Research map / Video / Video generation
Trajectory-guided motion control
Controls video generation motion with sparse trajectories, latent trajectory guidance and tracking signals.
184 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2018–2026) · an approach in Video generation, keyword cluster “video generation · video diffusion · text video”
State History Tried Baselines Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
13 papers in 2023–24 and 167 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 5.7% to 14.9%.
Open questions the data points to
- Papers report 360DVD on Panoramic video generation (Image Quality) anywhere from 0.436 to 0.705: comparisons against it depend on whose number you take.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2018 · Controllable Video Generation With Sparse Trajectories (CVPR 2018 · 96 citations)
- 2021 · Playable Video Generation (CVPR 2021 · 36 citations)
- 2022 · Show Me What and Tell Me How: Video Synthesis via Multimodal Conditioning (CVPR 2022 · 37 citations)
- 2023 · Tell Me What Happened: Unifying Text-Guided Video Completion via Multimodal Masked Video Generation (CVPR 2023 · 23 citations)
- 2024 · Follow Your Pose: Pose-Guided Text-to-Video Generation Using Pose-Free Videos (AAAI 2024 · 122 citations)
- 2025 · Motion Prompting: Controlling Video Generation with Motion Trajectories (CVPR 2025 · 32 citations)
- 2026 · OwlCap: Harmonizing Motion-Detail for Video Captioning via HMD-270K and Caption Set Equivalence Reward (AAAI 2026 · 2 citations)
Earliest papers here
- Controllable Video Generation With Sparse Trajectories (CVPR 2018 · 96 citations)
- Playable Video Generation (CVPR 2021 · 36 citations)
- Compositional Video Synthesis with Action Graphs (ICML 2021 · 5 citations)
What has been tried
67 papers here report 734 measured changes (a component added, removed or swapped, compared with a baseline): helped 72%, no clear effect 4%, hurt 23%, mixed 1%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use residual feature injection instead of concatenating pose condition into the input latent | Pose-guided video generation · Generation quality and pose-frame alignment | Helped | Follow Your Pose: Pose-Guided Text-to-Video Generation Using Pose-Free Videos (AAAI 2024 · 122 citations) | |
| Use WebVid and self-collected video data rather than WebVid alone | UCF-101 · FID | 56.16 -> 59.35 | No clear effect | Make Pixels Dance: High-Dynamic Video Generation (CVPR 2024 · 44 citations) |
| add cross-view attention | nuScenes validation · FVD | 247 -> 139 (108) | Helped | Panacea: Panoramic and Controllable Video Generation for Autonomous Driving (CVPR 2024 · 44 citations) |
| replace the 284M U-Net with a 500M FIT | internal dataset (64 × 36 px) · training throughput (ms/video/GPU) | 230.3 -> 69.5 | Helped | Snap Video: Scaled Spatiotemporal Transformers for Text-to-Video Synthesis (CVPR 2024 · 37 citations) |
| train with dense rather than sparse tracks | DAVIS ablation, N=4 · PSNR | 15.075 -> 15.638 | Helped | Motion Prompting: Controlling Video Generation with Motion Trajectories (CVPR 2025 · 32 citations) |
| remove camera conditioning | MSR-VTT · FVD | -8.53 | Helped | AC3D: Analyzing and Improving 3D Camera Control in Video Diffusion Transformers (CVPR 2025 · 26 citations) |
| remove context prompt | MagicBrush test · CLIPdir | 0.144 -> 0.151 | Hurt | UniReal: Universal Image Generation and Editing via Learning Real-world Dynamics (CVPR 2025 · 20 citations) |
| Add Spatial Subject Learning Module | Customized video generation · CLIP-T | 0.286 -> 0.294 | Helped | CustomCrafter: Customized Video Generation with Preserving Motion and Concept Composition Abilities (AAAI 2025 · 15 citations) |
| use CLIP instead of DINOv2 as the image encoder | MSRVTT-Personalization subject mode · Text-S | 0.257 -> 0.269 | Helped | Multi-subject Open-set Personalization in Video Generation (CVPR 2025 · 11 citations) |
| Replace pixel replication with latent feature replication for motion guidance | MoveBench · EPE | 3.7 -> 2.6 | Helped | Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance (NeurIPS 2025 · 8 citations) |
| Add layout encoder and relative region embedding | Video outpainting from 512×512 to 1440×810 · FVD | 774.1 -> 486.1 | Helped | Infinite-Canvas: Higher-Resolution Video Outpainting with Extensive Content Generation (AAAI 2025 · 7 citations) |
| Add Offset Shifting Denoising | Ablation study · Image Quality | 0.564 -> 0.587 | Helped | DynamicScaler: Seamless and Scalable Video Generation for Panoramic Scenes (CVPR 2025 · 7 citations) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| 360DVD | Panoramic video generation | Image Quality | 0.436 – 0.705 | 2 papers, e.g. CVPR 2025 and NeurIPS 2025 |
| DragAnything | VIPSeg validation | FVD | 268.47 – 294.91 | 2 papers, e.g. ICLR 2025 and AAAI 2025 |
| DragAnything | VIPSeg validation | FID | 28.16 – 30.81 | 2 papers, e.g. AAAI 2025 and ICLR 2025 |
What reviewers will expect
Datasets and baselines named in the 82 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- RealEstate10K · 10 of 82 papers
- DAVIS · 7 of 82 papers
- Panda-70M · 5 of 82 papers
- MSR-VTT · 4 of 82 papers
- OpenVid-1M · 4 of 82 papers
- VIPSeg · 3 of 82 papers
- WebVid · 3 of 82 papers
- Objaverse · 3 of 82 papers
Baselines
- CogVideoX-5B · 9 of 82 papers
- MotionCtrl · 8 of 82 papers
- CameraCtrl · 8 of 82 papers
- DragAnything · 6 of 82 papers
- Image Conductor · 5 of 82 papers
- SkyReels-A2 · 5 of 82 papers
- VACE · 5 of 82 papers
- CogVideoX · 4 of 82 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 180 papers; from OpenAlex and the papers' affiliations).
- Hong Kong University of Science and Technology · 8 papers
- Tencent · 7 papers
- Alibaba Group · 6 papers
- Adobe · 6 papers
- Shanghai Jiao Tong University · 5 papers
- Tsinghua University · 4 papers
- University of Science and Technology of China · 4 papers
- Meta · 4 papers
- ByteDance · 4 papers
- Snap Inc · 4 papers
Use it
Download BibTeX of 24 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Efficient video diffusion models (283 papers)
- Compositional text-to-video diffusion (159 papers)
- Autoregressive long video generation (156 papers)
- World-model video generation (110 papers)
Elsewhere in the topic
- Text-to-motion diffusion (140 papers)
- Diffusion-based human image animation (126 papers)
- Diffusion-based video editing (92 papers)
- Talking face generation (81 papers)
- Video-to-audio synthesis (76 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).