Choreographic Motif Planning
MLLM Planner / Anchors / Cues
Music and global intent are used to propose timeline anchors and local creativity cues.

SIGGRAPH Asia 2026
Customized 3D Dance Generation with Coarse-to-Fine Human-Centered Interactive Control
Abstract
With the rise of AI-generated content (AIGC) and advanced techniques for 3D human representation, the task of generating 3D dance movements has become an exciting area of research. Despite significant advancements, current methods often fail to provide comprehensive and distinct control over various multimodal inputs from users, such as music or specific descriptions of desired movements. As a result, the generated motions may be statistically plausible and technically correct, but they often lack depth, expressiveness, and alignment with the user's creative vision. To address this issue, we present CustomDance, a coarse-to-fine interactive system designed for customized 3D dance generation.
Inspired by real-world choreographic workflows, CustomDance combines music-aware planning, phrase-level retrieval, and diffusion-based completion and repair. It gives creators concrete, predictable choices while preserving iterative control over the final choreography.
Method

MLLM Planner / Anchors / Cues
Music and global intent are used to propose timeline anchors and local creativity cues.
Retriever / Dance Library / Recommendation
Creators explore ranked, real dance phrases using local text and body-part controls.
Completer / Diagnoser / Remaker
Generated transitions and local repair preserve the creator's phrase choices.
User Workflow
Results
Overview Video
Citation
@article{tang2026customdance,
title={CustomDance: Customized 3D Dance Generation with Coarse-to-Fine Human-Centered Interactive Control},
author={Tang, Xulong and Yang, Kaixing and Guo, Xiaohu and Balakrishnan, Prabhakaran and Alghofaili, Rawan},
journal={arXiv preprint arXiv:2608.06722},
year={2026}
}