Research map / Video / Video generation
CineTrans: Learning to Generate Videos with Cinematic Transitions via Masked Diffusion Models
Xiaoxue Wu, Bingjie Gao, Yu Qiao, Yaohui Wang, Xinyuan Chen
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 block-diagonal shot mask to attention (with fine-tuning) (ablation) | CineTrans-Unet (LaVie) · Multi-shot prompts (100) | Transition Control Score: 0.2398 -> 0.8598 | Improved |
| add block-diagonal shot mask to attention (with fine-tuning) (ablation) | CineTrans-DiT (Wan2.1-1.3B) · Multi-shot prompts (100) | Transition Control Score: 0.2112 -> 0.7003 | Improved |
| add mask without any fine-tuning (training-free) (ablation) | CineTrans-DiT · Multi-shot prompts (100) | Transition Control Score: 0.2051 -> 0.6564 | Improved |
| add mask without any fine-tuning (training-free) (ablation) | CineTrans-Unet · Multi-shot prompts (100) | Transition Control Score: 0 -> 0.6168 | Improved |
| add fine-tuning on Cine250K (with mask) (ablation) | CineTrans-Unet · Multi-shot prompts (100) | Transition Control Score: 0.6168 -> 0.8598 | Improved |
| add fine-tuning on Cine250K (with mask) (ablation) | CineTrans-DiT · Multi-shot prompts (100) | Transition Control Score: 0.6564 -> 0.7003 | Improved |
| add fine-tuning on Cine250K (with mask) (ablation) | CineTrans-DiT · Multi-shot prompts (100) | Inter-shot semantic consistency gap: 0.1772 -> 0.1552 | Improved |
| add fine-tuning on Cine250K (with mask) (ablation) | CineTrans-DiT · Multi-shot prompts (100) | Aesthetic Quality: 0.6556 -> 0.6508 | Hurt |
| add fine-tuning on Cine250K (with mask) (ablation) | CineTrans-Unet · Multi-shot prompts (100) | Inter-shot visual consistency score: 0.8186 -> 0.7247 | Mixed |
| fine-tune without mask (ablation) | CineTrans-DiT · Multi-shot prompts (100) | Transition Control Score: 0.2051 -> 0.2112 | No clear effect |
| use mask on some layers only vs all layers (ablation) | CineTrans-Unet and DiT · Mask layer selection | visual quality | Mixed |
| use CineTrans-DiT with mask+LoRA vs Cinematron customization model | CineTrans-DiT · Multi-shot prompts (100) | Transition Control Score: 0.3787 -> 0.7003 | Improved |
Closest papers by meaning
- ShotAdapter: Text-to-Multi-Shot Video Generation with Diffusion Models (CVPR 2025 · similarity 0.90)
- SEINE: Short-to-Long Video Diffusion Model for Generative Transition and Prediction (ICLR 2024 · similarity 0.89)
- CineTechBench: A Benchmark for Cinematographic Technique Understanding and Generation (NeurIPS 2025 · similarity 0.89)
- Stable Cinemetrics : Structured Taxonomy and Evaluation for Professional Video Generation (NeurIPS 2025 · similarity 0.89)
- CineOrchestra: Unified Entity-Centric Conditioning for Cinematic Video Generation (NeurIPS 2026 · similarity 0.89)
- DiTCtrl: Exploring Attention Control in Multi-Modal Diffusion Transformer for Tuning-Free Multi-Prompt Longer Video Generation (CVPR 2025 · similarity 0.89)
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.
