FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
mp3enc.c
Go to the documentation of this file.
1 /*
2  * MP3 muxer
3  * Copyright (c) 2003 Fabrice Bellard
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * FFmpeg is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
22 #include "avformat.h"
23 #include "avio_internal.h"
24 #include "id3v1.h"
25 #include "id3v2.h"
26 #include "rawenc.h"
27 #include "libavutil/avstring.h"
28 #include "libavcodec/mpegaudio.h"
31 #include "libavutil/intreadwrite.h"
32 #include "libavutil/opt.h"
33 #include "libavutil/dict.h"
34 #include "libavutil/avassert.h"
35 
36 static int id3v1_set_string(AVFormatContext *s, const char *key,
37  uint8_t *buf, int buf_size)
38 {
40  if ((tag = av_dict_get(s->metadata, key, NULL, 0)))
41  av_strlcpy(buf, tag->value, buf_size);
42  return !!tag;
43 }
44 
46 {
48  int i, count = 0;
49 
50  memset(buf, 0, ID3v1_TAG_SIZE); /* fail safe */
51  buf[0] = 'T';
52  buf[1] = 'A';
53  buf[2] = 'G';
54  /* we knowingly overspecify each tag length by one byte to compensate for the mandatory null byte added by av_strlcpy */
55  count += id3v1_set_string(s, "TIT2", buf + 3, 30 + 1); //title
56  count += id3v1_set_string(s, "TPE1", buf + 33, 30 + 1); //author|artist
57  count += id3v1_set_string(s, "TALB", buf + 63, 30 + 1); //album
58  count += id3v1_set_string(s, "TDRL", buf + 93, 4 + 1); //date
59  count += id3v1_set_string(s, "comment", buf + 97, 30 + 1);
60  if ((tag = av_dict_get(s->metadata, "TRCK", NULL, 0))) { //track
61  buf[125] = 0;
62  buf[126] = atoi(tag->value);
63  count++;
64  }
65  buf[127] = 0xFF; /* default to unknown genre */
66  if ((tag = av_dict_get(s->metadata, "TCON", NULL, 0))) { //genre
67  for(i = 0; i <= ID3v1_GENRE_MAX; i++) {
68  if (!av_strcasecmp(tag->value, ff_id3v1_genre_str[i])) {
69  buf[127] = i;
70  count++;
71  break;
72  }
73  }
74  }
75  return count;
76 }
77 
78 #define XING_NUM_BAGS 400
79 #define XING_TOC_SIZE 100
80 // maximum size of the xing frame: offset/Xing/flags/frames/size/TOC
81 #define XING_MAX_SIZE (32 + 4 + 4 + 4 + 4 + XING_TOC_SIZE)
82 
83 typedef struct MP3Context {
84  const AVClass *class;
89 
90  /* xing header */
91  int64_t xing_offset;
94  uint32_t want;
95  uint32_t seen;
96  uint32_t pos;
97  uint64_t bag[XING_NUM_BAGS];
100 
101  /* index of the audio stream */
103  /* number of attached pictures we still need to write */
105 
106  /* audio packets are queued here until we get all the attached pictures */
108 } MP3Context;
109 
110 static const uint8_t xing_offtbl[2][2] = {{32, 17}, {17, 9}};
111 
112 /*
113  * Write an empty XING header and initialize respective data.
114  */
116 {
117  MP3Context *mp3 = s->priv_data;
118  AVCodecContext *codec = s->streams[mp3->audio_stream_idx]->codec;
119  int bitrate_idx;
120  int best_bitrate_idx = -1;
121  int best_bitrate_error= INT_MAX;
122  int xing_offset;
125  int srate_idx, ver = 0, i, channels;
126  int needed;
127  const char *vendor = (codec->flags & CODEC_FLAG_BITEXACT) ? "Lavf" : LIBAVFORMAT_IDENT;
128 
129  if (!s->pb->seekable || !mp3->write_xing)
130  return 0;
131 
132  for (i = 0; i < FF_ARRAY_ELEMS(avpriv_mpa_freq_tab); i++) {
133  const uint16_t base_freq = avpriv_mpa_freq_tab[i];
134 
135  if (codec->sample_rate == base_freq) ver = 0x3; // MPEG 1
136  else if (codec->sample_rate == base_freq / 2) ver = 0x2; // MPEG 2
137  else if (codec->sample_rate == base_freq / 4) ver = 0x0; // MPEG 2.5
138  else continue;
139 
140  srate_idx = i;
141  break;
142  }
144  av_log(s, AV_LOG_WARNING, "Unsupported sample rate, not writing Xing header.\n");
145  return -1;
146  }
147 
148  switch (codec->channels) {
149  case 1: channels = MPA_MONO; break;
150  case 2: channels = MPA_STEREO; break;
151  default: av_log(s, AV_LOG_WARNING, "Unsupported number of channels, "
152  "not writing Xing header.\n");
153  return -1;
154  }
155 
156  /* dummy MPEG audio header */
157  header = 0xffU << 24; // sync
158  header |= (0x7 << 5 | ver << 3 | 0x1 << 1 | 0x1) << 16; // sync/audio-version/layer 3/no crc*/
159  header |= (srate_idx << 2) << 8;
160  header |= channels << 6;
161 
162  for (bitrate_idx=1; bitrate_idx<15; bitrate_idx++) {
163  int error;
164  avpriv_mpegaudio_decode_header(&c, header | (bitrate_idx << (4+8)));
165  error= FFABS(c.bit_rate - codec->bit_rate);
166  if(error < best_bitrate_error){
167  best_bitrate_error= error;
168  best_bitrate_idx = bitrate_idx;
169  }
170  }
171  av_assert0(best_bitrate_idx >= 0);
172 
173  for (bitrate_idx= best_bitrate_idx;; bitrate_idx++) {
174  if (15 == bitrate_idx)
175  return -1;
176  mask = bitrate_idx << (4+8);
177  header |= mask;
178  avpriv_mpegaudio_decode_header(&c, header);
179  xing_offset=xing_offtbl[c.lsf == 1][c.nb_channels == 1];
180  needed = 4 // header
181  + xing_offset
182  + 4 // xing tag
183  + 4 // frames/size/toc flags
184  + 4 // frames
185  + 4 // size
186  + XING_TOC_SIZE // toc
187  + 24
188  ;
189 
190  if (needed <= c.frame_size)
191  break;
192  header &= ~mask;
193  }
194 
195  avio_wb32(s->pb, header);
196 
197  ffio_fill(s->pb, 0, xing_offset);
198  mp3->xing_offset = avio_tell(s->pb);
199  ffio_wfourcc(s->pb, "Xing");
200  avio_wb32(s->pb, 0x01 | 0x02 | 0x04); // frames / size / TOC
201 
202  mp3->size = c.frame_size;
203  mp3->want=1;
204  mp3->seen=0;
205  mp3->pos=0;
206 
207  avio_wb32(s->pb, 0); // frames
208  avio_wb32(s->pb, 0); // size
209 
210  // toc
211  for (i = 0; i < XING_TOC_SIZE; ++i)
212  avio_w8(s->pb, (uint8_t)(255 * i / XING_TOC_SIZE));
213 
214  for (i = 0; i < strlen(vendor); ++i)
215  avio_w8(s->pb, vendor[i]);
216  for (; i < 21; ++i)
217  avio_w8(s->pb, 0);
218  avio_wb24(s->pb, FFMAX(codec->delay - 528 - 1, 0)<<12);
219 
220  ffio_fill(s->pb, 0, c.frame_size - needed);
221 
222  return 0;
223 }
224 
225 /*
226  * Add a frame to XING data.
227  * Following lame's "VbrTag.c".
228  */
230 {
231  int i;
232 
233  mp3->frames++;
234  mp3->seen++;
235  mp3->size += pkt->size;
236 
237  if (mp3->want == mp3->seen) {
238  mp3->bag[mp3->pos] = mp3->size;
239 
240  if (XING_NUM_BAGS == ++mp3->pos) {
241  /* shrink table to half size by throwing away each second bag. */
242  for (i = 1; i < XING_NUM_BAGS; i += 2)
243  mp3->bag[i >> 1] = mp3->bag[i];
244 
245  /* double wanted amount per bag. */
246  mp3->want *= 2;
247  /* adjust current position to half of table size. */
248  mp3->pos = XING_NUM_BAGS / 2;
249  }
250 
251  mp3->seen = 0;
252  }
253 }
254 
256 {
257  MP3Context *mp3 = s->priv_data;
258 
259  if (pkt->data && pkt->size >= 4) {
261  int av_unused base;
262  uint32_t head = AV_RB32(pkt->data);
263 
264  if (ff_mpa_check_header(head) < 0) {
265  av_log(s, AV_LOG_WARNING, "Audio packet of size %d (starting with %08X...) "
266  "is invalid, writing it anyway.\n", pkt->size, head);
267  return ff_raw_write_packet(s, pkt);
268  }
270 
271  if (!mp3->initial_bitrate)
272  mp3->initial_bitrate = c.bit_rate;
273  if ((c.bit_rate == 0) || (mp3->initial_bitrate != c.bit_rate))
274  mp3->has_variable_bitrate = 1;
275 
276 #ifdef FILTER_VBR_HEADERS
277  /* filter out XING and INFO headers. */
278  base = 4 + xing_offtbl[c.lsf == 1][c.nb_channels == 1];
279 
280  if (base + 4 <= pkt->size) {
281  uint32_t v = AV_RB32(pkt->data + base);
282 
283  if (MKBETAG('X','i','n','g') == v || MKBETAG('I','n','f','o') == v)
284  return 0;
285  }
286 
287  /* filter out VBRI headers. */
288  base = 4 + 32;
289 
290  if (base + 4 <= pkt->size && MKBETAG('V','B','R','I') == AV_RB32(pkt->data + base))
291  return 0;
292 #endif
293 
294  if (mp3->xing_offset)
295  mp3_xing_add_frame(mp3, pkt);
296  }
297 
298  return ff_raw_write_packet(s, pkt);
299 }
300 
302 {
303  MP3Context *mp3 = s->priv_data;
304  AVPacketList *pktl;
305  int ret = 0, write = 1;
306 
308  mp3_write_xing(s);
309 
310  while ((pktl = mp3->queue)) {
311  if (write && (ret = mp3_write_audio_packet(s, &pktl->pkt)) < 0)
312  write = 0;
313  av_free_packet(&pktl->pkt);
314  mp3->queue = pktl->next;
315  av_freep(&pktl);
316  }
317  mp3->queue_end = NULL;
318  return ret;
319 }
320 
322 {
323  MP3Context *mp3 = s->priv_data;
324  int i;
325 
326  /* replace "Xing" identification string with "Info" for CBR files. */
327  if (!mp3->has_variable_bitrate) {
328  avio_seek(s->pb, mp3->xing_offset, SEEK_SET);
329  ffio_wfourcc(s->pb, "Info");
330  }
331 
332  avio_seek(s->pb, mp3->xing_offset + 8, SEEK_SET);
333  avio_wb32(s->pb, mp3->frames);
334  avio_wb32(s->pb, mp3->size);
335 
336  avio_w8(s->pb, 0); // first toc entry has to be zero.
337 
338  for (i = 1; i < XING_TOC_SIZE; ++i) {
339  int j = i * mp3->pos / XING_TOC_SIZE;
340  int seek_point = 256LL * mp3->bag[j] / mp3->size;
341  avio_w8(s->pb, FFMIN(seek_point, 255));
342  }
343 
344  avio_seek(s->pb, 0, SEEK_END);
345 }
346 
348 {
350  MP3Context *mp3 = s->priv_data;
351 
352  if (mp3->pics_to_write) {
353  av_log(s, AV_LOG_WARNING, "No packets were sent for some of the "
354  "attached pictures.\n");
355  mp3_queue_flush(s);
356  }
357 
358  /* write the id3v1 tag */
359  if (mp3->write_id3v1 && id3v1_create_tag(s, buf) > 0) {
360  avio_write(s->pb, buf, ID3v1_TAG_SIZE);
361  }
362 
363  if (mp3->xing_offset)
364  mp3_update_xing(s);
365 
366  return 0;
367 }
368 
369 static int query_codec(enum AVCodecID id, int std_compliance)
370 {
372  while(cm->id != AV_CODEC_ID_NONE) {
373  if(id == cm->id)
374  return MKTAG('A', 'P', 'I', 'C');
375  cm++;
376  }
377  return -1;
378 }
379 
380 #if CONFIG_MP2_MUXER
381 AVOutputFormat ff_mp2_muxer = {
382  .name = "mp2",
383  .long_name = NULL_IF_CONFIG_SMALL("MP2 (MPEG audio layer 2)"),
384  .mime_type = "audio/x-mpeg",
385  .extensions = "mp2,m2a,mpa",
386  .audio_codec = AV_CODEC_ID_MP2,
387  .video_codec = AV_CODEC_ID_NONE,
388  .write_packet = ff_raw_write_packet,
389  .flags = AVFMT_NOTIMESTAMPS,
390 };
391 #endif
392 
393 #if CONFIG_MP3_MUXER
394 
395 static const AVOption options[] = {
396  { "id3v2_version", "Select ID3v2 version to write. Currently 3 and 4 are supported.",
397  offsetof(MP3Context, id3v2_version), AV_OPT_TYPE_INT, {.i64 = 4}, 0, 4, AV_OPT_FLAG_ENCODING_PARAM},
398  { "write_id3v1", "Enable ID3v1 writing. ID3v1 tags are written in UTF-8 which may not be supported by most software.",
399  offsetof(MP3Context, write_id3v1), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
400  { "write_xing", "Write the Xing header containing file duration.",
401  offsetof(MP3Context, write_xing), AV_OPT_TYPE_INT, {.i64 = 1}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
402  { NULL },
403 };
404 
405 static const AVClass mp3_muxer_class = {
406  .class_name = "MP3 muxer",
407  .item_name = av_default_item_name,
408  .option = options,
409  .version = LIBAVUTIL_VERSION_INT,
410 };
411 
412 static int mp3_write_packet(AVFormatContext *s, AVPacket *pkt)
413 {
414  MP3Context *mp3 = s->priv_data;
415 
416  if (pkt->stream_index == mp3->audio_stream_idx) {
417  if (mp3->pics_to_write) {
418  /* buffer audio packets until we get all the pictures */
419  AVPacketList *pktl = av_mallocz(sizeof(*pktl));
420  int ret;
421  if (!pktl)
422  return AVERROR(ENOMEM);
423 
424  ret = av_copy_packet(&pktl->pkt, pkt);
425  if (ret < 0) {
426  av_freep(&pktl);
427  return ret;
428  }
429 
430  if (mp3->queue_end)
431  mp3->queue_end->next = pktl;
432  else
433  mp3->queue = pktl;
434  mp3->queue_end = pktl;
435  } else
436  return mp3_write_audio_packet(s, pkt);
437  } else {
438  int ret;
439 
440  /* warn only once for each stream */
441  if (s->streams[pkt->stream_index]->nb_frames == 1) {
442  av_log(s, AV_LOG_WARNING, "Got more than one picture in stream %d,"
443  " ignoring.\n", pkt->stream_index);
444  }
445  if (!mp3->pics_to_write || s->streams[pkt->stream_index]->nb_frames >= 1)
446  return 0;
447 
448  if ((ret = ff_id3v2_write_apic(s, &mp3->id3, pkt)) < 0)
449  return ret;
450  mp3->pics_to_write--;
451 
452  /* flush the buffered audio packets */
453  if (!mp3->pics_to_write &&
454  (ret = mp3_queue_flush(s)) < 0)
455  return ret;
456  }
457 
458  return 0;
459 }
460 
461 /**
462  * Write an ID3v2 header at beginning of stream
463  */
464 
465 static int mp3_write_header(struct AVFormatContext *s)
466 {
467  MP3Context *mp3 = s->priv_data;
468  int ret, i;
469 
470  if (mp3->id3v2_version &&
471  mp3->id3v2_version != 3 &&
472  mp3->id3v2_version != 4) {
473  av_log(s, AV_LOG_ERROR, "Invalid ID3v2 version requested: %d. Only "
474  "3, 4 or 0 (disabled) are allowed.\n", mp3->id3v2_version);
475  return AVERROR(EINVAL);
476  }
477 
478  /* check the streams -- we want exactly one audio and arbitrary number of
479  * video (attached pictures) */
480  mp3->audio_stream_idx = -1;
481  for (i = 0; i < s->nb_streams; i++) {
482  AVStream *st = s->streams[i];
483  if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
484  if (mp3->audio_stream_idx >= 0 || st->codec->codec_id != AV_CODEC_ID_MP3) {
485  av_log(s, AV_LOG_ERROR, "Invalid audio stream. Exactly one MP3 "
486  "audio stream is required.\n");
487  return AVERROR(EINVAL);
488  }
489  mp3->audio_stream_idx = i;
490  } else if (st->codec->codec_type != AVMEDIA_TYPE_VIDEO) {
491  av_log(s, AV_LOG_ERROR, "Only audio streams and pictures are allowed in MP3.\n");
492  return AVERROR(EINVAL);
493  }
494  }
495  if (mp3->audio_stream_idx < 0) {
496  av_log(s, AV_LOG_ERROR, "No audio stream present.\n");
497  return AVERROR(EINVAL);
498  }
499  mp3->pics_to_write = s->nb_streams - 1;
500 
501  if (mp3->pics_to_write && !mp3->id3v2_version) {
502  av_log(s, AV_LOG_ERROR, "Attached pictures were requested, but the "
503  "ID3v2 header is disabled.\n");
504  return AVERROR(EINVAL);
505  }
506 
507  if (mp3->id3v2_version) {
509  ret = ff_id3v2_write_metadata(s, &mp3->id3);
510  if (ret < 0)
511  return ret;
512  }
513 
514  if (!mp3->pics_to_write) {
515  if (mp3->id3v2_version)
517  mp3_write_xing(s);
518  }
519 
520  return 0;
521 }
522 
523 AVOutputFormat ff_mp3_muxer = {
524  .name = "mp3",
525  .long_name = NULL_IF_CONFIG_SMALL("MP3 (MPEG audio layer 3)"),
526  .mime_type = "audio/x-mpeg",
527  .extensions = "mp3",
528  .priv_data_size = sizeof(MP3Context),
529  .audio_codec = AV_CODEC_ID_MP3,
530  .video_codec = AV_CODEC_ID_PNG,
531  .write_header = mp3_write_header,
532  .write_packet = mp3_write_packet,
536  .priv_class = &mp3_muxer_class,
537 };
538 #endif