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 #include "libavutil/crc.h"
36 #include "libavutil/mathematics.h"
37 #include "libavutil/replaygain.h"
38 
39 static int id3v1_set_string(AVFormatContext *s, const char *key,
40  uint8_t *buf, int buf_size)
41 {
43  if ((tag = av_dict_get(s->metadata, key, NULL, 0)))
44  av_strlcpy(buf, tag->value, buf_size);
45  return !!tag;
46 }
47 
49 {
51  int i, count = 0;
52 
53  memset(buf, 0, ID3v1_TAG_SIZE); /* fail safe */
54  buf[0] = 'T';
55  buf[1] = 'A';
56  buf[2] = 'G';
57  /* we knowingly overspecify each tag length by one byte to compensate for the mandatory null byte added by av_strlcpy */
58  count += id3v1_set_string(s, "TIT2", buf + 3, 30 + 1); //title
59  count += id3v1_set_string(s, "TPE1", buf + 33, 30 + 1); //author|artist
60  count += id3v1_set_string(s, "TALB", buf + 63, 30 + 1); //album
61  count += id3v1_set_string(s, "TDRL", buf + 93, 4 + 1); //date
62  count += id3v1_set_string(s, "comment", buf + 97, 30 + 1);
63  if ((tag = av_dict_get(s->metadata, "TRCK", NULL, 0))) { //track
64  buf[125] = 0;
65  buf[126] = atoi(tag->value);
66  count++;
67  }
68  buf[127] = 0xFF; /* default to unknown genre */
69  if ((tag = av_dict_get(s->metadata, "TCON", NULL, 0))) { //genre
70  for(i = 0; i <= ID3v1_GENRE_MAX; i++) {
71  if (!av_strcasecmp(tag->value, ff_id3v1_genre_str[i])) {
72  buf[127] = i;
73  count++;
74  break;
75  }
76  }
77  }
78  return count;
79 }
80 
81 #define XING_NUM_BAGS 400
82 #define XING_TOC_SIZE 100
83 // size of the XING/LAME data, starting from the Xing tag
84 #define XING_SIZE 156
85 
86 typedef struct MP3Context {
87  const AVClass *class;
92 
93  /* xing header */
94  // a buffer containing the whole XING/LAME frame
97 
98  AVCRC audio_crc; // CRC of the audio data
99  uint32_t audio_size; // total size of the audio data
100 
101  // offset of the XING/LAME frame in the file
103  // offset of the XING/INFO tag in the frame
105 
108  uint32_t want;
109  uint32_t seen;
110  uint32_t pos;
111  uint64_t bag[XING_NUM_BAGS];
114 
115  /* index of the audio stream */
117  /* number of attached pictures we still need to write */
119 
120  /* audio packets are queued here until we get all the attached pictures */
122 } MP3Context;
123 
124 static const uint8_t xing_offtbl[2][2] = {{32, 17}, {17, 9}};
125 
126 /*
127  * Write an empty XING header and initialize respective data.
128  */
130 {
131  MP3Context *mp3 = s->priv_data;
132  AVCodecContext *codec = s->streams[mp3->audio_stream_idx]->codec;
133  AVDictionaryEntry *enc = av_dict_get(s->streams[mp3->audio_stream_idx]->metadata, "encoder", NULL, 0);
134  AVIOContext *dyn_ctx;
135  int32_t header;
136  MPADecodeHeader mpah;
137  int srate_idx, i, channels;
138  int bitrate_idx;
139  int best_bitrate_idx = -1;
140  int best_bitrate_error = INT_MAX;
141  int ret;
142  int ver = 0;
143  int bytes_needed;
144 
145  if (!s->pb->seekable || !mp3->write_xing)
146  return 0;
147 
148  for (i = 0; i < FF_ARRAY_ELEMS(avpriv_mpa_freq_tab); i++) {
149  const uint16_t base_freq = avpriv_mpa_freq_tab[i];
150 
151  if (codec->sample_rate == base_freq) ver = 0x3; // MPEG 1
152  else if (codec->sample_rate == base_freq / 2) ver = 0x2; // MPEG 2
153  else if (codec->sample_rate == base_freq / 4) ver = 0x0; // MPEG 2.5
154  else continue;
155 
156  srate_idx = i;
157  break;
158  }
160  av_log(s, AV_LOG_WARNING, "Unsupported sample rate, not writing Xing header.\n");
161  return -1;
162  }
163 
164  switch (codec->channels) {
165  case 1: channels = MPA_MONO; break;
166  case 2: channels = MPA_STEREO; break;
167  default: av_log(s, AV_LOG_WARNING, "Unsupported number of channels, "
168  "not writing Xing header.\n");
169  return -1;
170  }
171 
172  /* dummy MPEG audio header */
173  header = 0xffU << 24; // sync
174  header |= (0x7 << 5 | ver << 3 | 0x1 << 1 | 0x1) << 16; // sync/audio-version/layer 3/no crc*/
175  header |= (srate_idx << 2) << 8;
176  header |= channels << 6;
177 
178  for (bitrate_idx = 1; bitrate_idx < 15; bitrate_idx++) {
179  int bit_rate = 1000 * avpriv_mpa_bitrate_tab[ver != 3][3 - 1][bitrate_idx];
180  int error = FFABS(bit_rate - codec->bit_rate);
181 
182  if (error < best_bitrate_error) {
183  best_bitrate_error = error;
184  best_bitrate_idx = bitrate_idx;
185  }
186  }
187  av_assert0(best_bitrate_idx >= 0);
188 
189  for (bitrate_idx = best_bitrate_idx; ; bitrate_idx++) {
190  int32_t mask = bitrate_idx << (4 + 8);
191  if (15 == bitrate_idx)
192  return -1;
193  header |= mask;
194 
195  avpriv_mpegaudio_decode_header(&mpah, header);
196  mp3->xing_offset = xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1] + 4;
197  bytes_needed = mp3->xing_offset + XING_SIZE;
198 
199  if (bytes_needed <= mpah.frame_size)
200  break;
201 
202  header &= ~mask;
203  }
204 
205  ret = avio_open_dyn_buf(&dyn_ctx);
206  if (ret < 0)
207  return ret;
208 
209  avio_wb32(dyn_ctx, header);
210 
211  ffio_fill(dyn_ctx, 0, mp3->xing_offset - 4);
212  ffio_wfourcc(dyn_ctx, "Xing");
213  avio_wb32(dyn_ctx, 0x01 | 0x02 | 0x04 | 0x08); // frames / size / TOC / vbr scale
214 
215  mp3->size = mpah.frame_size;
216  mp3->want=1;
217  mp3->seen=0;
218  mp3->pos=0;
219 
220  avio_wb32(dyn_ctx, 0); // frames
221  avio_wb32(dyn_ctx, 0); // size
222 
223  // TOC
224  for (i = 0; i < XING_TOC_SIZE; i++)
225  avio_w8(dyn_ctx, (uint8_t)(255 * i / XING_TOC_SIZE));
226 
227  // vbr quality
228  // we write it, because some (broken) tools always expect it to be present
229  avio_wb32(dyn_ctx, 0);
230 
231  // encoder short version string
232  if (enc) {
233  uint8_t encoder_str[9] = { 0 };
234  if ( strlen(enc->value) > sizeof(encoder_str)
235  && !strcmp("Lavc libmp3lame", enc->value)) {
236  memcpy(encoder_str, "Lavf lame", 9);
237  } else
238  memcpy(encoder_str, enc->value, FFMIN(strlen(enc->value), sizeof(encoder_str)));
239 
240  avio_write(dyn_ctx, encoder_str, sizeof(encoder_str));
241  } else
242  avio_write(dyn_ctx, "Lavf\0\0\0\0\0", 9);
243 
244  avio_w8(dyn_ctx, 0); // tag revision 0 / unknown vbr method
245  avio_w8(dyn_ctx, 0); // unknown lowpass filter value
246  ffio_fill(dyn_ctx, 0, 8); // empty replaygain fields
247  avio_w8(dyn_ctx, 0); // unknown encoding flags
248  avio_w8(dyn_ctx, 0); // unknown abr/minimal bitrate
249 
250  // encoder delay
251  if (codec->initial_padding - 528 - 1 >= 1 << 12) {
252  av_log(s, AV_LOG_WARNING, "Too many samples of initial padding.\n");
253  }
254  avio_wb24(dyn_ctx, FFMAX(codec->initial_padding - 528 - 1, 0)<<12);
255 
256  avio_w8(dyn_ctx, 0); // misc
257  avio_w8(dyn_ctx, 0); // mp3gain
258  avio_wb16(dyn_ctx, 0); // preset
259 
260  // audio length and CRCs (will be updated later)
261  avio_wb32(dyn_ctx, 0); // music length
262  avio_wb16(dyn_ctx, 0); // music crc
263  avio_wb16(dyn_ctx, 0); // tag crc
264 
265  ffio_fill(dyn_ctx, 0, mpah.frame_size - bytes_needed);
266 
267  mp3->xing_frame_size = avio_close_dyn_buf(dyn_ctx, &mp3->xing_frame);
268  mp3->xing_frame_offset = avio_tell(s->pb);
269  avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size);
270 
271  mp3->audio_size = mp3->xing_frame_size;
272 
273  return 0;
274 }
275 
276 /*
277  * Add a frame to XING data.
278  * Following lame's "VbrTag.c".
279  */
281 {
282  int i;
283 
284  mp3->frames++;
285  mp3->seen++;
286  mp3->size += pkt->size;
287 
288  if (mp3->want == mp3->seen) {
289  mp3->bag[mp3->pos] = mp3->size;
290 
291  if (XING_NUM_BAGS == ++mp3->pos) {
292  /* shrink table to half size by throwing away each second bag. */
293  for (i = 1; i < XING_NUM_BAGS; i += 2)
294  mp3->bag[i >> 1] = mp3->bag[i];
295 
296  /* double wanted amount per bag. */
297  mp3->want *= 2;
298  /* adjust current position to half of table size. */
299  mp3->pos = XING_NUM_BAGS / 2;
300  }
301 
302  mp3->seen = 0;
303  }
304 }
305 
307 {
308  MP3Context *mp3 = s->priv_data;
309 
310  if (pkt->data && pkt->size >= 4) {
311  MPADecodeHeader mpah;
312  int av_unused base;
313  uint32_t h;
314 
315  h = AV_RB32(pkt->data);
316  if (ff_mpa_check_header(h) == 0) {
318  if (!mp3->initial_bitrate)
319  mp3->initial_bitrate = mpah.bit_rate;
320  if ((mpah.bit_rate == 0) || (mp3->initial_bitrate != mpah.bit_rate))
321  mp3->has_variable_bitrate = 1;
322  } else {
323  av_log(s, AV_LOG_WARNING, "Audio packet of size %d (starting with %08X...) "
324  "is invalid, writing it anyway.\n", pkt->size, h);
325  }
326 
327 #ifdef FILTER_VBR_HEADERS
328  /* filter out XING and INFO headers. */
329  base = 4 + xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1];
330 
331  if (base + 4 <= pkt->size) {
332  uint32_t v = AV_RB32(pkt->data + base);
333 
334  if (MKBETAG('X','i','n','g') == v || MKBETAG('I','n','f','o') == v)
335  return 0;
336  }
337 
338  /* filter out VBRI headers. */
339  base = 4 + 32;
340 
341  if (base + 4 <= pkt->size && MKBETAG('V','B','R','I') == AV_RB32(pkt->data + base))
342  return 0;
343 #endif
344 
345  if (mp3->xing_offset) {
346  mp3_xing_add_frame(mp3, pkt);
347  mp3->audio_size += pkt->size;
349  mp3->audio_crc, pkt->data, pkt->size);
350  }
351  }
352 
353  return ff_raw_write_packet(s, pkt);
354 }
355 
357 {
358  MP3Context *mp3 = s->priv_data;
359  AVPacketList *pktl;
360  int ret = 0, write = 1;
361 
363  mp3_write_xing(s);
364 
365  while ((pktl = mp3->queue)) {
366  if (write && (ret = mp3_write_audio_packet(s, &pktl->pkt)) < 0)
367  write = 0;
368  av_free_packet(&pktl->pkt);
369  mp3->queue = pktl->next;
370  av_freep(&pktl);
371  }
372  mp3->queue_end = NULL;
373  return ret;
374 }
375 
377 {
378  MP3Context *mp3 = s->priv_data;
379  AVReplayGain *rg;
380  uint16_t tag_crc;
381  uint8_t *toc;
382  int i, rg_size;
383 
384  /* replace "Xing" identification string with "Info" for CBR files. */
385  if (!mp3->has_variable_bitrate)
386  AV_WL32(mp3->xing_frame + mp3->xing_offset, MKTAG('I', 'n', 'f', 'o'));
387 
388  AV_WB32(mp3->xing_frame + mp3->xing_offset + 8, mp3->frames);
389  AV_WB32(mp3->xing_frame + mp3->xing_offset + 12, mp3->size);
390 
391  toc = mp3->xing_frame + mp3->xing_offset + 16;
392  toc[0] = 0; // first toc entry has to be zero.
393  for (i = 1; i < XING_TOC_SIZE; ++i) {
394  int j = i * mp3->pos / XING_TOC_SIZE;
395  int seek_point = 256LL * mp3->bag[j] / mp3->size;
396  toc[i] = FFMIN(seek_point, 255);
397  }
398 
399  /* write replaygain */
401  &rg_size);
402  if (rg && rg_size >= sizeof(*rg)) {
403  uint16_t val;
404 
405  AV_WB32(mp3->xing_frame + mp3->xing_offset + 131,
406  av_rescale(rg->track_peak, 1 << 23, 100000));
407 
408  if (rg->track_gain != INT32_MIN) {
409  val = FFABS(rg->track_gain / 10000) & ((1 << 9) - 1);
410  val |= (rg->track_gain < 0) << 9;
411  val |= 1 << 13;
412  AV_WB16(mp3->xing_frame + mp3->xing_offset + 135, val);
413  }
414 
415  if (rg->album_gain != INT32_MIN) {
416  val = FFABS(rg->album_gain / 10000) & ((1 << 9) - 1);
417  val |= (rg->album_gain < 0) << 9;
418  val |= 1 << 14;
419  AV_WB16(mp3->xing_frame + mp3->xing_offset + 137, val);
420  }
421  }
422 
423  AV_WB32(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 8, mp3->audio_size);
424  AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 4, mp3->audio_crc);
425 
426  tag_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE), 0, mp3->xing_frame, 190);
427  AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 2, tag_crc);
428 
429  avio_seek(s->pb, mp3->xing_frame_offset, SEEK_SET);
430  avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size);
431  avio_seek(s->pb, 0, SEEK_END);
432 }
433 
435 {
437  MP3Context *mp3 = s->priv_data;
438 
439  if (mp3->pics_to_write) {
440  av_log(s, AV_LOG_WARNING, "No packets were sent for some of the "
441  "attached pictures.\n");
442  mp3_queue_flush(s);
443  }
444 
445  /* write the id3v1 tag */
446  if (mp3->write_id3v1 && id3v1_create_tag(s, buf) > 0) {
447  avio_write(s->pb, buf, ID3v1_TAG_SIZE);
448  }
449 
450  if (mp3->xing_offset)
451  mp3_update_xing(s);
452 
453  av_freep(&mp3->xing_frame);
454 
455  return 0;
456 }
457 
458 static int query_codec(enum AVCodecID id, int std_compliance)
459 {
461  while(cm->id != AV_CODEC_ID_NONE) {
462  if(id == cm->id)
463  return MKTAG('A', 'P', 'I', 'C');
464  cm++;
465  }
466  return -1;
467 }
468 
469 #if CONFIG_MP2_MUXER
470 AVOutputFormat ff_mp2_muxer = {
471  .name = "mp2",
472  .long_name = NULL_IF_CONFIG_SMALL("MP2 (MPEG audio layer 2)"),
473  .mime_type = "audio/mpeg",
474  .extensions = "mp2,m2a,mpa",
475  .audio_codec = AV_CODEC_ID_MP2,
476  .video_codec = AV_CODEC_ID_NONE,
477  .write_packet = ff_raw_write_packet,
478  .flags = AVFMT_NOTIMESTAMPS,
479 };
480 #endif
481 
482 #if CONFIG_MP3_MUXER
483 
484 static const AVOption options[] = {
485  { "id3v2_version", "Select ID3v2 version to write. Currently 3 and 4 are supported.",
486  offsetof(MP3Context, id3v2_version), AV_OPT_TYPE_INT, {.i64 = 4}, 0, 4, AV_OPT_FLAG_ENCODING_PARAM},
487  { "write_id3v1", "Enable ID3v1 writing. ID3v1 tags are written in UTF-8 which may not be supported by most software.",
488  offsetof(MP3Context, write_id3v1), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
489  { "write_xing", "Write the Xing header containing file duration.",
490  offsetof(MP3Context, write_xing), AV_OPT_TYPE_INT, {.i64 = 1}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
491  { NULL },
492 };
493 
494 static const AVClass mp3_muxer_class = {
495  .class_name = "MP3 muxer",
496  .item_name = av_default_item_name,
497  .option = options,
498  .version = LIBAVUTIL_VERSION_INT,
499 };
500 
501 static int mp3_write_packet(AVFormatContext *s, AVPacket *pkt)
502 {
503  MP3Context *mp3 = s->priv_data;
504 
505  if (pkt->stream_index == mp3->audio_stream_idx) {
506  if (mp3->pics_to_write) {
507  /* buffer audio packets until we get all the pictures */
508  AVPacketList *pktl = av_mallocz(sizeof(*pktl));
509  int ret;
510  if (!pktl) {
511  av_log(s, AV_LOG_WARNING, "Not enough memory to buffer audio. Skipping picture streams\n");
512  mp3->pics_to_write = 0;
513  mp3_queue_flush(s);
514  return mp3_write_audio_packet(s, pkt);
515  }
516 
517  ret = av_copy_packet(&pktl->pkt, pkt);
518  if (ret < 0) {
519  av_freep(&pktl);
520  return ret;
521  }
522 
523  if (mp3->queue_end)
524  mp3->queue_end->next = pktl;
525  else
526  mp3->queue = pktl;
527  mp3->queue_end = pktl;
528  } else
529  return mp3_write_audio_packet(s, pkt);
530  } else {
531  int ret;
532 
533  /* warn only once for each stream */
534  if (s->streams[pkt->stream_index]->nb_frames == 1) {
535  av_log(s, AV_LOG_WARNING, "Got more than one picture in stream %d,"
536  " ignoring.\n", pkt->stream_index);
537  }
538  if (!mp3->pics_to_write || s->streams[pkt->stream_index]->nb_frames >= 1)
539  return 0;
540 
541  if ((ret = ff_id3v2_write_apic(s, &mp3->id3, pkt)) < 0)
542  return ret;
543  mp3->pics_to_write--;
544 
545  /* flush the buffered audio packets */
546  if (!mp3->pics_to_write &&
547  (ret = mp3_queue_flush(s)) < 0)
548  return ret;
549  }
550 
551  return 0;
552 }
553 
554 /**
555  * Write an ID3v2 header at beginning of stream
556  */
557 
558 static int mp3_write_header(struct AVFormatContext *s)
559 {
560  MP3Context *mp3 = s->priv_data;
561  int ret, i;
562 
563  if (mp3->id3v2_version &&
564  mp3->id3v2_version != 3 &&
565  mp3->id3v2_version != 4) {
566  av_log(s, AV_LOG_ERROR, "Invalid ID3v2 version requested: %d. Only "
567  "3, 4 or 0 (disabled) are allowed.\n", mp3->id3v2_version);
568  return AVERROR(EINVAL);
569  }
570 
571  /* check the streams -- we want exactly one audio and arbitrary number of
572  * video (attached pictures) */
573  mp3->audio_stream_idx = -1;
574  for (i = 0; i < s->nb_streams; i++) {
575  AVStream *st = s->streams[i];
576  if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
577  if (mp3->audio_stream_idx >= 0 || st->codec->codec_id != AV_CODEC_ID_MP3) {
578  av_log(s, AV_LOG_ERROR, "Invalid audio stream. Exactly one MP3 "
579  "audio stream is required.\n");
580  return AVERROR(EINVAL);
581  }
582  mp3->audio_stream_idx = i;
583  } else if (st->codec->codec_type != AVMEDIA_TYPE_VIDEO) {
584  av_log(s, AV_LOG_ERROR, "Only audio streams and pictures are allowed in MP3.\n");
585  return AVERROR(EINVAL);
586  }
587  }
588  if (mp3->audio_stream_idx < 0) {
589  av_log(s, AV_LOG_ERROR, "No audio stream present.\n");
590  return AVERROR(EINVAL);
591  }
592  mp3->pics_to_write = s->nb_streams - 1;
593 
594  if (mp3->pics_to_write && !mp3->id3v2_version) {
595  av_log(s, AV_LOG_ERROR, "Attached pictures were requested, but the "
596  "ID3v2 header is disabled.\n");
597  return AVERROR(EINVAL);
598  }
599 
600  if (mp3->id3v2_version) {
602  ret = ff_id3v2_write_metadata(s, &mp3->id3);
603  if (ret < 0)
604  return ret;
605  }
606 
607  if (!mp3->pics_to_write) {
608  if (mp3->id3v2_version)
610  mp3_write_xing(s);
611  }
612 
613  return 0;
614 }
615 
616 AVOutputFormat ff_mp3_muxer = {
617  .name = "mp3",
618  .long_name = NULL_IF_CONFIG_SMALL("MP3 (MPEG audio layer 3)"),
619  .mime_type = "audio/mpeg",
620  .extensions = "mp3",
621  .priv_data_size = sizeof(MP3Context),
622  .audio_codec = AV_CODEC_ID_MP3,
623  .video_codec = AV_CODEC_ID_PNG,
624  .write_header = mp3_write_header,
625  .write_packet = mp3_write_packet,
629  .priv_class = &mp3_muxer_class,
630 };
631 #endif