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 // refer to: http://id3.org/ID3v1
49 static int id3v1_create_tag(AVFormatContext *s, uint8_t *buf)
51 AVDictionaryEntry *tag;
54 memset(buf, 0, ID3v1_TAG_SIZE); /* fail safe */
58 /* we knowingly overspecify each tag length by one byte to compensate for the mandatory null byte added by av_strlcpy */
59 count += id3v1_set_string(s, "TIT2", buf + 3, 30 + 1); //title
60 count += id3v1_set_string(s, "TPE1", buf + 33, 30 + 1); //author|artist
61 count += id3v1_set_string(s, "TALB", buf + 63, 30 + 1); //album
62 if ((tag = av_dict_get(s->metadata, "TYER", NULL, 0))) { //year
63 av_strlcpy(buf + 93, tag->value, 4 + 1);
65 } else if ((tag = av_dict_get(s->metadata, "TDRC", NULL, 0))) {
66 av_strlcpy(buf + 93, tag->value, 4 + 1);
68 } else if ((tag = av_dict_get(s->metadata, "TDAT", NULL, 0))) {
69 av_strlcpy(buf + 93, tag->value, 4 + 1);
73 count += id3v1_set_string(s, "comment", buf + 97, 30 + 1);
74 if ((tag = av_dict_get(s->metadata, "TRCK", NULL, 0))) { //track
76 buf[126] = atoi(tag->value);
79 buf[127] = 0xFF; /* default to unknown genre */
80 if ((tag = av_dict_get(s->metadata, "TCON", NULL, 0))) { //genre
81 for(i = 0; i <= ID3v1_GENRE_MAX; i++) {
82 if (!av_strcasecmp(tag->value, ff_id3v1_genre_str[i])) {
92 #define XING_NUM_BAGS 400
93 #define XING_TOC_SIZE 100
94 // size of the XING/LAME data, starting from the Xing tag
97 typedef struct MP3Context {
105 // a buffer containing the whole XING/LAME frame
109 AVCRC audio_crc; // CRC of the audio data
110 uint32_t audio_size; // total size of the audio data
112 // offset of the XING/LAME frame in the file
113 int64_t xing_frame_offset;
114 // offset of the XING/INFO tag in the frame
122 uint64_t bag[XING_NUM_BAGS];
124 int has_variable_bitrate;
128 /* index of the audio stream */
129 int audio_stream_idx;
130 /* number of attached pictures we still need to write */
133 /* audio packets are queued here until we get all the attached pictures */
134 AVPacketList *queue, *queue_end;
137 static const uint8_t xing_offtbl[2][2] = {{32, 17}, {17, 9}};
140 * Write an empty XING header and initialize respective data.
142 static int mp3_write_xing(AVFormatContext *s)
144 MP3Context *mp3 = s->priv_data;
145 AVCodecParameters *par = s->streams[mp3->audio_stream_idx]->codecpar;
146 AVDictionaryEntry *enc = av_dict_get(s->streams[mp3->audio_stream_idx]->metadata, "encoder", NULL, 0);
147 AVIOContext *dyn_ctx;
149 MPADecodeHeader mpah;
150 int srate_idx, i, channels;
152 int best_bitrate_idx = -1;
153 int best_bitrate_error = INT_MAX;
158 if (!(s->pb->seekable & AVIO_SEEKABLE_NORMAL) || !mp3->write_xing)
161 for (i = 0; i < FF_ARRAY_ELEMS(avpriv_mpa_freq_tab); i++) {
162 const uint16_t base_freq = avpriv_mpa_freq_tab[i];
164 if (par->sample_rate == base_freq) ver = 0x3; // MPEG 1
165 else if (par->sample_rate == base_freq / 2) ver = 0x2; // MPEG 2
166 else if (par->sample_rate == base_freq / 4) ver = 0x0; // MPEG 2.5
172 if (i == FF_ARRAY_ELEMS(avpriv_mpa_freq_tab)) {
173 av_log(s, AV_LOG_WARNING, "Unsupported sample rate, not writing Xing header.\n");
177 switch (par->channels) {
178 case 1: channels = MPA_MONO; break;
179 case 2: channels = MPA_STEREO; break;
180 default: av_log(s, AV_LOG_WARNING, "Unsupported number of channels, "
181 "not writing Xing header.\n");
185 /* dummy MPEG audio header */
186 header = 0xffU << 24; // sync
187 header |= (0x7 << 5 | ver << 3 | 0x1 << 1 | 0x1) << 16; // sync/audio-version/layer 3/no crc*/
188 header |= (srate_idx << 2) << 8;
189 header |= channels << 6;
191 for (bitrate_idx = 1; bitrate_idx < 15; bitrate_idx++) {
192 int bit_rate = 1000 * avpriv_mpa_bitrate_tab[ver != 3][3 - 1][bitrate_idx];
193 int error = FFABS(bit_rate - par->bit_rate);
195 if (error < best_bitrate_error) {
196 best_bitrate_error = error;
197 best_bitrate_idx = bitrate_idx;
200 av_assert0(best_bitrate_idx >= 0);
202 for (bitrate_idx = best_bitrate_idx; ; bitrate_idx++) {
203 int32_t mask = bitrate_idx << (4 + 8);
204 if (15 == bitrate_idx)
208 ret = avpriv_mpegaudio_decode_header(&mpah, header);
209 av_assert0(ret >= 0);
210 mp3->xing_offset = xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1] + 4;
211 bytes_needed = mp3->xing_offset + XING_SIZE;
213 if (bytes_needed <= mpah.frame_size)
219 ret = avio_open_dyn_buf(&dyn_ctx);
223 avio_wb32(dyn_ctx, header);
225 ffio_fill(dyn_ctx, 0, mp3->xing_offset - 4);
226 ffio_wfourcc(dyn_ctx, "Xing");
227 avio_wb32(dyn_ctx, 0x01 | 0x02 | 0x04 | 0x08); // frames / size / TOC / vbr scale
229 mp3->size = mpah.frame_size;
234 avio_wb32(dyn_ctx, 0); // frames
235 avio_wb32(dyn_ctx, 0); // size
238 for (i = 0; i < XING_TOC_SIZE; i++)
239 avio_w8(dyn_ctx, (uint8_t)(255 * i / XING_TOC_SIZE));
242 // we write it, because some (broken) tools always expect it to be present
243 avio_wb32(dyn_ctx, 0);
245 // encoder short version string
247 uint8_t encoder_str[9] = { 0 };
248 if ( strlen(enc->value) > sizeof(encoder_str)
249 && !strcmp("Lavc libmp3lame", enc->value)) {
250 memcpy(encoder_str, "Lavf lame", 9);
252 memcpy(encoder_str, enc->value, FFMIN(strlen(enc->value), sizeof(encoder_str)));
254 avio_write(dyn_ctx, encoder_str, sizeof(encoder_str));
256 avio_write(dyn_ctx, "Lavf\0\0\0\0\0", 9);
258 avio_w8(dyn_ctx, 0); // tag revision 0 / unknown vbr method
259 avio_w8(dyn_ctx, 0); // unknown lowpass filter value
260 ffio_fill(dyn_ctx, 0, 8); // empty replaygain fields
261 avio_w8(dyn_ctx, 0); // unknown encoding flags
262 avio_w8(dyn_ctx, 0); // unknown abr/minimal bitrate
263 avio_wb24(dyn_ctx, 0); // empty encoder delay/padding
265 avio_w8(dyn_ctx, 0); // misc
266 avio_w8(dyn_ctx, 0); // mp3gain
267 avio_wb16(dyn_ctx, 0); // preset
269 // audio length and CRCs (will be updated later)
270 avio_wb32(dyn_ctx, 0); // music length
271 avio_wb16(dyn_ctx, 0); // music crc
272 avio_wb16(dyn_ctx, 0); // tag crc
274 ffio_fill(dyn_ctx, 0, mpah.frame_size - bytes_needed);
276 mp3->xing_frame_size = avio_close_dyn_buf(dyn_ctx, &mp3->xing_frame);
277 mp3->xing_frame_offset = avio_tell(s->pb);
278 avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size);
280 mp3->audio_size = mp3->xing_frame_size;
286 * Add a frame to XING data.
287 * Following lame's "VbrTag.c".
289 static void mp3_xing_add_frame(MP3Context *mp3, AVPacket *pkt)
295 mp3->size += pkt->size;
297 if (mp3->want == mp3->seen) {
298 mp3->bag[mp3->pos] = mp3->size;
300 if (XING_NUM_BAGS == ++mp3->pos) {
301 /* shrink table to half size by throwing away each second bag. */
302 for (i = 1; i < XING_NUM_BAGS; i += 2)
303 mp3->bag[i >> 1] = mp3->bag[i];
305 /* double wanted amount per bag. */
307 /* adjust current position to half of table size. */
308 mp3->pos = XING_NUM_BAGS / 2;
315 static int mp3_write_audio_packet(AVFormatContext *s, AVPacket *pkt)
317 MP3Context *mp3 = s->priv_data;
319 if (pkt->data && pkt->size >= 4) {
320 MPADecodeHeader mpah;
325 h = AV_RB32(pkt->data);
326 ret = avpriv_mpegaudio_decode_header(&mpah, h);
328 if (!mp3->initial_bitrate)
329 mp3->initial_bitrate = mpah.bit_rate;
330 if ((mpah.bit_rate == 0) || (mp3->initial_bitrate != mpah.bit_rate))
331 mp3->has_variable_bitrate = 1;
333 av_log(s, AV_LOG_WARNING, "Audio packet of size %d (starting with %08"PRIX32"...) "
334 "is invalid, writing it anyway.\n", pkt->size, h);
337 #ifdef FILTER_VBR_HEADERS
338 /* filter out XING and INFO headers. */
339 base = 4 + xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1];
341 if (base + 4 <= pkt->size) {
342 uint32_t v = AV_RB32(pkt->data + base);
344 if (MKBETAG('X','i','n','g') == v || MKBETAG('I','n','f','o') == v)
348 /* filter out VBRI headers. */
351 if (base + 4 <= pkt->size && MKBETAG('V','B','R','I') == AV_RB32(pkt->data + base))
355 if (mp3->xing_offset) {
356 uint8_t *side_data = NULL;
359 mp3_xing_add_frame(mp3, pkt);
360 mp3->audio_size += pkt->size;
361 mp3->audio_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE),
362 mp3->audio_crc, pkt->data, pkt->size);
364 side_data = av_packet_get_side_data(pkt,
365 AV_PKT_DATA_SKIP_SAMPLES,
367 if (side_data && side_data_size >= 10) {
368 mp3->padding = FFMAX(AV_RL32(side_data + 4) + 528 + 1, 0);
370 mp3->delay = FFMAX(AV_RL32(side_data) - 528 - 1, 0);
377 return ff_raw_write_packet(s, pkt);
380 static int mp3_queue_flush(AVFormatContext *s)
382 MP3Context *mp3 = s->priv_data;
384 int ret = 0, write = 1;
386 ff_id3v2_finish(&mp3->id3, s->pb, s->metadata_header_padding);
390 ff_packet_list_get(&mp3->queue, &mp3->queue_end, &pkt);
391 if (write && (ret = mp3_write_audio_packet(s, &pkt)) < 0)
393 av_packet_unref(&pkt);
398 static void mp3_update_xing(AVFormatContext *s)
400 MP3Context *mp3 = s->priv_data;
405 int64_t old_pos = avio_tell(s->pb);
407 /* replace "Xing" identification string with "Info" for CBR files. */
408 if (!mp3->has_variable_bitrate)
409 AV_WL32(mp3->xing_frame + mp3->xing_offset, MKTAG('I', 'n', 'f', 'o'));
411 AV_WB32(mp3->xing_frame + mp3->xing_offset + 8, mp3->frames);
412 AV_WB32(mp3->xing_frame + mp3->xing_offset + 12, mp3->size);
414 toc = mp3->xing_frame + mp3->xing_offset + 16;
415 toc[0] = 0; // first toc entry has to be zero.
416 for (i = 1; i < XING_TOC_SIZE; ++i) {
417 int j = i * mp3->pos / XING_TOC_SIZE;
418 int seek_point = 256LL * mp3->bag[j] / mp3->size;
419 toc[i] = FFMIN(seek_point, 255);
422 /* write replaygain */
423 rg = (AVReplayGain*)av_stream_get_side_data(s->streams[0], AV_PKT_DATA_REPLAYGAIN,
425 if (rg && rg_size >= sizeof(*rg)) {
428 AV_WB32(mp3->xing_frame + mp3->xing_offset + 131,
429 av_rescale(rg->track_peak, 1 << 23, 100000));
431 if (rg->track_gain != INT32_MIN) {
432 val = FFABS(rg->track_gain / 10000) & ((1 << 9) - 1);
433 val |= (rg->track_gain < 0) << 9;
435 AV_WB16(mp3->xing_frame + mp3->xing_offset + 135, val);
438 if (rg->album_gain != INT32_MIN) {
439 val = FFABS(rg->album_gain / 10000) & ((1 << 9) - 1);
440 val |= (rg->album_gain < 0) << 9;
442 AV_WB16(mp3->xing_frame + mp3->xing_offset + 137, val);
446 /* write encoder delay/padding */
447 if (mp3->delay >= 1 << 12) {
448 mp3->delay = (1 << 12) - 1;
449 av_log(s, AV_LOG_WARNING, "Too many samples of initial padding.\n");
451 if (mp3->padding >= 1 << 12) {
452 mp3->padding = (1 << 12) - 1;
453 av_log(s, AV_LOG_WARNING, "Too many samples of trailing padding.\n");
455 AV_WB24(mp3->xing_frame + mp3->xing_offset + 141, (mp3->delay << 12) + mp3->padding);
457 AV_WB32(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 8, mp3->audio_size);
458 AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 4, mp3->audio_crc);
460 tag_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE), 0, mp3->xing_frame, 190);
461 AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 2, tag_crc);
463 avio_seek(s->pb, mp3->xing_frame_offset, SEEK_SET);
464 avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size);
465 avio_seek(s->pb, old_pos, SEEK_SET);
468 static int mp3_write_trailer(struct AVFormatContext *s)
470 uint8_t buf[ID3v1_TAG_SIZE];
471 MP3Context *mp3 = s->priv_data;
473 if (mp3->pics_to_write) {
474 av_log(s, AV_LOG_WARNING, "No packets were sent for some of the "
475 "attached pictures.\n");
479 /* write the id3v1 tag */
480 if (mp3->write_id3v1 && id3v1_create_tag(s, buf) > 0) {
481 avio_write(s->pb, buf, ID3v1_TAG_SIZE);
484 if (mp3->xing_offset)
490 static int query_codec(enum AVCodecID id, int std_compliance)
492 const CodecMime *cm= ff_id3v2_mime_tags;
493 while(cm->id != AV_CODEC_ID_NONE) {
495 return MKTAG('A', 'P', 'I', 'C');
501 static const AVOption options[] = {
502 { "id3v2_version", "Select ID3v2 version to write. Currently 3 and 4 are supported.",
503 offsetof(MP3Context, id3v2_version), AV_OPT_TYPE_INT, {.i64 = 4}, 0, 4, AV_OPT_FLAG_ENCODING_PARAM},
504 { "write_id3v1", "Enable ID3v1 writing. ID3v1 tags are written in UTF-8 which may not be supported by most software.",
505 offsetof(MP3Context, write_id3v1), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
506 { "write_xing", "Write the Xing header containing file duration.",
507 offsetof(MP3Context, write_xing), AV_OPT_TYPE_BOOL, {.i64 = 1}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
511 static const AVClass mp3_muxer_class = {
512 .class_name = "MP3 muxer",
513 .item_name = av_default_item_name,
515 .version = LIBAVUTIL_VERSION_INT,
518 static int mp3_write_packet(AVFormatContext *s, AVPacket *pkt)
520 MP3Context *mp3 = s->priv_data;
522 if (pkt->stream_index == mp3->audio_stream_idx) {
523 if (mp3->pics_to_write) {
524 /* buffer audio packets until we get all the pictures */
525 int ret = ff_packet_list_put(&mp3->queue, &mp3->queue_end, pkt, FF_PACKETLIST_FLAG_REF_PACKET);
528 av_log(s, AV_LOG_WARNING, "Not enough memory to buffer audio. Skipping picture streams\n");
529 mp3->pics_to_write = 0;
531 return mp3_write_audio_packet(s, pkt);
534 return mp3_write_audio_packet(s, pkt);
538 /* warn only once for each stream */
539 if (s->streams[pkt->stream_index]->nb_frames == 1) {
540 av_log(s, AV_LOG_WARNING, "Got more than one picture in stream %d,"
541 " ignoring.\n", pkt->stream_index);
543 if (!mp3->pics_to_write || s->streams[pkt->stream_index]->nb_frames >= 1)
546 if ((ret = ff_id3v2_write_apic(s, &mp3->id3, pkt)) < 0)
548 mp3->pics_to_write--;
550 /* flush the buffered audio packets */
551 if (!mp3->pics_to_write &&
552 (ret = mp3_queue_flush(s)) < 0)
560 * Write an ID3v2 header at beginning of stream
563 static int mp3_init(struct AVFormatContext *s)
565 MP3Context *mp3 = s->priv_data;
568 if (mp3->id3v2_version &&
569 mp3->id3v2_version != 3 &&
570 mp3->id3v2_version != 4) {
571 av_log(s, AV_LOG_ERROR, "Invalid ID3v2 version requested: %d. Only "
572 "3, 4 or 0 (disabled) are allowed.\n", mp3->id3v2_version);
573 return AVERROR(EINVAL);
576 /* check the streams -- we want exactly one audio and arbitrary number of
577 * video (attached pictures) */
578 mp3->audio_stream_idx = -1;
579 for (i = 0; i < s->nb_streams; i++) {
580 AVStream *st = s->streams[i];
581 if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
582 if (mp3->audio_stream_idx >= 0 || st->codecpar->codec_id != AV_CODEC_ID_MP3) {
583 av_log(s, AV_LOG_ERROR, "Invalid audio stream. Exactly one MP3 "
584 "audio stream is required.\n");
585 return AVERROR(EINVAL);
587 mp3->audio_stream_idx = i;
588 } else if (st->codecpar->codec_type != AVMEDIA_TYPE_VIDEO) {
589 av_log(s, AV_LOG_ERROR, "Only audio streams and pictures are allowed in MP3.\n");
590 return AVERROR(EINVAL);
593 if (mp3->audio_stream_idx < 0) {
594 av_log(s, AV_LOG_ERROR, "No audio stream present.\n");
595 return AVERROR(EINVAL);
597 mp3->pics_to_write = s->nb_streams - 1;
599 if (mp3->pics_to_write && !mp3->id3v2_version) {
600 av_log(s, AV_LOG_ERROR, "Attached pictures were requested, but the "
601 "ID3v2 header is disabled.\n");
602 return AVERROR(EINVAL);
608 static int mp3_write_header(struct AVFormatContext *s)
610 MP3Context *mp3 = s->priv_data;
613 if (mp3->id3v2_version) {
614 ff_id3v2_start(&mp3->id3, s->pb, mp3->id3v2_version, ID3v2_DEFAULT_MAGIC);
615 ret = ff_id3v2_write_metadata(s, &mp3->id3);
620 if (!mp3->pics_to_write) {
621 if (mp3->id3v2_version)
622 ff_id3v2_finish(&mp3->id3, s->pb, s->metadata_header_padding);
629 static void mp3_deinit(struct AVFormatContext *s)
631 MP3Context *mp3 = s->priv_data;
633 ff_packet_list_free(&mp3->queue, &mp3->queue_end);
634 av_freep(&mp3->xing_frame);
637 AVOutputFormat ff_mp3_muxer = {
639 .long_name = NULL_IF_CONFIG_SMALL("MP3 (MPEG audio layer 3)"),
640 .mime_type = "audio/mpeg",
642 .priv_data_size = sizeof(MP3Context),
643 .audio_codec = AV_CODEC_ID_MP3,
644 .video_codec = AV_CODEC_ID_PNG,
646 .write_header = mp3_write_header,
647 .write_packet = mp3_write_packet,
648 .write_trailer = mp3_write_trailer,
649 .deinit = mp3_deinit,
650 .query_codec = query_codec,
651 .flags = AVFMT_NOTIMESTAMPS,
652 .priv_class = &mp3_muxer_class,