VideoMemory: Toward Consistent Video Generation via Memory Integration

TL;DR

VideoMemory introduces an entity-centric framework with a dynamic memory bank to improve consistency of characters, props, and environments in long-form narrative video generation, addressing identity preservation challenges.

Contribution

The paper proposes VideoMemory, a novel memory-augmented approach that maintains entity consistency across shots in narrative videos, integrating structured planning with visual synthesis.

Findings

Achieves high entity-level coherence in multi-shot videos.

Outperforms existing models in preserving identity and appearance.

Demonstrates effectiveness on a new multi-shot consistency benchmark.

Abstract

Maintaining consistent characters, props, and environments across multiple shots is a central challenge in narrative video generation. Existing models can produce high-quality short clips but often fail to preserve entity identity and appearance when scenes change or when entities reappear after long temporal gaps. We present VideoMemory, an entity-centric framework that integrates narrative planning with visual generation through a Dynamic Memory Bank. Given a structured script, a multi-agent system decomposes the narrative into shots, retrieves entity representations from memory, and synthesizes keyframes and videos conditioned on these retrieved states. The Dynamic Memory Bank stores explicit visual and semantic descriptors for characters, props, and backgrounds, and is updated after each shot to reflect story-driven changes while preserving identity. This retrieval-update mechanism enables consistent portrayal of entities across distant shots and supports coherent long-form generation. To evaluate this setting, we construct a 54-case multi-shot consistency benchmark covering character-, prop-, and background-persistent scenarios. Extensive experiments show that VideoMemory achieves strong entity-level coherence and high perceptual quality across diverse narrative sequences.