Recherche avancée

Sur d’autres sites (15714)

• How to put a video inside a frame and record whole thing as a video ?

  7 octobre 2024, par crysis

  I have a video. I want to create a new video such that the video plays inside a frame similar to this screenshot. Loom recording does this. Most of the screen recorders do this.

  


  

  


  This screenshot is of the video. This is done in most of the screen recorders. Here is what I'm thinking

  


  

  • render video on a canvas with the background.

  


  • Play the video

  


  • Use mediarecorder API to record a new video.

  


  


  I don't have good understanding of frontend development. Is that how it is generally done or I need to use ffmpeg ? I would really appreciate any guidance here.

  


• Multiple live video outputs advice. Live stream/Record/Preview, FFMPEG, Windows, Decklink [closed]

  18 septembre 2024, par stroltz

  I am looking for advice on how best to achieve multiple live video outputs.

  


  The live source is a Decklink card on Windows. (We have a ffmpeg build working to access the card) We want 4 outputs ;

  


  


  • We want to run a preview window (low quality would be preferred) just so the user can see the video is working.

    


  


  • We want to be able to live stream - single bit rate, RTMP. (goes up to a CDN)

    


  


  • Independent from the streaming we want to be able to stop and start recording to file. Ideally using CRF. So a separate encode – but maybe we use the RTMP encode, not sure, and do 1 x encode only.

    


  


  • We also want to save a separate audio file. Stops and starts at the same time as the video file above (if required we could do this as a post process on the video file we make above)

    


  


  


  We want to keep CPU use down to as reasonable as possible. (so no high end hardware)

  


  We have had a suggestion of this with ffmpeg ;

  


  Input >> ffmpeg

  


  

  • split input to main and monitoring ;

  


  • scale monitoring stream to lower resolution

  


  • encode both streams

  


  • provide both outputs to local streaming server
ffmpeg >> local streaming server

  


  • use API to start and stop recordings (or web console, if you do it manually)

  


  • provide streams to CDN or/and provide access to your streams for end users

  


  


  recorded files >> another ffmpeg (controlled by some script that get
RECORDING COMPLETED event to start ffmpeg process)

  


  

  • extract audio from recorded file

  


  • save audio into file

  


  


  Which sounds possible, but if doing that, which local streaming server would work best (open source, API...)

  


  or open to other ideas as to the best way.

  


  https://trac.ffmpeg.org/wiki/Creating%20multiple%20outputs shows lots of ways, but I don't think you get to control the individual outputs independently.

  


• ffmpeg : record 1 hour segment, right on the hour, no/little frame lost [closed]

  2 septembre 2024, par Jean-Michel

  I need to record 24/7 an incoming SRT stream into 1 hour chunks, starting right on the hour.
The start needs to be as close as XX:00:00.000 as possible and little to no frame lost between the recordings.

  


  For now I am using a systemd service with that ExecStart command :

  


  /usr/bin/ffmpeg -fflags +discardcorrupt -i 'srt://:4010?pkt_size=1316&mode=listener' -c copy -map 0:v:0 -map 0:a? -f segment -strftime 1 -segment_time 3600 -reset_timestamps 1 -segment_format ts '/somefolder/recording_%%d%%m%%Y_%%H%%M00.ts'


  


  And using a crontab to restart the service on the hour to insure the next file properly starts at the expected time.

  


  0 * * * * cd /etc/systemd/system && systemctl restart srt_recorder_4010.service


  


  Unfortunately, the restart of the service takes 5 to 10 seconds and we are loosing that duration in the recorded file.
If I'm not using the crontab, I get 1 hour files but I have no control on the start time.

  


  Any suggestion to get something more accurate ?
Could there be a way to combine ffmpeg internal ability to exactly segment 1 hour file but enforce to do it on the hour ?
Or externally, if there was a way to send a "next file" command (api, signal) to a perpetual running ffmpeg to change recording file ?

  


  Thanks