Jitsi Meeting Recording Improvement

Hi all,
This is a question related to the Jitsi meeting recordings.

In my understanding about Jibri, it uses a temporary browser instance that is spinning up when a session recording is started. Then the browser window is recorded there. We can provide our own finalize_recording.sh script to specify what needs to be done after the recording stops. (Ex: Export to an S3 bucket.)

This approach has a limitation that only one meeting can be recorded at a time.

  • Is there any technical or functional limitation that prevents merging the video streams in the Jibri level using multimedia service like ffmpeg?
  • Are there any other workarounds that we can try?

Thanks.

as has been stated multiple times on this forum (there is a search function here…), the functional problem is duplicating the whole Jitsi-meet UI (written in Javascript) and in Java (the language used for Jitsi-meet server projects), and keeping them in sync. The approach that you are suggesting, and again you are about the 10000th to ask for it, was the previous one used by the project until they dropped it because it was too much bother to maintain.
Now there is an external project (it was announced guess where ? yes on this forum) that aims to do what you are searching for, it’s claimed it is very near to be functional, however you will have to see for yourself if the result is near enough to what users see on their screen to be considered acceptable, it’s here.

1 Like

Hi @gpatel-fr
I posted based on my understanding and your explanation was simple and clearer than the thread. I will try to read more using the information there.

Thanks.

Each jibri instance can record only one meeting but you may have multiple jibri instances (on the same server or on seperate servers)

1 Like

only by using containers, as you are doing in your installer if I remember correctly. Using Jibri in a container can even allow it to be used on a graphical workstation.

1 Like

Yes, that’s right. All that needs to be done on the host side is to add multiple snd-aloop devices.

1 Like

Thanks @gpatel-fr and @emrah,

So is this the way that the current jibri dan be improved for recording? Or it can be done extending the gstreamer pipelines project according to the mentioned post? Gstreamer IO

I have to set it up and see. (Not sure I understood the whole thread without setting it up.)

I have worked with Agora and have seen how it transcodes streams before pushing to a CDN. Which it handles from the server side. I was wondering why jitsi cannot have a similar technique in having multiple recording sessions if we have the streaming data in jibri.

It’s not ‘improved’, it’s just run several times on the same computer thanks to containers. @emrah’s script uses Linux containers (lxc), other use Docker. Whatever the container, it’s the very same software that can be run in a standard server.

I guess so, I have never tried it myself, but a stated aim is to do recording with less resources than Jibri.