“…a deep-dreamily progmented artisticial inteligenerative revisualization architexturizer for mindmelting visioneering.”
Bionic Art 👨🎨+🤖=🤯
The way I like to think about Beholder is that it “interprets” video through an artificial visual cortex configured with a highly restricted bespoke epistemology. Any image it sees, it attempts to see only in the context of what it is designed to know about, which is a model composed of visual concepts directed by the user.
How does it work?
I’ll be writing a lot more about how this works very soon, but until I have time to put that explanation together, what I can say is that Beholder currently uses VQGAN+CLIP heavily. If you are interested in exploring these libraries on your own, check out these two articles (not by me):