# What, When, and How: Audio Description as Constrained Global Optimization

arXiv: https://arxiv.org/abs/2609.30121 · HTML: https://mlsanity.com/paper/2609.30121/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 24 of its 46 references that are in the map.
- First posted on arXiv 24 Sep 2026.

## Builds on (most cited first, 24 of 24)

- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [Flamingo: a Visual Language Model for Few-Shot Learning](https://mlsanity.com/paper/2204.14198/) (2022)
- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [AuroraCap: Efficient, Performant Video Detailed Captioning and a New Benchmark](https://mlsanity.com/paper/2410.03051/) (2024)
- [ImageInWords: Unlocking Hyper-Detailed Image Descriptions](https://mlsanity.com/paper/2405.02793/) (2024)
- [AutoAD: Movie Description in Context](https://mlsanity.com/paper/2303.16899/) (2023)
- [AutoAD II: The Sequel -- Who, When, and What in Movie Audio Description](https://mlsanity.com/paper/2310.06838/) (2023)
- [MM-Narrator: Narrating Long-form Videos with Multimodal In-Context Learning](https://mlsanity.com/paper/2311.17435/) (2023)
- [Segment Any Text: A Universal Approach for Robust, Efficient and Adaptable Sentence Segmentation](https://mlsanity.com/paper/2406.16678/) (2024)
- [AutoAD III: The Prequel -- Back to the Pixels](https://mlsanity.com/paper/2404.14412/) (2024)
- [AutoAD-Zero: A Training-Free Framework for Zero-Shot Audio Description](https://mlsanity.com/paper/2407.15850/) (2024)
- [Rescribe: Authoring and Automatically Editing Audio Descriptions](https://mlsanity.com/paper/2010.03667/) (2020)
- [Learning Video Context as Interleaved Multimodal Sequences](https://mlsanity.com/paper/2407.21757/) (2024)
- [Contextual AD Narration with Interleaved Multimodal Sequence](https://mlsanity.com/paper/2403.12922/) (2024)
- [Shot-by-Shot: Film-Grammar-Aware Training-Free Audio Description Generation](https://mlsanity.com/paper/2504.01020/) (2025)
- [DistinctAD: Distinctive Audio Description Generation in Contexts](https://mlsanity.com/paper/2411.18180/) (2024)
- [VideoA11y: Method and Dataset for Accessible Video Description](https://mlsanity.com/paper/2502.20480/) (2025)
- [Look, Listen and Recognise: Character-Aware Audio-Visual Subtitling](https://mlsanity.com/paper/2401.12039/) (2024)
- [MovieSum: An Abstractive Summarization Dataset for Movie Screenplays](https://mlsanity.com/paper/2408.06281/) (2024)
- [More than a Moment: Towards Coherent Sequences of Audio Descriptions](https://mlsanity.com/paper/2510.25440/) (2025)
- [DANTE-AD: Dual-Vision Attention Network for Long-Term Audio Description](https://mlsanity.com/paper/2503.24096/) (2025)
- [Contrastive Learning with Narrative Twins for Modeling Story Salience](https://mlsanity.com/paper/2601.07765/) (2026)
- [REFRAMED: Towards Realistic Audio Description Generation for Movies](https://mlsanity.com/paper/2608.09765/) (2026)
- [From Visual Cues to Spoken Narration: Rethinking Audio Description](https://mlsanity.com/paper/2609.01725/) (2026)

## Abstract (arXiv)

Audio Description (AD) makes movies accessible to blind and visually impaired audiences by narrating visual information in gaps between dialogue. Existing automatic AD systems largely treat generation as a local video-to-text problem, assuming that the content to describe and its temporal location are already provided. Realistic AD instead requires coupled decisions about what visual information is narratively important, when it can be spoken without interfering with dialogue, and how it should be formulated to fit within the available time. We formalize AD generation as a constrained optimization problem over these three decisions. Our hybrid system uses large language models to propose and ground visual elements, estimate their salience to the narrative, and generate compressed realizations. A mixed-integer linear program then jointly selects and schedules descriptions across a scene subject to temporal constraints. When evaluated on REFRAMED, a benchmark for realistic AD of movies, our approach makes better decisions than prompted LLMs about what to describe and when to describe it, establishing a new SOTA on narrative QA and temporally grounded metrics. Ablations show that explicit temporal constraints drive gains in placement, while salience estimation controls how much narratively useful content is retained. Improvements are concentrated on temporal and narrative measures rather than n-gram overlap, although a significant gap to professional describers remains.