3 * Copyright (c) 2003 Fabrice Bellard
5 * This file is part of FFmpeg.
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.
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.
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
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 "libavcodec/packet_internal.h"
32 #include "libavutil/intreadwrite.h"
33 #include "libavutil/opt.h"
34 #include "libavutil/dict.h"
35 #include "libavutil/avassert.h"
36 #include "libavutil/crc.h"
37 #include "libavutil/mathematics.h"
38 #include "libavutil/replaygain.h"
40 static int id3v1_set_string(AVFormatContext *s, const char *key,
41 uint8_t *buf, int buf_size)
43 AVDictionaryEntry *tag;
44 if ((tag = av_dict_get(s->metadata, key, NULL, 0)))
45 av_strlcpy(buf, tag->value, buf_size);
49 // refer to: http://id3.org/ID3v1
50 static int id3v1_create_tag(AVFormatContext *s, uint8_t *buf)
52 AVDictionaryEntry *tag;
55 memset(buf, 0, ID3v1_TAG_SIZE); /* fail safe */
59 /* we knowingly overspecify each tag length by one byte to compensate for the mandatory null byte added by av_strlcpy */
60 count += id3v1_set_string(s, "TIT2", buf + 3, 30 + 1); //title
61 count += id3v1_set_string(s, "TPE1", buf + 33, 30 + 1); //author|artist
62 count += id3v1_set_string(s, "TALB", buf + 63, 30 + 1); //album
63 if ((tag = av_dict_get(s->metadata, "TYER", NULL, 0))) { //year
64 av_strlcpy(buf + 93, tag->value, 4 + 1);
66 } else if ((tag = av_dict_get(s->metadata, "TDRC", NULL, 0))) {
67 av_strlcpy(buf + 93, tag->value, 4 + 1);
69 } else if ((tag = av_dict_get(s->metadata, "TDAT", NULL, 0))) {
70 av_strlcpy(buf + 93, tag->value, 4 + 1);
74 count += id3v1_set_string(s, "comment", buf + 97, 30 + 1);
75 if ((tag = av_dict_get(s->metadata, "TRCK", NULL, 0))) { //track
77 buf[126] = atoi(tag->value);
80 buf[127] = 0xFF; /* default to unknown genre */
81 if ((tag = av_dict_get(s->metadata, "TCON", NULL, 0))) { //genre
82 for(i = 0; i <= ID3v1_GENRE_MAX; i++) {
83 if (!av_strcasecmp(tag->value, ff_id3v1_genre_str[i])) {
93 #define XING_NUM_BAGS 400
94 #define XING_TOC_SIZE 100
95 // size of the XING/LAME data, starting from the Xing tag
98 typedef struct MP3Context {
106 // a buffer containing the whole XING/LAME frame
110 AVCRC audio_crc; // CRC of the audio data
111 uint32_t audio_size; // total size of the audio data
113 // offset of the XING/LAME frame in the file
114 int64_t xing_frame_offset;
115 // offset of the XING/INFO tag in the frame
123 uint64_t bag[XING_NUM_BAGS];
125 int has_variable_bitrate;
129 /* index of the audio stream */
130 int audio_stream_idx;
131 /* number of attached pictures we still need to write */
134 /* audio packets are queued here until we get all the attached pictures */
135 AVPacketList *queue, *queue_end;
138 static const uint8_t xing_offtbl[2][2] = {{32, 17}, {17, 9}};
141 * Write an empty XING header and initialize respective data.
143 static int mp3_write_xing(AVFormatContext *s)
145 MP3Context *mp3 = s->priv_data;
146 AVCodecParameters *par = s->streams[mp3->audio_stream_idx]->codecpar;
147 AVDictionaryEntry *enc = av_dict_get(s->streams[mp3->audio_stream_idx]->metadata, "encoder", NULL, 0);
148 AVIOContext *dyn_ctx;
150 MPADecodeHeader mpah;
151 int srate_idx, i, channels;
153 int best_bitrate_idx = -1;
154 int best_bitrate_error = INT_MAX;
159 if (!(s->pb->seekable & AVIO_SEEKABLE_NORMAL) || !mp3->write_xing)
162 for (i = 0; i < FF_ARRAY_ELEMS(avpriv_mpa_freq_tab); i++) {
163 const uint16_t base_freq = avpriv_mpa_freq_tab[i];
165 if (par->sample_rate == base_freq) ver = 0x3; // MPEG 1
166 else if (par->sample_rate == base_freq / 2) ver = 0x2; // MPEG 2
167 else if (par->sample_rate == base_freq / 4) ver = 0x0; // MPEG 2.5
173 if (i == FF_ARRAY_ELEMS(avpriv_mpa_freq_tab)) {
174 av_log(s, AV_LOG_WARNING, "Unsupported sample rate, not writing Xing header.\n");
178 switch (par->channels) {
179 case 1: channels = MPA_MONO; break;
180 case 2: channels = MPA_STEREO; break;
181 default: av_log(s, AV_LOG_WARNING, "Unsupported number of channels, "
182 "not writing Xing header.\n");
186 /* dummy MPEG audio header */
187 header = 0xffU << 24; // sync
188 header |= (0x7 << 5 | ver << 3 | 0x1 << 1 | 0x1) << 16; // sync/audio-version/layer 3/no crc*/
189 header |= (srate_idx << 2) << 8;
190 header |= channels << 6;
192 for (bitrate_idx = 1; bitrate_idx < 15; bitrate_idx++) {
193 int bit_rate = 1000 * avpriv_mpa_bitrate_tab[ver != 3][3 - 1][bitrate_idx];
194 int error = FFABS(bit_rate - par->bit_rate);
196 if (error < best_bitrate_error) {
197 best_bitrate_error = error;
198 best_bitrate_idx = bitrate_idx;
201 av_assert0(best_bitrate_idx >= 0);
203 for (bitrate_idx = best_bitrate_idx; ; bitrate_idx++) {
204 int32_t mask = bitrate_idx << (4 + 8);
205 if (15 == bitrate_idx)
209 ret = avpriv_mpegaudio_decode_header(&mpah, header);
210 av_assert0(ret >= 0);
211 mp3->xing_offset = xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1] + 4;
212 bytes_needed = mp3->xing_offset + XING_SIZE;
214 if (bytes_needed <= mpah.frame_size)
220 ret = avio_open_dyn_buf(&dyn_ctx);
224 avio_wb32(dyn_ctx, header);
226 ffio_fill(dyn_ctx, 0, mp3->xing_offset - 4);
227 ffio_wfourcc(dyn_ctx, "Xing");
228 avio_wb32(dyn_ctx, 0x01 | 0x02 | 0x04 | 0x08); // frames / size / TOC / vbr scale
230 mp3->size = mpah.frame_size;
235 avio_wb32(dyn_ctx, 0); // frames
236 avio_wb32(dyn_ctx, 0); // size
239 for (i = 0; i < XING_TOC_SIZE; i++)
240 avio_w8(dyn_ctx, (uint8_t)(255 * i / XING_TOC_SIZE));
243 // we write it, because some (broken) tools always expect it to be present
244 avio_wb32(dyn_ctx, 0);
246 // encoder short version string
248 uint8_t encoder_str[9] = { 0 };
249 if ( strlen(enc->value) > sizeof(encoder_str)
250 && !strcmp("Lavc libmp3lame", enc->value)) {
251 memcpy(encoder_str, "Lavf lame", 9);
253 memcpy(encoder_str, enc->value, FFMIN(strlen(enc->value), sizeof(encoder_str)));
255 avio_write(dyn_ctx, encoder_str, sizeof(encoder_str));
257 avio_write(dyn_ctx, "Lavf\0\0\0\0\0", 9);
259 avio_w8(dyn_ctx, 0); // tag revision 0 / unknown vbr method
260 avio_w8(dyn_ctx, 0); // unknown lowpass filter value
261 ffio_fill(dyn_ctx, 0, 8); // empty replaygain fields
262 avio_w8(dyn_ctx, 0); // unknown encoding flags
263 avio_w8(dyn_ctx, 0); // unknown abr/minimal bitrate
264 avio_wb24(dyn_ctx, 0); // empty encoder delay/padding
266 avio_w8(dyn_ctx, 0); // misc
267 avio_w8(dyn_ctx, 0); // mp3gain
268 avio_wb16(dyn_ctx, 0); // preset
270 // audio length and CRCs (will be updated later)
271 avio_wb32(dyn_ctx, 0); // music length
272 avio_wb16(dyn_ctx, 0); // music crc
273 avio_wb16(dyn_ctx, 0); // tag crc
275 ffio_fill(dyn_ctx, 0, mpah.frame_size - bytes_needed);
277 mp3->xing_frame_size = avio_close_dyn_buf(dyn_ctx, &mp3->xing_frame);
278 mp3->xing_frame_offset = avio_tell(s->pb);
279 avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size);
281 mp3->audio_size = mp3->xing_frame_size;
287 * Add a frame to XING data.
288 * Following lame's "VbrTag.c".
290 static void mp3_xing_add_frame(MP3Context *mp3, AVPacket *pkt)
296 mp3->size += pkt->size;
298 if (mp3->want == mp3->seen) {
299 mp3->bag[mp3->pos] = mp3->size;
301 if (XING_NUM_BAGS == ++mp3->pos) {
302 /* shrink table to half size by throwing away each second bag. */
303 for (i = 1; i < XING_NUM_BAGS; i += 2)
304 mp3->bag[i >> 1] = mp3->bag[i];
306 /* double wanted amount per bag. */
308 /* adjust current position to half of table size. */
309 mp3->pos = XING_NUM_BAGS / 2;
316 static int mp3_write_audio_packet(AVFormatContext *s, AVPacket *pkt)
318 MP3Context *mp3 = s->priv_data;
320 if (pkt->data && pkt->size >= 4) {
321 MPADecodeHeader mpah;
326 h = AV_RB32(pkt->data);
327 ret = avpriv_mpegaudio_decode_header(&mpah, h);
329 if (!mp3->initial_bitrate)
330 mp3->initial_bitrate = mpah.bit_rate;
331 if ((mpah.bit_rate == 0) || (mp3->initial_bitrate != mpah.bit_rate))
332 mp3->has_variable_bitrate = 1;
334 av_log(s, AV_LOG_WARNING, "Audio packet of size %d (starting with %08"PRIX32"...) "
335 "is invalid, writing it anyway.\n", pkt->size, h);
338 #ifdef FILTER_VBR_HEADERS
339 /* filter out XING and INFO headers. */
340 base = 4 + xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1];
342 if (base + 4 <= pkt->size) {
343 uint32_t v = AV_RB32(pkt->data + base);
345 if (MKBETAG('X','i','n','g') == v || MKBETAG('I','n','f','o') == v)
349 /* filter out VBRI headers. */
352 if (base + 4 <= pkt->size && MKBETAG('V','B','R','I') == AV_RB32(pkt->data + base))
356 if (mp3->xing_offset) {
357 uint8_t *side_data = NULL;
360 mp3_xing_add_frame(mp3, pkt);
361 mp3->audio_size += pkt->size;
362 mp3->audio_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE),
363 mp3->audio_crc, pkt->data, pkt->size);
365 side_data = av_packet_get_side_data(pkt,
366 AV_PKT_DATA_SKIP_SAMPLES,
368 if (side_data && side_data_size >= 10) {
369 mp3->padding = FFMAX(AV_RL32(side_data + 4) + 528 + 1, 0);
371 mp3->delay = FFMAX(AV_RL32(side_data) - 528 - 1, 0);
378 return ff_raw_write_packet(s, pkt);
381 static int mp3_queue_flush(AVFormatContext *s)
383 MP3Context *mp3 = s->priv_data;
385 int ret = 0, write = 1;
387 ff_id3v2_finish(&mp3->id3, s->pb, s->metadata_header_padding);
391 avpriv_packet_list_get(&mp3->queue, &mp3->queue_end, &pkt);
392 if (write && (ret = mp3_write_audio_packet(s, &pkt)) < 0)
394 av_packet_unref(&pkt);
399 static void mp3_update_xing(AVFormatContext *s)
401 MP3Context *mp3 = s->priv_data;
406 int64_t old_pos = avio_tell(s->pb);
408 /* replace "Xing" identification string with "Info" for CBR files. */
409 if (!mp3->has_variable_bitrate)
410 AV_WL32(mp3->xing_frame + mp3->xing_offset, MKTAG('I', 'n', 'f', 'o'));
412 AV_WB32(mp3->xing_frame + mp3->xing_offset + 8, mp3->frames);
413 AV_WB32(mp3->xing_frame + mp3->xing_offset + 12, mp3->size);
415 toc = mp3->xing_frame + mp3->xing_offset + 16;
416 toc[0] = 0; // first toc entry has to be zero.
417 for (i = 1; i < XING_TOC_SIZE; ++i) {
418 int j = i * mp3->pos / XING_TOC_SIZE;
419 int seek_point = 256LL * mp3->bag[j] / mp3->size;
420 toc[i] = FFMIN(seek_point, 255);
423 /* write replaygain */
424 rg = (AVReplayGain*)av_stream_get_side_data(s->streams[0], AV_PKT_DATA_REPLAYGAIN,
426 if (rg && rg_size >= sizeof(*rg)) {
429 AV_WB32(mp3->xing_frame + mp3->xing_offset + 131,
430 av_rescale(rg->track_peak, 1 << 23, 100000));
432 if (rg->track_gain != INT32_MIN) {
433 val = FFABS(rg->track_gain / 10000) & ((1 << 9) - 1);
434 val |= (rg->track_gain < 0) << 9;
436 AV_WB16(mp3->xing_frame + mp3->xing_offset + 135, val);
439 if (rg->album_gain != INT32_MIN) {
440 val = FFABS(rg->album_gain / 10000) & ((1 << 9) - 1);
441 val |= (rg->album_gain < 0) << 9;
443 AV_WB16(mp3->xing_frame + mp3->xing_offset + 137, val);
447 /* write encoder delay/padding */
448 if (mp3->delay >= 1 << 12) {
449 mp3->delay = (1 << 12) - 1;
450 av_log(s, AV_LOG_WARNING, "Too many samples of initial padding.\n");
452 if (mp3->padding >= 1 << 12) {
453 mp3->padding = (1 << 12) - 1;
454 av_log(s, AV_LOG_WARNING, "Too many samples of trailing padding.\n");
456 AV_WB24(mp3->xing_frame + mp3->xing_offset + 141, (mp3->delay << 12) + mp3->padding);
458 AV_WB32(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 8, mp3->audio_size);
459 AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 4, mp3->audio_crc);
461 tag_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE), 0, mp3->xing_frame, 190);
462 AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 2, tag_crc);
464 avio_seek(s->pb, mp3->xing_frame_offset, SEEK_SET);
465 avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size);
466 avio_seek(s->pb, old_pos, SEEK_SET);
469 static int mp3_write_trailer(struct AVFormatContext *s)
471 uint8_t buf[ID3v1_TAG_SIZE];
472 MP3Context *mp3 = s->priv_data;
474 if (mp3->pics_to_write) {
475 av_log(s, AV_LOG_WARNING, "No packets were sent for some of the "
476 "attached pictures.\n");
480 /* write the id3v1 tag */
481 if (mp3->write_id3v1 && id3v1_create_tag(s, buf) > 0) {
482 avio_write(s->pb, buf, ID3v1_TAG_SIZE);
485 if (mp3->xing_offset)
491 static int query_codec(enum AVCodecID id, int std_compliance)
493 const CodecMime *cm= ff_id3v2_mime_tags;
494 while(cm->id != AV_CODEC_ID_NONE) {
496 return MKTAG('A', 'P', 'I', 'C');
502 static const AVOption options[] = {
503 { "id3v2_version", "Select ID3v2 version to write. Currently 3 and 4 are supported.",
504 offsetof(MP3Context, id3v2_version), AV_OPT_TYPE_INT, {.i64 = 4}, 0, 4, AV_OPT_FLAG_ENCODING_PARAM},
505 { "write_id3v1", "Enable ID3v1 writing. ID3v1 tags are written in UTF-8 which may not be supported by most software.",
506 offsetof(MP3Context, write_id3v1), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
507 { "write_xing", "Write the Xing header containing file duration.",
508 offsetof(MP3Context, write_xing), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
512 static const AVClass mp3_muxer_class = {
513 .class_name = "MP3 muxer",
514 .item_name = av_default_item_name,
516 .version = LIBAVUTIL_VERSION_INT,
519 static int mp3_write_packet(AVFormatContext *s, AVPacket *pkt)
521 MP3Context *mp3 = s->priv_data;
523 if (pkt->stream_index == mp3->audio_stream_idx) {
524 if (mp3->pics_to_write) {
525 /* buffer audio packets until we get all the pictures */
526 int ret = avpriv_packet_list_put(&mp3->queue, &mp3->queue_end, pkt, av_packet_ref, 0);
529 av_log(s, AV_LOG_WARNING, "Not enough memory to buffer audio. Skipping picture streams\n");
530 mp3->pics_to_write = 0;
532 return mp3_write_audio_packet(s, pkt);
535 return mp3_write_audio_packet(s, pkt);
539 /* warn only once for each stream */
540 if (s->streams[pkt->stream_index]->nb_frames == 1) {
541 av_log(s, AV_LOG_WARNING, "Got more than one picture in stream %d,"
542 " ignoring.\n", pkt->stream_index);
544 if (!mp3->pics_to_write || s->streams[pkt->stream_index]->nb_frames >= 1)
547 if ((ret = ff_id3v2_write_apic(s, &mp3->id3, pkt)) < 0)
549 mp3->pics_to_write--;
551 /* flush the buffered audio packets */
552 if (!mp3->pics_to_write &&
553 (ret = mp3_queue_flush(s)) < 0)
561 * Write an ID3v2 header at beginning of stream
564 static int mp3_init(struct AVFormatContext *s)
566 MP3Context *mp3 = s->priv_data;
569 if (mp3->id3v2_version &&
570 mp3->id3v2_version != 3 &&
571 mp3->id3v2_version != 4) {
572 av_log(s, AV_LOG_ERROR, "Invalid ID3v2 version requested: %d. Only "
573 "3, 4 or 0 (disabled) are allowed.\n", mp3->id3v2_version);
574 return AVERROR(EINVAL);
577 /* check the streams -- we want exactly one audio and arbitrary number of
578 * video (attached pictures) */
579 mp3->audio_stream_idx = -1;
580 for (i = 0; i < s->nb_streams; i++) {
581 AVStream *st = s->streams[i];
582 if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
583 if (mp3->audio_stream_idx >= 0 || st->codecpar->codec_id != AV_CODEC_ID_MP3) {
584 av_log(s, AV_LOG_ERROR, "Invalid audio stream. Exactly one MP3 "
585 "audio stream is required.\n");
586 return AVERROR(EINVAL);
588 mp3->audio_stream_idx = i;
589 } else if (st->codecpar->codec_type != AVMEDIA_TYPE_VIDEO) {
590 av_log(s, AV_LOG_ERROR, "Only audio streams and pictures are allowed in MP3.\n");
591 return AVERROR(EINVAL);
594 if (mp3->audio_stream_idx < 0) {
595 av_log(s, AV_LOG_ERROR, "No audio stream present.\n");
596 return AVERROR(EINVAL);
598 mp3->pics_to_write = s->nb_streams - 1;
600 if (mp3->pics_to_write && !mp3->id3v2_version) {
601 av_log(s, AV_LOG_ERROR, "Attached pictures were requested, but the "
602 "ID3v2 header is disabled.\n");
603 return AVERROR(EINVAL);
609 static int mp3_write_header(struct AVFormatContext *s)
611 MP3Context *mp3 = s->priv_data;
614 if (mp3->id3v2_version) {
615 ff_id3v2_start(&mp3->id3, s->pb, mp3->id3v2_version, ID3v2_DEFAULT_MAGIC);
616 ret = ff_id3v2_write_metadata(s, &mp3->id3);
621 if (!mp3->pics_to_write) {
622 if (mp3->id3v2_version)
623 ff_id3v2_finish(&mp3->id3, s->pb, s->metadata_header_padding);
630 static void mp3_deinit(struct AVFormatContext *s)
632 MP3Context *mp3 = s->priv_data;
634 avpriv_packet_list_free(&mp3->queue, &mp3->queue_end);
635 av_freep(&mp3->xing_frame);
638 AVOutputFormat ff_mp3_muxer = {
640 .long_name = NULL_IF_CONFIG_SMALL("MP3 (MPEG audio layer 3)"),
641 .mime_type = "audio/mpeg",
643 .priv_data_size = sizeof(MP3Context),
644 .audio_codec = AV_CODEC_ID_MP3,
645 .video_codec = AV_CODEC_ID_PNG,
647 .write_header = mp3_write_header,
648 .write_packet = mp3_write_packet,
649 .write_trailer = mp3_write_trailer,
650 .deinit = mp3_deinit,
651 .query_codec = query_codec,
652 .flags = AVFMT_NOTIMESTAMPS,
653 .priv_class = &mp3_muxer_class,