UniMotion: A Unified Framework for Motion-Text-Vision Understanding and Generation
Official repository for the ECCV 2026 paper "UniMotion: A Unified Framework for Motion-Text-Vision Understanding and Generation".
Ziyi Wang*, Xinshun Wang*, Shuang Chen*, Yang Cong, and Mengyuan Liu#
- 2026-06: UniMotion is accepted by ECCV 2026.
- The full source code, model checkpoints, data preparation scripts, and inference/training instructions are expected to be released within one month.
UniMotion is a unified framework for motion-text-vision understanding, generation, and editing. It supports tri-modal any-to-any tasks across motion, text, and visual inputs.
More examples are available on the project page.
If you find this work useful, please consider citing:
@inproceedings{wang2026unimotion,
title = {UniMotion: A Unified Framework for Motion-Text-Vision Understanding and Generation},
author = {Wang, Ziyi and Wang, Xinshun and Chen, Shuang and Cong, Yang and Liu, Mengyuan},
booktitle = {European Conference on Computer Vision},
year = {2026}
}For questions, please contact the authors through the project page or open an issue in this repository.