Recherche avancée

Sur d’autres sites (6685)

• FFmpeg starting manually but not with Systemd on boot

  23 juin 2021, par eKrajnak

  On Raspberry Pi 4 B 4GB with official Debian 10 image, I have /home/pi/run.sh script with following :

  


  #!/bin/bash
ffmpeg -nostdin -framerate 15 -video_size 1280x720 -input_format yuyv422  -i /dev/video0 -f alsa -i hw:Device \
    -af acompressor=threshold=-14dB:ratio=9:attack=10:release=1000 -c:a aac -ac 2 -ar 48000 -ab 160k \
    -c:v libx264 -pix_fmt yuv420p -b:v 3M -bf 1 -g 20 -flags +ilme+ildct -preset ultrafast \
    -streamid 0:0x101 -streamid 1:0x100 -mpegts_pmt_start_pid 4096 -mpegts_start_pid 0x259 -metadata:s:a:0 language="" -mpegts_service_id 131 -mpegts_transport_stream_id 9217 -metadata provider_name="Doesnt matter" -metadata service_name="Doesnt matter" \
    -minrate 3500 -maxrate 3500k -bufsize 4500k -muxrate 4000k  -f mpegts "udp://@239.1.67.13:1234?pkt_size=1316&bitrate=4000000&dscp=34" -loglevel debug < /dev/null > /tmp/ff3.log 2>&1


  


  Script is starting from console without problems. It takes audio from USB sound card and video from USB camera and creates UDP stream to IPTV. Then I created Systemd service :

  


  [Unit]
Description=Streamer
After=multi-user.target sound.target network.target

[Service]
ExecStart=/home/pi/run.sh
KillMode=control-group
Restart=on-failure
TimeoutSec=1

[Install]
WantedBy=multi-user.target
Alias=streaming.service


  


  After restarting Raspberry, script has started, but FFmpeg hangs on error failures in log :

  


  cur_dts is invalid st:0 (257) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
cur_dts is invalid st:1 (256) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
cur_dts is invalid st:0 (257) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
cur_dts is invalid st:1 (256) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
cur_dts is invalid st:0 (257) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
cur_dts is invalid st:1 (256) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
cur_dts is invalid st:0 (257) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)
cur_dts is invalid st:1 (256) [init:1 i_done:0 finish:0] (this is harmless if it occurs once at the start per stream)


  


  and will not start streaming to UDP target. But, if I manually login to SSH and issue systemctl stop streaming and then systemctl start streaming Ffmpeg starts successfully. What's different with service auto-start on boot ?

  


  Setting the "sleep timeout" at script begginging will not help. However, removing audio stream from FFmpeg config looks to solve auto-start on boot.

  


• Pass embedded closed captions with FFmpeg

  4 décembre 2020, par Andrew

  I am taking an HLS stream and outputting it through a DeckLink card to convert it to SDI. I need to pass the CEA-708 closed captions that are embedded in the stream, and it has to happen live. This is the command I'm currently using :

  



  ffmpeg.exe -loglevel info -re -i https://my-url.m3u8 -c:v wrapped_avframe -pix_fmt uyvy422 -vf scale=1280x720 -c:a pcm_s16le -ar 48000 -ac 2 -f decklink -s 1280x720 -r 60000/1001 -b:v 4M "DeckLink SDI (2)"

  



  I have tried adding the -teletext all command to no avail.

  



  Here is the output of FFprobe :

  



  Input #0, hls, from 'https://my-url.m3u8':
  Duration: N/A, start: 83023.890600, bitrate: N/A
  Program 0
    Metadata:
      variant_bitrate : 998000
    Stream #0:0: Video: h264 (High) ([27][0][0][0] / 0x001B), yuv420p(progressive), 640x360, Closed Captions, 29.97 fps, 29.97 tbr, 90k tbn, 59.94 tbc
    Metadata:
      variant_bitrate : 998000
    Stream #0:1(eng): Audio: aac (LC) ([15][0][0][0] / 0x000F), 44100 Hz, stereo, fltp, 158 kb/s (default)
    Metadata:
      variant_bitrate : 998000
      comment         : Master Audio
    Stream #0:2: Data: timed_id3 (ID3  / 0x20334449)
    Metadata:
      variant_bitrate : 998000
  Program 1
    Metadata:
      variant_bitrate : 658000
    Stream #0:3: Video: h264 (High) ([27][0][0][0] / 0x001B), yuv420p(progressive), 480x270, Closed Captions, 29.97 fps, 29.97 tbr, 90k tbn, 59.94 tbc
    Metadata:
      variant_bitrate : 658000
    Stream #0:4(eng): Audio: aac (LC) ([15][0][0][0] / 0x000F), 44100 Hz, stereo, fltp, 158 kb/s (default)
    Metadata:
      variant_bitrate : 658000
      comment         : Master Audio
    Stream #0:5: Data: timed_id3 (ID3  / 0x20334449)
    Metadata:
      variant_bitrate : 658000
  Program 2
    Metadata:
      variant_bitrate : 1958000
    Stream #0:6: Video: h264 (High) ([27][0][0][0] / 0x001B), yuv420p(progressive), 960x540, Closed Captions, 29.97 fps, 29.97 tbr, 90k tbn, 59.94 tbc
    Metadata:
      variant_bitrate : 1958000
    Stream #0:7(eng): Audio: aac (LC) ([15][0][0][0] / 0x000F), 44100 Hz, stereo, fltp, 158 kb/s (default)
    Metadata:
      variant_bitrate : 1958000
      comment         : Master Audio
    Stream #0:8: Data: timed_id3 (ID3  / 0x20334449)
    Metadata:
      variant_bitrate : 1958000
  Program 3
    Metadata:
      variant_bitrate : 2658000
    Stream #0:9: Video: h264 (High) ([27][0][0][0] / 0x001B), yuv420p(progressive), 1280x720, Closed Captions, 29.97 fps, 29.97 tbr, 90k tbn, 59.94 tbc
    Metadata:
      variant_bitrate : 2658000
    Stream #0:10(eng): Audio: aac (LC) ([15][0][0][0] / 0x000F), 44100 Hz, stereo, fltp, 158 kb/s (default)
    Metadata:
      variant_bitrate : 2658000
      comment         : Master Audio
    Stream #0:11: Data: timed_id3 (ID3  / 0x20334449)
    Metadata:
      variant_bitrate : 2658000
Unsupported codec with id 100358 for input stream 2
Unsupported codec with id 100358 for input stream 5
Unsupported codec with id 100358 for input stream 8
Unsupported codec with id 100358 for input stream 11


  



  As you can see, closed captions are embedded in each of the VBR streams, but are getting stripped during the transcode to SDI. How can I pass these to view them ?

  


• AVFrame with NV12 format breaks using D3D11 HW acceleration

  5 décembre 2024, par faith0058

  I have a ffmpeg-autogen video decoder, im trying to use a D3D11 as HW accelerator, but the frame i get in result(with NV12 format) is broken. If im doing the same with Vulkan as accelerator, everything works fine, but vulkan uses more graphic card resources..

  


  For rendering i use OpenTK, received frames i convert to RGB with textures and shaders.

  


  Below my decoder initialization.

  


  public VideoStreamDecoder(AVCodecParameters* parameters, AVCodec* codec)
{
    AVCodec* _codec = codec;
    
    _pCodecContext = ffmpeg.avcodec_alloc_context3(_codec);

    ffmpeg.av_hwdevice_ctx_create(&_pCodecContext->hw_device_ctx, AVHWDeviceType.AV_HWDEVICE_TYPE_D3D11VA, null, null, 0).ThrowExceptionIfError();

    ffmpeg.avcodec_parameters_to_context(_pCodecContext, parameters).ThrowExceptionIfError();
    ffmpeg.avcodec_open2(_pCodecContext, _codec, null).ThrowExceptionIfError();

    CodecName = ffmpeg.avcodec_get_name(_codec->id);
    FrameSize = new Size(_pCodecContext->width, _pCodecContext->height);
    PixelFormat = _pCodecContext->pix_fmt;

    _pFrame = ffmpeg.av_frame_alloc();
    _receivedFrame = ffmpeg.av_frame_alloc();
}


  


  And frames reading function

  


  public bool TryReadNextFrame(out AVFrame frame, AVPacket packet)
{
    int error;

    do
    {
        ffmpeg.avcodec_send_packet(_pCodecContext, &packet).ThrowExceptionIfError();

        error = ffmpeg.avcodec_receive_frame(_pCodecContext, _pFrame);

    } while (error == ffmpeg.AVERROR(ffmpeg.EAGAIN));

    error.ThrowExceptionIfError();

    ffmpeg.av_hwframe_transfer_data(_receivedFrame, _pFrame, 0);
    //here i get NV12 frame

    var clonedFrame = ffmpeg.av_frame_clone(_receivedFrame);

    frame = *clonedFrame;

    return true;
}


  


  And here is result i get with both accelerators :

  


  Vulkan
normally working

  


  The same camera, but D3D11
broken

  


  One more intresting moment, is that cameras, that dont work have yuv420p format, the only one working with D3D11 has yuvj420.

  


  I tried changing format using sws_scale from NV12 to NV12, and it worked, but it uses too much CPU.
Also tried changin format to rgb before av_hwframe_transfer_data function and rendering it using Vulkan rendering, it sort of works with Vulkan, but with D3D11 it doesnt.

  


  UPDATE

  


  I noticed, that frame linesize was bigger, than width and tried

  


  _receivedFrame->linesize[0] = _receivedFrame->width;


  


  And it helped, now image looks better, but still not perfect.

  


  UPDATE 2

  


  I triead also
_receivedFrame->linesize[1] = _receivedFrame->width; and now everything works just as it should, took me whole day to write two lines of code :)