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

Source: https://huggingface.co/papers/2604.23789

Abstract

MuSS is a large-scale dual-track dataset designed for multi-shot video generation that addresses narrative logic, spatiotemporal alignment, and copy-paste issues in subject-to-video generation through a progressive captioning pipeline and cross-shot matching mechanism.

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: authenticnarrative logic,spatiotemporal text-video alignmentconflicts, and the “copy-paste” dilemma prevalent inSubject-to-Video (S2V) generation. To bridge this gap, we introduce MuSS, a large-scale,dual-track datasettailored for multi-shot video and S2V generation. Sourced from over 3,000 movies, MuSS explicitly supports both complex montage transitions and subject-centric narratives. To construct this dataset, we pioneer aprogressive captioning pipelinethat eliminates contextual conflicts by ensuring local shot-level accuracy before enforcing global narrative coherence. Crucially, we implement across-shot matching mechanismto fundamentally eradicate the S2V copy-paste shortcut. Alongside the dataset, we propose theCinematic Narrative Benchmark, featuring a visual-logic-driven paradigm and a novelAnti-Copy-Paste Variance (ACP-Var) metricto rigorously assesscontinuous storytellingand3D structural consistency. Extensive experiments demonstrate that while current baselines struggle with continuousnarrative logicor degenerate into trivial 2D sticker generators, our MuSS-augmented model achieves state-of-the-art narrative effectiveness and cross-shot identity preservation.

View arXiv pageView PDFGitHub5Add to collection

Get this paper in your agent:

hf papers read 2604\.23789

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2604.23789 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2604.23789 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2604.23789 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.