Animated comedy shorts and interactive experiments — keyframed, generated, lip-synced, and assembled with real pipeline work. No render farms, no crews: AI models doing the heavy lifting, every step documented.
Hand-specified, model-built web experiments. Everything runs client-side — no server, no build step.
Procedural geometry on a parametric spine — two strands wound half a turn out of phase, base-pair rungs connecting the correct points at every step. Mobile-first.
Launch →Eight planets, Saturn's rings, a moon tracking Earth, visible orbits — nested motion and a camera that won't flip or trap itself. Smooth on phone and desktop.
Launch →Complete playable Tetris — collision, rotation, line clears, scoring, levels, touch controls — built from a single cold spec for about five cents.
Launch →Character scenes generated frame-by-frame and stitched into sequences. Click a poster to watch — hosted on YouTube so this page stays feather-light.
Every clip runs the same assembly line. The models change; the steps don't.
Character and scene stills generated first, then reviewed and picked by eye.
Image-to-video models bring each keyframe to motion, shot by shot.
TTS voices meet sync models — dialogue lands on the frames that matter.
Concat, color-check, final cut. Human eyes on every second before it ships.