Recording your screen on (SUSE) Linux

It is 2020 and this is still not as easy as it should be. Somehow recording my screen with myself on webcam is a pain. Possibly the easiest way is actually to use Jitsi or similar. That is, to use your web browser; Chromium or FireFox.

It is now 2022 and I am now using OBS Studio with wlrobs on Sway/wlroots/Wayland. This is much better than the other options listed here.

However I have this strange desire to do things locally without FireFox. At least for now anyway. FireFox is an interesting virtual machine and development environment xD.

I’m using openSuSE Tumbleweed, but the package names are the same on most RPM based distros or similar on Debian.

Outline

vokoscreenNG Recording the screen and web cam
H.264 encoding for video and Vorbis for audio
VLC for video playback
ffmpeg for clipping, cropping, concatenating and re-encoding the videos

Alternatively to vokoscreenNG:

guvcview for displaying the webcam video (and crashing X, probably)
Simplescreenrecorder for recording the screen

Setup

Codecs are the bane of Linux distributions everywhere and openSUSE is no exception. H.264 is by far the best that I have tried and ofcourse it is on the naughtly list. So the first thing you must do is enable the packman or VLC repository in Yast or with Zypper and force an update of any existing multimedia pacakges to that repository.

Then we need to add some packages like:

sudo zypper in gstreamer-plugins-bad-orig-addon gstreamer-plugins-ugly-orig-addon
sudo zypper in simplescreenrecorder vlc guvcview ffmpeg

I’m actually not sure which package contains the H.264 gubbins which Simplescreenrecorder uses because I installed so much crap in the process of discovery.

Recording

I think vokoscreenNG (or Simplescreenrecorder and guvcview) are fairly self explanatory. Just start them and click around or look at the --help if you can’t stand GUIs. Just note that vokoscreenNG/Simplescreenrecorder may not default to H.264.

Previously on my videos I was trying to do them in one shot, like a live presentation. Which is good practice, but time consuming (and I was experiencing some crashes), so I now set Simplescreenrecorder/vokoscreenNG to save the video as chunks and I clip them and stitch them together afterwards.

For webcam I simply have ~~guvcview~~ vokoscreenNG display the output on the desktop. guvcview works OK except that the graphics driver occasionally throws a wobbly due to a buffer being filled or something and crashes X. Probably I should do something about that…

I also tried recording with ffmpeg, but performance seemed worse, audio and video were out of sync and so on. vokoscreenNG is not without problems either, at least not on the i3 desktop where it is unable to select a subsection of the screen for recording. So I had to crop the video afterwards.

Editing

I really didn’t want to lay ruin upon my package manager by installing 5 different video editing packages to see what worked. So I just use a combination of VLC to watch the video and ffmpeg to concatenate the clips together.

Failed attempt

This failed because inpoint and endpoint had no appreciable effect

As befits a staple, bloated, all things to all people tool like ffmpeg there are at least three different ways to join some videos together. I’m not doing anything fancy, so I use the simplest.

tee vids.txt <<EOF
file vid1.mkv
inpoint 00:00:01.000
outpoint 00:00:05.000
file vid2.mkv
file vid3.mkv
outpoint 00:00:30.000
EOF
ffmpeg -f concat -i vids -c copy final.mkv

~~I can clip the videos with inpoint and endpoint~~. I’m not interested in making everything perfect, but I think it is best not to waste too much of the viewers time with stuff that can be easily cut.

Successfull attempt

The simplest way (IMO), didn’t work so I had to resort -filter_complex. The name is accurate; it appears ffmpeg implements its own stream processing language to connect filters together. After some time I was able to figure out how to chain the concat filter with crop.

The below concatenates the three input video segments and crops the bottom 1% of the screen.

ffmpeg -ss 1 -t 1:03 -i p0.mkv \
       -ss 2 -t 13:30 -i p1.mkv \ 
       -ss 3 -t 07:27 -i p2.mkv \
       -filter_complex 'concat=a=1:n=3 [v][a]; [v] crop=in_w:in_h*0.99:0:0' \
       -map '[a]' \
       -c:a libvorbis -c:v libx264 fzsync.mkv

Clipping was easier to figure out, for that you just need to specify -ss and -t on each input (-i) file.

Using the filter means we must re-encode the video, which is slow, but had the added benefit of halving the overall size. So you probably shouldn’t use the copy codec with concat demuxer anyway as it won’t have the best end result.

OBS

If you are on Tumbledweed then you will still need the packman repository. Also you will need to install obs-studio-devel, wayland-devel, gcc and meson.

To record a Wayland session we need the wlrobs obs plugin. It’s relatively easy to compile and install as it has few dependencies. However there is no package on Tumbleweed as far as I know.

OBS Studio is what FOSDEM recomends speakers use to record their presentations and it works pretty well.