![[2602.18873] BiMotion: B-spline Motion for Text-guided Dynamic 3D Character Generation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.18873] BiMotion: B-spline Motion for Text-guided Dynamic 3D Character Generation
Summary
BiMotion introduces a novel approach to dynamic 3D character generation using B-spline curves, enhancing motion quality and alignment with text prompts.
Why It Matters
This research addresses significant challenges in generating realistic 3D character motions from textual descriptions, which is crucial for applications in gaming, animation, and virtual reality. By improving motion coherence and expressiveness, it advances the capabilities of generative AI in computer vision.
Key Takeaways
- BiMotion uses continuous B-spline curves for improved motion generation.
- The method overcomes limitations of fixed-length inputs and discrete representations.
- A new dataset, BIMO, supports training with diverse 3D motion sequences.
- BiMotion achieves higher quality and faster generation than existing methods.
- The framework enhances alignment of generated motions with textual prompts.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.18873 (cs) [Submitted on 21 Feb 2026] Title:BiMotion: B-spline Motion for Text-guided Dynamic 3D Character Generation Authors:Miaowei Wang, Qingxuan Yan, Zhi Cao, Yayuan Li, Oisin Mac Aodha, Jason J. Corso, Amir Vaxman View a PDF of the paper titled BiMotion: B-spline Motion for Text-guided Dynamic 3D Character Generation, by Miaowei Wang and 6 other authors View PDF HTML (experimental) Abstract:Text-guided dynamic 3D character generation has advanced rapidly, yet producing high-quality motion that faithfully reflects rich textual descriptions remains challenging. Existing methods tend to generate limited sub-actions or incoherent motion due to fixed-length temporal inputs and discrete frame-wise representations that fail to capture rich motion semantics. We address 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. Specifically, our closed-form, Laplacian-regularized B-spline solver efficiently compresses variable-length motion sequences into compact representations with a fixed number of control points. Further, we introduce a normal-fusion strategy for input shape adherence along with correspondence-aware and local-rigidity losses for motion-restoration quality. To train our model, we collate BIMO, a new dataset containing diverse variable-length 3D mot...