Paper Overview
Field: Computer Vision (CV) Authors: Yifan Wang, Tong He Posted: 2026-05-14 arXiv: 2605.15182
Original Abstract (English)
Camera-controlled video generation has made substantial progress, enabling generated videos to follow prescribed viewpoint trajectories. However, existing methods usually learn camera-specific conditioning through camera encoders, control branches, or attention and positional-encoding modifications, which often require post-training on large-scale camera-annotated videos. Training-free alternatives avoid such post-training, but often shift the cost to test-time optimization or extra denoising-time guidance. We propose Warp-as-History, a simple interface that turns camera-induced warps into camera-warped pseudo-history with target-frame positional alignment and visible-token selection. Given a target camera trajectory, we construct camera-warped pseudo-history from past observations and fee...
*(Abstract truncated at source; the full text is available on arXiv.)*
Key Points (from available abstract)
- Problem: Camera-controlled video generation methods typically rely on camera encoders, control branches, or attention/positional-encoding modifications, requiring costly post-training on large camera-annotated video datasets.
- Training-free alternatives avoid post-training but shift the computational cost to test-time optimization or additional denoising-time guidance.
- Proposed method: Warp-as-History — a simple interface that converts camera-induced warps into *camera-warped pseudo-history*, using target-frame positional alignment and visible-token selection.
- Mechanism: Given a target camera trajectory, the model constructs camera-warped pseudo-history from past observations, aiming for generalizable camera control.
*Auto-collected on 2026-05-15.*