Something on the screen is going right. Put a video on his monitor and that is what he is playing — partway through he turns round and tells you about it. Drop in a voice clip and he says yours — the mouth follows the recording. Add a tune if you want one under it. Nothing leaves your device.
The monitor starts empty — drop any video on it and that is what plays, with its own
sound. He stops typing once the shot is about the screen, so you can hear it, and the two sliders set
how loud it is on the screen and off it. The export runs in real time rather than as fast as it can, because
that is the only way the video on the monitor is really playing in the file. The desk and the man are
generated in Blender by hype-src/build.py; there is no .blend anywhere.