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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add Multi-Shot Narrative RoPE (ablation) | MultiShotMaster · Multi-shot T2V | transition deviation: 4.68 -> 1.72 | Improved |
| add Multi-Shot Narrative RoPE (ablation) | MultiShotMaster · Multi-shot T2V | narrative coherence: 0.645 -> 0.695 | Improved |
| add Multi-Shot Narrative RoPE (ablation) | MultiShotMaster · Multi-shot T2V | semantic consistency: 0.702 -> 0.697 | Hurt |
| add Multi-Shot Narrative RoPE (ablation) | MultiShotMaster · Multi-shot T2V | scene consistency: 0.455 -> 0.447 | Hurt |
| average subject token copies instead of random copy (ablation) | MultiShotMaster · Multi-shot R2V | subject consistency: 0.482 -> 0.493 | Improved |
| add multi-shot multi-reference attention mask (ablation) | MultiShotMaster · Multi-shot R2V | aesthetic score: 3.72 -> 3.86 | Improved |
| add multi-shot multi-reference attention mask (ablation) | MultiShotMaster · Multi-shot R2V | scene reference consistency: 0.414 -> 0.456 | Improved |
| add Spatiotemporal Position-Aware RoPE (ablation) | MultiShotMaster · Multi-shot R2V | subject reference consistency: 0.425 -> 0.493 | Improved |
| add Spatiotemporal Position-Aware RoPE (ablation) | MultiShotMaster · Multi-shot R2V | narrative coherence: 0.761 -> 0.825 | Improved |
| train reference injection on single-shot data first, then multi-shot (ablation) | MultiShotMaster · Multi-shot R2V | grounding consistency: 0.477 -> 0.583 | Improved |
| add subject-focused post-training (2x subject loss weight) (ablation) | MultiShotMaster · Multi-shot R2V | subject reference consistency: 0.485 -> 0.493 | Improved |
| pretrain reference injection on single-shot data instead of multi-shot only first (ablation) | MultiShotMaster · Multi-shot R2V | subject reference consistency: 0.472 -> 0.485 | Improved |
Closest papers by meaning
- ShotAdapter: Text-to-Multi-Shot Video Generation with Diffusion Models (CVPR 2025 · similarity 0.88)
- IM-Zero: Instance-level Motion Controllable Video Generation in a Zero-shot Manner (CVPR 2025 · similarity 0.87)
- ONE-SHOT: Compositional Human-Environment Video Synthesis via Spatial-Decoupled Motion Injection and Hybrid Context Integration (NeurIPS 2026 · similarity 0.87)
- CineOrchestra: Unified Entity-Centric Conditioning for Cinematic Video Generation (NeurIPS 2026 · similarity 0.87)
- EchoShot: Multi-Shot Portrait Video Generation (NeurIPS 2025 · similarity 0.86)
- DiTCtrl: Exploring Attention Control in Multi-Modal Diffusion Transformer for Tuning-Free Multi-Prompt Longer Video Generation (CVPR 2025 · similarity 0.86)
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.
