Research map / Video / Video generation
BiMotion: B-spline Motion for Text-guided Dynamic 3D Character Generation
Miaowei Wang, Qingxuan Yan, Zhi Cao, Yayuan Li, Oisin Mac Aodha, Jason J Corso, Amir Vaxman
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 |
|---|---|---|---|
| represent motion with B-spline control points instead of uniformly downsampled frames with linear interpolation (ablation) | BiMotion VAE · BIMO validation | mean L1 error (x10^-2): 2.674 -> 1.078 | Improved |
| add multi-level control-point embedding instead of frequency positional encoding (ablation) | BiMotion VAE · BIMO validation | mean L1 error (x10^-2): 1.648 -> 1.078 | Improved |
| add normal fusion (ablation) | BiMotion VAE · BIMO validation | mean L1 error (x10^-2): 1.328 -> 1.078 | Improved |
| add correspondence loss (ablation) | BiMotion VAE · BIMO validation | mean L1 error (x10^-2): 1.303 -> 1.078 | Improved |
| add local rigidity loss (ablation) | BiMotion VAE · BIMO validation | mean L1 error (x10^-2): 1.349 -> 1.078 | Improved |
| use Laplacian regularizer instead of Ridge for short sequences (T < k) (ablation) | B-spline solver · qualitative, T=10 mesh sequence | interpolation naturalness (qualitative) | Improved |
| switch from AnimateAnyMesh to BiMotion generator | text-guided mesh motion generation · VBench dynamic degree | DD: 0.100 -> 0.800 | Improved |
| switch from AnimateAnyMesh to BiMotion generator | text-guided mesh motion generation · VBench subject consistency | SC: 0.951 -> 0.948 | Hurt |
| switch from AnimateAnyMesh to BiMotion generator | text-guided mesh motion generation · 20 meshes, user study | text-to-motion agreement (1-5): 2.314 -> 4.095 | Improved |
| switch from AnimateAnyMesh to BiMotion generator | text-guided mesh motion generation · A100 efficiency | generation time: 16.847s -> 4.383s | Improved |
| switch from AnimateAnyMesh to BiMotion as vertices grow from 9K to 24K | text-guided mesh motion generation · scaling with mesh size | generation time: AnimateAnyMesh 6.4 s -> 20.0 s vs ours 3.7 s -> 4.6 s | 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.
- AnimateAnyMesh — AnimateAnyMesh: A Feed-Forward 4D Foundation Model for Text-Driven Universal Mesh Animation (ICCV 2025) · better on 7 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
- Controllable Dynamic 3D Shape Generation via 3D Trajectories and Text (NeurIPS 2026 · similarity 0.87)
- How to Move Your Dragon: Text-to-Motion Synthesis for Large-Vocabulary Objects (ICML 2025 · similarity 0.86)
- Enhancing Motion in Text-to-Video Generation with Decomposed Encoding and Conditioning (NeurIPS 2024 · similarity 0.85)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.85)
- Make-It-Vivid: Dressing Your Animatable Biped Cartoon Characters from Text (CVPR 2024 · similarity 0.85)
- TapMo: Shape-aware Motion Generation of Skeleton-free Characters (ICLR 2024 · similarity 0.84)
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.
