Contextual AD Narration with Interleaved Multimodal Sequence Analysis

Uni-AD proposes a unified framework for Audio Description (AD) generation, leveraging multimodal inputs and contextual information to enhance performance.