movenc.c

来自「现在关于h.264的源码很多」· C语言 代码 · 共 1,855 行 · 第 1/4 页

        len = strlen(s->title)+1;
        put_be16(pb, len*2+10);             /* size */
        put_be32(pb, 0x01);                 /* type */
        put_be16(pb, language_code("und")); /* language */
        put_be16(pb, 0x01);                 /* ? */
        ascii_to_wc (pb, s->title, len);
        size += len*2+10;

        // size
        curpos = url_ftell(pb);
        url_fseek(pb, pos, SEEK_SET);
        put_be32(pb, size);
        url_fseek(pb, pos+24, SEEK_SET);
        put_be32(pb, size-24);
        url_fseek(pb, curpos, SEEK_SET);
    }

    return size;
}

static int mov_write_moov_tag(ByteIOContext *pb, MOVContext *mov,
                              AVFormatContext *s)
{
    int i;
    offset_t pos = url_ftell(pb);
    put_be32(pb, 0); /* size placeholder*/
    put_tag(pb, "moov");
    mov->timescale = globalTimescale;

    for (i=0; i<MAX_STREAMS; i++) {
        if(mov->tracks[i].entry <= 0) continue;

        if(mov->tracks[i].enc->codec_type == CODEC_TYPE_VIDEO) {
            mov->tracks[i].timescale = mov->tracks[i].enc->time_base.den;
            mov->tracks[i].sampleDuration = mov->tracks[i].enc->time_base.num;
        }
        else if(mov->tracks[i].enc->codec_type == CODEC_TYPE_AUDIO) {
            /* If AMR, track timescale = 8000, AMR_WB = 16000 */
            if(mov->tracks[i].enc->codec_id == CODEC_ID_AMR_NB) {
                mov->tracks[i].sampleDuration = 160;  // Bytes per chunk
                mov->tracks[i].timescale = 8000;
            }
            else {
                mov->tracks[i].timescale = mov->tracks[i].enc->sample_rate;
                mov->tracks[i].sampleDuration = mov->tracks[i].enc->frame_size;
            }
        }

        mov->tracks[i].trackDuration =
            (int64_t)mov->tracks[i].sampleCount * mov->tracks[i].sampleDuration;
        mov->tracks[i].time = mov->time;
        mov->tracks[i].trackID = i+1;
    }

    mov_write_mvhd_tag(pb, mov);
    //mov_write_iods_tag(pb, mov);
    for (i=0; i<MAX_STREAMS; i++) {
        if(mov->tracks[i].entry > 0) {
            mov_write_trak_tag(pb, &(mov->tracks[i]));
        }
    }

    if (mov->mode == MODE_PSP)
        mov_write_uuidusmt_tag(pb, s);
    else
    mov_write_udta_tag(pb, mov, s);

    return updateSize(pb, pos);
}

int mov_write_mdat_tag(ByteIOContext *pb, MOVContext* mov)
{
    put_be32(pb, 8);    // placeholder for extended size field (64 bit)
    put_tag(pb, "wide");

    mov->mdat_pos = url_ftell(pb);
    put_be32(pb, 0); /* size placeholder*/
    put_tag(pb, "mdat");
    return 0;
}

/* TODO: This needs to be more general */
static void mov_write_ftyp_tag (ByteIOContext *pb, AVFormatContext *s)
{
    MOVContext *mov = s->priv_data;

    put_be32(pb, 0x14 ); /* size */
    put_tag(pb, "ftyp");

    if ( mov->mode == MODE_3GP )
        put_tag(pb, "3gp4");
    else if ( mov->mode == MODE_3G2 )
        put_tag(pb, "3g2a");
    else if ( mov->mode == MODE_PSP )
        put_tag(pb, "MSNV");
    else if ( mov->mode == MODE_MP4 )
        put_tag(pb, "isom");
    else
        put_tag(pb, "qt  ");

    put_be32(pb, 0x200 );

    if ( mov->mode == MODE_3GP )
        put_tag(pb, "3gp4");
    else if ( mov->mode == MODE_3G2 )
        put_tag(pb, "3g2a");
    else if ( mov->mode == MODE_PSP )
        put_tag(pb, "MSNV");
    else if ( mov->mode == MODE_MP4 )
        put_tag(pb, "mp41");
    else
        put_tag(pb, "qt  ");
}

static void mov_write_uuidprof_tag(ByteIOContext *pb, AVFormatContext *s)
{
    AVCodecContext *VideoCodec = s->streams[0]->codec;
    AVCodecContext *AudioCodec = s->streams[1]->codec;
    int AudioRate = AudioCodec->sample_rate;
    int FrameRate = ((VideoCodec->time_base.den) * (0x10000))/ (VideoCodec->time_base.num);
    int audio_kbitrate= AudioCodec->bit_rate / 1000;
    int video_kbitrate= FFMIN(VideoCodec->bit_rate / 1000, 800 - audio_kbitrate);

    put_be32(pb, 0x94 ); /* size */
    put_tag(pb, "uuid");
    put_tag(pb, "PROF");

    put_be32(pb, 0x21d24fce ); /* 96 bit UUID */
    put_be32(pb, 0xbb88695c );
    put_be32(pb, 0xfac9c740 );

    put_be32(pb, 0x0 );  /* ? */
    put_be32(pb, 0x3 );  /* 3 sections ? */

    put_be32(pb, 0x14 ); /* size */
    put_tag(pb, "FPRF");
    put_be32(pb, 0x0 );  /* ? */
    put_be32(pb, 0x0 );  /* ? */
    put_be32(pb, 0x0 );  /* ? */

    put_be32(pb, 0x2c );  /* size */
    put_tag(pb, "APRF");   /* audio */
    put_be32(pb, 0x0 );
    put_be32(pb, 0x2 );   /* TrackID */
    put_tag(pb, "mp4a");
    put_be32(pb, 0x20f );
    put_be32(pb, 0x0 );
    put_be32(pb, audio_kbitrate);
    put_be32(pb, audio_kbitrate);
    put_be32(pb, AudioRate );
    put_be32(pb, AudioCodec->channels );

    put_be32(pb, 0x34 );  /* size */
    put_tag(pb, "VPRF");   /* video */
    put_be32(pb, 0x0 );
    put_be32(pb, 0x1 );    /* TrackID */
    if (VideoCodec->codec_id == CODEC_ID_H264) {
        put_tag(pb, "avc1");
        put_be16(pb, 0x014D );
        put_be16(pb, 0x0015 );
    } else {
        put_tag(pb, "mp4v");
        put_be16(pb, 0x0000 );
        put_be16(pb, 0x0103 );
    }
    put_be32(pb, 0x0 );
    put_be32(pb, video_kbitrate);
    put_be32(pb, video_kbitrate);
    put_be32(pb, FrameRate);
    put_be32(pb, FrameRate);
    put_be16(pb, VideoCodec->width);
    put_be16(pb, VideoCodec->height);
    put_be32(pb, 0x010001); /* ? */
}

static int mov_write_header(AVFormatContext *s)
{
    ByteIOContext *pb = &s->pb;
    MOVContext *mov = s->priv_data;
    int i;

    /* Default mode == MP4 */
    mov->mode = MODE_MP4;

    if (s->oformat != NULL) {
        if (!strcmp("3gp", s->oformat->name)) mov->mode = MODE_3GP;
        else if (!strcmp("3g2", s->oformat->name)) mov->mode = MODE_3G2;
        else if (!strcmp("mov", s->oformat->name)) mov->mode = MODE_MOV;
        else if (!strcmp("psp", s->oformat->name)) mov->mode = MODE_PSP;

        mov_write_ftyp_tag(pb,s);
        if ( mov->mode == MODE_PSP ) {
            if ( s->nb_streams != 2 ) {
                av_log(s, AV_LOG_ERROR, "PSP mode need one video and one audio stream\n");
                return -1;
            }
            mov_write_uuidprof_tag(pb,s);
        }
    }

    for(i=0; i<s->nb_streams; i++){
        AVCodecContext *c= s->streams[i]->codec;

        if(c->codec_type == CODEC_TYPE_VIDEO){
            av_set_pts_info(s->streams[i], 64, 1, c->time_base.den);
            if (!codec_get_tag(codec_movvideo_tags, c->codec_id)){
                if(!codec_get_tag(codec_bmp_tags, c->codec_id))
                    return -1;
                else
                    av_log(s, AV_LOG_INFO, "Warning, using MS style video codec tag, the file may be unplayable!\n");
            }
        }else if(c->codec_type == CODEC_TYPE_AUDIO){
            av_set_pts_info(s->streams[i], 64, 1, c->sample_rate);
            if (!codec_get_tag(codec_movaudio_tags, c->codec_id)){
                if(!codec_get_tag(codec_wav_tags, c->codec_id))
                    return -1;
                else
                    av_log(s, AV_LOG_INFO, "Warning, using MS style audio codec tag, the file may be unplayable!\n");
            }
        }
        mov->tracks[i].language = ff_mov_iso639_to_lang(s->streams[i]->language, mov->mode != MODE_MOV);
    }

    for (i=0; i<MAX_STREAMS; i++) {
        mov->tracks[i].mode = mov->mode;
    }

    put_flush_packet(pb);

    return 0;
}

static int mov_write_packet(AVFormatContext *s, AVPacket *pkt)
{
    MOVContext *mov = s->priv_data;
    ByteIOContext *pb = &s->pb;
    AVCodecContext *enc = s->streams[pkt->stream_index]->codec;
    MOVTrack* trk = &mov->tracks[pkt->stream_index];
    int cl, id;
    unsigned int samplesInChunk = 0;
    int size= pkt->size;

    if (url_is_streamed(&s->pb)) return 0; /* Can't handle that */
    if (!size) return 0; /* Discard 0 sized packets */

    if (enc->codec_type == CODEC_TYPE_VIDEO ) {
        samplesInChunk = 1;
    }
    else if (enc->codec_type == CODEC_TYPE_AUDIO ) {
        if( enc->codec_id == CODEC_ID_AMR_NB) {
            /* We must find out how many AMR blocks there are in one packet */
            static uint16_t packed_size[16] =
                {13, 14, 16, 18, 20, 21, 27, 32, 6, 0, 0, 0, 0, 0, 0, 0};
            int len = 0;

            while (len < size && samplesInChunk < 100) {
                len += packed_size[(pkt->data[len] >> 3) & 0x0F];
                samplesInChunk++;
            }
        }
        else if(enc->codec_id == CODEC_ID_PCM_ALAW) {
            samplesInChunk = size/enc->channels;
        }
        else if(enc->codec_id == CODEC_ID_PCM_S16BE || enc->codec_id == CODEC_ID_PCM_S16LE) {
            samplesInChunk = size/(2*enc->channels);
        }
        else {
            samplesInChunk = 1;
        }
    }

    if ((enc->codec_id == CODEC_ID_MPEG4 || enc->codec_id == CODEC_ID_AAC)
        && trk->vosLen == 0) {
//        assert(enc->extradata_size);

        trk->vosLen = enc->extradata_size;
        trk->vosData = av_malloc(trk->vosLen);
        memcpy(trk->vosData, enc->extradata, trk->vosLen);
    }

    if (enc->codec_id == CODEC_ID_H264) {
        if (trk->vosLen == 0) {
            /* copy extradata */
            trk->vosLen = enc->extradata_size;
            trk->vosData = av_malloc(trk->vosLen);
            memcpy(trk->vosData, enc->extradata, trk->vosLen);
        }
        if (*(uint8_t *)trk->vosData != 1) {
            /* from x264 or from bytestream h264 */
            /* nal reformating needed */
            avc_parse_nal_units(&pkt->data, &pkt->size);
            assert(pkt->size);
            size = pkt->size;
        }
    }

    cl = trk->entry / MOV_INDEX_CLUSTER_SIZE;
    id = trk->entry % MOV_INDEX_CLUSTER_SIZE;

    if (trk->ents_allocated <= trk->entry) {
        trk->cluster = av_realloc(trk->cluster, (cl+1)*sizeof(void*));
        if (!trk->cluster)
            return -1;
        trk->cluster[cl] = av_malloc(MOV_INDEX_CLUSTER_SIZE*sizeof(MOVIentry));
        if (!trk->cluster[cl])
            return -1;
        trk->ents_allocated += MOV_INDEX_CLUSTER_SIZE;
    }
    if (mov->mdat_written == 0) {
        mov_write_mdat_tag(pb, mov);
        mov->mdat_written = 1;
        mov->time = s->timestamp + 0x7C25B080; //1970 based -> 1904 based
    }

    trk->cluster[cl][id].pos = url_ftell(pb);
    trk->cluster[cl][id].samplesInChunk = samplesInChunk;
    trk->cluster[cl][id].size = size;
    trk->cluster[cl][id].entries = samplesInChunk;
    if(enc->codec_type == CODEC_TYPE_VIDEO) {
        if (pkt->dts != pkt->pts)
            trk->hasBframes = 1;
        trk->cluster[cl][id].cts = pkt->pts - pkt->dts;
        trk->cluster[cl][id].key_frame = !!(pkt->flags & PKT_FLAG_KEY);
        if(trk->cluster[cl][id].key_frame)
            trk->hasKeyframes = 1;
    }
    trk->enc = enc;
    trk->entry++;
    trk->sampleCount += samplesInChunk;
    trk->mdat_size += size;

    put_buffer(pb, pkt->data, size);

    put_flush_packet(pb);
    return 0;
}

static int mov_write_trailer(AVFormatContext *s)
{
    MOVContext *mov = s->priv_data;
    ByteIOContext *pb = &s->pb;
    int res = 0;
    int i;
    uint64_t j;

    offset_t moov_pos = url_ftell(pb);

    /* Write size of mdat tag */
    for (i=0, j=0; i<MAX_STREAMS; i++) {
        if(mov->tracks[i].ents_allocated > 0) {
            j += mov->tracks[i].mdat_size;
        }
    }
    if (j+8 <= UINT32_MAX) {
        url_fseek(pb, mov->mdat_pos, SEEK_SET);
        put_be32(pb, j+8);
    } else {
        /* overwrite 'wide' placeholder atom */
        url_fseek(pb, mov->mdat_pos - 8, SEEK_SET);
        put_be32(pb, 1); /* special value: real atom size will be 64 bit value after tag field */
        put_tag(pb, "mdat");
        put_be64(pb, j+16);
    }
    url_fseek(pb, moov_pos, SEEK_SET);

    mov_write_moov_tag(pb, mov, s);

    for (i=0; i<MAX_STREAMS; i++) {
        for (j=0; j<mov->tracks[i].ents_allocated/MOV_INDEX_CLUSTER_SIZE; j++) {
            av_free(mov->tracks[i].cluster[j]);
        }
        av_free(mov->tracks[i].cluster);
        if( mov->tracks[i].vosLen ) av_free( mov->tracks[i].vosData );

        mov->tracks[i].cluster = NULL;
        mov->tracks[i].ents_allocated = mov->tracks[i].entry = 0;
    }

    put_flush_packet(pb);

    return res;
}

static AVOutputFormat mov_oformat = {
    "mov",
    "mov format",
    NULL,
    "mov",
    sizeof(MOVContext),
    CODEC_ID_AAC,
    CODEC_ID_MPEG4,
    mov_write_header,
    mov_write_packet,
    mov_write_trailer,
    .flags = AVFMT_GLOBALHEADER,
};

static AVOutputFormat _3gp_oformat = {
    "3gp",
    "3gp format",
    NULL,
    "3gp",
    sizeof(MOVContext),
    CODEC_ID_AMR_NB,
    CODEC_ID_H263,
    mov_write_header,
    mov_write_packet,
    mov_write_trailer,
    .flags = AVFMT_GLOBALHEADER,
};

static AVOutputFormat mp4_oformat = {
    "mp4",
    "mp4 format",
    "application/mp4",
    "mp4,m4a",
    sizeof(MOVContext),
    CODEC_ID_AAC,
    CODEC_ID_MPEG4,
    mov_write_header,
    mov_write_packet,
    mov_write_trailer,
    .flags = AVFMT_GLOBALHEADER,
};

static AVOutputFormat psp_oformat = {
    "psp",
    "psp mp4 format",
    NULL,
    "mp4,psp",
    sizeof(MOVContext),
    CODEC_ID_AAC,
    CODEC_ID_MPEG4,
    mov_write_header,
    mov_write_packet,
    mov_write_trailer,
    .flags = AVFMT_GLOBALHEADER,
};

static AVOutputFormat _3g2_oformat = {
    "3g2",
    "3gp2 format",
    NULL,
    "3g2",
    sizeof(MOVContext),
    CODEC_ID_AMR_NB,
    CODEC_ID_H263,
    mov_write_header,
    mov_write_packet,
    mov_write_trailer,
    .flags = AVFMT_GLOBALHEADER,
};

int movenc_init(void)
{
    av_register_output_format(&mov_oformat);
    av_register_output_format(&_3gp_oformat);
    av_register_output_format(&mp4_oformat);
    av_register_output_format(&psp_oformat);
    av_register_output_format(&_3g2_oformat);
    return 0;
}