Very cool Pilou !
.
I did some research and found out that what I am looking for is
called "Algorithmic Paredolia" performed by digital neural net-
works. The first algorithm of this kind was created by Google and
is called "Deepdream"
.
I wonder if it is the same / modified or a new algorithm used
for the "remix" function of Midjourney which generates pretty
amazing results.
.
Now I am searching for such an offline open-source tool to
merge imagery and expect to find some sophisticated stuff
which is capable to "learn" what I expect it to do by iterative
processes.