ATAS tops all six human-rating criteria for long-form video

"Long-form agentic video generation must turn a screenplay into shots that remain spatially and temporally consistent across nonconsecutive requests." That is the opening of a new arxiv preprint describing ATAS, short for Any Time, Any Space.

The system compiles a screenplay into shot requests with explicit location, camera, and event-time references. Before each shot, it retrieves from two memories: a scene memory of registered static geometry, and an event memory of dynamic observations indexed by scene, event, and time. After the shot renders, ATAS reconstructs and registers the clip, and "updates memory only when review and state checks accept the observation, while film delivery remains independent of memory admission."

The headline numbers are backbone-agnostic. Across four video generators, ATAS "improves all reported cross-shot consistency measures and achieves the highest aggregate benchmark score for each backbone." It also outperforms the compared agents and 3D-guided methods on target-image spatial revisit, leads three event-recapture criteria, and receives the highest human ratings on all six criteria.

The abstract does not name those six criteria, does not publish per-generator numbers, and does not identify which four backbones were tested. Code and video results are posted at github.com/ziplab/ATAS.