RESONANCE
an interface for steering generative video: closed human-in-the-loop generation.
┌──────────────────────────────────────┐
│ │
▼ │
render ──► generate ──► re-ingest ──┘
(canvas) (video model) (next frame
conditions
the next)Most generative video is open-loop: you write a prompt, wait, and accept or discard what returns. Resonance closes the loop. What is on the canvas is fed back as the condition for the next generation, so the output becomes something you steer continuously rather than re-roll. The interface exists to make that steering direct. You adjust the surface, and the next generation inherits it.
This one is an experiment, not an instrument. Nothing here is measured or benchmarked, and no claim is made about output quality. It demonstrates a control loop and stops there.
Where this one demonstrates, Latent calibrates against real film stock and Teardown № 1 publishes every number with its source file. Those two carry the evidence; this one carries the idea.