We used an open-source motion model to process Muybridge’s images because the closed-source models, with their superior quality and ‘realism,’ rejected the images, deeming them pornographic.
We tasked the animDiff model with creating perfect 16-frame loops of motion based on the provided inputs—much like a human animator performing the basic exercise of creating a walk cycle. The model reanimated the photographic traces of long-perished bodies in a wonderfully clumsy, inhuman way, as if it were an animator who had never seen an actual person move and had to imagine it.
We believe there’s a poetic justice in this process. Over 200 years ago, Muybridge classified and froze bodies against a grid to study their movements, but through our efforts and the model’s interpretation, we have made them mysterious and fluid once again.