MotionPyramid: Controllable Motion Synthesis via Stylized Phase Manifolds

Jingyuan Li¹, Peizhuo Li¹, Andreas Aristidou^2,3, Olga Sorkine-Hornung¹

¹ETH Zurich, Switzerland
²University of Cyprus, Cyprus
³CYENS Centre of Excellence, Cyprus

Computer Graphics Forum (Proceedings of SCA 2026)

Overview

This repository contains the implementation for MotionPyramid, a framework for controllable motion synthesis via stylized phase manifolds.

Workflow

1. Data processing in Unity: import and preprocess motion data with the Unity tools described in Unity/README.md.
2. Python environment setup: install the required Python dependencies and prepare the training environment following Python/README.md.
3. VQ embedding training: train the VQ phase embedding model with the Python module described in Python/README.md.
4. Average-pose checking: inspect the learned phase representation and average-pose behavior with the Unity tools described in Unity/README.md.
5. Text-to-phase and phase-to-motion training: train the diffusion models for text-conditioned phase generation and phase-to-motion synthesis with the Python module described in Python/README.md.
6. Inference: run the trained Python models to generate motion results following Python/README.md.
7. Visualization in Unity: load and inspect the generated results with the Unity scenes and tools described in Unity/README.md.

Repository Structure

MotionPyramid_private/
|-- Python/    # Python training and inference code
|-- Unity/     # Unity visualization and demo framework
`-- README.md  # Project overview

Citation

If you use this code or project resources, please cite:

@article{li2026motionpyramid,
  title   = {MotionPyramid: Controllable Motion Synthesis via Stylized Phase Manifolds},
  author  = {Li, Jingyuan and Li, Peizhuo and Aristidou, Andreas and Sorkine-Hornung, Olga},
  journal = {Computer Graphics Forum},
  note    = {Proceedings of SCA 2026},
  year    = {2026}
}

Acknowledgements

This work was supported in part by the European Research Council (ERC) under the European Union's Horizon 2020 Research and Innovation Programme (ERC Consolidator Grant, agreement No. 101003104, MYCLOTH) and by the EU Commission's Horizon Europe program (grant No. 101178362). Open access publishing was facilitated by ETH Zurich, as part of the Wiley-ETH Zurich agreement via the Consortium Of Swiss Academic Libraries.