Research map / Video / Video generation
MoStGAN-V: Video Generation With Temporal Motion Styles
Xiaoqian Shen, Xiang Li, Mohamed Elhoseiny
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 time-dependent motion styles with MoStAtt | StyleGAN-V · FaceForensics 256² | FVD16: 47.4 -> 39.7 | Improved |
| add time-dependent motion styles with MoStAtt | StyleGAN-V · SkyTimelapse 256² | FVD16: 79.5 -> 65.3 | Improved |
| add time-dependent motion styles with MoStAtt | StyleGAN-V · RainbowJelly 256² | FVD16: 195.4 -> 70.1 | Improved |
| add time-dependent motion styles with MoStAtt | StyleGAN-V · CelebV-HQ 256² | FVD16: 68.0 -> 56.1 | Improved |
| use one motion style without MoStAtt (ablation) | MoStGAN-V · FaceForensics 256² | FVD16: 39.7 -> 50.9 | Hurt |
| use one motion style without MoStAtt (ablation) | MoStGAN-V · RainbowJelly 256² | FVD16: 70.1 -> 248.7 | Hurt |
| use 6 motion styles instead of 8 (ablation) | MoStGAN-V · RainbowJelly 256² | FVD16: 70.1 -> 96.1 | Hurt |
| use 10 motion styles instead of 8 (ablation) | MoStGAN-V · FaceForensics 256² | FVD16: 39.7 -> 40.4 | Hurt |
| apply motion-style modulation before content-style modulation (ablation) | MoStGAN-V · FaceForensics 256² | FVD16: 39.76 -> 65.07 | Hurt |
| add frame differences as discriminator inputs (ablation) | StyleGAN-V · FaceForensics 256² | FVD16: 47.4 -> 58.1 | Hurt |
| add frame differences as discriminator inputs (ablation) | MoStGAN-V · FaceForensics 256² | FVD16: 49.1 -> 39.7 | Improved |
| add motion diversity loss Ldiv (ablation) | MoStGAN-V · FaceForensics 256² | FVD128: 79.8 -> 72.6 | Improved |
Later papers that compared with this method
3 later accepted papers report MoStGAN-V as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Controllable Video Generation with Provable Disentanglement (ICLR 2026 · better on 8 of 8 shared results)
- AR-Diffusion: Asynchronous Video Generation with Auto-Regressive Diffusion (CVPR 2025 · better on 0 of 1 shared result)
- LEO: Generative Latent Image Animator for Human Video Synthesis (IJCV 2024 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- StyleGAN-V — StyleGAN-V: A Continuous Video Generator With the Price, Image Quality and Perks of StyleGAN2 (CVPR 2022) · better on 8 of 8 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
- Make Pixels Dance: High-Dynamic Video Generation (CVPR 2024 · similarity 0.88)
- Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video Synthesis (CVPR 2024 · similarity 0.87)
- MotionDreamer: One-to-Many Motion Synthesis with Localized Generative Masked Transformer (ICLR 2025 · similarity 0.86)
- Decouple Content and Motion for Conditional Image-to-Video Generation (AAAI 2024 · similarity 0.86)
- ByTheWay: Boost Your Text-to-Video Generation Model to Higher Quality in a Training-free Way (CVPR 2025 · similarity 0.86)
- FlowMo: Variance-Based Flow Guidance for Coherent Motion in Video Generation (NeurIPS 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.
