BiMotion: B-spline Motion for Text-guided Dynamic 3D Character Generation

BiMotion generates expressive, high-quality motions for 3D characters from text descriptions using continuous B-spline representations.

Abstract

Text-guided dynamic 3D character generation has advanced rapidly, yet producing high-quality motion that faithfully reflects rich textual descriptions remains challenging. BiMotion addresses these limitations by representing motion with continuous differentiable B-spline curves, enabling more effective motion generation without modifying the capabilities of the underlying generative model.

Key Contributions:

B-spline Representation: A closed-form, Laplacian-regularized B-spline solver that efficiently compresses variable-length motion sequences into compact fixed-length control points.
Enhanced Geometry Adherence: Introduces a normal-fusion strategy for input shape adherence along with correspondence-aware and local-rigidity losses.
BIMO Dataset: A new dataset featuring diverse variable-length 3D motion sequences with rich, high-quality text annotations.

Video Demo

Framework

Overview. BiMotion uses a B-spline representation for motion generation. During training (red arrow), vertex differences are converted into control points and encoded into motion latents. During inference (black arrow), the initial mesh and the text prompt generate motion latents that are decoded into control points and converted into the generated mesh sequence via B-spline reprojection.

Results

BibTeX

@inproceedings{wang2026bimotion,
  title     = {BiMotion: B-spline Motion for Text-guided Dynamic 3D Character Generation},
  author    = {Wang, Miaowei and Yan, Qingxuan and Cao, Zhi and Li, Yayuan and Mac Aodha, Oisin and Corso, Jason J. and Vaxman, Amir},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year      = {2026},
}