
Artificial Matt
ABOUT
Eight original characters, eight single takes, about ten seconds each. A short sampler of talking performance set to an original score. An old man in a white shearling collar. A porcelain figure in cobalt-painted armor. A woman wrapped in a patterned blue shawl. A bearded man with a red, inked face. A fisherman in a yellow oilskin. A horned, ash-pale creature. A red-masked warlord in a moonlit garden. A gaunt man in black. Each one speaks, listens or turns in one continuous shot, with nothing cut away to hide the seams.
SELECTED FRAMES — 8 OF 8
PROCESS
This piece was a test of one thing: whether a generated character can hold a close-up while speaking. Every shot is a single continuous take, framed as a studio portrait, so there is no edit to cover a bad mouth, a warped hand or a drifting face. Camera direction was limited to one slow, deliberate move per shot: a push-in, a slight drift or a held frame. Anything more and faces and costume detail started to warp. The masked and horned characters were the hardest, because there is less face for the model to work with and the mouth has to read through make-up and prosthetic-style detail. Lip synched dialogue was generated in Seedance. I threw away any take where the mouth drifted off the words, the eyes went dead between lines, or the head turned faster than a person would. Lighting was kept to one soft key and dark, plain backgrounds so the faces carry the frame. The fisherman is the single exception: the yellow backdrop is there to break the rhythm halfway through. The score is original, made in Suno.


