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;
115 /* index of the audio stream */
116 int audio_stream_idx;
117 /* number of attached pictures we still need to write */
120 /* audio packets are queued here until we get all the attached pictures */
121 AVPacketList *queue, *queue_end;
124 static const uint8_t xing_offtbl[2][2] = {{32, 17}, {17, 9}};
127 * Write an empty XING header and initialize respective data.
129 static int mp3_write_xing(AVFormatContext *s)
131 MP3Context *mp3 = s->priv_data;
132 AVCodecParameters *par = s->streams[mp3->audio_stream_idx]->codecpar;
133 AVDictionaryEntry *enc = av_dict_get(s->streams[mp3->audio_stream_idx]->metadata, "encoder", NULL, 0);
134 AVIOContext *dyn_ctx;
136 MPADecodeHeader mpah;
137 int srate_idx, i, channels;
139 int best_bitrate_idx = -1;
140 int best_bitrate_error = INT_MAX;
145 if (!s->pb->seekable || !mp3->write_xing)
148 for (i = 0; i < FF_ARRAY_ELEMS(avpriv_mpa_freq_tab); i++) {
149 const uint16_t base_freq = avpriv_mpa_freq_tab[i];
151 if (par->sample_rate == base_freq) ver = 0x3; // MPEG 1
152 else if (par->sample_rate == base_freq / 2) ver = 0x2; // MPEG 2
153 else if (par->sample_rate == base_freq / 4) ver = 0x0; // MPEG 2.5
159 if (i == FF_ARRAY_ELEMS(avpriv_mpa_freq_tab)) {
160 av_log(s, AV_LOG_WARNING, "Unsupported sample rate, not writing Xing header.\n");
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 = 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;
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 - par->bit_rate);
182 if (error < best_bitrate_error) {
183 best_bitrate_error = error;
184 best_bitrate_idx = bitrate_idx;
187 av_assert0(best_bitrate_idx >= 0);
189 for (bitrate_idx = best_bitrate_idx; ; bitrate_idx++) {
190 int32_t mask = bitrate_idx << (4 + 8);
191 if (15 == bitrate_idx)
195 ret = avpriv_mpegaudio_decode_header(&mpah, header);
196 av_assert0(ret >= 0);
197 mp3->xing_offset = xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1] + 4;
198 bytes_needed = mp3->xing_offset + XING_SIZE;
200 if (bytes_needed <= mpah.frame_size)
206 ret = avio_open_dyn_buf(&dyn_ctx);
210 avio_wb32(dyn_ctx, header);
212 ffio_fill(dyn_ctx, 0, mp3->xing_offset - 4);
213 ffio_wfourcc(dyn_ctx, "Xing");
214 avio_wb32(dyn_ctx, 0x01 | 0x02 | 0x04 | 0x08); // frames / size / TOC / vbr scale
216 mp3->size = mpah.frame_size;
221 avio_wb32(dyn_ctx, 0); // frames
222 avio_wb32(dyn_ctx, 0); // size
225 for (i = 0; i < XING_TOC_SIZE; i++)
226 avio_w8(dyn_ctx, (uint8_t)(255 * i / XING_TOC_SIZE));
229 // we write it, because some (broken) tools always expect it to be present
230 avio_wb32(dyn_ctx, 0);
232 // encoder short version string
234 uint8_t encoder_str[9] = { 0 };
235 if ( strlen(enc->value) > sizeof(encoder_str)
236 && !strcmp("Lavc libmp3lame", enc->value)) {
237 memcpy(encoder_str, "Lavf lame", 9);
239 memcpy(encoder_str, enc->value, FFMIN(strlen(enc->value), sizeof(encoder_str)));
241 avio_write(dyn_ctx, encoder_str, sizeof(encoder_str));
243 avio_write(dyn_ctx, "Lavf\0\0\0\0\0", 9);
245 avio_w8(dyn_ctx, 0); // tag revision 0 / unknown vbr method
246 avio_w8(dyn_ctx, 0); // unknown lowpass filter value
247 ffio_fill(dyn_ctx, 0, 8); // empty replaygain fields
248 avio_w8(dyn_ctx, 0); // unknown encoding flags
249 avio_w8(dyn_ctx, 0); // unknown abr/minimal bitrate
252 if (par->initial_padding - 528 - 1 >= 1 << 12) {
253 av_log(s, AV_LOG_WARNING, "Too many samples of initial padding.\n");
255 avio_wb24(dyn_ctx, FFMAX(par->initial_padding - 528 - 1, 0)<<12);
257 avio_w8(dyn_ctx, 0); // misc
258 avio_w8(dyn_ctx, 0); // mp3gain
259 avio_wb16(dyn_ctx, 0); // preset
261 // audio length and CRCs (will be updated later)
262 avio_wb32(dyn_ctx, 0); // music length
263 avio_wb16(dyn_ctx, 0); // music crc
264 avio_wb16(dyn_ctx, 0); // tag crc
266 ffio_fill(dyn_ctx, 0, mpah.frame_size - bytes_needed);
268 mp3->xing_frame_size = avio_close_dyn_buf(dyn_ctx, &mp3->xing_frame);
269 mp3->xing_frame_offset = avio_tell(s->pb);
270 avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size);
272 mp3->audio_size = mp3->xing_frame_size;
278 * Add a frame to XING data.
279 * Following lame's "VbrTag.c".
281 static void mp3_xing_add_frame(MP3Context *mp3, AVPacket *pkt)
287 mp3->size += pkt->size;
289 if (mp3->want == mp3->seen) {
290 mp3->bag[mp3->pos] = mp3->size;
292 if (XING_NUM_BAGS == ++mp3->pos) {
293 /* shrink table to half size by throwing away each second bag. */
294 for (i = 1; i < XING_NUM_BAGS; i += 2)
295 mp3->bag[i >> 1] = mp3->bag[i];
297 /* double wanted amount per bag. */
299 /* adjust current position to half of table size. */
300 mp3->pos = XING_NUM_BAGS / 2;
307 static int mp3_write_audio_packet(AVFormatContext *s, AVPacket *pkt)
309 MP3Context *mp3 = s->priv_data;
311 if (pkt->data && pkt->size >= 4) {
312 MPADecodeHeader mpah;
317 h = AV_RB32(pkt->data);
318 ret = avpriv_mpegaudio_decode_header(&mpah, h);
320 if (!mp3->initial_bitrate)
321 mp3->initial_bitrate = mpah.bit_rate;
322 if ((mpah.bit_rate == 0) || (mp3->initial_bitrate != mpah.bit_rate))
323 mp3->has_variable_bitrate = 1;
325 av_log(s, AV_LOG_WARNING, "Audio packet of size %d (starting with %08X...) "
326 "is invalid, writing it anyway.\n", pkt->size, h);
329 #ifdef FILTER_VBR_HEADERS
330 /* filter out XING and INFO headers. */
331 base = 4 + xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1];
333 if (base + 4 <= pkt->size) {
334 uint32_t v = AV_RB32(pkt->data + base);
336 if (MKBETAG('X','i','n','g') == v || MKBETAG('I','n','f','o') == v)
340 /* filter out VBRI headers. */
343 if (base + 4 <= pkt->size && MKBETAG('V','B','R','I') == AV_RB32(pkt->data + base))
347 if (mp3->xing_offset) {
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);
355 return ff_raw_write_packet(s, pkt);
358 static int mp3_queue_flush(AVFormatContext *s)
360 MP3Context *mp3 = s->priv_data;
362 int ret = 0, write = 1;
364 ff_id3v2_finish(&mp3->id3, s->pb, s->metadata_header_padding);
367 while ((pktl = mp3->queue)) {
368 if (write && (ret = mp3_write_audio_packet(s, &pktl->pkt)) < 0)
370 av_packet_unref(&pktl->pkt);
371 mp3->queue = pktl->next;
374 mp3->queue_end = NULL;
378 static void mp3_update_xing(AVFormatContext *s)
380 MP3Context *mp3 = s->priv_data;
386 /* replace "Xing" identification string with "Info" for CBR files. */
387 if (!mp3->has_variable_bitrate)
388 AV_WL32(mp3->xing_frame + mp3->xing_offset, MKTAG('I', 'n', 'f', 'o'));
390 AV_WB32(mp3->xing_frame + mp3->xing_offset + 8, mp3->frames);
391 AV_WB32(mp3->xing_frame + mp3->xing_offset + 12, mp3->size);
393 toc = mp3->xing_frame + mp3->xing_offset + 16;
394 toc[0] = 0; // first toc entry has to be zero.
395 for (i = 1; i < XING_TOC_SIZE; ++i) {
396 int j = i * mp3->pos / XING_TOC_SIZE;
397 int seek_point = 256LL * mp3->bag[j] / mp3->size;
398 toc[i] = FFMIN(seek_point, 255);
401 /* write replaygain */
402 rg = (AVReplayGain*)av_stream_get_side_data(s->streams[0], AV_PKT_DATA_REPLAYGAIN,
404 if (rg && rg_size >= sizeof(*rg)) {
407 AV_WB32(mp3->xing_frame + mp3->xing_offset + 131,
408 av_rescale(rg->track_peak, 1 << 23, 100000));
410 if (rg->track_gain != INT32_MIN) {
411 val = FFABS(rg->track_gain / 10000) & ((1 << 9) - 1);
412 val |= (rg->track_gain < 0) << 9;
414 AV_WB16(mp3->xing_frame + mp3->xing_offset + 135, val);
417 if (rg->album_gain != INT32_MIN) {
418 val = FFABS(rg->album_gain / 10000) & ((1 << 9) - 1);
419 val |= (rg->album_gain < 0) << 9;
421 AV_WB16(mp3->xing_frame + mp3->xing_offset + 137, val);
425 AV_WB32(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 8, mp3->audio_size);
426 AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 4, mp3->audio_crc);
428 tag_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE), 0, mp3->xing_frame, 190);
429 AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 2, tag_crc);
431 avio_seek(s->pb, mp3->xing_frame_offset, SEEK_SET);
432 avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size);
433 avio_seek(s->pb, 0, SEEK_END);
436 static int mp3_write_trailer(struct AVFormatContext *s)
438 uint8_t buf[ID3v1_TAG_SIZE];
439 MP3Context *mp3 = s->priv_data;
441 if (mp3->pics_to_write) {
442 av_log(s, AV_LOG_WARNING, "No packets were sent for some of the "
443 "attached pictures.\n");
447 /* write the id3v1 tag */
448 if (mp3->write_id3v1 && id3v1_create_tag(s, buf) > 0) {
449 avio_write(s->pb, buf, ID3v1_TAG_SIZE);
452 if (mp3->xing_offset)
455 av_freep(&mp3->xing_frame);
460 static int query_codec(enum AVCodecID id, int std_compliance)
462 const CodecMime *cm= ff_id3v2_mime_tags;
463 while(cm->id != AV_CODEC_ID_NONE) {
465 return MKTAG('A', 'P', 'I', 'C');
472 AVOutputFormat ff_mp2_muxer = {
474 .long_name = NULL_IF_CONFIG_SMALL("MP2 (MPEG audio layer 2)"),
475 .mime_type = "audio/mpeg",
476 .extensions = "mp2,m2a,mpa",
477 .audio_codec = AV_CODEC_ID_MP2,
478 .video_codec = AV_CODEC_ID_NONE,
479 .write_packet = ff_raw_write_packet,
480 .flags = AVFMT_NOTIMESTAMPS,
486 static const AVOption options[] = {
487 { "id3v2_version", "Select ID3v2 version to write. Currently 3 and 4 are supported.",
488 offsetof(MP3Context, id3v2_version), AV_OPT_TYPE_INT, {.i64 = 4}, 0, 4, AV_OPT_FLAG_ENCODING_PARAM},
489 { "write_id3v1", "Enable ID3v1 writing. ID3v1 tags are written in UTF-8 which may not be supported by most software.",
490 offsetof(MP3Context, write_id3v1), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
491 { "write_xing", "Write the Xing header containing file duration.",
492 offsetof(MP3Context, write_xing), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
496 static const AVClass mp3_muxer_class = {
497 .class_name = "MP3 muxer",
498 .item_name = av_default_item_name,
500 .version = LIBAVUTIL_VERSION_INT,
503 static int mp3_write_packet(AVFormatContext *s, AVPacket *pkt)
505 MP3Context *mp3 = s->priv_data;
507 if (pkt->stream_index == mp3->audio_stream_idx) {
508 if (mp3->pics_to_write) {
509 /* buffer audio packets until we get all the pictures */
510 AVPacketList *pktl = av_mallocz(sizeof(*pktl));
513 av_log(s, AV_LOG_WARNING, "Not enough memory to buffer audio. Skipping picture streams\n");
514 mp3->pics_to_write = 0;
516 return mp3_write_audio_packet(s, pkt);
519 ret = av_copy_packet(&pktl->pkt, pkt);
526 mp3->queue_end->next = pktl;
529 mp3->queue_end = pktl;
531 return mp3_write_audio_packet(s, pkt);
535 /* warn only once for each stream */
536 if (s->streams[pkt->stream_index]->nb_frames == 1) {
537 av_log(s, AV_LOG_WARNING, "Got more than one picture in stream %d,"
538 " ignoring.\n", pkt->stream_index);
540 if (!mp3->pics_to_write || s->streams[pkt->stream_index]->nb_frames >= 1)
543 if ((ret = ff_id3v2_write_apic(s, &mp3->id3, pkt)) < 0)
545 mp3->pics_to_write--;
547 /* flush the buffered audio packets */
548 if (!mp3->pics_to_write &&
549 (ret = mp3_queue_flush(s)) < 0)
557 * Write an ID3v2 header at beginning of stream
560 static int mp3_write_header(struct AVFormatContext *s)
562 MP3Context *mp3 = s->priv_data;
565 if (mp3->id3v2_version &&
566 mp3->id3v2_version != 3 &&
567 mp3->id3v2_version != 4) {
568 av_log(s, AV_LOG_ERROR, "Invalid ID3v2 version requested: %d. Only "
569 "3, 4 or 0 (disabled) are allowed.\n", mp3->id3v2_version);
570 return AVERROR(EINVAL);
573 /* check the streams -- we want exactly one audio and arbitrary number of
574 * video (attached pictures) */
575 mp3->audio_stream_idx = -1;
576 for (i = 0; i < s->nb_streams; i++) {
577 AVStream *st = s->streams[i];
578 if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
579 if (mp3->audio_stream_idx >= 0 || st->codecpar->codec_id != AV_CODEC_ID_MP3) {
580 av_log(s, AV_LOG_ERROR, "Invalid audio stream. Exactly one MP3 "
581 "audio stream is required.\n");
582 return AVERROR(EINVAL);
584 mp3->audio_stream_idx = i;
585 } else if (st->codecpar->codec_type != AVMEDIA_TYPE_VIDEO) {
586 av_log(s, AV_LOG_ERROR, "Only audio streams and pictures are allowed in MP3.\n");
587 return AVERROR(EINVAL);
590 if (mp3->audio_stream_idx < 0) {
591 av_log(s, AV_LOG_ERROR, "No audio stream present.\n");
592 return AVERROR(EINVAL);
594 mp3->pics_to_write = s->nb_streams - 1;
596 if (mp3->pics_to_write && !mp3->id3v2_version) {
597 av_log(s, AV_LOG_ERROR, "Attached pictures were requested, but the "
598 "ID3v2 header is disabled.\n");
599 return AVERROR(EINVAL);
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 AVOutputFormat ff_mp3_muxer = {
620 .long_name = NULL_IF_CONFIG_SMALL("MP3 (MPEG audio layer 3)"),
621 .mime_type = "audio/mpeg",
623 .priv_data_size = sizeof(MP3Context),
624 .audio_codec = AV_CODEC_ID_MP3,
625 .video_codec = AV_CODEC_ID_PNG,
626 .write_header = mp3_write_header,
627 .write_packet = mp3_write_packet,
628 .write_trailer = mp3_write_trailer,
629 .query_codec = query_codec,
630 .flags = AVFMT_NOTIMESTAMPS,
631 .priv_class = &mp3_muxer_class,