Research map / Video / Video generation
TC4D: Trajectory-Conditioned Text-to-4D Generation
Sherwin Bahmani, Xian Liu, Wang Yifan, Ivan Skorokhodov, Victor Rong, Ziwei Liu, Xihui Liu, Jeong Joon Park, Sergey Tulyakov, Gordon Wetzstein, Andrea Tagliasacchi, David B Lindell
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 |
|---|---|---|---|
| factor motion into global trajectory plus local deformation instead of deformation only (4D-fy with post-hoc trajectory) | TC4D · User study vs 4D-fy | overall preference: 85% preferred TC4D | Improved |
| add local deformation on top of rigid trajectory motion (ablation) | TC4D · User study ablation | overall preference: 97% preferred full model | Improved |
| add trajectory conditioning during training rather than only after training (ablation) | TC4D · User study ablation | overall preference: 89% preferred full model | Improved |
| add trajectory-aware VSDS segment sampling (ablation) | TC4D · User study ablation | overall preference: 85% preferred full model | Improved |
| add diffusion timestep annealing (ablation) | TC4D · User study ablation | overall preference: 84% preferred full model | Improved |
| add smoothness penalty on the deformation field (ablation) | TC4D · User study ablation | overall preference: 74% preferred full model | Improved |
| add trajectory conditioning, appearance and structure comparison against 4D-fy | TC4D · User study vs 4D-fy | structure quality: 52% (not significant) | No clear effect |
| use CLIP score to evaluate motion differences | TC4D vs 4D-fy · all prompts | CLIP score: 31.90 -> 32.01 | No clear effect |
Closest papers by meaning
- 4Real: Towards Photorealistic 4D Scene Generation via Video Diffusion Models (NeurIPS 2024 · similarity 0.91)
- A Unified Approach for Text- and Image-guided 4D Scene Generation (CVPR 2024 · similarity 0.91)
- Align Your Gaussians: Text-to-4D with Dynamic 3D Gaussians and Composed Diffusion Models (CVPR 2024 · similarity 0.89)
- 4D-fy: Text-to-4D Generation Using Hybrid Score Distillation Sampling (CVPR 2024 · similarity 0.89)
- Controllable Dynamic 3D Shape Generation via 3D Trajectories and Text (NeurIPS 2026 · similarity 0.89)
- Enhancing Motion in Text-to-Video Generation with Decomposed Encoding and Conditioning (NeurIPS 2024 · 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.
