filmica.

image to text to image

image to text to image

3 min.

United States of America

A sequel of sorts to Road Show, image to text to image is a journey through the latent space of an AI diffusion model using 21 photos as stepping stones. Whenever the image is still, it is a real photograph. Whenever there is movement, it is a computationally hallucinated wormhole through the machine memory of five billion images.

    Crew