MVAgent: Multi-Agent Video Generation via Consistent Condition Construction and Shot-Level Policy Optimization

2026-09-24

AI Summary

This study addresses the challenge of cross-shot inconsistencies in character appearance, spatial layout, and state within multi-shot video generation by proposing a multi-agent collaborative framework. Methodologically, it achieves consistency control through typed conditional inputs and ensures spatiotemporal coherence via camera-traversal-based spatial anchoring and continuity memory mechanisms. Furthermore, the framework introduces a novel Trunk-GDPO strategy integrated with a frozen generator architecture for optimization. Experimental results demonstrate that the proposed method achieves state-of-the-art performance on the ViMax-Bench benchmark in terms of both cross-shot consistency and narrative planning quality, while also securing significant advantages in human preference evaluations.

Abstract

Multi-shot agentic video generation requires consistent character appearance, stable spatial layout across camera angles, and continuous character state between shots. When every shot is a separate request to a frozen generator, repeated text does not determine appearance, layout or state. We therefore recast the problem as condition construction and present MVAgent, a multi-agent pipeline whose agents collaborate through typed conditioning inputs. Because an environment image shows one viewpoint, a Spatial Grounding agent samples views from generated camera-traversal clips and anchors each shot to the view matching its framing. As generated shots drift from the plan, an Observer records how each shot ends in a continuity memory, from which a Transition agent builds character action and spatial references for the next shot. An Orchestrator composes these inputs into each request. Since a request reveals its effect only after rendering, we train it by agentic reinforcement learning with Trunk-GDPO, which compares rendered candidates at every shot rather than once per video and continues the best as the trunk. With generator and judges frozen, MVAgent attains the highest cross-shot consistency and narrative-planning quality among the compared methods on ViMax-Bench and is preferred over the strongest agentic baseline in human evaluation.

Problem

• multi-shot video generation
• cross-shot consistency
• character appearance
• spatial layout
• condition construction

Innovation

• Multi-Agent Video Generation
• Condition Construction
• Shot-Level Policy Optimization
• Trunk-GDPO
• Agentic Reinforcement Learning