Rotoscoping at 16K without losing your mind

The roto problem

Rotoscoping is the art of cutting a subject out of its background, frame by frame. At 24 frames per second, a ten-second shot is 240 frames. At 16K, each frame is roughly 150 megapixels.

Manual roto at that scale is not feasible. A single shot can take a week of an artist's time, and the result is only as good as the artist's patience on frame 237.

Edge propagation

ZHiCKROTO's core innovation is per-frame edge propagation. You roto one frame — the keyframe — and the engine propagates that matte across the sequence by tracking the edges of the subject.

The tracking is motion-blur aware. When an edge moves quickly through a frame, it smears across multiple pixels. ZHiCKROTO recognises this smear as a single edge in motion, not a fuzzy boundary, and tracks it accordingly.

It is also occlusion aware. When a subject passes behind an object, the edge disappears and reappears. The engine maintains the matte through the occlusion by reasoning about depth — the subject is still there, it is just hidden.

Surviving 16K

A matte that looks correct at 1080p can fall apart at 16K because every edge artefact is magnified sixteen times. The key to surviving 16K is sub-pixel precision: the matte edge must be placed within a fraction of a pixel of the true edge, not approximately on it.

ZHiCKROTO achieves this by working in float space internally. The matte is not a binary mask; it is a coverage map with values between 0 and 1. The edge is a gradient, not a cliff, and that gradient composites cleanly at any resolution.

The zero-keystroke promise

The result is a matte that needs no roto-paint fixes. You keyframe once, propagate, and the sequence is done. That is the promise, and after two years of production use it holds for roughly 94% of shots — the remaining 6% still need a human touch, but even those are minutes of work, not weeks.