AI Paper Atlas

Research map / Video / Video generation

MultiShotMaster: A Controllable Multi-Shot Video Generation Framework

Qinghe Wang, Xiaoyu Shi, Baolu Li, Weikang Bian, Quande Liu, Huchuan Lu, Xintao Wang, Pengfei Wan, Kun Gai, Xu Jia

CVPR 2026 · Video › Video generation · approach: Trajectory-guided motion control · 30 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add Multi-Shot Narrative RoPE (ablation)MultiShotMaster · Multi-shot T2Vtransition deviation: 4.68 -> 1.72Improved
add Multi-Shot Narrative RoPE (ablation)MultiShotMaster · Multi-shot T2Vnarrative coherence: 0.645 -> 0.695Improved
add Multi-Shot Narrative RoPE (ablation)MultiShotMaster · Multi-shot T2Vsemantic consistency: 0.702 -> 0.697Hurt
add Multi-Shot Narrative RoPE (ablation)MultiShotMaster · Multi-shot T2Vscene consistency: 0.455 -> 0.447Hurt
average subject token copies instead of random copy (ablation)MultiShotMaster · Multi-shot R2Vsubject consistency: 0.482 -> 0.493Improved
add multi-shot multi-reference attention mask (ablation)MultiShotMaster · Multi-shot R2Vaesthetic score: 3.72 -> 3.86Improved
add multi-shot multi-reference attention mask (ablation)MultiShotMaster · Multi-shot R2Vscene reference consistency: 0.414 -> 0.456Improved
add Spatiotemporal Position-Aware RoPE (ablation)MultiShotMaster · Multi-shot R2Vsubject reference consistency: 0.425 -> 0.493Improved
add Spatiotemporal Position-Aware RoPE (ablation)MultiShotMaster · Multi-shot R2Vnarrative coherence: 0.761 -> 0.825Improved
train reference injection on single-shot data first, then multi-shot (ablation)MultiShotMaster · Multi-shot R2Vgrounding consistency: 0.477 -> 0.583Improved
add subject-focused post-training (2x subject loss weight) (ablation)MultiShotMaster · Multi-shot R2Vsubject reference consistency: 0.485 -> 0.493Improved
pretrain reference injection on single-shot data instead of multi-shot only first (ablation)MultiShotMaster · Multi-shot R2Vsubject reference consistency: 0.472 -> 0.485Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy