Recording screencasts and lectures on OpenBSD and Linux

On Monday 2020-03-16 the buildings of the danish public sector were closed as an emergency response to COVID-19. This includes Aarhus University where I teach two undergraduate courses. The university asks lecturers to move their teaching to digital platforms. As many times before, this requires creative thinking for those of us who do not use Microsoft and Apple products.

I needed a way to record my pdf slideshows while talking over the presentation. Ideally, I also wanted the ability to show the video of my webcam as an overlay in an attempt to make the presentation a bit more engaging when explaining complex parts.

Fortunately, ffmpeg(1) makes it easy to record the screen and laptop audio. I want to keep the fan noise low during recording by applying minimal compression and encoding. The following shell script serves the purpose of starting and stopping recording:

#!/bin/sh
lockfile=/tmp/screenrecord.pid

startrecording() {
    out="$HOME/screenrecord-$(date '+%Y-%m-%d_%H:%M:%S').mkv"
    ffmpeg -y \
        -f x11grab \
        -framerate 60 \
        -s "$(xdpyinfo | grep dimensions | awk '{print $2}')" \
        -i $DISPLAY \
        -f sndio -i default \
        -r 30 \
        -c:v libx264rgb -crf 0 -preset ultrafast -c:a flac \
        "$out" >/dev/null 2>&1 &
    printf '%s' "$!" > "$lockfile"

    sleep 1
    if [ ! -f "$out" ]; then
        echo 'error: ffmpeg recording did not start' >&2
        notify-send -u CRITICAL "${0##*/}" 'ffmpeg recording did not start'
        rm -f "$lockfile"
        exit 1
    fi
}

stoprecording() {
    kill "$(cat "$lockfile")"
    rm -f "$lockfile"
    notify-send "${0##*/}" 'recording ended'
}

if [ -f "$lockfile" ]; then
    stoprecording
else
    startrecording
fi

I have bound the above script to the key binding Alt+r which makes it easy to start and stop recording in my X session.

On Linux systems, the sound driver sndio should be replaced by alsa in the above ffmpeg(1) command. Audio recording is disabled by default on OpenBSD, but can be permanently enabled with the following commands:

# sysctl kern.audio.record=1
# echo kern.audio.record=1 >> /etc/sysctl.conf

On OpenBSD I can show the webcam video feed with the video(1) command. The following script toggles the video feed:

#!/bin/sh
# remember to `chown $USER /dev/video0`
if pgrep video >/dev/null 2>&1; then
    pkill video
else
    nohup video -s 320 >/dev/null 2>&1 &
fi

On Linux, the command mpv /dev/video0 can take place of the video(1) command above. I have the above script bound to the keybinding Alt+v so I can quickly show and hide my face while recording.

I set dwm(1), my window manager, to open the video feed as a floating window on the bottom right of the screen. The full dwm configuration can be found here.

When I am done recording a lecture, I encode and compress the video file to save bandwidth during upload. The following script encodes all input files and reduces file size to roughly 15% without concievable loss in quality:

#!/bin/sh

encode() {
    ffmpeg -y -i "$1" \
        -c:v libx264 -threads 0 -preset faster -pix_fmt yuv420p \
        -c:a aac -crf 10 \
        "${1%.*}_out.mp4"
}

for f in "$@"; do
    encode "$f"
done

If there is a delay between video and audio, this can also be adjusted using ffmpeg(1). I correct for a 0.3 s delay that I encounter when recording on my laptop:

#!/bin/sh

synchronize() {
    ffmpeg -y -i "$1" \
        -itsoffset 0.300 \
        -i "$1" \
        -map 0:v -map 1:a \
        -c copy \
        "${1%.*}_out.${1##*.}"
}

for f in "$@"; do
    synchronize "$f"
done