Research map / Video / Video generation
3D Scene Prompting for Scene-Consistent Camera-Controllable Video Generation
JoungBin Lee, Jaewoo Jung, Jisang Han, Takuya Narihira, Kazumi Fukuda, Junyoung Seo, Sunghwan Hong, Yuki Mitsufuji, Seungryong Kim
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 dynamic masking before building the static point cloud (ablation) | 3DScenePrompt · DynPose-100K | PSNR: 12.2304 -> 13.0468 | Improved |
| add spatial prompts (n=7 projected views) to temporal-only conditioning (ablation) | 3DScenePrompt · DynPose-100K | PSNR: 11.9555 -> 13.0468 | Improved |
| add spatial prompts (n=7) to temporal-only conditioning (ablation) | 3DScenePrompt · DynPose-100K camera control | mRotErr: 3.4142 -> 2.3772 | Improved |
| increase spatial frames n beyond 7 (to all frames L) (ablation) | 3DScenePrompt · DynPose-100K | PSNR / MEt3R: PSNR 13.0468 -> 13.0206; MEt3R 0.124189 -> 0.123507 | No clear effect |
| increase temporal window from w=1 to w=9 (ablation) | 3DScenePrompt · DynPose-100K camera control | mCamMC: 8.9785 -> 8.6352 | No clear effect |
| increase temporal window from w=1 to w=9 (ablation) | 3DScenePrompt · DynPose-100K | Motion Smooth: 0.9253 -> 0.9335 | Improved |
| add SAM2 propagation to L1 flow mask (ablation) | dynamic masking pipeline · DynPose-100K subset | PSNR: 12.1083 -> 13.5957 | Improved |
| add backward point tracking before SAM2 (ablation) | dynamic masking pipeline · DynPose-100K subset | MEt3R: 0.157610 -> 0.122859 | Improved |
| add dynamic mask pipeline (ablation) | 3DScenePrompt · DynPose-100K subset | MEt3R: 0.104002 -> 0.122859 | Mixed |
| replace MegaSAM with DepthAnything v3 for 3D memory (ablation) | 3DScenePrompt · DynPose-100K | PSNR: 13.0468 -> 13.4534 | Improved |
| use spatial 3D memory instead of DFoT history guidance | scene-consistent generation · RealEstate10K | MEt3R: 0.181164 -> 0.040843 | Improved |
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.
- FloVD — FloVD: Optical Flow Meets Video Diffusion Model for Enhanced Camera-Controlled Video Synthesis (CVPR 2025) · better on 1 of 1 shared result
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
- GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control (CVPR 2025 · similarity 0.91)
- Geometry-as-context: Modulating Explicit 3D in Scene-consistent Video Generation to Geometry Context (CVPR 2026 · similarity 0.91)
- Video World Models with Long-term Spatial Memory (NeurIPS 2025 · similarity 0.89)
- VD3D: Taming Large Video Diffusion Transformers for 3D Camera Control (ICLR 2025 · similarity 0.88)
- Motion Prompting: Controlling Video Generation with Motion Trajectories (CVPR 2025 · similarity 0.88)
- AC3D: Analyzing and Improving 3D Camera Control in Video Diffusion Transformers (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.
