Recherche avancée

Sur d’autres sites (6474)

• how to capture html5 video programmatically (headless)

  25 octobre 2019, par ygk

  I am trying to capture html5 video with js/css effects on it programmatically. I tried a couple of methods ;

  1. First I find this great blog post and implemented it. Everything was perfect till I found that phantomjs is not supporting html5 video tag so can not capture the video.

  2. Second option was to use headless chrome to take continuous screenshots and feed these screenshots into ffmpeg to create the video. Although it worked to some level headless chromes screenshots was taking some time.. I couldn’t create a smooth video..

  3. On my third try I gave a chance to chrome’s Page.startScreencast api. It could get video capture but frame rates was really problematic. The reason is that..

  4. Now I am working on xvfb + chrome/firefox + ffmpeg combination for capturing video as mentioned on that comment. Theoretically it is promising but I couldn’t managed to capture a video. Instead I have black screen..

  My setup is below :

  • light-http server having a simple video (web) within html5 video tag ; on localhost
  • start xvfb with Firefox and navigate to localhost/index.html ( video is there )
    xvfb-run --listen-tcp --server-num 44 --auth-file /tmp/xvfb.auth -s "-ac -screen 0 1440x685x24" firefox --headless http://localhost

  • start ffmpeg with x11grab parameter to grab frames from xvfb

    ffmpeg -f x11grab -video_size 1440x685 -i :44 -codec:v libx264 -r 12 ./output.mp4

  • the result is black video :)

  what should be the problem, how can I debug the problem ?

  ---

  ps : there is one more possible solution which I didn’t tried yet. As phantomjs has capability to capture canvas ; it may be possible to

  • put video into canvas
  • then capture the canvas

  It seems like a dirty workaround that is why not tried yet..

  ---

  UPDATE-1

  Tried to get screenshot with xwd -root -silent -display :44 -out screen.xwdand than convert to jpeg convert screen.xwd shot.jpg the result is black jpg..

• Background version of uploaded video creation in Elixir with Arc and FFmpeg

  12 décembre 2017, par Levit

  I tried to make video uploading on Elixir phoenix project using Arc. I need to have different size versions of video. So I used FFmpeg to do this :

  def transform(:large, {file, scope}) do
  
    {:ffmpeg, fn(input, output) -> "-i #{input} -c:a copy -s 1920x1080 -f mp4 #{output}" end, :mp4}
  
  end

  However, it takes a lot of time to create this versions even to small videos. So, I decided to make version creation on background. I want Arc to upload video, then I give the response with uploaded original while other versions are creating.

  I was hoping to find such option at https://github.com/stavro/arc but I didn’t succed. I only found "To disable asynchronous processing, add @async false to your upload definition" but I don’t want to disable it, right ?

  I tried to use Exq https://github.com/akira/exq for background processing but I didn’t menage to use it in uploader.

  Could anybody tell me how it should be made in a proper way or give me some dirty hack advice to make it work. Thanks.

  I tried Task as was advised in comments, but I am not sure how to use them in this case. When I try

  {:ffmpeg, fn(input, output) -> Task.async(fn -> "-i #{input} -c:a copy -s 1920x1080 -f mp4 #{output}" end) end, :mp4}

  or

  {:ffmpeg, Task.async(fn -> fn(input, output) ->  "-i #{input} -c:a copy -s 1920x1080 -f mp4 #{output}" end end), :mp4}

  I got "protocol String.Chars not implemented for %Task".

  When I try

  {:ffmpeg, Task.async(fn(input, output) ->  "-i #{input} -c:a copy -s 1920x1080 -f mp4 #{output}" end), :mp4}

  I got "#Function<20.83953603/2 in MyWebSite.Content.transform/2> with arity 2 called with no arguments". I tried to pass function as an argument with "&" but it fails as well.

  My uploader :

  defmodule MyWebSite.Content do
  
  use Arc.Definition
  
  use Arc.Ecto.Definition
  
  
  
  @acl :public_read
  
  
  
  @versions [:original,  :huge]
  
  
  
  def transform(:huge, {file, scope}) do
  
    {:ffmpeg, Task.async(fn(input, output) ->  "-i #{input} -c:a copy -s 1920x1080 -f mp4 #{output}" end), :mp4}
  
  end
  
  
  
  def s3_object_headers(version, {file, scope}) do
  
    [timeout: 3_000_00, content_type: Plug.MIME.path(file.file_name)]
  
  end
  
  end

• How to use scale filter in doc/examples/transcoding.c

  19 janvier 2018, par siods333333

  I tried to replace the "null" filter with the "scale=320x180" filter in the doc/examples/transcoding.c file, it only resulted into the this error message :

  [libx264 @ 03303ee0] Input picture width (640) is greater than stride (256)
  
  Error occurred: Generic error in an external library

  What is wrong ? Look that init_filters happens after open_output_file, the encoder is already set before it even knows the resolution of the output.

  How to do this properly ?

  Look at this piece of code, I don’t get what it’s talking about, fiters ain’t going to magically set the correct resolution :

          /* In this example, we transcode to same properties (picture size,
  
           * sample rate etc.). These properties can be changed for output
  
           * streams easily using filters */
  
          if (dec_ctx->codec_type == AVMEDIA_TYPE_VIDEO) {
  
              enc_ctx->height = dec_ctx->height;
  
              enc_ctx->width = dec_ctx->width;

  Found this in the ffmpeg.c itself :

      enc_ctx->width  = av_buffersink_get_w(ost->filter->filter);
  
      enc_ctx->height = av_buffersink_get_h(ost->filter->filter);