Research map / Video / Video generation
Generating, Fast and Slow: Scalable Parallel Video Generation with Video Interface Networks
Bhishma Dedhia, David Bourgin, Krishna Kumar Singh, Yuheng Li, Yan Kang, Zhan Xu, Niraj K. Jha, Yuchen Liu
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 global tokens (ablation) | VIN-DiT · VIN ablation | MAWE: 1.69 -> 1.09 | Improved |
| add global tokens (ablation) | VIN-DiT · VIN ablation | Scene cuts: 0.33 -> 0.21 | Improved |
| use early token fusion instead of no fusion (ablation) | VIN-DiT · VIN ablation | Scene cuts: 1.00 -> 0.21 | Improved |
| use early token fusion instead of no fusion (ablation) | VIN-DiT · VIN ablation | MAWE: 1.13 -> 1.09 | Improved |
| use early fusion instead of late fusion (ablation) | VIN-DiT · VIN ablation | MAWE: 1.22 -> 1.09 | Improved |
| use early fusion instead of mid fusion (ablation) | VIN-DiT · VIN ablation | Scene cuts: 0.33 -> 0.21 | Improved |
| expand local frames at inference to 12 from 8 (ablation) | VIN-DiT · VIN ablation | MAWE: 1.51 -> 1.09 | Improved |
| expand local frames at inference to 12 from 10 (ablation) | VIN-DiT · VIN ablation | MAWE: 1.17 -> 1.09 | Improved |
| sample keyframes more densely (0.2s instead of 1.0s) (ablation) | VIN-DiT · VIN ablation | MAWE: 1.09 -> 1.21 | No clear effect |
| use VIN parallel generation instead of full attention | VIN-DiT · 256-frame VBench | Dynamic degree: 68.15% -> 87.32% | Improved |
| use VIN instead of full attention on efficiency | VIN-DiT · Efficiency analysis | FLOPs: 25-40% fewer FLOPs | Improved |
Closest papers by meaning
- Q-VDiT: Towards Accurate Quantization and Distillation of Video-Generation Diffusion Transformers (ICML 2025 · similarity 0.89)
- FreeLong: Training-Free Long Video Generation with SpectralBlend Temporal Attention (NeurIPS 2024 · similarity 0.89)
- Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video Synthesis (CVPR 2024 · similarity 0.89)
- Towards Precise Scaling Laws for Video Diffusion Transformers (CVPR 2025 · similarity 0.89)
- From Slow Bidirectional to Fast Autoregressive Video Diffusion Models (CVPR 2025 · similarity 0.89)
- OSV: One Step is Enough for High-Quality Image to 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.
