3 * Copyright (c) 2003 Fabrice Bellard
5 * This file is part of Libav.
7 * Libav 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.
12 * Libav 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.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #include "avio_internal.h"
27 #include "libavutil/avstring.h"
28 #include "libavcodec/mpegaudio.h"
29 #include "libavcodec/mpegaudiodata.h"
30 #include "libavcodec/mpegaudiodecheader.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"
39 static int id3v1_set_string(AVFormatContext *s, const char *key,
40 uint8_t *buf, int buf_size)
42 AVDictionaryEntry *tag;
43 if ((tag = av_dict_get(s->metadata, key, NULL, 0)))
44 av_strlcpy(buf, tag->value, buf_size);
48 static int id3v1_create_tag(AVFormatContext *s, uint8_t *buf)
50 AVDictionaryEntry *tag;
53 memset(buf, 0, ID3v1_TAG_SIZE); /* fail safe */
57 count += id3v1_set_string(s, "TIT2", buf + 3, 30); //title
58 count += id3v1_set_string(s, "TPE1", buf + 33, 30); //author|artist
59 count += id3v1_set_string(s, "TALB", buf + 63, 30); //album
60 count += id3v1_set_string(s, "TDRL", buf + 93, 4); //date
61 count += id3v1_set_string(s, "comment", buf + 97, 30);
62 if ((tag = av_dict_get(s->metadata, "TRCK", NULL, 0))) { //track
64 buf[126] = atoi(tag->value);
67 buf[127] = 0xFF; /* default to unknown genre */
68 if ((tag = av_dict_get(s->metadata, "TCON", NULL, 0))) { //genre
69 for(i = 0; i <= ID3v1_GENRE_MAX; i++) {
70 if (!av_strcasecmp(tag->value, ff_id3v1_genre_str[i])) {
80 #define XING_NUM_BAGS 400
81 #define XING_TOC_SIZE 100
82 // size of the XING/LAME data, starting from the Xing tag
85 typedef struct MP3Context {
93 // a buffer containing the whole XING/LAME frame
97 AVCRC audio_crc; // CRC of the audio data
98 uint32_t audio_size; // total size of the audio data
100 // offset of the XING/LAME frame in the file
101 int64_t xing_frame_offset;
102 // offset of the XING/INFO tag in the frame
110 uint64_t bag[XING_NUM_BAGS];
112 int has_variable_bitrate;
114 /* index of the audio stream */
115 int audio_stream_idx;
116 /* number of attached pictures we still need to write */
119 /* audio packets are queued here until we get all the attached pictures */
120 AVPacketList *queue, *queue_end;
123 static const uint8_t xing_offtbl[2][2] = {{32, 17}, {17, 9}};
126 * Write an empty XING header and initialize respective data.
128 static void mp3_write_xing(AVFormatContext *s)
130 MP3Context *mp3 = s->priv_data;
131 AVCodecParameters *par = s->streams[mp3->audio_stream_idx]->codecpar;
132 AVDictionaryEntry *enc = av_dict_get(s->streams[mp3->audio_stream_idx]->metadata, "encoder", NULL, 0);
133 AVIOContext *dyn_ctx;
135 MPADecodeHeader mpah;
136 int srate_idx, i, channels;
138 int best_bitrate_idx;
139 int best_bitrate_error = INT_MAX;
142 int lsf, bytes_needed;
144 if (!(s->pb->seekable & AVIO_SEEKABLE_NORMAL) || !mp3->write_xing)
147 for (i = 0; i < FF_ARRAY_ELEMS(avpriv_mpa_freq_tab); i++) {
148 const uint16_t base_freq = avpriv_mpa_freq_tab[i];
150 if (par->sample_rate == base_freq) ver = 0x3; // MPEG 1
151 else if (par->sample_rate == base_freq / 2) ver = 0x2; // MPEG 2
152 else if (par->sample_rate == base_freq / 4) ver = 0x0; // MPEG 2.5
158 if (i == FF_ARRAY_ELEMS(avpriv_mpa_freq_tab)) {
159 av_log(s, AV_LOG_WARNING, "Unsupported sample rate, not writing Xing "
164 switch (par->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");
172 /* dummy MPEG audio header */
173 header = 0xff << 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;
178 lsf = !((header & (1 << 20) && header & (1 << 19)));
180 mp3->xing_offset = xing_offtbl[ver != 3][channels == 1] + 4;
181 bytes_needed = mp3->xing_offset + XING_SIZE;
183 for (bitrate_idx = 1; bitrate_idx < 15; bitrate_idx++) {
184 int bit_rate = 1000 * avpriv_mpa_bitrate_tab[lsf][3 - 1][bitrate_idx];
185 int error = FFABS(bit_rate - par->bit_rate);
187 if (error < best_bitrate_error){
188 best_bitrate_error = error;
189 best_bitrate_idx = bitrate_idx;
193 for (bitrate_idx = best_bitrate_idx; bitrate_idx < 15; bitrate_idx++) {
194 int32_t mask = bitrate_idx << (4 + 8);
197 avpriv_mpegaudio_decode_header(&mpah, header);
199 if (bytes_needed <= mpah.frame_size)
205 ret = avio_open_dyn_buf(&dyn_ctx);
209 avio_wb32(dyn_ctx, header);
211 avpriv_mpegaudio_decode_header(&mpah, header);
213 av_assert0(mpah.frame_size >= bytes_needed);
215 ffio_fill(dyn_ctx, 0, mp3->xing_offset - 4);
216 ffio_wfourcc(dyn_ctx, "Xing");
217 avio_wb32(dyn_ctx, 0x01 | 0x02 | 0x04 | 0x08); // frames / size / TOC / vbr scale
219 mp3->size = mpah.frame_size;
222 avio_wb32(dyn_ctx, 0); // frames
223 avio_wb32(dyn_ctx, 0); // size
226 for (i = 0; i < XING_TOC_SIZE; i++)
227 avio_w8(dyn_ctx, 255 * i / XING_TOC_SIZE);
230 // we write it, because some (broken) tools always expect it to be present
231 avio_wb32(dyn_ctx, 0);
233 // encoder short version string
235 uint8_t encoder_str[9] = { 0 };
236 memcpy(encoder_str, enc->value, FFMIN(strlen(enc->value), sizeof(encoder_str)));
237 avio_write(dyn_ctx, encoder_str, sizeof(encoder_str));
239 ffio_fill(dyn_ctx, 0, 9);
241 avio_w8(dyn_ctx, 0); // tag revision 0 / unknown vbr method
242 avio_w8(dyn_ctx, 0); // unknown lowpass filter value
243 ffio_fill(dyn_ctx, 0, 8); // empty replaygain fields
244 avio_w8(dyn_ctx, 0); // unknown encoding flags
245 avio_w8(dyn_ctx, 0); // unknown abr/minimal bitrate
248 if (par->initial_padding >= 1 << 12 ||
249 par->trailing_padding >= 1 << 12) {
250 av_log(s, AV_LOG_WARNING, "Too many samples of padding.\n");
251 avio_wb24(dyn_ctx, 0);
253 avio_wb24(dyn_ctx, par->initial_padding << 12 | par->trailing_padding);
256 avio_w8(dyn_ctx, 0); // misc
257 avio_w8(dyn_ctx, 0); // mp3gain
258 avio_wb16(dyn_ctx, 0); // preset
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
265 ffio_fill(dyn_ctx, 0, mpah.frame_size - bytes_needed);
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);
271 mp3->audio_size = mp3->xing_frame_size;
275 * Add a frame to XING data.
276 * Following lame's "VbrTag.c".
278 static void mp3_xing_add_frame(MP3Context *mp3, AVPacket *pkt)
284 mp3->size += pkt->size;
286 if (mp3->want == mp3->seen) {
287 mp3->bag[mp3->pos] = mp3->size;
289 if (XING_NUM_BAGS == ++mp3->pos) {
290 /* shrink table to half size by throwing away each second bag. */
291 for (i = 1; i < XING_NUM_BAGS; i += 2)
292 mp3->bag[i / 2] = mp3->bag[i];
294 /* double wanted amount per bag. */
296 /* adjust current position to half of table size. */
297 mp3->pos = XING_NUM_BAGS / 2;
304 static int mp3_write_audio_packet(AVFormatContext *s, AVPacket *pkt)
306 MP3Context *mp3 = s->priv_data;
308 if (mp3->xing_offset && pkt->size >= 4) {
313 h = AV_RB32(pkt->data);
314 ret = avpriv_mpegaudio_decode_header(&c, h);
316 if (!mp3->initial_bitrate)
317 mp3->initial_bitrate = c.bit_rate;
318 if ((c.bit_rate == 0) || (mp3->initial_bitrate != c.bit_rate))
319 mp3->has_variable_bitrate = 1;
322 mp3_xing_add_frame(mp3, pkt);
324 if (mp3->xing_offset) {
325 mp3->audio_size += pkt->size;
326 mp3->audio_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE),
327 mp3->audio_crc, pkt->data, pkt->size);
331 return ff_raw_write_packet(s, pkt);
334 static int mp3_queue_flush(AVFormatContext *s)
336 MP3Context *mp3 = s->priv_data;
338 int ret = 0, write = 1;
340 ff_id3v2_finish(&mp3->id3, s->pb);
343 while ((pktl = mp3->queue)) {
344 if (write && (ret = mp3_write_audio_packet(s, &pktl->pkt)) < 0)
346 av_packet_unref(&pktl->pkt);
347 mp3->queue = pktl->next;
350 mp3->queue_end = NULL;
354 static void mp3_update_xing(AVFormatContext *s)
356 MP3Context *mp3 = s->priv_data;
362 /* replace "Xing" identification string with "Info" for CBR files. */
363 if (!mp3->has_variable_bitrate)
364 AV_WL32(mp3->xing_frame + mp3->xing_offset, MKTAG('I', 'n', 'f', 'o'));
366 AV_WB32(mp3->xing_frame + mp3->xing_offset + 8, mp3->frames);
367 AV_WB32(mp3->xing_frame + mp3->xing_offset + 12, mp3->size);
369 toc = mp3->xing_frame + mp3->xing_offset + 16;
370 toc[0] = 0; // first toc entry has to be zero.
371 for (i = 1; i < XING_TOC_SIZE; ++i) {
372 int j = i * mp3->pos / XING_TOC_SIZE;
373 int seek_point = 256LL * mp3->bag[j] / mp3->size;
374 toc[i] = FFMIN(seek_point, 255);
377 /* write replaygain */
378 rg = (AVReplayGain*)av_stream_get_side_data(s->streams[0], AV_PKT_DATA_REPLAYGAIN,
380 if (rg && rg_size >= sizeof(*rg)) {
383 AV_WB32(mp3->xing_frame + mp3->xing_offset + 131,
384 av_rescale(rg->track_peak, 1 << 23, 100000));
386 if (rg->track_gain != INT32_MIN) {
387 val = FFABS(rg->track_gain / 10000) & ((1 << 9) - 1);
388 val |= (rg->track_gain < 0) << 9;
390 AV_WB16(mp3->xing_frame + mp3->xing_offset + 135, val);
393 if (rg->album_gain != INT32_MIN) {
394 val = FFABS(rg->album_gain / 10000) & ((1 << 9) - 1);
395 val |= (rg->album_gain < 0) << 9;
397 AV_WB16(mp3->xing_frame + mp3->xing_offset + 137, val);
401 AV_WB32(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 8, mp3->audio_size);
402 AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 4, mp3->audio_crc);
404 tag_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE), 0, mp3->xing_frame, 190);
405 AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 2, tag_crc);
407 avio_seek(s->pb, mp3->xing_frame_offset, SEEK_SET);
408 avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size);
409 avio_seek(s->pb, 0, SEEK_END);
412 static int mp3_write_trailer(struct AVFormatContext *s)
414 uint8_t buf[ID3v1_TAG_SIZE];
415 MP3Context *mp3 = s->priv_data;
417 if (mp3->pics_to_write) {
418 av_log(s, AV_LOG_WARNING, "No packets were sent for some of the "
419 "attached pictures.\n");
423 /* write the id3v1 tag */
424 if (mp3->write_id3v1 && id3v1_create_tag(s, buf) > 0) {
425 avio_write(s->pb, buf, ID3v1_TAG_SIZE);
428 if (mp3->xing_offset)
431 av_freep(&mp3->xing_frame);
436 static const AVOption options[] = {
437 { "id3v2_version", "Select ID3v2 version to write. Currently 3 and 4 are supported.",
438 offsetof(MP3Context, id3v2_version), AV_OPT_TYPE_INT, {.i64 = 4}, 0, 4, AV_OPT_FLAG_ENCODING_PARAM},
439 { "write_id3v1", "Enable ID3v1 writing. ID3v1 tags are written in UTF-8 which may not be supported by most software.",
440 offsetof(MP3Context, write_id3v1), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
441 { "write_xing", "Write the Xing header containing file duration.",
442 offsetof(MP3Context, write_xing), AV_OPT_TYPE_INT, {.i64 = 1}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
446 static const AVClass mp3_muxer_class = {
447 .class_name = "MP3 muxer",
448 .item_name = av_default_item_name,
450 .version = LIBAVUTIL_VERSION_INT,
453 static int mp3_write_packet(AVFormatContext *s, AVPacket *pkt)
455 MP3Context *mp3 = s->priv_data;
457 if (pkt->stream_index == mp3->audio_stream_idx) {
458 if (mp3->pics_to_write) {
459 /* buffer audio packets until we get all the pictures */
460 AVPacketList *pktl = av_mallocz(sizeof(*pktl));
462 return AVERROR(ENOMEM);
465 pktl->pkt.buf = av_buffer_ref(pkt->buf);
466 if (!pktl->pkt.buf) {
468 return AVERROR(ENOMEM);
472 mp3->queue_end->next = pktl;
475 mp3->queue_end = pktl;
477 return mp3_write_audio_packet(s, pkt);
481 /* warn only once for each stream */
482 if (s->streams[pkt->stream_index]->nb_frames == 1) {
483 av_log(s, AV_LOG_WARNING, "Got more than one picture in stream %d,"
484 " ignoring.\n", pkt->stream_index);
486 if (!mp3->pics_to_write || s->streams[pkt->stream_index]->nb_frames >= 1)
489 if ((ret = ff_id3v2_write_apic(s, &mp3->id3, pkt)) < 0)
491 mp3->pics_to_write--;
493 /* flush the buffered audio packets */
494 if (!mp3->pics_to_write &&
495 (ret = mp3_queue_flush(s)) < 0)
503 * Write an ID3v2 header at beginning of stream
506 static int mp3_write_header(struct AVFormatContext *s)
508 MP3Context *mp3 = s->priv_data;
511 if (mp3->id3v2_version &&
512 mp3->id3v2_version != 3 &&
513 mp3->id3v2_version != 4) {
514 av_log(s, AV_LOG_ERROR, "Invalid ID3v2 version requested: %d. Only "
515 "3, 4 or 0 (disabled) are allowed.\n", mp3->id3v2_version);
516 return AVERROR(EINVAL);
519 /* check the streams -- we want exactly one audio and arbitrary number of
520 * video (attached pictures) */
521 mp3->audio_stream_idx = -1;
522 for (i = 0; i < s->nb_streams; i++) {
523 AVStream *st = s->streams[i];
524 if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
525 if (mp3->audio_stream_idx >= 0 || st->codecpar->codec_id != AV_CODEC_ID_MP3) {
526 av_log(s, AV_LOG_ERROR, "Invalid audio stream. Exactly one MP3 "
527 "audio stream is required.\n");
528 return AVERROR(EINVAL);
530 mp3->audio_stream_idx = i;
531 } else if (st->codecpar->codec_type != AVMEDIA_TYPE_VIDEO) {
532 av_log(s, AV_LOG_ERROR, "Only audio streams and pictures are allowed in MP3.\n");
533 return AVERROR(EINVAL);
536 if (mp3->audio_stream_idx < 0) {
537 av_log(s, AV_LOG_ERROR, "No audio stream present.\n");
538 return AVERROR(EINVAL);
540 mp3->pics_to_write = s->nb_streams - 1;
542 if (mp3->pics_to_write && !mp3->id3v2_version) {
543 av_log(s, AV_LOG_ERROR, "Attached pictures were requested, but the "
544 "ID3v2 header is disabled.\n");
545 return AVERROR(EINVAL);
548 if (mp3->id3v2_version) {
549 ff_id3v2_start(&mp3->id3, s->pb, mp3->id3v2_version, ID3v2_DEFAULT_MAGIC);
550 ret = ff_id3v2_write_metadata(s, &mp3->id3);
555 if (!mp3->pics_to_write) {
556 if (mp3->id3v2_version)
557 ff_id3v2_finish(&mp3->id3, s->pb);
564 AVOutputFormat ff_mp3_muxer = {
566 .long_name = NULL_IF_CONFIG_SMALL("MP3 (MPEG audio layer 3)"),
567 .mime_type = "audio/mpeg",
569 .priv_data_size = sizeof(MP3Context),
570 .audio_codec = AV_CODEC_ID_MP3,
571 .video_codec = AV_CODEC_ID_PNG,
572 .write_header = mp3_write_header,
573 .write_packet = mp3_write_packet,
574 .write_trailer = mp3_write_trailer,
575 .flags = AVFMT_NOTIMESTAMPS,
576 .priv_class = &mp3_muxer_class,