Research map / Video / Video generation
FlashMotion: Few-Step Controllable Video Generation with Trajectory Guidance
Quanhao Li, Zhen Xing, Rui Wang, Haidong Cao, Qi Dai, Daoguo Dong, Zuxuan Wu
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 FastAdapter fine-tuning stage instead of applying SlowAdapter to FastGenerator (ablation) | FlashMotion (ControlNet) · FlashBench | FVD: 171.83 -> 96.08 | Improved |
| add adversarial (GAN) loss to adapter fine-tuning (ablation) | FlashMotion (ResNet) · FlashBench | FVD: 206.75 -> 108.96 | Improved |
| add adversarial (GAN) loss to adapter fine-tuning (ablation) | FlashMotion (ControlNet) · FlashBench | Mask IoU (%): 71.56 -> 69.15 | Hurt |
| add diffusion loss to adapter fine-tuning (ablation) | FlashMotion (ResNet) · FlashBench | Mask IoU (%): 52.04 -> 63.96 | Improved |
| add diffusion loss to adapter fine-tuning (ablation) | FlashMotion (ControlNet) · FlashBench | Mask IoU (%): 55.91 -> 69.15 | Improved |
| add dynamic diffusion loss scaling instead of fixed scale 1 (ablation) | FlashMotion (ResNet) · FlashBench | FID: 26.32 -> 15.81 | Improved |
| add dynamic diffusion loss scaling instead of fixed scale 1 (ablation) | FlashMotion (ControlNet) · FlashBench | FVD: 155.55 -> 96.08 | Improved |
| add Semantic Self-Attention and Trajectory Cross-Attention to discriminator head (ablation) | FlashMotion discriminator (ControlNet) · FlashBench | Mask IoU (%): 63.04 -> 69.15 | Improved |
| FlashMotion few-step versus multi-step MagicMotion | Wan2.2-TI2V-5B ControlNet adapter · FlashBench | FVD: 138.83 -> 96.08 | Improved |
| FlashMotion with ResNet adapter versus multi-step MagicMotion | Wan2.2-TI2V-5B ResNet adapter · FlashBench | Mask IoU (%): 68.10 -> 63.96; FVD 138.83 -> 108.96 | Mixed |
Later papers that compared with this method
1 later accepted paper reports FlashMotion as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- GraphVid: Interactive Graph-Controllable Video Generation (ECCV 2026 · 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
- Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance (NeurIPS 2025 · similarity 0.90)
- SG-I2V: Self-Guided Trajectory Control in Image-to-Video Generation (ICLR 2025 · similarity 0.89)
- MotionPro: A Precise Motion Controller for Image-to-Video Generation (CVPR 2025 · similarity 0.89)
- Trajectory attention for fine-grained video motion control (ICLR 2025 · similarity 0.89)
- Motion-Zero: A Zero-Shot Trajectory Control Framework of Moving Object for Diffusion-Based Video Generation (AAAI 2025 · similarity 0.88)
- Motion Prompting: Controlling Video Generation with Motion Trajectories (CVPR 2025 · similarity 0.88)
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.
