Recherche avancée

Sur d’autres sites (8793)

• Revision 28201 : - chaine le langue manquante

  28 avril 2009, par marcimat@… — Log

   chaine le langue manquante

• Huge memory leak when filtering video with libavfilter

  29 mai 2017, par Captain Jack

  I have a relatively simple FFMPEG C program, to which a video frame is fed, processed via filter graph and sent to frame renderer.

  Here are some code snippets :

  /* Filter graph here */
  
  char args[512];
  
  enum AVPixelFormat pix_fmts[] = {AV_PIX_FMT_RGB32 };    
  
  AVFilterGraph   *filter_graph;
  
  avfilter_register_all();
  
  AVFilter *buffersrc  = avfilter_get_by_name("buffer");
  
  AVFilter *buffersink = avfilter_get_by_name("ffbuffersink");
  
  AVBufferSinkParams *buffersink_params;
  
  AVFilterInOut *outputs = avfilter_inout_alloc();
  
  AVFilterInOut *inputs  = avfilter_inout_alloc();
  
  filter_graph = avfilter_graph_alloc();
  
  
  
  snprintf(args, sizeof(args),
  
          "video_size=%dx%d:pix_fmt=%d:time_base=%d/%d:pixel_aspect=%d/%d",
  
          av->codec_ctx->width, av->codec_ctx->height, av->codec_ctx->pix_fmt,
  
          av->codec_ctx->time_base.num, av->codec_ctx->time_base.den,
  
          av->codec_ctx->sample_aspect_ratio.num, av->codec_ctx->sample_aspect_ratio.den);
  
  
  
  if(avfilter_graph_create_filter(&av->buffersrc_ctx, buffersrc, "in",args, NULL, filter_graph) < 0) 
  
  {
  
      fprintf(stderr, "Cannot create buffer source\n");
  
      return(0);
  
  }
  
  
  
   /* buffer video sink: to terminate the filter chain. */
  
  buffersink_params = av_buffersink_params_alloc();
  
  buffersink_params->pixel_fmts = pix_fmts;
  
  
  
  if(avfilter_graph_create_filter(&av->buffersink_ctx, buffersink, "out",NULL, buffersink_params, filter_graph) < 0) 
  
  {
  
      printf("Cannot create buffer sink\n");
  
      return(HACKTV_ERROR);
  
  }
  
  
  
    /* Endpoints for the filter graph. */
  
      outputs->name       = av_strdup("in");
  
      outputs->filter_ctx = av->buffersrc_ctx;
  
      outputs->pad_idx    = 0;
  
      outputs->next       = NULL;
  
  
  
      inputs->name       = av_strdup("out");
  
      inputs->filter_ctx = av->buffersink_ctx;
  
      inputs->pad_idx    = 0;
  
      inputs->next       = NULL;
  
  
  
  const char *filter_descr = "vflip";
  
  
  
      if (avfilter_graph_parse_ptr(filter_graph, filter_descr, &inputs, &outputs, NULL) < 0)
  
  {
  
      printf("Cannot parse filter graph\n");
  
      return(0);
  
  }
  
  
  
   if (avfilter_graph_config(filter_graph, NULL) < 0) 
  
  {
  
      printf("Cannot configure filter graph\n");
  
      return(0);
  
  }
  
  
  
  av_free(buffersink_params);
  
  avfilter_inout_free(&inputs);
  
  avfilter_inout_free(&outputs);

  The above code is called by these elsewhere :

  av->frame_in->pts = av_frame_get_best_effort_timestamp(av->frame_in);
  
  
  
  /* push the decoded frame into the filtergraph*/
  
  if (av_buffersrc_add_frame(av->buffersrc_ctx, av->frame_in) < 0) 
  
  {
  
      printf( "Error while feeding the filtdergraph\n");
  
      break;
  
   }
  
  
  
   /* pull filtered pictures from the filtergraph */ 
  
   if(av_buffersink_get_frame(av->buffersink_ctx, av->frame_out) < 0) 
  
   {
  
        printf( "Error while sourcing the filtergraph\n");
  
         break;
  
    }  
  
  
  
  /* do stuff with frame */

  Now, the code works absolutely fine and the video comes out the way I expect it to (vertically flipped for testing purposes).

  The biggest issue I have is that there is a massive memory leak. An high res video will consume 2Gb in a matter of seconds and crash the program. I traced the leak to this piece of code :

  /* push the decoded frame into the filtergraph*/
  
  if (av_buffersrc_add_frame(av->buffersrc_ctx, av->frame_in) < 0) 

  If I bypass the filter by doing av->frame_out=av->frame_in; without pushing the frame into it (and obviously not pulling from it), there is no leak and memory usage is stable.

  Now, I am very new to C, so be gentle, but it seems like I should be clearing out the buffersrc_ctx somehow but no idea how. I’ve looked in official documentations but couldn’t find anything.

  Can someone advise ?

• Increased File Size When Converting MP4 to WebM using FFmpeg

  23 décembre 2024, par kimgijeong

  I am using FFmpeg to convert MP4 to WebM with the following command :

  


  ffmpeg -y -hide_banner -nostats \
-f mov,mp4,m4a,3gp,3g2,mj2 -i "http://127.0.0.1:80/lotteon-low-bitrate.mp4" \
-threads auto -f webm -acodec libopus -b:a 96.059k -vcodec libsvtav1 -preset 11 -pix_fmt yuv420p \
-vf "scale='min(-1, iw)':'min(-1,ih)':force_original_aspect_ratio=decrease,crop=trunc(iw/2)*2:trunc(ih/2)*2" \
"/usr/local/m2/m2temp/xcdrtmp/2052_1.webm"


  


  However, the output webm file size is larger than the source MP4 file. For example :

  


  


  • Source MP4 : 4.6 MB (bit rate : 994,053 bps)

    


  


  • Output WebM : 16 MB (bit rate : 3,902,037 bps)

    


  


  


  I know SVT-AV1 encoder defaults to CRF mode. Due to not specifying the bitrate explicitly, the SVT-AV1 encoder automatically sets the bit_rate. It appears that the encoder is setting it to a much higher value (3,323,104 bps), causing the increase in file size compared to the source MP4 (994,053 bps). Here are the methods i tried to reduce the WebM file size compared to the source MP4 :

  


  

  1. -b:v 994k

  


  


  I tried to match the target bitrate with the source MP4's bitrate to reduce the output size, but encountered the following error :

  


  Svt[error]: Instance 1: Force key frames is not supported for VBR mode Last message r
epeated 2 times [libsvtav1 @ 0x239dd100] Error setting encoder parameters: bad parameter (0x80001005)


  


  Looking at the official documentation, this mode change (from CRF to VBR when setting a target bitrate) appears to be expected behavior. However, the error is puzzling since I haven't set any force keyframe parameters in the FFmpeg command.

  


  

  2. svtav1-params "mbr=994k"

  


  


  The second method i tried was using the svtav1-params "mbr=994k" option to set the maxrate while maintaining CRF mode This method showed some improvement, but the output file size was still larger than the source MP4.

  


  

  • Output WebM : 5MB (bit rate : 1,209,877 bps)

  


  


  The more critical reason why we can't adopt the second method (using svtav1-params "mbr=994k") is that even for the same MP4 source file, the output file size varies slightly with each encoding.

  


  

  3. -b:v 994k -svtav1-params “rc=2:pred-struct=1”(CBR, low delay)

  


  


  The final method I tried was setting the target bitrate while using CBR (Constant Bit Rate) and low-delay mode The default value for pred-structure is 2(random access), but I had to use low-delay mode due to the following error :

  


  Svt[error]: CBR Rate control is currently not supported for SVT_AV1_PRED_RANDOM_ACCESS, use VBR mode


  


  This way was the only approach among those i tried that successfully reduced the output size.

  


  

  • Output WebM : 4.3MB (bit rate : 1,032,373 bps)

  


  


  In summary, I have three questions about this MP4 to WebM conversion issue :

  


  


  1. When setting the target bitrate with -b:v 994k, the switch to VBR mode is expected behavior. However, why does the force keyframe error occur when we haven't explicitly set any force keyframe parameters ?

     


  


  2. Why does the output WebM file size fluctuate when setting maxrate either through -maxrate or svtav1-params "mbr=994k", even when using the same MP4 source file ?

     


  


  3. Besides using -b:v 994k -svtav1-params "rc=2:pred-struct=1" (CBR with low delay), are there any other methods that can guarantee a WebM output size smaller than the source MP4 when converting from MP4 to WebM ?

     


  


  


  I am using a recent version of the SVT-AV1 codec :

  


  Svt[info]: SVT [version]:       SVT-AV1 Encoder Lib 58146ca
Svt[info]: SVT [build]  :       GCC 11.5.0 20240719 (Red Hat 11.5.0-2)   64 bit
Svt[info]: LIB Build date: Oct 28 2024 07:40:59
ffmpeg video svt-av1