Recherche avancée

Sur d’autres sites (6306)

• fftools/ffmpeg : add InputStream.index

  19 mai 2023, par Anton Khirnov

  fftools/ffmpeg : add InputStream.index
  This allows to avoid access to the underlying AVStream in many places.
  

  • [DH] fftools/ffmpeg.c
  • [DH] fftools/ffmpeg.h
  • [DH] fftools/ffmpeg_dec.c
  • [DH] fftools/ffmpeg_demux.c
  • [DH] fftools/ffmpeg_filter.c
  • [DH] fftools/ffmpeg_mux_init.c

• Using libavformat to mux H.264 frames into RTP

  18 septembre 2021, par DanielB6

  I have an encoder that produces a series of H.264 I-frames and P-frames. I'm trying to use libavformat to mux and transmit these frames over RTP, but I'm stuck.

  



  My program sends RTP data, but the RTP timestamp increments by 1 each successive frame, instead of 90000/fps. It also doesn't look like it's doing the proper framing for H.264 NAL, since I can't decode the stream as H.264 in Wireshark.

  



  I suspect that I'm not setting up the codec information properly, but it appears in many places in the output format context, so it's unclear what exactly needs to be setup. The examples seem to all copy codec context info from encoders, which isn't my use case.

  



  This is what I'm trying :

  



  int main() {
    AVFormatContext context = avformat_alloc_context();

    if (!context) {
        printf("avformat_alloc_context failed\n");
        return;
    }

    AVOutputFormat *format = av_guess_format("rtp", NULL, NULL);

    if (!format) {
        printf("av_guess_format failed\n");
        return;
    }

    context->oformat = format;

    snprintf(context->filename, sizeof(context->filename), "rtp://%s:%d", "192.168.2.16", 10000);

    if (avio_open(&(context->pb), context->filename, AVIO_FLAG_READ_WRITE) < 0) {
        printf("avio_open failed\n");
        return;
    }

    stream = avformat_new_stream(context, NULL);

    if (!stream) {
        printf("avformat_new_stream failed\n");
        return;
    }

    stream->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
    stream->codecpar->codec_id = AV_CODEC_ID_H264;
    stream->codecpar->width = 1920;
    stream->codecpar->height = 1080;

    avformat_write_header(context, NULL);

    ...
    write packets
    ...
}


  



  Example write packet :

  



  int write_packet(uint8_t *data, int size) {
    AVPacket p;
    av_init_packet(&p);
    p.data = buffer;
    p.size = size;
    p.stream_index = stream->index;

    av_interleaved_write_frame(context, &p);
}


  



  I've even went so far to build in libx264, find the encoder, and copy the codec context info from there into the stream codecpar, with the same result. My goal is to build without libx264, and any other libs that aren't required, but it isn't clear whether libx264 is required for defaults such as time base.

  



  How can the libavformat RTP muxer be initialized to properly send H.264 frames over RTCP+RTP ?

  


• Using libavformat to mux H.264 frames into RTP

  22 novembre 2016, par DanielB6

  I have an encoder that produces a series of H.264 I-frames and P-frames. I’m trying to use libavformat to mux and transmit these frames over RTP, but I’m stuck.

  My program sends RTP data, but the RTP timestamp increments by 1 each successive frame, instead of 90000/fps. It also doesn’t look like it’s doing the proper framing for H.264 NAL, since I can’t decode the stream as H.264 in Wireshark.

  I suspect that I’m not setting up the codec information properly, but it appears in many places in the output format context, so it’s unclear what exactly needs to be setup. The examples seem to all copy codec context info from encoders, which isn’t my use case.

  This is what I’m trying :

  int main() {
  
      AVFormatContext context = avformat_alloc_context();
  
  
  
      if (!context) {
  
          printf("avformat_alloc_context failed\n");
  
          return;
  
      }
  
  
  
      AVOutputFormat *format = av_guess_format("rtp", NULL, NULL);
  
  
  
      if (!format) {
  
          printf("av_guess_format failed\n");
  
          return;
  
      }
  
  
  
      context->oformat = format;
  
  
  
      snprintf(context->filename, sizeof(context->filename), "rtp://%s:%d", "192.168.2.16", 10000);
  
  
  
      if (avio_open(&(context->pb), context->filename, AVIO_FLAG_READ_WRITE) < 0) {
  
          printf("avio_open failed\n");
  
          return;
  
      }
  
  
  
      stream = avformat_new_stream(context, NULL);
  
  
  
      if (!stream) {
  
          printf("avformat_new_stream failed\n");
  
          return;
  
      }
  
  
  
      stream->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
  
      stream->codecpar->codec_id = AV_CODEC_ID_H264;
  
      stream->codecpar->width = 1920;
  
      stream->codecpar->height = 1080;
  
  
  
      avformat_write_header(context, NULL);
  
  
  
      ...
  
      write packets
  
      ...
  
  }

  Example write packet :

  int write_packet(uint8_t *data, int size) {
  
      AVPacket p;
  
      av_init_packet(&p);
  
      p.data = buffer;
  
      p.size = size;
  
      p.stream_index = stream->index;
  
  
  
      av_interleaved_write_frame(context, &p);
  
  }

  I’ve even went so far to build in libx264, find the encoder, and copy the codec context info from there into the stream codecpar, with the same result. My goal is to build without libx264, and any other libs that aren’t required, but it isn’t clear whether libx264 is required for defaults such as time base.

  How can the libavformat RTP muxer be initialized to properly send H.264 frames over RTCP+RTP ?