A Large-Scale Dataset and Cinematic Narrative Benchmark for Multi-Shot Subject-to-Video Generation

arXiv 2026
Haojie Zhang, Di Wu, Bingyan Liu, Linjie Zhong, Yuancheng Wei, Xingsong Ye, Nanqing Liu, Yaling Liang

TL;DR: MuSS is a large-scale cinematic dataset and benchmark for multi-shot video generation and Subject-to-Video generation. It targets authentic narrative logic, shot-level text-video alignment, and cross-shot identity preservation beyond isolated single-shot generation.

MuSS overview

MuSS focuses on two complementary settings: Complex Cinematic Narrative for montage, shot transitions, and multi-character storytelling; and Subject-Centric Narrative for preserving the same subject across disjoint shots and viewpoints.

Abstract


While video foundation models excel at single-shot generation, real-world cinematic storytelling inherently relies on complex multi-shot sequencing. Further progress is constrained by the absence of datasets that address three core challenges: authentic narrative logic, spatiotemporal text-video alignment conflicts, and the copy-paste dilemma prevalent in Subject-to-Video generation. To bridge this gap, we introduce MuSS, a large-scale, dual-track dataset tailored for multi-shot video and Subject-to-Video generation. Sourced from over 3,000 movies, MuSS explicitly supports both complex montage transitions and subject-centric narratives. Alongside the dataset, we propose the Cinematic Narrative Benchmark, featuring a visual-logic-driven paradigm and a novel Anti-Copy-Paste Variance metric to rigorously assess continuous storytelling and 3D structural consistency.

3,000+ movies as cinematic source material
30,000+ professionally captioned multi-shot clips
1,000h+ high-quality video content

Dataset


MuSS dataset statistics

MuSS provides large-scale cinematic material with diverse clip durations, caption lengths, visual concepts, and source videos.

MuSS dataset construction pipeline

The construction pipeline first turns raw cinematic footage into high-quality physical shots with coherent captions, then builds cross-shot Subject-to-Video pairs by sampling reference subjects from disjoint shot contexts.

Benchmark


Cinematic Narrative Benchmark pipeline

The Cinematic Narrative Benchmark combines shot boundary parsing, expert perception models, and large multimodal model based visual-logic assessment.

Track Evaluation Goal Metrics
Track 1: Narrative Effectiveness Shot-level alignment, transition precision, scene continuity, and visual logic. Txt.Align, Trans.Dev, Scene.Con, Con.Gap, Scene.Logic, Casting.Logic, Act.Logic, Spat.Logic
Track 2: Subject Consistency Cross-shot identity preservation, subject grounding, motion strength, and anti-copy-paste behavior. Subj.Recall, Ref-Sub.Con, Inter-Sub.Con, Act.Str, ACP-Var, CP-Rate
Qualitative benchmark comparison

Qualitative benchmark results highlight structural limitations of existing methods and the role of MuSS in improving multi-shot consistency and 3D identity preservation.

Data Examples


Complex Cinematic Narrative

Track 1 data examples

Progressive multi-shot captions are aligned to physical shots, capturing shot transitions, scene changes, and multi-character narrative flow.

Subject-Centric Narrative

Track 2 data examples

A reference subject is extracted from a disjoint shot, while the target sequence preserves identity across different viewpoints and contexts.

BibTeX

@article{zhang2026muss,
  title   = {MuSS: A Large-Scale Dataset and Cinematic Narrative Benchmark for Multi-Shot Subject-to-Video Generation},
  author  = {Zhang, Haojie and Wu, Di and Liu, Bingyan and Zhong, Linjie and Wei, Yuancheng and Ye, Xingsong and Liu, Nanqing and Liang, Yaling},
  journal = {arXiv preprint arXiv:2604.23789},
  year    = {2026}
}