AI Trained on Photos He Took Himself, Not the Internet
Instead of indiscriminately training AI on internet images, one creator trains it using images he shoots himself. SHOWstudio founder Nick Knight is building his own dataset by repeatedly photographing models and clothing directly. Rather than leaving the volume of data to the open internet, he controls which models and which clothes become training material. In effect, the creator controls the starting point of the AI's output.
What he is aiming for is not a tool that plausibly replicates existing photographs. Knight turns still images into moving images, experimenting with a new image medium that sits between photography and film. This is also why he believes AI images should not become direct copies of existing work. His argument is that the novelty of generative technology should be found not in remaking someone else's output, but in an expressive approach that extends photography into movement.
In this experiment, data and rights are not separated. Knight emphasizes that fashion models themselves should hold the rights to their digital likeness, meaning the digital portrait by which their face and body can be identified. Who is being used to train the images, what those images are turned into, and who holds the rights to that outcome are, in his view, the core conditions of AI production.
