Research map / Video / Video segmentation and tracking
Fast-Vid2Vid: Spatial-Temporal Compression for Video-to-Video Synthesis
Long Zhuo, Guangcong Wang, Shikai Li, Wayne Wu, Ziwei 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 |
|---|---|---|---|
| apply spatial-temporal compression (Fast-Vid2Vid) to vid2vid | vid2vid · Sketch2Face | FID: 34.17 -> 29.02 | Improved |
| apply spatial-temporal compression | vid2vid · Sketch2Face | FPS: 2.77 -> 18.56 | Improved |
| apply spatial-temporal compression | vid2vid · Pose2Body | FVD: 12.31 -> 10.03 | Improved |
| apply spatial-temporal compression | vid2vid · Segmentation2City | FPS: 4.27 -> 24.77 | Improved |
| use Spatial KD instead of NAS for spatial compression with Temporal KD (ablation) | Fast-Vid2Vid · Sketch2Face | FID: 32.41 -> 29.02 | Improved |
| use Spatial KD instead of CA (ablation) | Fast-Vid2Vid · Sketch2Face | FVD: 6.76 -> 5.79 | Improved |
| use motion compensation instead of linear interpolation (ablation) | Fast-Vid2Vid · Sketch2Face | FVD: 6.45 -> 5.79 | Improved |
| select key-frames instead of fixed time gap (ablation) | Fast-Vid2Vid · Sketch2Face | FVD: 6.85 -> 5.79 | Improved |
| select key-frames instead of random time gap (ablation) | Fast-Vid2Vid · Sketch2Face | FID: 31.43 -> 29.02 | Improved |
| add local temporal knowledge distillation loss (ablation) | Fast-Vid2Vid · Sketch2Face | FVD | Improved |
| add global temporal KD on top of local KD (ablation) | Fast-Vid2Vid · Sketch2Face | FVD | Improved |
| increase key-frame selection sliding window (ablation) | Fast-Vid2Vid · Sketch2Face, Segmentation2City, Pose2Body | FVD | Mixed |
Later papers that compared with this method
2 later accepted papers report Fast-Vid2Vid as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- DiffPerformer: Iterative Learning of Consistent Latent Guidance for Diffusion-based Human Video Generation (CVPR 2024 · better on 7 of 7 shared results)
- Shortcut-V2V: Compression Framework for Video-to-Video Translation Based on Temporal Redundancy Reduction (ICCV 2023 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video Synthesis (CVPR 2024 · similarity 0.89)
- FlowVid: Taming Imperfect Optical Flows for Consistent Video-to-Video Synthesis (CVPR 2024 · similarity 0.89)
- VideoMAR: Autoregressive Video Generation with Continuous Tokens (NeurIPS 2025 · similarity 0.88)
- FastVID: Dynamic Density Pruning for Fast Video Large Language Models (NeurIPS 2025 · similarity 0.87)
- PVC: Progressive Visual Token Compression for Unified Image and Video Processing in Large Vision-Language Models (CVPR 2025 · similarity 0.87)
- Encapsulated Composition of Text-to-Image and Text-to-Video Models for High-Quality Video Synthesis (CVPR 2025 · similarity 0.87)
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.
