Co-Director: Agentic Generative Video Storytelling

TL;DR

Co-Director introduces a hierarchical multi-agent framework for coherent video storytelling using diffusion models, balancing exploration and exploitation to improve narrative consistency.

Contribution

It formalizes video storytelling as a global optimization problem with hierarchical parameterization, addressing semantic drift and cascading failures in agentic pipelines.

Findings

Outperforms state-of-the-art baselines in experiments.

Introduces GenAD-Bench, a new dataset for personalized advertising.

Demonstrates effective generalization to cinematic narratives.

Abstract

While diffusion models generate high-fidelity video clips, transforming them into coherent storytelling engines remains challenging. Current agentic pipelines automate this via chained modules but suffer from semantic drift and cascading failures due to independent, handcrafted prompting. We present Co-Director, a hierarchical multi-agent framework formalizing video storytelling as a global optimization problem. To ensure semantic coherence, we introduce hierarchical parameterization: a multi-armed bandit globally identifies promising creative directions, while a local multimodal self-refinement loop mitigates identity drift and ensures sequence-level consistency. This balances the exploration of novel narrative strategies with the exploitation of effective creative configurations. For evaluation, we introduce GenAD-Bench, a 400-scenario dataset of fictional products for personalized advertising. Experiments demonstrate that Co-Director significantly outperforms state-of-the-art baselines, offering a principled approach that seamlessly generalizes to broader cinematic narratives. Project Page: https://co-director-agent.github.io/