Recherche avancée

Sur d’autres sites (5852)

• How can I spawn many ffmpeg processes that takes screenshots of stream every X seconds without killing my CPU ?

  2 novembre 2020, par kinx

  I'm spawning hundreds of ffmpeg processes that takes RTSP streams and outputs images every X seconds. I can handle up to 300 processes of this with child_process.fork and child_process.spawn.

  


  const imageStream = (streamUrl, streamId) => {
  return new Promise((resolve, reject) => {
    let initialBuffer = true;
    let retries = 0;
    let maxRetries = 10;
    const setOptions = (streamUrl, streamId) => {
      return [
        "-rtsp_transport",
        "tcp",
        "-i",
        streamUrl,
        "-update",
        1,
        "-r",
        "1/5",
        "-f",
        "image2",
        "-progress",
        "pipe:1",
        `./public/posters/${streamId}.jpg`,
        "-y",
      ];
    };
    const initStream = (streamUrl, streamId) => {
      const command = spawn("ffmpeg", setOptions(streamUrl, streamId), {
        detached: true,
      });
      command.setMaxListeners(0);
      command.on("exit", (code, signal) => {
        if (code === 1 && initialBuffer) {
          if (retries === maxRetries) {
            command.kill("SIGINT");
          }
          retries++;
          initStream(streamUrl, streamId);
        }
      });
      command.stdout.on("data", (chunk) => {
        if (initialBuffer) {
          initialBuffer = false;
          resolve(true);
        }
        exited = true;
        command.kill("SIGINT");
      });
      process.on("SIGINT", () => {
        command.kill("SIGINT");
        command.removeAllListeners();
        process.exit();
      });
      process.on("exit", () => {
        command.kill("SIGINT");
        command.removeAllListeners();
      });
    };
    initStream(streamUrl, streamId);
  });


  


  I'm using progress buffer to listen for data to tell my app that ffmpeg has connected and to start manipulating that data.

  


  The issue is that this is just too much for me and I need to handle more. I don't really need to stream these as live video, I just need to take image captures every X seconds.

  


  However, it seems like there's a limited amount of child_proocesses I can use, and when I do spawn more ffmpeg processes, my Ryzen 9 CPU goes up to 100% and it freezes until I can kill the terminal.

  


  What's the best way to do this ?

  


• (FFMPEG Error)Real-time buffer [USB cam's name] too full or near too full

  9 juin 2020, par ZzangWoo

  I am doing project about making rtsp stream of my USB Camera.
And my project environment is 
- OS : windows server 2019
- CPU : AMD Ryzen 7 3700X
- RAM : 64GB
- GPU : NVIDIA GeForce RTX 2070 SUPER

  



  My project goal is to detect object with YOLO and show original cam video, detected video to client. So I need to change my USB Camera to RTSP Stream. But I had an error below picture.
enter image description here

  



  And this is my command line.

  



  ffmpeg -re -f dshow -i video="JOYTRON HD20" -pix_fmt yuv420p -vsync 1 -threads 0 -vcodec libx264 -r 30 -g 60 -sc_threshold 0 -b:v 640k -bufsize 768k -maxrate 800k -preset veryfast -profile:v baseline -tune film -acodec aac -b:a 128k -ac 2 -ar 48000 -f rtsp rtsp ://localhost:8888/test

  



  I saw an answer that the problem is bandwidth. So I added parameter. It didnt help.

  



  I also added the -rtbufsize parameter and -thread_queue_size parameter. But it didnt help anything.

  



  What should I do ??

  


• SVT-AV1 : After encoding the video seeking is very bad in any video player (even HTML5)

  18 juin 2020, par Viktor Machnik

  i am using SVT-AV1 and FFMPEG to encode Videos into the AV1 video and opus audio codec (.webm), it works fine, except that the video seeking do not work really (extremly bad). When i seek, the CPU useage jumps up and it takes up to minutes until the seeking process finishes.

  



  Here is how i encode the videos :

  



  

  1. Convert any video into yuv : ffmpeg -i  -preset veryslow -level 6.2 .yuv

  


  2. First AV1 run svt-av1 -i '.yuv' -w  -h  --fps  --rc 0 -q 30 --preset 8 -b '.\output1.ivf' --output-stat-file '.\stat_file.stat' --keyint 1 --enable-restoration-filtering 1

  


  3. Second AV1 run svt-av1 -i '.yuv' -w  -h  --fps  --rc 0 -q 30 --preset 3 -b '.\output.ivf' --input-stat-file '.\stat_file.stat' --keyint 1 --enable-restoration-filtering 1

  


  4. Get source video audio in opus codec ffmpeg -i  -c:a libopus -vn -preset veryslow -level 6.2 output.ogg

  


  5. Get final .webm video ffmpeg -i output.ivf -i output.ogg -c copy output.webm

  


  



  I have allready tryed to play with the —keyin option, also just letting it away and use the encoder default, but the results are always the same. (--keyin 1 seems to work better than without this option, but also very very bad)

  



  What am i doing wrong ?

  



  Extras : I am using Windows 10 with downloaded version of SVT-AV1 and FFMPEG (I just renamed the SVT-AV1 Encoder .exe file to be svt-av1.exe). Used CPU is Ryzen 9 3900X