample rate for audio) */
  610. put_be32(pb, track->trackDuration); /* duration */
  611. put_be16(pb, track->language); /* language */
  612. put_be16(pb, 0); /* reserved (quality) */
  613. return 32;
  614. }
  615. static int mov_write_mdia_tag(ByteIOContext *pb, MOVTrack* track)
  616. {
  617. offset_t pos = url_ftell(pb);
  618. put_be32(pb, 0); /* size */
  619. put_tag(pb, "mdia");
  620. mov_write_mdhd_tag(pb, track);
  621. mov_write_hdlr_tag(pb, track);
  622. mov_write_minf_tag(pb, track);
  623. return updateSize(pb, pos);
  624. }
  625. static int mov_write_tkhd_tag(ByteIOContext *pb, MOVTrack* track)
  626. {
  627. put_be32(pb, 0x5c); /* size (always 0x5c) */
  628. put_tag(pb, "tkhd");
  629. put_be32(pb, 0xf); /* version & flags (track enabled) */
  630. put_be32(pb, track->time); /* creation time */
  631. put_be32(pb, track->time); /* modification time */
  632. put_be32(pb, track->trackID); /* track-id */
  633. put_be32(pb, 0); /* reserved */
  634. put_be32(pb, av_rescale_rnd(track->trackDuration, globalTimescale, track->timescale, AV_ROUND_UP)); /* duration */
  635. put_be32(pb, 0); /* reserved */
  636. put_be32(pb, 0); /* reserved */
  637. put_be32(pb, 0x0); /* reserved (Layer & Alternate group) */
  638. /* Volume, only for audio */
  639. if(track->enc->codec_type == CODEC_TYPE_AUDIO)
  640. put_be16(pb, 0x0100);
  641. else
  642. put_be16(pb, 0);
  643. put_be16(pb, 0); /* reserved */
  644. /* Matrix structure */
  645. put_be32(pb, 0x00010000); /* reserved */
  646. put_be32(pb, 0x0); /* reserved */
  647. put_be32(pb, 0x0); /* reserved */
  648. put_be32(pb, 0x0); /* reserved */
  649. put_be32(pb, 0x00010000); /* reserved */
  650. put_be32(pb, 0x0); /* reserved */
  651. put_be32(pb, 0x0); /* reserved */
  652. put_be32(pb, 0x0); /* reserved */
  653. put_be32(pb, 0x40000000); /* reserved */
  654. /* Track width and height, for visual only */
  655. if(track->enc->codec_type == CODEC_TYPE_VIDEO) {
  656. double sample_aspect_ratio = av_q2d(track->enc->sample_aspect_ratio);
  657. if( !sample_aspect_ratio ) sample_aspect_ratio = 1;
  658. put_be32(pb, sample_aspect_ratio * track->enc->width*0x10000);
  659. put_be32(pb, track->enc->height*0x10000);
  660. }
  661. else {
  662. put_be32(pb, 0);
  663. put_be32(pb, 0);
  664. }
  665. return 0x5c;
  666. }
  667. // This box seems important for the psp playback ... without it the movie seems to hang
  668. static int mov_write_edts_tag(ByteIOContext *pb, MOVTrack *track)
  669. {
  670. put_be32(pb, 0x24); /* size */
  671. put_tag(pb, "edts");
  672. put_be32(pb, 0x1c); /* size */
  673. put_tag(pb, "elst");
  674. put_be32(pb, 0x0);
  675. put_be32(pb, 0x1);
  676. put_be32(pb, av_rescale_rnd(track->trackDuration, globalTimescale, track->timescale, AV_ROUND_UP)); /* duration ... doesn't seem to effect psp */
  677. put_be32(pb, 0x0);
  678. put_be32(pb, 0x00010000);
  679. return 0x24;
  680. }
  681. // goes at the end of each track! ... Critical for PSP playback ("Incompatible data" without it)
  682. static int mov_write_uuid_tag_psp(ByteIOContext *pb, MOVTrack *mov)
  683. {
  684. put_be32(pb, 0x34); /* size ... reports as 28 in mp4box! */
  685. put_tag(pb, "uuid");
  686. put_tag(pb, "USMT");
  687. put_be32(pb, 0x21d24fce);
  688. put_be32(pb, 0xbb88695c);
  689. put_be32(pb, 0xfac9c740);
  690. put_be32(pb, 0x1c); // another size here!
  691. put_tag(pb, "MTDT");
  692. put_be32(pb, 0x00010012);
  693. put_be32(pb, 0x0a);
  694. put_be32(pb, 0x55c40000);
  695. put_be32(pb, 0x1);
  696. put_be32(pb, 0x0);
  697. return 0x34;
  698. }
  699. static int mov_write_trak_tag(ByteIOContext *pb, MOVTrack* track)
  700. {
  701. offset_t pos = url_ftell(pb);
  702. put_be32(pb, 0); /* size */
  703. put_tag(pb, "trak");
  704. mov_write_tkhd_tag(pb, track);
  705. if (track->mode == MODE_PSP)
  706. mov_write_edts_tag(pb, track); // PSP Movies require edts box
  707. mov_write_mdia_tag(pb, track);
  708. if (track->mode == MODE_PSP)
  709. mov_write_uuid_tag_psp(pb,track); // PSP Movies require this uuid box
  710. return updateSize(pb, pos);
  711. }
  712. #if 0
  713. /* TODO: Not sorted out, but not necessary either */
  714. static int mov_write_iods_tag(ByteIOContext *pb, MOVContext *mov)
  715. {
  716. put_be32(pb, 0x15); /* size */
  717. put_tag(pb, "iods");
  718. put_be32(pb, 0); /* version & flags */
  719. put_be16(pb, 0x1007);
  720. put_byte(pb, 0);
  721. put_be16(pb, 0x4fff);
  722. put_be16(pb, 0xfffe);
  723. put_be16(pb, 0x01ff);
  724. return 0x15;
  725. }
  726. #endif
  727. static int mov_write_mvhd_tag(ByteIOContext *pb, MOVContext *mov)
  728. {
  729. int maxTrackID = 1, i;
  730. int64_t maxTrackLenTemp, maxTrackLen = 0;
  731. put_be32(pb, 0x6c); /* size (always 0x6c) */
  732. put_tag(pb, "mvhd");
  733. put_be32(pb, 0); /* version & flags */
  734. put_be32(pb, mov->time); /* creation time */
  735. put_be32(pb, mov->time); /* modification time */
  736. put_be32(pb, mov->timescale); /* timescale */
  737. for (i=0; i<MAX_STREAMS; i++) {
  738. if(mov->tracks[i].entry > 0) {
  739. maxTrackLenTemp = av_rescale_rnd(mov->tracks[i].trackDuration, globalTimescale, mov->tracks[i].timescale, AV_ROUND_UP);
  740. if(maxTrackLen < maxTrackLenTemp)
  741. maxTrackLen = maxTrackLenTemp;
  742. if(maxTrackID < mov->tracks[i].trackID)
  743. maxTrackID = mov->tracks[i].trackID;
  744. }
  745. }
  746. put_be32(pb, maxTrackLen); /* duration of longest track */
  747. put_be32(pb, 0x00010000); /* reserved (preferred rate) 1.0 = normal */
  748. put_be16(pb, 0x0100); /* reserved (preferred volume) 1.0 = normal */
  749. put_be16(pb, 0); /* reserved */
  750. put_be32(pb, 0); /* reserved */
  751. put_be32(pb, 0); /* reserved */
  752. /* Matrix structure */
  753. put_be32(pb, 0x00010000); /* reserved */
  754. put_be32(pb, 0x0); /* reserved */
  755. put_be32(pb, 0x0); /* reserved */
  756. put_be32(pb, 0x0); /* reserved */
  757. put_be32(pb, 0x00010000); /* reserved */
  758. put_be32(pb, 0x0); /* reserved */
  759. put_be32(pb, 0x0); /* reserved */
  760. put_be32(pb, 0x0); /* reserved */
  761. put_be32(pb, 0x40000000); /* reserved */
  762. put_be32(pb, 0); /* reserved (preview time) */
  763. put_be32(pb, 0); /* reserved (preview duration) */
  764. put_be32(pb, 0); /* reserved (poster time) */
  765. put_be32(pb, 0); /* reserved (selection time) */
  766. put_be32(pb, 0); /* reserved (selection duration) */
  767. put_be32(pb, 0); /* reserved (current time) */
  768. put_be32(pb, maxTrackID+1); /* Next track id */
  769. return 0x6c;
  770. }
  771. static int mov_write_itunes_hdlr_tag(ByteIOContext *pb, MOVContext* mov,
  772. AVFormatContext *s)
  773. {
  774. offset_t pos = url_ftell(pb);
  775. put_be32(pb, 0); /* size */
  776. put_tag(pb, "hdlr");
  777. put_be32(pb, 0);
  778. put_be32(pb, 0);
  779. put_tag(pb, "mdir");
  780. put_tag(pb, "appl");
  781. put_be32(pb, 0);
  782. put_be32(pb, 0);
  783. put_be16(pb, 0);
  784. return updateSize(pb, pos);
  785. }
  786. /* helper function to write a data tag with the specified string as data */
  787. static int mov_write_string_data_tag(ByteIOContext *pb, MOVContext* mov,
  788. AVFormatContext *s, const char *data)
  789. {
  790. offset_t pos = url_ftell(pb);
  791. put_be32(pb, 0); /* size */
  792. put_tag(pb, "data");
  793. put_be32(pb, 1);
  794. put_be32(pb, 0);
  795. put_buffer(pb, data, strlen(data));
  796. return updateSize(pb, pos);
  797. }
  798. /* iTunes name of the song/movie */
  799. static int mov_write_nam_tag(ByteIOContext *pb, MOVContext* mov,
  800. AVFormatContext *s)
  801. {
  802. int size = 0;
  803. if ( s->title[0] ) {
  804. offset_t pos = url_ftell(pb);
  805. put_be32(pb, 0); /* size */
  806. put_tag(pb, "\251nam");
  807. mov_write_string_data_tag(pb, mov, s, s->title);
  808. size = updateSize(pb, pos);
  809. }
  810. return size;
  811. }
  812. /* iTunes name of the artist/performer */
  813. static int mov_write_ART_tag(ByteIOContext *pb, MOVContext* mov,
  814. AVFormatContext *s)
  815. {
  816. int size = 0;
  817. if ( s->author[0] ) {
  818. offset_t pos = url_ftell(pb);
  819. put_be32(pb, 0); /* size */
  820. put_tag(pb, "\251ART");
  821. // we use the author here as this is the only thing that we have...
  822. mov_write_string_data_tag(pb, mov, s, s->author);
  823. size = updateSize(pb, pos);
  824. }
  825. return size;
  826. }
  827. /* iTunes name of the writer */
  828. static int mov_write_wrt_tag(ByteIOContext *pb, MOVContext* mov,
  829. AVFormatContext *s)
  830. {
  831. int size = 0;
  832. if ( s->author[0] ) {
  833. offset_t pos = url_ftell(pb);
  834. put_be32(pb, 0); /* size */
  835. put_tag(pb, "\251wrt");
  836. mov_write_string_data_tag(pb, mov, s, s->author);
  837. size = updateSize(pb, pos);
  838. }
  839. return size;
  840. }
  841. /* iTunes name of the album */
  842. static int mov_write_alb_tag(ByteIOContext *pb, MOVContext* mov,
  843. AVFormatContext *s)
  844. {
  845. int size = 0;
  846. if ( s->album[0] ) {
  847. offset_t pos = url_ftell(pb);
  848. put_be32(pb, 0); /* size */
  849. put_tag(pb, "\251alb");
  850. mov_write_string_data_tag(pb, mov, s, s->album);
  851. size = updateSize(pb, pos);
  852. }
  853. return size;
  854. }
  855. /* iTunes year */
  856. static int mov_write_day_tag(ByteIOContext *pb, MOVContext* mov,
  857. AVFormatContext *s)
  858. {
  859. char year[5];
  860. int size = 0;
  861. if ( s->year ) {
  862. offset_t pos = url_ftell(pb);
  863. put_be32(pb, 0); /* size */
  864. put_tag(pb, "\251day");
  865. snprintf(year, 5, "%04d", s->year);
  866. mov_write_string_data_tag(pb, mov, s, year);
  867. size = updateSize(pb, pos);
  868. }
  869. return size;
  870. }
  871. /* iTunes tool used to create the file */
  872. static int mov_write_too_tag(ByteIOContext *pb, MOVContext* mov,
  873. AVFormatContext *s)
  874. {
  875. offset_t pos = url_ftell(pb);
  876. put_be32(pb, 0); /* size */
  877. put_tag(pb, "\251too");
  878. mov_write_string_data_tag(pb, mov, s, LIBAVFORMAT_IDENT);
  879. return updateSize(pb, pos);
  880. }
  881. /* iTunes comment */
  882. static int mov_write_cmt_tag(ByteIOContext *pb, MOVContext* mov,
  883. AVFormatContext *s)
  884. {
  885. int size = 0;
  886. if ( s->comment[0] ) {
  887. offset_t pos = url_ftell(pb);
  888. put_be32(pb, 0); /* size */
  889. put_tag(pb, "\251cmt");
  890. mov_write_string_data_tag(pb, mov, s, s->comment);
  891. size = updateSize(pb, pos);
  892. }
  893. return size;
  894. }
  895. /* iTunes custom genre */
  896. static int mov_write_gen_tag(ByteIOContext *pb, MOVContext* mov,
  897. AVFormatContext *s)
  898. {
  899. int size = 0;
  900. if ( s->genre[0] ) {
  901. offset_t pos = url_ftell(pb);
  902. put_be32(pb, 0); /* size */
  903. put_tag(pb, "\251gen");
  904. mov_write_string_data_tag(pb, mov, s, s->genre);
  905. size = updateSize(pb, pos);
  906. }
  907. return size;
  908. }
  909. /* iTunes track number */
  910. static int mov_write_trkn_tag(ByteIOContext *pb, MOVContext* mov,
  911. AVFormatContext *s)
  912. {
  913. int size = 0;
  914. if ( s->track ) {
  915. offset_t pos = url_ftell(pb);
  916. put_be32(pb, 0); /* size */
  917. put_tag(pb, "trkn");
  918. {
  919. offset_t pos = url_ftell(pb);
  920. put_be32(pb, 0); /* size */
  921. put_tag(pb, "data");
  922. put_be32(pb, 0); // 8 bytes empty
  923. put_be32(pb, 0);
  924. put_be16(pb, 0); // empty
  925. put_be16(pb, s->track); // track number
  926. put_be16(pb, 0); // total track number
  927. put_be16(pb, 0); // empty
  928. updateSize(pb, pos);
  929. }
  930. size = updateSize(pb, pos);
  931. }
  932. return size;
  933. }
  934. /* iTunes meta data list */
  935. static int mov_write_ilst_tag(ByteIOContext *pb, MOVContext* mov,
  936. AVFormatContext *s)
  937. {
  938. offset_t pos = url_ftell(pb);
  939. put_be32(pb, 0); /* size */
  940. put_tag(pb, "ilst");
  941. mov_write_nam_tag(pb, mov, s);
  942. mov_write_ART_tag(pb, mov, s);
  943. mov_write_wrt_tag(pb, mov, s);
  944. mov_write_alb_tag(pb, mov, s);
  945. mov_write_day_tag(pb, mov, s);
  946. mov_write_too_tag(pb, mov, s);
  947. mov_write_cmt_tag(pb, mov, s);
  948. mov_write_gen_tag(pb, mov, s);
  949. mov_write_trkn_tag(pb, mov, s);
  950. return updateSize(pb, pos);
  951. }
  952. /* iTunes meta data tag */
  953. static int mov_write_meta_tag(ByteIOContext *pb, MOVContext* mov,
  954. AVFormatContext *s)
  955. {
  956. int size = 0;
  957. // only save meta tag if required
  958. if ( s->title[0] || s->author[0] || s->album[0] || s->year ||
  959. s->comment[0] || s->genre[0] || s->track ) {
  960. offset_t pos = url_ftell(pb);
  961. put_be32(pb, 0); /* size */
  962. put_tag(pb, "meta");
  963. put_be32(pb, 0);
  964. mov_write_itunes_hdlr_tag(pb, mov, s);
  965. mov_write_ilst_tag(pb, mov, s);
  966. size = updateSize(pb, pos);
  967. }
  968. return size;
  969. }
  970. static int mov_write_udta_tag(ByteIOContext *pb, MOVContext* mov,
  971. AVFormatContext *s)
  972. {
  973. offset_t pos = url_ftell(pb);
  974. int i;
  975. put_be32(pb, 0); /* size */
  976. put_tag(pb, "udta");
  977. /* iTunes meta data */
  978. mov_write_meta_tag(pb, mov, s);
  979. /* Requirements */
  980. for (i=0; i<MAX_STREAMS; i++) {
  981. if(mov->tracks[i].entry <= 0) continue;
  982. if (mov->tracks[i].enc->codec_id == CODEC_ID_AAC ||
  983. mov->tracks[i].enc->codec_id == CODEC_ID_MPEG4) {
  984. offset_t pos = url_ftell(pb);
  985. put_be32(pb, 0); /* size */
  986. put_tag(pb, "\251req");
  987. put_be16(pb, sizeof("QuickTime 6.0 or greater") - 1);
  988. put_be16(pb, 0);
  989. put_buffer(pb, "QuickTime 6.0 or greater",
  990. sizeof("QuickTime 6.0 or greater") - 1);
  991. updateSize(pb, pos);
  992. break;
  993. }
  994. }
  995. /* Encoder */
  996. if(mov->tracks[0].enc && !(mov->tracks[0].enc->flags & CODEC_FLAG_BITEXACT))
  997. {
  998. offset_t pos = url_ftell(pb);
  999. put_be32(pb, 0); /* size */
  1000. put_tag(pb, "\251enc");
  1001. put_be16(pb, sizeof(LIBAVFORMAT_IDENT) - 1); /* string length */
  1002. put_be16(pb, 0);
  1003. put_buffer(pb, LIBAVFORMAT_IDENT, sizeof(LIBAVFORMAT_IDENT) - 1);
  1004. updateSize(pb, pos);
  1005. }
  1006. if( s->title[0] )
  1007. {
  1008. offset_t pos = url_ftell(pb);
  1009. put_be32(pb, 0); /* size */
  1010. put_tag(pb, "\251nam");
  1011. put_be16(pb, strlen(s->title)); /* string length */
  1012. put_be16(pb, 0);
  1013. put_buffer(pb, s->title, strlen(s->title));
  1014. updateSize(pb, pos);
  1015. }
  1016. if( s->author[0] )
  1017. {
  1018. offset_t pos = url_ftell(pb);
  1019. put_be32(pb, 0); /* size */
  1020. put_tag(pb, /*"\251aut"*/ "\251day" );
  1021. put_be16(pb, strlen(s->author)); /* string length */
  1022. put_be16(pb, 0);
  1023. put_buffer(pb, s->author, strlen(s->author));
  1024. updateSize(pb, pos);
  1025. }
  1026. if( s->comment[0] )
  1027. {
  1028. offset_t pos = url_ftell(pb);
  1029. put_be32(pb, 0); /* size */
  1030. put_tag(pb, "\251des");
  1031. put_be16(pb, strlen(s->comment)); /* string length */
  1032. put_be16(pb, 0);
  1033. put_buffer(pb, s->comment, strlen(s->comment));
  1034. updateSize(pb, pos);
  1035. }
  1036. return updateSize(pb, pos);
  1037. }
  1038. static size_t ascii_to_wc (ByteIOContext *pb, char *b, size_t n)
  1039. {
  1040. size_t i;
  1041. unsigned char c;
  1042. for (i = 0; i < n - 1; i++) {
  1043. c = b[i];
  1044. if (! (0x20 <= c && c <= 0x7f ))
  1045. c = 0x3f; /* '?' */
  1046. put_be16(pb, c);
  1047. }
  1048. put_be16(pb, 0x00);
  1049. return 2*n;
  1050. }
  1051. static uint16_t language_code (char *str)
  1052. {
  1053. return ((((str[0]-'a') & 0x1F)<<10) + (((str[1]-'a') & 0x1F)<<5) + ((str[2]-'a') & 0x1F));
  1054. }
  1055. static int mov_write_uuidusmt_tag (ByteIOContext *pb, AVFormatContext *s)
  1056. {
  1057. size_t len, size;
  1058. offset_t pos, curpos;
  1059. size = 0;
  1060. if (s->title[0]) {
  1061. pos = url_ftell(pb);
  1062. put_be32(pb, 0); /* size placeholder*/
  1063. put_tag(pb, "uuid");
  1064. put_tag(pb, "USMT");
  1065. put_be32(pb, 0x21d24fce ); /* 96 bit UUID */
  1066. put_be32(pb, 0xbb88695c );
  1067. put_be32(pb, 0xfac9c740 );
  1068. size += 24;
  1069. put_be32(pb, 0); /* size placeholder*/
  1070. put_tag(pb, "MTDT");
  1071. put_be16(pb, 1);
  1072. size += 10;
  1073. // Title
  1074. len = strlen(s->title)+1;
  1075. put_be16(pb, len*2+10); /* size */
  1076. put_be32(pb, 0x01); /* type */
  1077. put_be16(pb, language_code("und")); /* language */
  1078. put_be16(pb, 0x01); /* ? */
  1079. ascii_to_wc (pb, s->title, len);
  1080. size += len*2+10;
  1081. // size
  1082. curpos = url_ftell(pb);
  1083. url_fseek(pb, pos, SEEK_SET);
  1084. put_be32(pb, size);
  1085. url_fseek(pb, pos+24, SEEK_SET);
  1086. put_be32(pb, size-24);
  1087. url_fseek(pb, curpos, SEEK_SET);
  1088. }
  1089. return size;
  1090. }
  1091. static int mov_write_moov_tag(ByteIOContext *pb, MOVContext *mov,
  1092. AVFormatContext *s)
  1093. {
  1094. int i;
  1095. offset_t pos = url_ftell(pb);
  1096. put_be32(pb, 0); /* size placeholder*/
  1097. put_tag(pb, "moov");
  1098. mov->timescale = globalTimescale;
  1099. for (i=0; i<MAX_STREAMS; i++) {
  1100. if(mov->tracks[i].entry <= 0) continue;
  1101. if(mov->tracks[i].enc->codec_type == CODEC_TYPE_VIDEO) {
  1102. mov->tracks[i].timescale = mov->tracks[i].enc->time_base.den;
  1103. mov->tracks[i].sampleDuration = mov->tracks[i].enc->time_base.num;
  1104. }
  1105. else if(mov->tracks[i].enc->codec_type == CODEC_TYPE_AUDIO) {
  1106. /* If AMR, track timescale = 8000, AMR_WB = 16000 */
  1107. if(mov->tracks[i].enc->codec_id == CODEC_ID_AMR_NB) {
  1108. mov->tracks[i].sampleDuration = 160; // Bytes per chunk
  1109. mov->tracks[i].timescale = 8000;
  1110. }
  1111. else {
  1112. mov->tracks[i].timescale = mov->tracks[i].enc->sample_rate;
  1113. mov->tracks[i].sampleDuration = mov->tracks[i].enc->frame_size;
  1114. }
  1115. }
  1116. mov->tracks[i].trackDuration =
  1117. mov->tracks[i].sampleCount * mov->tracks[i].sampleDuration;
  1118. mov->tracks[i].time = mov->time;
  1119. mov->tracks[i].trackID = i+1;
  1120. }
  1121. mov_write_mvhd_tag(pb, mov);
  1122. //mov_write_iods_tag(pb, mov);
  1123. for (i=0; i<MAX_STREAMS; i++) {
  1124. if(mov->tracks[i].entry > 0) {
  1125. mov_write_trak_tag(pb, &(mov->tracks[i]));
  1126. }
  1127. }
  1128. if (mov->mode == MODE_PSP)
  1129. mov_write_uuidusmt_tag(pb, s);
  1130. else
  1131. mov_write_udta_tag(pb, mov, s);
  1132. return updateSize(pb, pos);
  1133. }
  1134. int mov_write_mdat_tag(ByteIOContext *pb, MOVContext* mov)
  1135. {
  1136. put_be32(pb, 8); // placeholder for extended size field (64 bit)
  1137. put_tag(pb, "wide");
  1138. mov->mdat_pos = url_ftell(pb);
  1139. put_be32(pb, 0); /* size placeholder*/
  1140. put_tag(pb, "mdat");
  1141. return 0;
  1142. }
  1143. /* TODO: This needs to be more general */
  1144. static void mov_write_ftyp_tag (ByteIOContext *pb, AVFormatContext *s)
  1145. {
  1146. MOVContext *mov = s->priv_data;
  1147. put_be32(pb, 0x14 ); /* size */
  1148. put_tag(pb, "ftyp");
  1149. if ( mov->mode == MODE_3GP )
  1150. put_tag(pb, "3gp4");
  1151. else if ( mov->mode == MODE_3G2 )
  1152. put_tag(pb, "3g2a");
  1153. else if ( mov->mode == MODE_PSP )
  1154. put_tag(pb, "MSNV");
  1155. else
  1156. put_tag(pb, "isom");
  1157. put_be32(pb, 0x200 );
  1158. if ( mov->mode == MODE_3GP )
  1159. put_tag(pb, "3gp4");
  1160. else if ( mov->mode == MODE_3G2 )
  1161. put_tag(pb, "3g2a");
  1162. else if ( mov->mode == MODE_PSP )
  1163. put_tag(pb, "MSNV");
  1164. else
  1165. put_tag(pb, "mp41");
  1166. }
  1167. static void mov_write_uuidprof_tag(ByteIOContext *pb, AVFormatContext *s)
  1168. {
  1169. AVCodecContext *VideoCodec = s->streams[0]->codec;
  1170. AVCodecContext *AudioCodec = s->streams[1]->codec;
  1171. int AudioRate = AudioCodec->sample_rate;
  1172. int FrameRate = ((VideoCodec->time_base.den) * (0x10000))/ (VideoCodec->time_base.num);
  1173. put_be32(pb, 0x94 ); /* size */
  1174. put_tag(pb, "uuid");
  1175. put_tag(pb, "PROF");
  1176. put_be32(pb, 0x21d24fce ); /* 96 bit UUID */
  1177. put_be32(pb, 0xbb88695c );
  1178. put_be32(pb, 0xfac9c740 );
  1179. put_be32(pb, 0x0 ); /* ? */
  1180. put_be32(pb, 0x3 ); /* 3 sections ? */
  1181. put_be32(pb, 0x14 ); /* size */
  1182. put_tag(pb, "FPRF");
  1183. put_be32(pb, 0x0 ); /* ? */
  1184. put_be32(pb, 0x0 ); /* ? */
  1185. put_be32(pb, 0x0 ); /* ? */
  1186. put_be32(pb, 0x2c ); /* size */
  1187. put_tag(pb, "APRF"); /* audio */
  1188. put_be32(pb, 0x0 );
  1189. put_be32(pb, 0x2 ); /* TrackID */
  1190. put_tag(pb, "mp4a");
  1191. put_be32(pb, 0x20f );
  1192. put_be32(pb, 0x0 );
  1193. put_be32(pb, AudioCodec->bit_rate / 1000);
  1194. put_be32(pb, AudioCodec->bit_rate / 1000);
  1195. put_be32(pb, AudioRate );
  1196. put_be32(pb, AudioCodec->channels );
  1197. put_be32(pb, 0x34 ); /* size */
  1198. put_tag(pb, "VPRF"); /* video */
  1199. put_be32(pb, 0x0 );
  1200. put_be32(pb, 0x1 ); /* TrackID */
  1201. put_tag(pb, "mp4v");
  1202. put_be32(pb, 0x103 );
  1203. put_be32(pb, 0x0 );
  1204. put_be32(pb, VideoCodec->bit_rate / 1000);
  1205. put_be32(pb, VideoCodec->bit_rate / 1000);
  1206. put_be32(pb, FrameRate);
  1207. put_be32(pb, FrameRate);
  1208. put_be16(pb, VideoCodec->width);
  1209. put_be16(pb, VideoCodec->height);
  1210. put_be32(pb, 0x010001); /* ? */
  1211. }
  1212. static int mov_write_header(AVFormatContext *s)
  1213. {
  1214. ByteIOContext *pb = &s->pb;
  1215. MOVContext *mov = s->priv_data;
  1216. int i;
  1217. for(i=0; i<s->nb_streams; i++){
  1218. AVCodecContext *c= s->streams[i]->codec;
  1219. if (c->codec_type == CODEC_TYPE_VIDEO){
  1220. if (!codec_get_tag(codec_movvideo_tags, c->codec_id)){
  1221. if(!codec_get_tag(codec_bmp_tags, c->codec_id))
  1222. return -1;
  1223. else
  1224. av_log(s, AV_LOG_INFO, "Warning, using MS style video codec tag, the file may be unplayable!\n");
  1225. }
  1226. }else if(c->codec_type == CODEC_TYPE_AUDIO){
  1227. if (!codec_get_tag(codec_movaudio_tags, c->codec_id)){
  1228. if(!codec_get_tag(codec_wav_tags, c->codec_id))
  1229. return -1;
  1230. else
  1231. av_log(s, AV_LOG_INFO, "Warning, using MS style audio codec tag, the file may be unplayable!\n");
  1232. }
  1233. }
  1234. /* don't know yet if mp4 or not */
  1235. mov->tracks[i].language = ff_mov_iso639_to_lang(s->streams[i]->language, 1);
  1236. }
  1237. /* Default mode == MP4 */
  1238. mov->mode = MODE_MP4;
  1239. if (s->oformat != NULL) {
  1240. if (!strcmp("3gp", s->oformat->name)) mov->mode = MODE_3GP;
  1241. else if (!strcmp("3g2", s->oformat->name)) mov->mode = MODE_3G2;
  1242. else if (!strcmp("mov", s->oformat->name)) mov->mode = MODE_MOV;
  1243. else if (!strcmp("psp", s->oformat->name)) mov->mode = MODE_PSP;
  1244. if ( mov->mode == MODE_3GP || mov->mode == MODE_3G2 ||
  1245. mov->mode == MODE_MP4 || mov->mode == MODE_PSP )
  1246. mov_write_ftyp_tag(pb,s);
  1247. if ( mov->mode == MODE_PSP ) {
  1248. if ( s->nb_streams != 2 ) {
  1249. av_log(s, AV_LOG_ERROR, "PSP mode need one video and one audio stream\n");
  1250. return -1;
  1251. }
  1252. mov_write_uuidprof_tag(pb,s);
  1253. }
  1254. }
  1255. for (i=0; i<MAX_STREAMS; i++) {
  1256. mov->tracks[i].mode = mov->mode;
  1257. }
  1258. put_flush_packet(pb);
  1259. return 0;
  1260. }
  1261. static int mov_write_packet(AVFormatContext *s, AVPacket *pkt)
  1262. {
  1263. MOVContext *mov = s->priv_data;
  1264. ByteIOContext *pb = &s->pb;
  1265. AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
  1266. MOVTrack* trk = &mov->tracks[pkt->stream_index];
  1267. int cl, id;
  1268. unsigned int samplesInChunk = 0;
  1269. int size= pkt->size;
  1270. if (url_is_streamed(&s->pb)) return 0; /* Can't handle that */
  1271. if (!size) return 0; /* Discard 0 sized packets */
  1272. if (enc->codec_type == CODEC_TYPE_VIDEO ) {
  1273. samplesInChunk = 1;
  1274. }
  1275. else if (enc->codec_type == CODEC_TYPE_AUDIO ) {
  1276. if( enc->codec_id == CODEC_ID_AMR_NB) {
  1277. /* We must find out how many AMR blocks there are in one packet */
  1278. static uint16_t packed_size[16] =
  1279. {13, 14, 16, 18, 20, 21, 27, 32, 6, 0, 0, 0, 0, 0, 0, 0};
  1280. int len = 0;
  1281. while (len < size && samplesInChunk < 100) {
  1282. len += packed_size[(pkt->data[len] >> 3) & 0x0F];
  1283. samplesInChunk++;
  1284. }
  1285. }
  1286. else if(enc->codec_id == CODEC_ID_PCM_ALAW) {
  1287. samplesInChunk = size/enc->channels;
  1288. }
  1289. else if(enc->codec_id == CODEC_ID_PCM_S16BE || enc->codec_id == CODEC_ID_PCM_S16LE) {
  1290. samplesInChunk = size/(2*enc->channels);
  1291. }
  1292. else {
  1293. samplesInChunk = 1;
  1294. }
  1295. }
  1296. if ((enc->codec_id == CODEC_ID_MPEG4 || enc->codec_id == CODEC_ID_AAC)
  1297. && trk->vosLen == 0) {
  1298. // assert(enc->extradata_size);
  1299. trk->vosLen = enc->extradata_size;
  1300. trk->vosData = av_malloc(trk->vosLen);
  1301. memcpy(trk->vosData, enc->extradata, trk->vosLen);
  1302. }
  1303. cl = trk->entry / MOV_INDEX_CLUSTER_SIZE;
  1304. id = trk->entry % MOV_INDEX_CLUSTER_SIZE;
  1305. if (trk->ents_allocated <= trk->entry) {
  1306. trk->cluster = av_realloc(trk->cluster, (cl+1)*sizeof(void*));
  1307. if (!trk->cluster)
  1308. return -1;
  1309. trk->cluster[cl] = av_malloc(MOV_INDEX_CLUSTER_SIZE*sizeof(MOVIentry));
  1310. if (!trk->cluster[cl])
  1311. return -1;
  1312. trk->ents_allocated += MOV_INDEX_CLUSTER_SIZE;
  1313. }
  1314. if (mov->mdat_written == 0) {
  1315. mov_write_mdat_tag(pb, mov);
  1316. mov->mdat_written = 1;
  1317. mov->time = s->timestamp + 0x7C25B080; //1970 based -> 1904 based
  1318. }
  1319. trk->cluster[cl][id].pos = url_ftell(pb);
  1320. trk->cluster[cl][id].samplesInChunk = samplesInChunk;
  1321. trk->cluster[cl][id].size = size;
  1322. trk->cluster[cl][id].entries = samplesInChunk;
  1323. if(enc->codec_type == CODEC_TYPE_VIDEO) {
  1324. trk->cluster[cl][id].key_frame = !!(pkt->flags & PKT_FLAG_KEY);
  1325. if(trk->cluster[cl][id].key_frame)
  1326. trk->hasKeyframes = 1;
  1327. }
  1328. trk->enc = enc;
  1329. trk->entry++;
  1330. trk->sampleCount += samplesInChunk;
  1331. trk->mdat_size += size;
  1332. put_buffer(pb, pkt->data, size);
  1333. put_flush_packet(pb);
  1334. return 0;
  1335. }
  1336. static int mov_write_trailer(AVFormatContext *s)
  1337. {
  1338. MOVContext *mov = s->priv_data;
  1339. ByteIOContext *pb = &s->pb;
  1340. int res = 0;
  1341. int i;
  1342. uint64_t j;
  1343. offset_t moov_pos = url_ftell(pb);
  1344. /* Write size of mdat tag */
  1345. for (i=0, j=0; i<MAX_STREAMS; i++) {
  1346. if(mov->tracks[i].ents_allocated > 0) {
  1347. j += mov->tracks[i].mdat_size;
  1348. }
  1349. }
  1350. if (j+8 <= UINT32_MAX) {
  1351. url_fseek(pb, mov->mdat_pos, SEEK_SET);
  1352. put_be32(pb, j+8);
  1353. } else {
  1354. /* overwrite 'wide' placeholder atom */
  1355. url_fseek(pb, mov->mdat_pos - 8, SEEK_SET);
  1356. put_be32(pb, 1); /* special value: real atom size will be 64 bit value after tag field */
  1357. put_tag(pb, "mdat");
  1358. put_be64(pb, j+16);
  1359. }
  1360. url_fseek(pb, moov_pos, SEEK_SET);
  1361. mov_write_moov_tag(pb, mov, s);
  1362. for (i=0; i<MAX_STREAMS; i++) {
  1363. for (j=0; j<mov->tracks[i].ents_allocated/MOV_INDEX_CLUSTER_SIZE; j++) {
  1364. av_free(mov->tracks[i].cluster[j]);
  1365. }
  1366. av_free(mov->tracks[i].cluster);
  1367. if( mov->tracks[i].vosLen ) av_free( mov->tracks[i].vosData );
  1368. mov->tracks[i].cluster = NULL;
  1369. mov->tracks[i].ents_allocated = mov->tracks[i].entry = 0;
  1370. }
  1371. put_flush_packet(pb);
  1372. return res;
  1373. }
  1374. static AVOutputFormat mov_oformat = {
  1375. "mov",
  1376. "mov format",
  1377. NULL,
  1378. "mov",
  1379. sizeof(MOVContext),
  1380. CODEC_ID_AAC,
  1381. CODEC_ID_MPEG4,
  1382. mov_write_header,
  1383. mov_write_packet,
  1384. mov_write_trailer,
  1385. .flags = AVFMT_GLOBALHEADER,
  1386. };
  1387. static AVOutputFormat _3gp_oformat = {
  1388. "3gp",
  1389. "3gp format",
  1390. NULL,
  1391. "3gp",
  1392. sizeof(MOVContext),
  1393. CODEC_ID_AMR_NB,
  1394. CODEC_ID_H263,
  1395. mov_write_header,
  1396. mov_write_packet,
  1397. mov_write_trailer,
  1398. .flags = AVFMT_GLOBALHEADER,
  1399. };
  1400. static AVOutputFormat mp4_oformat = {
  1401. "mp4",
  1402. "mp4 format",
  1403. "application/mp4",
  1404. "mp4,m4a",
  1405. sizeof(MOVContext),
  1406. CODEC_ID_AAC,
  1407. CODEC_ID_MPEG4,
  1408. mov_write_header,
  1409. mov_write_packet,
  1410. mov_write_trailer,
  1411. .flags = AVFMT_GLOBALHEADER,
  1412. };
  1413. static AVOutputFormat psp_oformat = {
  1414. "psp",
  1415. "psp mp4 format",
  1416. NULL,
  1417. "mp4,psp",
  1418. sizeof(MOVContext),
  1419. CODEC_ID_AAC,
  1420. CODEC_ID_MPEG4,
  1421. mov_write_header,
  1422. mov_write_packet,
  1423. mov_write_trailer,
  1424. .flags = AVFMT_GLOBALHEADER,
  1425. };
  1426. static AVOutputFormat _3g2_oformat = {
  1427. "3g2",
  1428. "3gp2 format",
  1429. NULL,
  1430. "3g2",
  1431. sizeof(MOVContext),
  1432. CODEC_ID_AMR_NB,
  1433. CODEC_ID_H263,
  1434. mov_write_header,
  1435. mov_write_packet,
  1436. mov_write_trailer,
  1437. .flags = AVFMT_GLOBALHEADER,
  1438. };
  1439. int movenc_init(void)
  1440. {
  1441. av_register_output_format(&mov_oformat);
  1442. av_register_output_format(&_3gp_oformat);
  1443. av_register_output_format(&mp4_oformat);
  1444. av_register_output_format(&psp_oformat);
  1445. av_register_output_format(&_3g2_oformat);
  1446. return 0;
  1447. }