Recherche avancée

Sur d’autres sites (18427)

• swscale : Set alpha to opaque for internal palettes.

  5 avril 2014, par Justin Ruggles

  swscale : Set alpha to opaque for internal palettes.
  Fixes conversion of pal8 to rgb formats with alpha.
  Updated references for 2 FATE tests which previously encoded fully
  
  transparent images.
  Based on a patch by Baptiste Coudurier <baptiste.coudurier@gmail.com>
  

  • [DH] libswscale/swscale_unscaled.c
  • [DH] tests/ref/fate/targa-conformance-CCM8
  • [DH] tests/ref/fate/targa-conformance-UCM8

• Best way to transcode with FFMPEG and RADEO(AMDGPU)

  2 mars 2020, par Colin Bitterfield

  Question :

  1. Is this a correct transcoding command for FFMPEG ?
  2. Why is the error message showing up ?
  3. Does anyone have their Radeon system transcoding large amounts of Videos.

  [ I had this working just fine with a NVIDEO 1080 using cuda and could use multiple ffmpegs at the same time]

  Recently, I am trying to use a RADEON 580X (8GB) for transcoding.
  Every other video has a problem with
  ```amdgpu : The CS has been cancelled because the context is lost.``

  I can’t run more than one ffmpeg transcode without this error message and I can’t run an X Server without this problem.

  OS : Ubuntu 18.04 LTS (Fully Patched)
  RADEON : Pro Drivers installed
  FFMPEG : ffmpeg version N-96728-ge007059 (Compiled on the machine)

  Video that fails quickly :

  Input #0, mov,mp4,m4a,3gp,3g2,mj2, from 'input.mp4':
  
    Metadata:
  
      major_brand     : isom
  
      minor_version   : 512
  
      compatible_brands: isomiso2avc1mp41
  
      creation_time   : 2011-10-07T04:44:57.000000Z
  
      encoder         : Lavf51.12.1
  
    Duration: 00:32:05.21, start: 0.000000, bitrate: 15148 kb/s
  
      Stream #0:0(eng): Video: h264 (Constrained Baseline) (avc1 / 0x31637661), yuv420p, 1920x1080 [SAR 1:1 DAR 16:9], 15039 kb/s, 29.97 fps, 29.97 tbr, 2997 tbn, 48 tbc (default)
  
      Metadata:
  
        creation_time   : 2011-10-07T04:55:41.000000Z
  
        handler_name    : VideoHandler
  
      Stream #0:1(eng): Audio: aac (LC) (mp4a / 0x6134706D), 44100 Hz, stereo, fltp, 105 kb/s (default)
  
      Metadata:
  
        creation_time   : 2011-10-07T04:55:41.000000Z
  
        handler_name    : SoundHandler

  Transcoding Command :

  ~/bin/ffmpeg -hwaccel vaapi  -threads 16 -i input.mp4  -vaapi_device /dev/dri/renderD128 -vcodec h264_vaapi -vf format='nv12|vaapi,hwupload','deinterlace_vaapi=rate=field:auto=1,scale_vaapi=w=-1:h=480,hwupload' -b:v 2M output_480p.mp4 -y

  From this part, I am inferring (decode) is not hardware accelerated :

  Stream mapping:
  
    Stream #0:0 -> #0:0 (h264 (native) -> h264 (h264_vaapi))
  
    Stream #0:1 -> #0:1 (aac (native) -> aac (native))

• Segmentation Fault when calling FFMPEG on Lambda - but not locally

  17 avril 2024, par thiagogps

  I am trying to extract the next 30 seconds of audio from a live video stream on YouTube using AWS Lambda. However, I'm facing an issue where Lambda does not wait for an FFmpeg subprocess to complete, unlike when running the same script locally. Below is a simplified Python script illustrating the problem :

  


  import subprocess
from datetime import datetime

def lambda_handler(event, context, streaming_url):
    ffmpeg_command = [
        "ffmpeg", 
        "-loglevel", "error", 
        "-i", streaming_url, 
        "-t", "30", 
        "-acodec", "pcm_s16le", 
        "-ar", "44100", 
        "-ac", "2", 
        "/tmp/output.wav"
    ]

    print("Starting subprocess...")
    print(f"Start time: {datetime.now()}")
    subprocess.run(ffmpeg_command, capture_output=True)
    print(f"End time: {datetime.now()}")


  


  In this script, I am using FFmpeg to capture audio from the specified streaming_url, converting it to a .wav format, and saving it as output.wav. When executed locally, this script waits until the subprocess finishes, evidenced by the significant time difference between the start and end print statements. However, when run on AWS Lambda, it proceeds almost immediately without waiting for the subprocess to complete, resulting in incomplete audio files.

  


  Question : How can I ensure that AWS Lambda waits for the FFmpeg subprocess to fully execute before the function exits ? I assume I'm not understanding correctly how Lambda handles subprocesses. I even tried adding a time.sleep(30) after the subprocess.run, but that didn't help. Is there a specific configuration or method to handle subprocesses in Lambda correctly ?

  


  EDIT : With the help of the comments, I understood that in fact it's returning quickly in Lambda because of a segmentation fault, since it gives me a returncode of -11, so I edited the question and its title accordingly. Locally, there is no such error. I found out this is a similar situation to Using FFmpeg with URL input causes SIGSEGV in AWS Lambda (Python runtime), but I'm still unable to solve it.