Mystic Substance and Technology Studies

This work is based on the Generative Adversarial Network StyleGAN3 to explore the nostalgic, psychedelic, and both strange and familiar feelings found in the dreamcore aesthetic. Instead of recreating real scenes, I wanted to see how a machine could reorganize the visual structure of memories.

I built a dataset of images showing abandoned children’s places from the early 2000s, such as swimming pools, amusement parks, and schools. I trained the model from scratch so it could learn the visual features of these places during generation. The result shows that a surreal atmosphere is created, placing each scene somewhere between memory, dream, and hallucination.

2023
Video 1′04″