StoryTeller: Training-Free Narrative Grounding for Long-Form Audio Description

2026-07-13

AI Summary

This work addresses the challenge of maintaining narrative coherence across scenes when generating audio descriptions for long videos—a critical issue that hinders visually impaired users’ comprehension of overall storylines. The authors propose a training-free inference framework that leverages a narrative memory bank validated by vision-language models (VLMs) and integrates publicly available movie metadata to generate story-aware audio descriptions without relying on scripts, subtitles, or model fine-tuning. Key innovations include the first fully training-free approach for coherent long-video description generation, a semantic filtering and VLM-based verification mechanism to ensure metadata consistency, and the introduction of StoryAD-QA, a new benchmark for evaluating story understanding. Experiments demonstrate that the method significantly outperforms strong baselines across automatic metrics, question-answering tasks, and human evaluations, markedly improving coherence, factual accuracy, and narrative support.

Abstract

Long-form audio description (AD) requires more than describing visible actions: it must preserve characters, events, relationships, and story context across scenes so that blind and low-vision (BLV) audiences can follow a film. Modern video-language models (VLMs) are effective on short clips, but they often treat each moment independently, producing descriptions that miss who characters are, why events matter, and how the current scene connects to earlier narrative context. We propose StoryTeller, a training-free framework for story-aware long-form AD. Instead of relying only on local visual cues, StoryTeller maintains a verified narrative memory that carries forward story-relevant information across scenes, enabling later descriptions to remain coherent, grounded, and contextually informative. Given only raw video and a movie title, StoryTeller can optionally retrieve public movie metadata to resolve names and story context, while accepting only facts that are supported by the video through semantic filtering and VLM verification. The method requires no subtitles, scripts, AD transcripts, aligned captions, character banks, precomputed face identities, or task-specific fine-tuning. To evaluate whether generated AD preserves narrative information, we introduce StoryAD-QA, a question-answering benchmark that tests whether a language model can answer story-context questions using only the generated descriptions. Experiments on standard AD benchmarks and diverse long-form videos show that StoryTeller consistently improves narrative coherence, factual grounding, and story comprehension over strong baselines in automatic, QA-based, and human evaluations.

Problem

• long-form audio description
• narrative grounding
• video-language models
• story coherence
• blind and low-vision accessibility

Innovation

• training-free
• narrative grounding
• long-form audio description
• verified narrative memory
• video-language models