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 "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 /* 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, "TDRC", 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
65 buf[126] = atoi(tag->value);
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])) {
81 #define XING_NUM_BAGS 400
82 #define XING_TOC_SIZE 100
83 // size of the XING/LAME data, starting from the Xing tag
86 typedef struct MP3Context {
94 // a buffer containing the whole XING/LAME frame
98 AVCRC audio_crc; // CRC of the audio data
99 uint32_t audio_size; // total size of the audio data
101 // offset of the XING/LAME frame in the file
102 int64_t xing_frame_offset;
103 // offset of the XING/INFO tag in the frame
111 uint64_t bag[XING_NUM_BAGS];
113 int has_variable_bitrate;
117 /* index of the audio stream */
118 int audio_stream_idx;
119 /* number of attached pictures we still need to write */
122 /* audio packets are queued here until we get all the attached pictures */
123 AVPacketList *queue, *queue_end;
126 static const uint8_t xing_offtbl[2][2] = {{32, 17}, {17, 9}};
129 * Write an empty XING header and initialize respective data.
131 static int mp3_write_xing(AVFormatContext *s)
133 MP3Context *mp3 = s->priv_data;
134 AVCodecParameters *par = s->streams[mp3->audio_stream_idx]->codecpar;
135 AVDictionaryEntry *enc = av_dict_get(s->streams[mp3->audio_stream_idx]->metadata, "encoder", NULL, 0);
136 AVIOContext *dyn_ctx;
138 MPADecodeHeader mpah;
139 int srate_idx, i, channels;
141 int best_bitrate_idx = -1;
142 int best_bitrate_error = INT_MAX;
147 if (!(s->pb->seekable & AVIO_SEEKABLE_NORMAL) || !mp3->write_xing)
150 for (i = 0; i < FF_ARRAY_ELEMS(avpriv_mpa_freq_tab); i++) {
151 const uint16_t base_freq = avpriv_mpa_freq_tab[i];
153 if (par->sample_rate == base_freq) ver = 0x3; // MPEG 1
154 else if (par->sample_rate == base_freq / 2) ver = 0x2; // MPEG 2
155 else if (par->sample_rate == base_freq / 4) ver = 0x0; // MPEG 2.5
161 if (i == FF_ARRAY_ELEMS(avpriv_mpa_freq_tab)) {
162 av_log(s, AV_LOG_WARNING, "Unsupported sample rate, not writing Xing header.\n");
166 switch (par->channels) {
167 case 1: channels = MPA_MONO; break;
168 case 2: channels = MPA_STEREO; break;
169 default: av_log(s, AV_LOG_WARNING, "Unsupported number of channels, "
170 "not writing Xing header.\n");
174 /* dummy MPEG audio header */
175 header = 0xffU << 24; // sync
176 header |= (0x7 << 5 | ver << 3 | 0x1 << 1 | 0x1) << 16; // sync/audio-version/layer 3/no crc*/
177 header |= (srate_idx << 2) << 8;
178 header |= channels << 6;
180 for (bitrate_idx = 1; bitrate_idx < 15; bitrate_idx++) {
181 int bit_rate = 1000 * avpriv_mpa_bitrate_tab[ver != 3][3 - 1][bitrate_idx];
182 int error = FFABS(bit_rate - par->bit_rate);
184 if (error < best_bitrate_error) {
185 best_bitrate_error = error;
186 best_bitrate_idx = bitrate_idx;
189 av_assert0(best_bitrate_idx >= 0);
191 for (bitrate_idx = best_bitrate_idx; ; bitrate_idx++) {
192 int32_t mask = bitrate_idx << (4 + 8);
193 if (15 == bitrate_idx)
197 ret = avpriv_mpegaudio_decode_header(&mpah, header);
198 av_assert0(ret >= 0);
199 mp3->xing_offset = xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1] + 4;
200 bytes_needed = mp3->xing_offset + XING_SIZE;
202 if (bytes_needed <= mpah.frame_size)
208 ret = avio_open_dyn_buf(&dyn_ctx);
212 avio_wb32(dyn_ctx, header);
214 ffio_fill(dyn_ctx, 0, mp3->xing_offset - 4);
215 ffio_wfourcc(dyn_ctx, "Xing");
216 avio_wb32(dyn_ctx, 0x01 | 0x02 | 0x04 | 0x08); // frames / size / TOC / vbr scale
218 mp3->size = mpah.frame_size;
223 avio_wb32(dyn_ctx, 0); // frames
224 avio_wb32(dyn_ctx, 0); // size
227 for (i = 0; i < XING_TOC_SIZE; i++)
228 avio_w8(dyn_ctx, (uint8_t)(255 * i / XING_TOC_SIZE));
231 // we write it, because some (broken) tools always expect it to be present
232 avio_wb32(dyn_ctx, 0);
234 // encoder short version string
236 uint8_t encoder_str[9] = { 0 };
237 if ( strlen(enc->value) > sizeof(encoder_str)
238 && !strcmp("Lavc libmp3lame", enc->value)) {
239 memcpy(encoder_str, "Lavf lame", 9);
241 memcpy(encoder_str, enc->value, FFMIN(strlen(enc->value), sizeof(encoder_str)));
243 avio_write(dyn_ctx, encoder_str, sizeof(encoder_str));
245 avio_write(dyn_ctx, "Lavf\0\0\0\0\0", 9);
247 avio_w8(dyn_ctx, 0); // tag revision 0 / unknown vbr method
248 avio_w8(dyn_ctx, 0); // unknown lowpass filter value
249 ffio_fill(dyn_ctx, 0, 8); // empty replaygain fields
250 avio_w8(dyn_ctx, 0); // unknown encoding flags
251 avio_w8(dyn_ctx, 0); // unknown abr/minimal bitrate
252 avio_wb24(dyn_ctx, 0); // empty encoder delay/padding
254 avio_w8(dyn_ctx, 0); // misc
255 avio_w8(dyn_ctx, 0); // mp3gain
256 avio_wb16(dyn_ctx, 0); // preset
258 // audio length and CRCs (will be updated later)
259 avio_wb32(dyn_ctx, 0); // music length
260 avio_wb16(dyn_ctx, 0); // music crc
261 avio_wb16(dyn_ctx, 0); // tag crc
263 ffio_fill(dyn_ctx, 0, mpah.frame_size - bytes_needed);
265 mp3->xing_frame_size = avio_close_dyn_buf(dyn_ctx, &mp3->xing_frame);
266 mp3->xing_frame_offset = avio_tell(s->pb);
267 avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size);
269 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 >> 1] = 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 (pkt->data && pkt->size >= 4) {
309 MPADecodeHeader mpah;
314 h = AV_RB32(pkt->data);
315 ret = avpriv_mpegaudio_decode_header(&mpah, h);
317 if (!mp3->initial_bitrate)
318 mp3->initial_bitrate = mpah.bit_rate;
319 if ((mpah.bit_rate == 0) || (mp3->initial_bitrate != mpah.bit_rate))
320 mp3->has_variable_bitrate = 1;
322 av_log(s, AV_LOG_WARNING, "Audio packet of size %d (starting with %08"PRIX32"...) "
323 "is invalid, writing it anyway.\n", pkt->size, h);
326 #ifdef FILTER_VBR_HEADERS
327 /* filter out XING and INFO headers. */
328 base = 4 + xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1];
330 if (base + 4 <= pkt->size) {
331 uint32_t v = AV_RB32(pkt->data + base);
333 if (MKBETAG('X','i','n','g') == v || MKBETAG('I','n','f','o') == v)
337 /* filter out VBRI headers. */
340 if (base + 4 <= pkt->size && MKBETAG('V','B','R','I') == AV_RB32(pkt->data + base))
344 if (mp3->xing_offset) {
345 uint8_t *side_data = NULL;
346 int side_data_size = 0;
348 mp3_xing_add_frame(mp3, pkt);
349 mp3->audio_size += pkt->size;
350 mp3->audio_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE),
351 mp3->audio_crc, pkt->data, pkt->size);
353 side_data = av_packet_get_side_data(pkt,
354 AV_PKT_DATA_SKIP_SAMPLES,
356 if (side_data && side_data_size >= 10) {
357 mp3->padding = FFMAX(AV_RL32(side_data + 4) + 528 + 1, 0);
359 mp3->delay = FFMAX(AV_RL32(side_data) - 528 - 1, 0);
366 return ff_raw_write_packet(s, pkt);
369 static int mp3_queue_flush(AVFormatContext *s)
371 MP3Context *mp3 = s->priv_data;
373 int ret = 0, write = 1;
375 ff_id3v2_finish(&mp3->id3, s->pb, s->metadata_header_padding);
379 ff_packet_list_get(&mp3->queue, &mp3->queue_end, &pkt);
380 if (write && (ret = mp3_write_audio_packet(s, &pkt)) < 0)
382 av_packet_unref(&pkt);
387 static void mp3_update_xing(AVFormatContext *s)
389 MP3Context *mp3 = s->priv_data;
394 int64_t old_pos = avio_tell(s->pb);
396 /* replace "Xing" identification string with "Info" for CBR files. */
397 if (!mp3->has_variable_bitrate)
398 AV_WL32(mp3->xing_frame + mp3->xing_offset, MKTAG('I', 'n', 'f', 'o'));
400 AV_WB32(mp3->xing_frame + mp3->xing_offset + 8, mp3->frames);
401 AV_WB32(mp3->xing_frame + mp3->xing_offset + 12, mp3->size);
403 toc = mp3->xing_frame + mp3->xing_offset + 16;
404 toc[0] = 0; // first toc entry has to be zero.
405 for (i = 1; i < XING_TOC_SIZE; ++i) {
406 int j = i * mp3->pos / XING_TOC_SIZE;
407 int seek_point = 256LL * mp3->bag[j] / mp3->size;
408 toc[i] = FFMIN(seek_point, 255);
411 /* write replaygain */
412 rg = (AVReplayGain*)av_stream_get_side_data(s->streams[0], AV_PKT_DATA_REPLAYGAIN,
414 if (rg && rg_size >= sizeof(*rg)) {
417 AV_WB32(mp3->xing_frame + mp3->xing_offset + 131,
418 av_rescale(rg->track_peak, 1 << 23, 100000));
420 if (rg->track_gain != INT32_MIN) {
421 val = FFABS(rg->track_gain / 10000) & ((1 << 9) - 1);
422 val |= (rg->track_gain < 0) << 9;
424 AV_WB16(mp3->xing_frame + mp3->xing_offset + 135, val);
427 if (rg->album_gain != INT32_MIN) {
428 val = FFABS(rg->album_gain / 10000) & ((1 << 9) - 1);
429 val |= (rg->album_gain < 0) << 9;
431 AV_WB16(mp3->xing_frame + mp3->xing_offset + 137, val);
435 /* write encoder delay/padding */
436 if (mp3->delay >= 1 << 12) {
437 mp3->delay = (1 << 12) - 1;
438 av_log(s, AV_LOG_WARNING, "Too many samples of initial padding.\n");
440 if (mp3->padding >= 1 << 12) {
441 mp3->padding = (1 << 12) - 1;
442 av_log(s, AV_LOG_WARNING, "Too many samples of trailing padding.\n");
444 AV_WB24(mp3->xing_frame + mp3->xing_offset + 141, (mp3->delay << 12) + mp3->padding);
446 AV_WB32(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 8, mp3->audio_size);
447 AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 4, mp3->audio_crc);
449 tag_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE), 0, mp3->xing_frame, 190);
450 AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 2, tag_crc);
452 avio_seek(s->pb, mp3->xing_frame_offset, SEEK_SET);
453 avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size);
454 avio_seek(s->pb, old_pos, SEEK_SET);
457 static int mp3_write_trailer(struct AVFormatContext *s)
459 uint8_t buf[ID3v1_TAG_SIZE];
460 MP3Context *mp3 = s->priv_data;
462 if (mp3->pics_to_write) {
463 av_log(s, AV_LOG_WARNING, "No packets were sent for some of the "
464 "attached pictures.\n");
468 /* write the id3v1 tag */
469 if (mp3->write_id3v1 && id3v1_create_tag(s, buf) > 0) {
470 avio_write(s->pb, buf, ID3v1_TAG_SIZE);
473 if (mp3->xing_offset)
479 static int query_codec(enum AVCodecID id, int std_compliance)
481 const CodecMime *cm= ff_id3v2_mime_tags;
482 while(cm->id != AV_CODEC_ID_NONE) {
484 return MKTAG('A', 'P', 'I', 'C');
490 static const AVOption options[] = {
491 { "id3v2_version", "Select ID3v2 version to write. Currently 3 and 4 are supported.",
492 offsetof(MP3Context, id3v2_version), AV_OPT_TYPE_INT, {.i64 = 4}, 0, 4, AV_OPT_FLAG_ENCODING_PARAM},
493 { "write_id3v1", "Enable ID3v1 writing. ID3v1 tags are written in UTF-8 which may not be supported by most software.",
494 offsetof(MP3Context, write_id3v1), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
495 { "write_xing", "Write the Xing header containing file duration.",
496 offsetof(MP3Context, write_xing), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
500 static const AVClass mp3_muxer_class = {
501 .class_name = "MP3 muxer",
502 .item_name = av_default_item_name,
504 .version = LIBAVUTIL_VERSION_INT,
507 static int mp3_write_packet(AVFormatContext *s, AVPacket *pkt)
509 MP3Context *mp3 = s->priv_data;
511 if (pkt->stream_index == mp3->audio_stream_idx) {
512 if (mp3->pics_to_write) {
513 /* buffer audio packets until we get all the pictures */
514 int ret = ff_packet_list_put(&mp3->queue, &mp3->queue_end, pkt, FF_PACKETLIST_FLAG_REF_PACKET);
517 av_log(s, AV_LOG_WARNING, "Not enough memory to buffer audio. Skipping picture streams\n");
518 mp3->pics_to_write = 0;
520 return mp3_write_audio_packet(s, pkt);
523 return mp3_write_audio_packet(s, pkt);
527 /* warn only once for each stream */
528 if (s->streams[pkt->stream_index]->nb_frames == 1) {
529 av_log(s, AV_LOG_WARNING, "Got more than one picture in stream %d,"
530 " ignoring.\n", pkt->stream_index);
532 if (!mp3->pics_to_write || s->streams[pkt->stream_index]->nb_frames >= 1)
535 if ((ret = ff_id3v2_write_apic(s, &mp3->id3, pkt)) < 0)
537 mp3->pics_to_write--;
539 /* flush the buffered audio packets */
540 if (!mp3->pics_to_write &&
541 (ret = mp3_queue_flush(s)) < 0)
549 * Write an ID3v2 header at beginning of stream
552 static int mp3_init(struct AVFormatContext *s)
554 MP3Context *mp3 = s->priv_data;
557 if (mp3->id3v2_version &&
558 mp3->id3v2_version != 3 &&
559 mp3->id3v2_version != 4) {
560 av_log(s, AV_LOG_ERROR, "Invalid ID3v2 version requested: %d. Only "
561 "3, 4 or 0 (disabled) are allowed.\n", mp3->id3v2_version);
562 return AVERROR(EINVAL);
565 /* check the streams -- we want exactly one audio and arbitrary number of
566 * video (attached pictures) */
567 mp3->audio_stream_idx = -1;
568 for (i = 0; i < s->nb_streams; i++) {
569 AVStream *st = s->streams[i];
570 if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
571 if (mp3->audio_stream_idx >= 0 || st->codecpar->codec_id != AV_CODEC_ID_MP3) {
572 av_log(s, AV_LOG_ERROR, "Invalid audio stream. Exactly one MP3 "
573 "audio stream is required.\n");
574 return AVERROR(EINVAL);
576 mp3->audio_stream_idx = i;
577 } else if (st->codecpar->codec_type != AVMEDIA_TYPE_VIDEO) {
578 av_log(s, AV_LOG_ERROR, "Only audio streams and pictures are allowed in MP3.\n");
579 return AVERROR(EINVAL);
582 if (mp3->audio_stream_idx < 0) {
583 av_log(s, AV_LOG_ERROR, "No audio stream present.\n");
584 return AVERROR(EINVAL);
586 mp3->pics_to_write = s->nb_streams - 1;
588 if (mp3->pics_to_write && !mp3->id3v2_version) {
589 av_log(s, AV_LOG_ERROR, "Attached pictures were requested, but the "
590 "ID3v2 header is disabled.\n");
591 return AVERROR(EINVAL);
597 static int mp3_write_header(struct AVFormatContext *s)
599 MP3Context *mp3 = s->priv_data;
602 if (mp3->id3v2_version) {
603 ff_id3v2_start(&mp3->id3, s->pb, mp3->id3v2_version, ID3v2_DEFAULT_MAGIC);
604 ret = ff_id3v2_write_metadata(s, &mp3->id3);
609 if (!mp3->pics_to_write) {
610 if (mp3->id3v2_version)
611 ff_id3v2_finish(&mp3->id3, s->pb, s->metadata_header_padding);
618 static void mp3_deinit(struct AVFormatContext *s)
620 MP3Context *mp3 = s->priv_data;
622 ff_packet_list_free(&mp3->queue, &mp3->queue_end);
623 av_freep(&mp3->xing_frame);
626 AVOutputFormat ff_mp3_muxer = {
628 .long_name = NULL_IF_CONFIG_SMALL("MP3 (MPEG audio layer 3)"),
629 .mime_type = "audio/mpeg",
631 .priv_data_size = sizeof(MP3Context),
632 .audio_codec = AV_CODEC_ID_MP3,
633 .video_codec = AV_CODEC_ID_PNG,
635 .write_header = mp3_write_header,
636 .write_packet = mp3_write_packet,
637 .write_trailer = mp3_write_trailer,
638 .deinit = mp3_deinit,
639 .query_codec = query_codec,
640 .flags = AVFMT_NOTIMESTAMPS,
641 .priv_class = &mp3_muxer_class,