Recherche avancée

Sur d’autres sites (14347)

• How can I efficiently process a video on a per-request basis ?

  20 décembre 2015, par mipadi

  I’m working on a web application in which a watermark must be applied to a video before it is sent to the user. Currently this watermark is static, and is created using ffmpeg when a video is updated. However, the application is changing so that a unique watermark will be added to the video for every request made for the video. This prevents a problem, as the video files may be fairly large and adding a watermark may be time-consuming (e.g., in some cases it may take over a minute to add a watermark), but the watermarks cannot be added on upload.

  I figured that streaming video could be a solution and implemented a solution using the nginx-rtmp-module, but several problems cropped up :

  1. RTMP solutions are a no-go as they appear to require Flash. This application must be supported on devices that don’t support Flash at all, or don’t (and won’t) have it installed.
  2. I have considered using MPEG-DASH, but that enjoys only limited support. Namely, it is not supported on versions of Firefox targeted by the application, nor is it supported on iOS or some versions of Safari.
  3. I have considered HLS, but that enjoys even more limited support than MPEG-DASH.
  4. Regardless, I haven’t actually been able to get Dash.js (the reference player for MPEG-DASH streams) to work, although that may be due to an encoding issue, I’m not sure.

  I wondered if there is a better (perhaps simpler) solution to this problem ; perhaps streaming video isn’t the way to go at all ? Is there an efficient way to transcode a video file on-the-fly and start sending it to the browser quickly ?

  I am not against using solutions like node.js or other platforms/frameworks, and solutions can use HTML5 <video></video> if necessary.

• Ffmpeg hardware acceleration unsupported formats betwen transpose and auto_scaler_0

  11 août 2021, par Eduard6421

  I am trying to develop a transoding service which makes use of nvidia hardware acceleration capabilities ( The gpu used in this process is a Tesla T4) ; I want to generate a mpeg-dash playlist for my video so that i can stream it ;

  


  ffmpeg -y -hwaccel cuda -hwaccel_output_format cuda -i mobil1.mp4 -c:v h264_nvenc -c:a aac  \
-map v:0 -b:v:0 1000k -maxrate:0 1500k -bufsize:0 2500k -filter:v:0 "scale_npp=1920:1080:force_original_aspect_ratio=decrease" -map 0:a -b:a 128k \
-f dash dash.mpd


  


  But when mobile videos are uploaded ( which have rotation metadata ) I encounter the following error :

  


  Impossible to convert between the formats supported by the filter 'transpose' and the filter 'auto_scaler_0'
Error reinitializing filters!
Failed to inject frame into filter network: Function not implemented
Error while processing the decoded data for stream #0:0


  


  How can i solve this issue ? I am using the following docker image :
jrottenberg/ffmpeg:4.4-nvidia

  


• Python to generate a video mpd file (even a node package) [on hold]

  14 mars 2017, par Junior

  I am in the midst of developing a video streaming app and I’m currently stuck at the Adaptive Bitrate Streaming part.

  Is there a python or node module that can (run server-side and) generate a mpd file of a video after it is uploaded, this way it is easy to link the file to Dash.js.

  So to clarify, I would just upload my video and have the server generate an mpd file for the video. The next time I use Dash.js, i just attach this file.

  Thanks.