X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavformat%2Fmp3enc.c;h=1e2ff1aa9462cd6912d9516578e3dc0658ef3766;hb=90bc423212396e96a02edc1118982ab7f7766a63;hp=9a54e01d9e985a33343fb64e09fc3ffa6db47917;hpb=e6153f173a49e5bfa70b0c04d2f82930533597b9;p=ffmpeg diff --git a/libavformat/mp3enc.c b/libavformat/mp3enc.c index 9a54e01d9e9..1e2ff1aa946 100644 --- a/libavformat/mp3enc.c +++ b/libavformat/mp3enc.c @@ -32,6 +32,9 @@ #include "libavutil/opt.h" #include "libavutil/dict.h" #include "libavutil/avassert.h" +#include "libavutil/crc.h" +#include "libavutil/mathematics.h" +#include "libavutil/replaygain.h" static int id3v1_set_string(AVFormatContext *s, const char *key, uint8_t *buf, int buf_size) @@ -74,12 +77,39 @@ static int id3v1_create_tag(AVFormatContext *s, uint8_t *buf) return count; } +#define XING_NUM_BAGS 400 +#define XING_TOC_SIZE 100 +// size of the XING/LAME data, starting from the Xing tag +#define XING_SIZE 156 + typedef struct MP3Context { const AVClass *class; ID3v2EncContext id3; int id3v2_version; int write_id3v1; - int64_t nb_frames_offset; + int write_xing; + + /* xing header */ + // a buffer containing the whole XING/LAME frame + uint8_t *xing_frame; + int xing_frame_size; + + AVCRC audio_crc; // CRC of the audio data + uint32_t audio_size; // total size of the audio data + + // offset of the XING/LAME frame in the file + int64_t xing_frame_offset; + // offset of the XING/INFO tag in the frame + int xing_offset; + + int32_t frames; + int32_t size; + uint32_t want; + uint32_t seen; + uint32_t pos; + uint64_t bag[XING_NUM_BAGS]; + int initial_bitrate; + int has_variable_bitrate; /* index of the audio stream */ int audio_stream_idx; @@ -90,53 +120,215 @@ typedef struct MP3Context { AVPacketList *queue, *queue_end; } MP3Context; -/* insert a dummy frame containing number of frames */ +static const uint8_t xing_offtbl[2][2] = {{32, 17}, {17, 9}}; + +/* + * Write an empty XING header and initialize respective data. + */ static void mp3_write_xing(AVFormatContext *s) { MP3Context *mp3 = s->priv_data; - AVCodecContext *codec = s->streams[mp3->audio_stream_idx]->codec; - int bitrate_idx = 1; // 32 kbps - int64_t xing_offset = (codec->channels == 2) ? 32 : 17; + AVCodecParameters *par = s->streams[mp3->audio_stream_idx]->codecpar; + AVDictionaryEntry *enc = av_dict_get(s->streams[mp3->audio_stream_idx]->metadata, "encoder", NULL, 0); + AVIOContext *dyn_ctx; int32_t header; MPADecodeHeader mpah; int srate_idx, i, channels; - - if (!s->pb->seekable) + int bitrate_idx; + int best_bitrate_idx; + int best_bitrate_error = INT_MAX; + int ret; + int ver = 0; + int lsf, bytes_needed; + + if (!s->pb->seekable || !mp3->write_xing) return; - for (i = 0; i < FF_ARRAY_ELEMS(avpriv_mpa_freq_tab); i++) - if (avpriv_mpa_freq_tab[i] == codec->sample_rate) { - srate_idx = i; - break; - } + for (i = 0; i < FF_ARRAY_ELEMS(avpriv_mpa_freq_tab); i++) { + const uint16_t base_freq = avpriv_mpa_freq_tab[i]; + + if (par->sample_rate == base_freq) ver = 0x3; // MPEG 1 + else if (par->sample_rate == base_freq / 2) ver = 0x2; // MPEG 2 + else if (par->sample_rate == base_freq / 4) ver = 0x0; // MPEG 2.5 + else continue; + + srate_idx = i; + break; + } if (i == FF_ARRAY_ELEMS(avpriv_mpa_freq_tab)) { - av_log(s, AV_LOG_ERROR, "Unsupported sample rate.\n"); + av_log(s, AV_LOG_WARNING, "Unsupported sample rate, not writing Xing " + "header.\n"); return; } - switch (codec->channels) { + switch (par->channels) { case 1: channels = MPA_MONO; break; case 2: channels = MPA_STEREO; break; - default: av_log(s, AV_LOG_ERROR, "Unsupported number of channels.\n"); return; + default: av_log(s, AV_LOG_WARNING, "Unsupported number of channels, " + "not writing Xing header.\n"); + return; } /* dummy MPEG audio header */ header = 0xff << 24; // sync - header |= (0x7 << 5 | 0x3 << 3 | 0x1 << 1 | 0x1) << 16; // sync/mpeg-1/layer 3/no crc*/ - header |= (bitrate_idx << 4 | srate_idx << 2) << 8; + header |= (0x7 << 5 | ver << 3 | 0x1 << 1 | 0x1) << 16; // sync/audio-version/layer 3/no crc*/ + header |= (srate_idx << 2) << 8; header |= channels << 6; - avio_wb32(s->pb, header); + + lsf = !((header & (1 << 20) && header & (1 << 19))); + + mp3->xing_offset = xing_offtbl[ver != 3][channels == 1] + 4; + bytes_needed = mp3->xing_offset + XING_SIZE; + + for (bitrate_idx = 1; bitrate_idx < 15; bitrate_idx++) { + int bit_rate = 1000 * avpriv_mpa_bitrate_tab[lsf][3 - 1][bitrate_idx]; + int error = FFABS(bit_rate - par->bit_rate); + + if (error < best_bitrate_error){ + best_bitrate_error = error; + best_bitrate_idx = bitrate_idx; + } + } + + for (bitrate_idx = best_bitrate_idx; bitrate_idx < 15; bitrate_idx++) { + int32_t mask = bitrate_idx << (4 + 8); + header |= mask; + + avpriv_mpegaudio_decode_header(&mpah, header); + + if (bytes_needed <= mpah.frame_size) + break; + + header &= ~mask; + } + + ret = avio_open_dyn_buf(&dyn_ctx); + if (ret < 0) + return; + + avio_wb32(dyn_ctx, header); avpriv_mpegaudio_decode_header(&mpah, header); - ffio_fill(s->pb, 0, xing_offset); - ffio_wfourcc(s->pb, "Xing"); - avio_wb32(s->pb, 0x1); // only number of frames - mp3->nb_frames_offset = avio_tell(s->pb); - avio_wb32(s->pb, 0); + av_assert0(mpah.frame_size >= bytes_needed); + + ffio_fill(dyn_ctx, 0, mp3->xing_offset - 4); + ffio_wfourcc(dyn_ctx, "Xing"); + avio_wb32(dyn_ctx, 0x01 | 0x02 | 0x04 | 0x08); // frames / size / TOC / vbr scale + + mp3->size = mpah.frame_size; + mp3->want = 1; + + avio_wb32(dyn_ctx, 0); // frames + avio_wb32(dyn_ctx, 0); // size + + // TOC + for (i = 0; i < XING_TOC_SIZE; i++) + avio_w8(dyn_ctx, 255 * i / XING_TOC_SIZE); + + // vbr quality + // we write it, because some (broken) tools always expect it to be present + avio_wb32(dyn_ctx, 0); + + // encoder short version string + if (enc) { + uint8_t encoder_str[9] = { 0 }; + memcpy(encoder_str, enc->value, FFMIN(strlen(enc->value), sizeof(encoder_str))); + avio_write(dyn_ctx, encoder_str, sizeof(encoder_str)); + } else + ffio_fill(dyn_ctx, 0, 9); + + avio_w8(dyn_ctx, 0); // tag revision 0 / unknown vbr method + avio_w8(dyn_ctx, 0); // unknown lowpass filter value + ffio_fill(dyn_ctx, 0, 8); // empty replaygain fields + avio_w8(dyn_ctx, 0); // unknown encoding flags + avio_w8(dyn_ctx, 0); // unknown abr/minimal bitrate + + // encoder delay + if (par->initial_padding >= 1 << 12 || + par->trailing_padding >= 1 << 12) { + av_log(s, AV_LOG_WARNING, "Too many samples of padding.\n"); + avio_wb24(dyn_ctx, 0); + } else { + avio_wb24(dyn_ctx, par->initial_padding << 12 | par->trailing_padding); + } + + avio_w8(dyn_ctx, 0); // misc + avio_w8(dyn_ctx, 0); // mp3gain + avio_wb16(dyn_ctx, 0); // preset + + // audio length and CRCs (will be updated later) + avio_wb32(dyn_ctx, 0); // music length + avio_wb16(dyn_ctx, 0); // music crc + avio_wb16(dyn_ctx, 0); // tag crc + + ffio_fill(dyn_ctx, 0, mpah.frame_size - bytes_needed); + + mp3->xing_frame_size = avio_close_dyn_buf(dyn_ctx, &mp3->xing_frame); + mp3->xing_frame_offset = avio_tell(s->pb); + avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size); + + mp3->audio_size = mp3->xing_frame_size; +} + +/* + * Add a frame to XING data. + * Following lame's "VbrTag.c". + */ +static void mp3_xing_add_frame(MP3Context *mp3, AVPacket *pkt) +{ + int i; + + mp3->frames++; + mp3->seen++; + mp3->size += pkt->size; - mpah.frame_size -= 4 + xing_offset + 4 + 4 + 4; - ffio_fill(s->pb, 0, mpah.frame_size); + if (mp3->want == mp3->seen) { + mp3->bag[mp3->pos] = mp3->size; + + if (XING_NUM_BAGS == ++mp3->pos) { + /* shrink table to half size by throwing away each second bag. */ + for (i = 1; i < XING_NUM_BAGS; i += 2) + mp3->bag[i / 2] = mp3->bag[i]; + + /* double wanted amount per bag. */ + mp3->want *= 2; + /* adjust current position to half of table size. */ + mp3->pos = XING_NUM_BAGS / 2; + } + + mp3->seen = 0; + } +} + +static int mp3_write_audio_packet(AVFormatContext *s, AVPacket *pkt) +{ + MP3Context *mp3 = s->priv_data; + + if (mp3->xing_offset && pkt->size >= 4) { + MPADecodeHeader c; + int ret; + uint32_t h; + + h = AV_RB32(pkt->data); + ret = avpriv_mpegaudio_decode_header(&c, h); + if (ret >= 0) { + if (!mp3->initial_bitrate) + mp3->initial_bitrate = c.bit_rate; + if ((c.bit_rate == 0) || (mp3->initial_bitrate != c.bit_rate)) + mp3->has_variable_bitrate = 1; + } + + mp3_xing_add_frame(mp3, pkt); + + if (mp3->xing_offset) { + mp3->audio_size += pkt->size; + mp3->audio_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE), + mp3->audio_crc, pkt->data, pkt->size); + } + } + + return ff_raw_write_packet(s, pkt); } static int mp3_queue_flush(AVFormatContext *s) @@ -149,9 +341,9 @@ static int mp3_queue_flush(AVFormatContext *s) mp3_write_xing(s); while ((pktl = mp3->queue)) { - if (write && (ret = ff_raw_write_packet(s, &pktl->pkt)) < 0) + if (write && (ret = mp3_write_audio_packet(s, &pktl->pkt)) < 0) write = 0; - av_free_packet(&pktl->pkt); + av_packet_unref(&pktl->pkt); mp3->queue = pktl->next; av_freep(&pktl); } @@ -159,54 +351,95 @@ static int mp3_queue_flush(AVFormatContext *s) return ret; } +static void mp3_update_xing(AVFormatContext *s) +{ + MP3Context *mp3 = s->priv_data; + AVReplayGain *rg; + uint16_t tag_crc; + uint8_t *toc; + int i, rg_size; + + /* replace "Xing" identification string with "Info" for CBR files. */ + if (!mp3->has_variable_bitrate) + AV_WL32(mp3->xing_frame + mp3->xing_offset, MKTAG('I', 'n', 'f', 'o')); + + AV_WB32(mp3->xing_frame + mp3->xing_offset + 8, mp3->frames); + AV_WB32(mp3->xing_frame + mp3->xing_offset + 12, mp3->size); + + toc = mp3->xing_frame + mp3->xing_offset + 16; + toc[0] = 0; // first toc entry has to be zero. + for (i = 1; i < XING_TOC_SIZE; ++i) { + int j = i * mp3->pos / XING_TOC_SIZE; + int seek_point = 256LL * mp3->bag[j] / mp3->size; + toc[i] = FFMIN(seek_point, 255); + } + + /* write replaygain */ + rg = (AVReplayGain*)av_stream_get_side_data(s->streams[0], AV_PKT_DATA_REPLAYGAIN, + &rg_size); + if (rg && rg_size >= sizeof(*rg)) { + uint16_t val; + + AV_WB32(mp3->xing_frame + mp3->xing_offset + 131, + av_rescale(rg->track_peak, 1 << 23, 100000)); + + if (rg->track_gain != INT32_MIN) { + val = FFABS(rg->track_gain / 10000) & ((1 << 9) - 1); + val |= (rg->track_gain < 0) << 9; + val |= 1 << 13; + AV_WB16(mp3->xing_frame + mp3->xing_offset + 135, val); + } + + if (rg->album_gain != INT32_MIN) { + val = FFABS(rg->album_gain / 10000) & ((1 << 9) - 1); + val |= (rg->album_gain < 0) << 9; + val |= 1 << 14; + AV_WB16(mp3->xing_frame + mp3->xing_offset + 137, val); + } + } + + AV_WB32(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 8, mp3->audio_size); + AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 4, mp3->audio_crc); + + tag_crc = av_crc(av_crc_get_table(AV_CRC_16_ANSI_LE), 0, mp3->xing_frame, 190); + AV_WB16(mp3->xing_frame + mp3->xing_offset + XING_SIZE - 2, tag_crc); + + avio_seek(s->pb, mp3->xing_frame_offset, SEEK_SET); + avio_write(s->pb, mp3->xing_frame, mp3->xing_frame_size); + avio_seek(s->pb, 0, SEEK_END); +} + static int mp3_write_trailer(struct AVFormatContext *s) { uint8_t buf[ID3v1_TAG_SIZE]; MP3Context *mp3 = s->priv_data; - if (mp3 && mp3->pics_to_write) { + if (mp3->pics_to_write) { av_log(s, AV_LOG_WARNING, "No packets were sent for some of the " "attached pictures.\n"); mp3_queue_flush(s); } /* write the id3v1 tag */ - if (mp3 && mp3->write_id3v1 && id3v1_create_tag(s, buf) > 0) { + if (mp3->write_id3v1 && id3v1_create_tag(s, buf) > 0) { avio_write(s->pb, buf, ID3v1_TAG_SIZE); } - /* write number of frames */ - if (mp3 && mp3->nb_frames_offset) { - avio_seek(s->pb, mp3->nb_frames_offset, SEEK_SET); - avio_wb32(s->pb, s->streams[mp3->audio_stream_idx]->nb_frames); - avio_seek(s->pb, 0, SEEK_END); - } + if (mp3->xing_offset) + mp3_update_xing(s); - avio_flush(s->pb); + av_freep(&mp3->xing_frame); return 0; } -#if CONFIG_MP2_MUXER -AVOutputFormat ff_mp2_muxer = { - .name = "mp2", - .long_name = NULL_IF_CONFIG_SMALL("MP2 (MPEG audio layer 2)"), - .mime_type = "audio/x-mpeg", - .extensions = "mp2,m2a", - .audio_codec = AV_CODEC_ID_MP2, - .video_codec = AV_CODEC_ID_NONE, - .write_packet = ff_raw_write_packet, - .write_trailer = mp3_write_trailer, -}; -#endif - -#if CONFIG_MP3_MUXER - static const AVOption options[] = { { "id3v2_version", "Select ID3v2 version to write. Currently 3 and 4 are supported.", - offsetof(MP3Context, id3v2_version), AV_OPT_TYPE_INT, {.i64 = 4}, 3, 4, AV_OPT_FLAG_ENCODING_PARAM}, + offsetof(MP3Context, id3v2_version), AV_OPT_TYPE_INT, {.i64 = 4}, 0, 4, AV_OPT_FLAG_ENCODING_PARAM}, { "write_id3v1", "Enable ID3v1 writing. ID3v1 tags are written in UTF-8 which may not be supported by most software.", offsetof(MP3Context, write_id3v1), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM}, + { "write_xing", "Write the Xing header containing file duration.", + offsetof(MP3Context, write_xing), AV_OPT_TYPE_INT, {.i64 = 1}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM}, { NULL }, }; @@ -229,7 +462,11 @@ static int mp3_write_packet(AVFormatContext *s, AVPacket *pkt) return AVERROR(ENOMEM); pktl->pkt = *pkt; - pkt->destruct = NULL; + pktl->pkt.buf = av_buffer_ref(pkt->buf); + if (!pktl->pkt.buf) { + av_freep(&pktl); + return AVERROR(ENOMEM); + } if (mp3->queue_end) mp3->queue_end->next = pktl; @@ -237,7 +474,7 @@ static int mp3_write_packet(AVFormatContext *s, AVPacket *pkt) mp3->queue = pktl; mp3->queue_end = pktl; } else - return ff_raw_write_packet(s, pkt); + return mp3_write_audio_packet(s, pkt); } else { int ret; @@ -271,19 +508,27 @@ static int mp3_write_header(struct AVFormatContext *s) MP3Context *mp3 = s->priv_data; int ret, i; + if (mp3->id3v2_version && + mp3->id3v2_version != 3 && + mp3->id3v2_version != 4) { + av_log(s, AV_LOG_ERROR, "Invalid ID3v2 version requested: %d. Only " + "3, 4 or 0 (disabled) are allowed.\n", mp3->id3v2_version); + return AVERROR(EINVAL); + } + /* check the streams -- we want exactly one audio and arbitrary number of * video (attached pictures) */ mp3->audio_stream_idx = -1; for (i = 0; i < s->nb_streams; i++) { AVStream *st = s->streams[i]; - if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) { - if (mp3->audio_stream_idx >= 0 || st->codec->codec_id != AV_CODEC_ID_MP3) { + if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) { + if (mp3->audio_stream_idx >= 0 || st->codecpar->codec_id != AV_CODEC_ID_MP3) { av_log(s, AV_LOG_ERROR, "Invalid audio stream. Exactly one MP3 " "audio stream is required.\n"); return AVERROR(EINVAL); } mp3->audio_stream_idx = i; - } else if (st->codec->codec_type != AVMEDIA_TYPE_VIDEO) { + } else if (st->codecpar->codec_type != AVMEDIA_TYPE_VIDEO) { av_log(s, AV_LOG_ERROR, "Only audio streams and pictures are allowed in MP3.\n"); return AVERROR(EINVAL); } @@ -294,13 +539,22 @@ static int mp3_write_header(struct AVFormatContext *s) } mp3->pics_to_write = s->nb_streams - 1; - ff_id3v2_start(&mp3->id3, s->pb, mp3->id3v2_version, ID3v2_DEFAULT_MAGIC); - ret = ff_id3v2_write_metadata(s, &mp3->id3); - if (ret < 0) - return ret; + if (mp3->pics_to_write && !mp3->id3v2_version) { + av_log(s, AV_LOG_ERROR, "Attached pictures were requested, but the " + "ID3v2 header is disabled.\n"); + return AVERROR(EINVAL); + } + + if (mp3->id3v2_version) { + ff_id3v2_start(&mp3->id3, s->pb, mp3->id3v2_version, ID3v2_DEFAULT_MAGIC); + ret = ff_id3v2_write_metadata(s, &mp3->id3); + if (ret < 0) + return ret; + } if (!mp3->pics_to_write) { - ff_id3v2_finish(&mp3->id3, s->pb); + if (mp3->id3v2_version) + ff_id3v2_finish(&mp3->id3, s->pb); mp3_write_xing(s); } @@ -310,7 +564,7 @@ static int mp3_write_header(struct AVFormatContext *s) AVOutputFormat ff_mp3_muxer = { .name = "mp3", .long_name = NULL_IF_CONFIG_SMALL("MP3 (MPEG audio layer 3)"), - .mime_type = "audio/x-mpeg", + .mime_type = "audio/mpeg", .extensions = "mp3", .priv_data_size = sizeof(MP3Context), .audio_codec = AV_CODEC_ID_MP3, @@ -321,4 +575,3 @@ AVOutputFormat ff_mp3_muxer = { .flags = AVFMT_NOTIMESTAMPS, .priv_class = &mp3_muxer_class, }; -#endif