UIKA

This repository hosts the released model weights and auxiliary assets for UIKA: Fast Universal Head Avatar from Pose-Free Images.

UIKA is a feed-forward model for creating animatable 3D Gaussian head avatars from pose-free inputs. It supports a single reference image, multiple images of the same identity, or smartphone-style video captures. Use these files together with the official PyTorch code:

• Code: https://github.com/ant-research/UIKA
• Paper: https://arxiv.org/abs/2601.07603
• Project page: https://zijian-wu.github.io/uika-page/

Files

• uika.safetensors - main UIKA checkpoint. The official inference config expects it at model_zoo/uika/uika.safetensors.
• fuvt_15k.safetensors - FUVT UV-estimation module checkpoint used by UIKA.
• human_parametric_models.tar - human parametric model assets required by the pipeline. Third-party assets may have separate license terms.
• ref_motion_example.tar - example reference/motion assets for testing inference.

License

The UIKA-authored code and the released UIKA/FUVT checkpoints are released under the MIT License. Third-party model weights, datasets, and human parametric model assets remain subject to their own licenses and access terms. Review the official repository's third-party notices before redistribution or commercial use.

Citation

If you use UIKA in your work, please cite:

@inproceedings{wu2026uika,
    title     = {UIKA: Fast Universal Head Avatar from Pose-Free Images},
    author    = {Wu, Zijian and Zhou, Boyao and Hu, Liangxiao and Liu, Hongyu and Sun, Yuan and Wang, Xuan and Cao, Xun and Shen, Yujun and Zhu, Hao},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
    year      = {2026}
}