Hello guix,

Here is a revised version of the graph I came up with. It is not
styled yet, because it is much bigger, and had no time to do that yet.
I believe it catches the concepts well, however I will add some notes here:

in the screen_recording_cutting_information we can specify to create
an imageset first, then recompose that to get the slideshow like video
instead.
We can simply shortcut the overlayer then to produce a video without
narration video overlay. We can shortcut the compositor to get a
version without audio, and so on.

Most of these transformers are only a few-liners, so don't get overwhelmed :)

Also, we could use a makefile, or some other dependency resolution
mechanisms to control the whole stuff, a

g_bor