Mark Hernandez-Motaghy

This model is built from footage of my grandmother shaping masa with an heirloom textile. After years of being copied from one hard drive to another, my video files began to corrupt and decay. As the material is processed through a conditional GAN (pix2pix), the model reconstructs that fragmented, bit-rotting image from learned patterns. Can a model “remember” without retrieving the original? The output is a memory first degraded by hardware, then regenerated by the system.