Monocular world-space human recovery methods perform well on standard benchmarks, but those benchmarks rarely include fast camera motion. To test this gap, the authors introduce the K-Pop stress test, using broadcast dance videos that combine rapid, large camera movements with dense multi-person choreography. Leading methods produce plausible per-frame meshes on these videos, yet their world-space trajectories drift far from the performers' actual stage positions.
Through systematic diagnosis, the authors show that the camera and human estimates each remain self-consistent. The drift appears when they are combined at mismatched scales: each is recovered at its own scale, so a mismatch between the two can turn camera motion into human drift that grows with camera displacement. This becomes severe in broadcast dance videos, where camera movement is large.
The authors derive a closed-form correction that estimates the scale mismatch and applies a single scale to all performers, removing the drift while keeping their relative positions in one shared world frame. The correction requires no retraining and works with existing backends. Combined with physical constraints, it yields stable world-space reconstruction for choreography modeling, scene editing, data generation, and virtual reshooting.