CoNFies: Controllable Neural Face Avatars

(FG 2023)

1Carnegie Mellon University   2Fujitsu Research of America  


Neural Radiance Fields (NeRF) are compelling tech- niques for modeling dynamic 3D scenes from 2D image collections. These volumetric representations would be well suited for synthesizing novel facial expressions but for two problems. First, deformable NeRFs are object agnostic and model holistic movement of the scene: they can replay how the motion changes over time, but they cannot alter it in an interpretable way. Second, control- lable volumetric representations typically require either time-consuming manual annotations or 3D supervision to provide semantic meaning to the scene. We propose a controllable neural representation for face self-portraits (CoNFies), that solves both of these problems within a common framework, and it can rely on automated processing. We use automated facial action recognition (AFAR) to characterize facial expressions as a combi- nation of action units (AU) and their intensities. AUs provide both the semantic locations and control labels for the system. CoNFies outperformed competing methods for novel view and expression synthesis in terms of visual and anatomic fidelity of expressions.


Localized Control Results

    Control eye opening (Left: CoNeRF. Right: ours.)
    Control mouth opening (Left: CoNeRF. Right: ours.)

Demo Video


  title={CoNFies: Controllable Neural Face Avatars},
  author={Yu, Heng and Niinuma, Koichiro and Jeni, Laszlo A},
  journal={arXiv preprint arXiv:2211.08610},


This research was supported by Fujitsu. We thank Joel Julin from University of Pittsburgh for helping with data collection and comments that greatly improved the manuscript. We thank Xuxin Cheng from Carnegie Mellon University who provided the data acquisition equipment. We would also like to show our gratitude to Nian-Hsuan Tsai from Carnegie Mellon University who helped build the OpenFace system.