ffmpeg和代码生成的碎片化的mp4文件不同

Question

目前,使用libavformat从代码生成碎片化MP4文件时遇到问题.我的文件可以使用VLC播放,但无法在(Chrome)浏览器中流式传输(通过WebSocket)和播放(通过MediaSource).(我用它来通过WebSocket测试流式分段MP4文件到浏览器).

注意:下面的文件由Baseline profile 4级编码.因此,您应该将MIME类型(在index.html中)更改为const mimeCodec ='video/mp4; 编解码器= "avc1.42C028"'; 能够发挥它们.

我检查并发现我生成的Mp4文件与使用ffmpeg工具生成的文件略有不同.

这就是我所做的:

我有一个.h264 文件

1. 第一种方法,我使用ffmpeg生成碎片化的MP4文件.

   ffmpeg -i temp.h264 -vcodec copy -f mp4 -movflags empty_moov+default_base_moof+frag_keyframe ffmpeg.mp4
   

   生成的文件可以由Quicktime播放器和VLC播放器播放

2. 第二种方法,我使用libavformat以 programmaticaly方式生成碎片化的Mp4文件

   首先我初始化上下文,codecin代码是AVCodecContext*输入流的

    av_register_all();
    avcodec_register_all();
    int ret;
    AVOutputFormat* fmt = av_guess_format("mp4", 0, 0);
    if(!fmt) {
      return;
    }
   
    AVFormatContext* ctx = avformat_alloc_context();
    // Create AVIO context to capture generated Mp4 contain
    uint8_t *avio_ctx_buffer = NULL;
    size_t avio_ctx_buffer_size = 50000;
    IOOutput buffer = {};
    const size_t bd_buf_size = 50000;
    avio_ctx_buffer = (uint8_t*)av_malloc(avio_ctx_buffer_size);
    buffer.buff = (uint8_t*)av_malloc(bd_buf_size);
   
    buffer.size = bd_buf_size;
    AVIOContext* ioCtx = avio_alloc_context(avio_ctx_buffer, (int)avio_ctx_buffer_size, 1, &buffer, NULL, MP4Formatter::onWritePacket, NULL);
   
    ctx->oformat = fmt;
    if (ctx->oformat->flags & AVFMT_GLOBALHEADER)
      ctx->flags |= CODEC_FLAG_GLOBAL_HEADER;
    ctx->pb = ioCtx;
    av_opt_set(ctx, "movflags", "frag_keyframe+empty_moov+default_base_moof", 0);
   
    AVStream* st = avformat_new_stream(ctx, codec->codec);
    if (!st) {
      return;
    }
   
    st->id = (ctx->nb_streams - 1);
    avcodec_parameters_from_context(st->codecpar, codec);
    st->time_base = codec->time_base;
    ioCtx->seekable = false;
   

   其次,我实现了onWritePacket回调

    int MP4Formatter::onWritePacket(void *opaque, uint8_t* buf, int buf_size) {
      file.write((char*)buf, buf_size);
    }
   

   第三,在h264文件的每个数据包上,我都是用它来编写的 av_interleaved_write_frame

    if (firstFrame) {
      AVDictionary *opts = NULL;
      av_dict_set(&opts, "movflags", "frag_keyframe+empty_moov+default_base_moof", 0);
      if(!parseSPSPPS(data, length)) {
        return;
      }
      cout << "spslen " << spslen << " ppslen " << ppslen << endl;
      auto c = st->codecpar;
      // Extradata contains PPS & SPS for AVCC format
      int extradata_len = 8 + spslen + 1 + 2 + ppslen;
      c->extradata = (uint8_t*)av_mallocz(extradata_len);
      c->extradata_size = extradata_len;
      c->extradata[0] = 0x01;
      c->extradata[1] = sps[1];
      c->extradata[2] = sps[2];
      c->extradata[3] = sps[3];
      c->extradata[4] = 0xFC | 3;
      c->extradata[5] = 0xE0 | 1;
      int tmp = spslen;
      c->extradata[6] = (tmp >> 8) & 0x00ff;
      c->extradata[7] = tmp & 0x00ff;
      int i = 0;
      for (i=0; i<tmp; i++) {
        c->extradata[8 + i] = sps[i];
      }
      c->extradata[8 + tmp] = 0x01;
      int tmp2 = ppslen;
      c->extradata[8 + tmp + 1] = (tmp2 >> 8) & 0x00ff;
      c->extradata[8 + tmp + 2] = tmp2 & 0x00ff;
      for (i=0; i<tmp2; i++) {
        c->extradata[8 + tmp + 3 + i] = pps[i];
      }
      int ret = avformat_write_header(ctx, &opts);
      if(ret < 0) {
        return;
      }
      firstFrame = false;
    }
    AVPacket pkt;
    av_init_packet(&pkt);
    pkt.buf = av_buffer_alloc(length);
    memcpy(pkt.buf->data, data, length);
    pkt.buf->size = length;
    pkt.data = pkt.buf->data;
    pkt.size = pkt.buf->size;
    pkt.pts = ts;
    pkt.dts = ts;
   
    if (keyFrame) {
      pkt.flags |= AV_PKT_FLAG_KEY;
    }
    else {
      pkt.flags = 0;
    }
   
    pkt.stream_index = st->id;
   
    av_interleaved_write_frame(ctx, &pkt);
    av_buffer_unref(&pkt.buf);
    av_packet_unref(&pkt);
   

你们可以看看我的档案,看看有什么不对吗？

Answer 1

也许你可以尝试这个：

AVDictionary* opts = NULL;
av_dict_set(&opts, "movflags", "frag_keyframe+empty_moov", 0);

进而：

avformat_write_header(formatContext, &opts);

正如这个问题 如何以编程方式生成碎片 mp4 文件