Paper recorded by Signals 4 on 2026-08-30 in cs.AI. Abstract reproduced from arXiv; link to the original below.
Published 2026-08-30 on arXiv · recorded by Signals 4 on 2026-09-01
Category: cs.AI · 人工智能 · first seen 2026-09-01
Long-horizon story-driven video generation requires a production agent to coordinate narrative decomposition, state tracking, shot design, prompt construction, rendering, and revision across interdependent scenes. Existing adaptive video systems primarily refine requests or reusable skills, leaving recurring production failures disconnected from persistent, stage-targeted improvements across stori