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;
88 
89  /* xing header */
90  int64_t xing_offset;
93  uint32_t want;
94  uint32_t seen;
95  uint32_t pos;
96  uint64_t bag[XING_NUM_BAGS];
99 
100  /* index of the audio stream */
102  /* number of attached pictures we still need to write */
104 
105  /* audio packets are queued here until we get all the attached pictures */
107 } MP3Context;
108 
109 static const uint8_t xing_offtbl[2][2] = {{32, 17}, {17, 9}};
110 
111 /*
112  * Write an empty XING header and initialize respective data.
113  */
115 {
116  MP3Context *mp3 = s->priv_data;
117  AVCodecContext *codec = s->streams[mp3->audio_stream_idx]->codec;
118  int bitrate_idx;
119  int best_bitrate_idx = -1;
120  int best_bitrate_error= INT_MAX;
121  int xing_offset;
122  int32_t header, mask;
124  int srate_idx, ver = 0, i, channels;
125  int needed;
126  const char *vendor = (codec->flags & CODEC_FLAG_BITEXACT) ? "Lavf" : LIBAVFORMAT_IDENT;
127 
128  if (!s->pb->seekable)
129  return 0;
130 
131  for (i = 0; i < FF_ARRAY_ELEMS(avpriv_mpa_freq_tab); i++) {
132  const uint16_t base_freq = avpriv_mpa_freq_tab[i];
133 
134  if (codec->sample_rate == base_freq) ver = 0x3; // MPEG 1
135  else if (codec->sample_rate == base_freq / 2) ver = 0x2; // MPEG 2
136  else if (codec->sample_rate == base_freq / 4) ver = 0x0; // MPEG 2.5
137  else continue;
138 
139  srate_idx = i;
140  break;
141  }
143  av_log(s, AV_LOG_WARNING, "Unsupported sample rate, not writing Xing header.\n");
144  return -1;
145  }
146 
147  switch (codec->channels) {
148  case 1: channels = MPA_MONO; break;
149  case 2: channels = MPA_STEREO; break;
150  default: av_log(s, AV_LOG_WARNING, "Unsupported number of channels, "
151  "not writing Xing header.\n");
152  return -1;
153  }
154 
155  /* dummy MPEG audio header */
156  header = 0xffU << 24; // sync
157  header |= (0x7 << 5 | ver << 3 | 0x1 << 1 | 0x1) << 16; // sync/audio-version/layer 3/no crc*/
158  header |= (srate_idx << 2) << 8;
159  header |= channels << 6;
160 
161  for (bitrate_idx=1; bitrate_idx<15; bitrate_idx++) {
162  int error;
163  avpriv_mpegaudio_decode_header(&c, header | (bitrate_idx << (4+8)));
164  error= FFABS(c.bit_rate - codec->bit_rate);
165  if(error < best_bitrate_error){
166  best_bitrate_error= error;
167  best_bitrate_idx = bitrate_idx;
168  }
169  }
170  av_assert0(best_bitrate_idx >= 0);
171 
172  for (bitrate_idx= best_bitrate_idx;; bitrate_idx++) {
173  if (15 == bitrate_idx)
174  return -1;
175  mask = bitrate_idx << (4+8);
176  header |= mask;
177  avpriv_mpegaudio_decode_header(&c, header);
178  xing_offset=xing_offtbl[c.lsf == 1][c.nb_channels == 1];
179  needed = 4 // header
180  + xing_offset
181  + 4 // xing tag
182  + 4 // frames/size/toc flags
183  + 4 // frames
184  + 4 // size
185  + XING_TOC_SIZE // toc
186  + 24
187  ;
188 
189  if (needed <= c.frame_size)
190  break;
191  header &= ~mask;
192  }
193 
194  avio_wb32(s->pb, header);
195 
196  ffio_fill(s->pb, 0, xing_offset);
197  mp3->xing_offset = avio_tell(s->pb);
198  ffio_wfourcc(s->pb, "Xing");
199  avio_wb32(s->pb, 0x01 | 0x02 | 0x04); // frames / size / TOC
200 
201  mp3->size = c.frame_size;
202  mp3->want=1;
203  mp3->seen=0;
204  mp3->pos=0;
205 
206  avio_wb32(s->pb, 0); // frames
207  avio_wb32(s->pb, 0); // size
208 
209  // toc
210  for (i = 0; i < XING_TOC_SIZE; ++i)
211  avio_w8(s->pb, (uint8_t)(255 * i / XING_TOC_SIZE));
212 
213  for (i = 0; i < strlen(vendor); ++i)
214  avio_w8(s->pb, vendor[i]);
215  for (; i < 21; ++i)
216  avio_w8(s->pb, 0);
217  avio_wb24(s->pb, FFMAX(codec->delay - 528 - 1, 0)<<12);
218 
219  ffio_fill(s->pb, 0, c.frame_size - needed);
220 
221  return 0;
222 }
223 
224 /*
225  * Add a frame to XING data.
226  * Following lame's "VbrTag.c".
227  */
229 {
230  int i;
231 
232  mp3->frames++;
233  mp3->seen++;
234  mp3->size += pkt->size;
235 
236  if (mp3->want == mp3->seen) {
237  mp3->bag[mp3->pos] = mp3->size;
238 
239  if (XING_NUM_BAGS == ++mp3->pos) {
240  /* shrink table to half size by throwing away each second bag. */
241  for (i = 1; i < XING_NUM_BAGS; i += 2)
242  mp3->bag[i >> 1] = mp3->bag[i];
243 
244  /* double wanted amount per bag. */
245  mp3->want *= 2;
246  /* adjust current position to half of table size. */
247  mp3->pos = XING_NUM_BAGS / 2;
248  }
249 
250  mp3->seen = 0;
251  }
252 }
253 
255 {
256  MP3Context *mp3 = s->priv_data;
257 
258  if (pkt->data && pkt->size >= 4) {
260  int av_unused base;
261  uint32_t head = AV_RB32(pkt->data);
262 
263  if (ff_mpa_check_header(head) < 0) {
264  av_log(s, AV_LOG_WARNING, "Audio packet of size %d (starting with %08X...) "
265  "is invalid, writing it anyway.\n", pkt->size, head);
266  return ff_raw_write_packet(s, pkt);
267  }
269 
270  if (!mp3->initial_bitrate)
271  mp3->initial_bitrate = c.bit_rate;
272  if ((c.bit_rate == 0) || (mp3->initial_bitrate != c.bit_rate))
273  mp3->has_variable_bitrate = 1;
274 
275 #ifdef FILTER_VBR_HEADERS
276  /* filter out XING and INFO headers. */
277  base = 4 + xing_offtbl[c.lsf == 1][c.nb_channels == 1];
278 
279  if (base + 4 <= pkt->size) {
280  uint32_t v = AV_RB32(pkt->data + base);
281 
282  if (MKBETAG('X','i','n','g') == v || MKBETAG('I','n','f','o') == v)
283  return 0;
284  }
285 
286  /* filter out VBRI headers. */
287  base = 4 + 32;
288 
289  if (base + 4 <= pkt->size && MKBETAG('V','B','R','I') == AV_RB32(pkt->data + base))
290  return 0;
291 #endif
292 
293  if (mp3->xing_offset)
294  mp3_xing_add_frame(mp3, pkt);
295  }
296 
297  return ff_raw_write_packet(s, pkt);
298 }
299 
301 {
302  MP3Context *mp3 = s->priv_data;
303  AVPacketList *pktl;
304  int ret = 0, write = 1;
305 
306  ff_id3v2_finish(&mp3->id3, s->pb);
307  mp3_write_xing(s);
308 
309  while ((pktl = mp3->queue)) {
310  if (write && (ret = mp3_write_audio_packet(s, &pktl->pkt)) < 0)
311  write = 0;
312  av_free_packet(&pktl->pkt);
313  mp3->queue = pktl->next;
314  av_freep(&pktl);
315  }
316  mp3->queue_end = NULL;
317  return ret;
318 }
319 
321 {
322  MP3Context *mp3 = s->priv_data;
323  int i;
324 
325  /* replace "Xing" identification string with "Info" for CBR files. */
326  if (!mp3->has_variable_bitrate) {
327  avio_seek(s->pb, mp3->xing_offset, SEEK_SET);
328  ffio_wfourcc(s->pb, "Info");
329  }
330 
331  avio_seek(s->pb, mp3->xing_offset + 8, SEEK_SET);
332  avio_wb32(s->pb, mp3->frames);
333  avio_wb32(s->pb, mp3->size);
334 
335  avio_w8(s->pb, 0); // first toc entry has to be zero.
336 
337  for (i = 1; i < XING_TOC_SIZE; ++i) {
338  int j = i * mp3->pos / XING_TOC_SIZE;
339  int seek_point = 256LL * mp3->bag[j] / mp3->size;
340  avio_w8(s->pb, FFMIN(seek_point, 255));
341  }
342 
343  avio_seek(s->pb, 0, SEEK_END);
344 }
345 
347 {
349  MP3Context *mp3 = s->priv_data;
350 
351  if (mp3->pics_to_write) {
352  av_log(s, AV_LOG_WARNING, "No packets were sent for some of the "
353  "attached pictures.\n");
354  mp3_queue_flush(s);
355  }
356 
357  /* write the id3v1 tag */
358  if (mp3->write_id3v1 && id3v1_create_tag(s, buf) > 0) {
359  avio_write(s->pb, buf, ID3v1_TAG_SIZE);
360  }
361 
362  if (mp3->xing_offset)
363  mp3_update_xing(s);
364 
365  return 0;
366 }
367 
368 static int query_codec(enum AVCodecID id, int std_compliance)
369 {
371  while(cm->id != AV_CODEC_ID_NONE) {
372  if(id == cm->id)
373  return MKTAG('A', 'P', 'I', 'C');
374  cm++;
375  }
376  return -1;
377 }
378 
379 #if CONFIG_MP2_MUXER
380 AVOutputFormat ff_mp2_muxer = {
381  .name = "mp2",
382  .long_name = NULL_IF_CONFIG_SMALL("MP2 (MPEG audio layer 2)"),
383  .mime_type = "audio/x-mpeg",
384  .extensions = "mp2,m2a,mpa",
385  .audio_codec = AV_CODEC_ID_MP2,
386  .video_codec = AV_CODEC_ID_NONE,
387  .write_packet = ff_raw_write_packet,
388  .flags = AVFMT_NOTIMESTAMPS,
389 };
390 #endif
391 
392 #if CONFIG_MP3_MUXER
393 
394 static const AVOption options[] = {
395  { "id3v2_version", "Select ID3v2 version to write. Currently 3 and 4 are supported.",
396  offsetof(MP3Context, id3v2_version), AV_OPT_TYPE_INT, {.i64 = 4}, 3, 4, AV_OPT_FLAG_ENCODING_PARAM},
397  { "write_id3v1", "Enable ID3v1 writing. ID3v1 tags are written in UTF-8 which may not be supported by most software.",
398  offsetof(MP3Context, write_id3v1), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
399  { NULL },
400 };
401 
402 static const AVClass mp3_muxer_class = {
403  .class_name = "MP3 muxer",
404  .item_name = av_default_item_name,
405  .option = options,
406  .version = LIBAVUTIL_VERSION_INT,
407 };
408 
409 static int mp3_write_packet(AVFormatContext *s, AVPacket *pkt)
410 {
411  MP3Context *mp3 = s->priv_data;
412 
413  if (pkt->stream_index == mp3->audio_stream_idx) {
414  if (mp3->pics_to_write) {
415  /* buffer audio packets until we get all the pictures */
416  AVPacketList *pktl = av_mallocz(sizeof(*pktl));
417  if (!pktl)
418  return AVERROR(ENOMEM);
419 
420  pktl->pkt = *pkt;
421  pktl->pkt.buf = av_buffer_ref(pkt->buf);
422  if (!pktl->pkt.buf) {
423  av_freep(&pktl);
424  return AVERROR(ENOMEM);
425  }
426 
427  if (mp3->queue_end)
428  mp3->queue_end->next = pktl;
429  else
430  mp3->queue = pktl;
431  mp3->queue_end = pktl;
432  } else
433  return mp3_write_audio_packet(s, pkt);
434  } else {
435  int ret;
436 
437  /* warn only once for each stream */
438  if (s->streams[pkt->stream_index]->nb_frames == 1) {
439  av_log(s, AV_LOG_WARNING, "Got more than one picture in stream %d,"
440  " ignoring.\n", pkt->stream_index);
441  }
442  if (!mp3->pics_to_write || s->streams[pkt->stream_index]->nb_frames >= 1)
443  return 0;
444 
445  if ((ret = ff_id3v2_write_apic(s, &mp3->id3, pkt)) < 0)
446  return ret;
447  mp3->pics_to_write--;
448 
449  /* flush the buffered audio packets */
450  if (!mp3->pics_to_write &&
451  (ret = mp3_queue_flush(s)) < 0)
452  return ret;
453  }
454 
455  return 0;
456 }
457 
458 /**
459  * Write an ID3v2 header at beginning of stream
460  */
461 
462 static int mp3_write_header(struct AVFormatContext *s)
463 {
464  MP3Context *mp3 = s->priv_data;
465  int ret, i;
466 
467  /* check the streams -- we want exactly one audio and arbitrary number of
468  * video (attached pictures) */
469  mp3->audio_stream_idx = -1;
470  for (i = 0; i < s->nb_streams; i++) {
471  AVStream *st = s->streams[i];
472  if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
473  if (mp3->audio_stream_idx >= 0 || st->codec->codec_id != AV_CODEC_ID_MP3) {
474  av_log(s, AV_LOG_ERROR, "Invalid audio stream. Exactly one MP3 "
475  "audio stream is required.\n");
476  return AVERROR(EINVAL);
477  }
478  mp3->audio_stream_idx = i;
479  } else if (st->codec->codec_type != AVMEDIA_TYPE_VIDEO) {
480  av_log(s, AV_LOG_ERROR, "Only audio streams and pictures are allowed in MP3.\n");
481  return AVERROR(EINVAL);
482  }
483  }
484  if (mp3->audio_stream_idx < 0) {
485  av_log(s, AV_LOG_ERROR, "No audio stream present.\n");
486  return AVERROR(EINVAL);
487  }
488  mp3->pics_to_write = s->nb_streams - 1;
489 
491  ret = ff_id3v2_write_metadata(s, &mp3->id3);
492  if (ret < 0)
493  return ret;
494 
495  if (!mp3->pics_to_write) {
496  ff_id3v2_finish(&mp3->id3, s->pb);
497  mp3_write_xing(s);
498  }
499 
500  return 0;
501 }
502 
503 AVOutputFormat ff_mp3_muxer = {
504  .name = "mp3",
505  .long_name = NULL_IF_CONFIG_SMALL("MP3 (MPEG audio layer 3)"),
506  .mime_type = "audio/x-mpeg",
507  .extensions = "mp3",
508  .priv_data_size = sizeof(MP3Context),
509  .audio_codec = AV_CODEC_ID_MP3,
510  .video_codec = AV_CODEC_ID_PNG,
511  .write_header = mp3_write_header,
512  .write_packet = mp3_write_packet,
516  .priv_class = &mp3_muxer_class,
517 };
518 #endif