VideoAuteur: Towards Long Narrative Video Generation

TL;DR

This paper introduces a new dataset and a novel model for generating long, coherent narrative videos in the cooking domain, addressing the challenge of maintaining visual and semantic consistency over extended sequences.

Contribution

It presents a large-scale cooking video dataset and a Long Narrative Video Director model that improves coherence and alignment in long-form video generation.

Findings

Enhanced visual fidelity and semantic coherence in generated videos

Significant improvements in keyframe quality and alignment

Effective use of text-image embedding integration

Abstract

Recent video generation models have shown promising results in producing high-quality video clips lasting several seconds. However, these models face challenges in generating long sequences that convey clear and informative events, limiting their ability to support coherent narrations. In this paper, we present a large-scale cooking video dataset designed to advance long-form narrative generation in the cooking domain. We validate the quality of our proposed dataset in terms of visual fidelity and textual caption accuracy using state-of-the-art Vision-Language Models (VLMs) and video generation models, respectively. We further introduce a Long Narrative Video Director to enhance both visual and semantic coherence in generated videos and emphasize the role of aligning visual embeddings to achieve improved overall video quality. Our method demonstrates substantial improvements in generating visually detailed and semantically aligned keyframes, supported by finetuning techniques that integrate text and image embeddings within the video generation process. Project page: https://videoauteur.github.io/