AdaMesh

AdaMesh: Personalized Facial Expressions and Head Poses for Adaptive Speech-Driven 3D Facial Animation

---

Setup

• pip install -r requirements.txt

• Our method relies on 3D face reconstruction. If you want to extract facial coefficients or visualize them, please refer to the following repository to install the environment: Spectre for 3D face reconstruction, Voca for visualization

Training

Please note:

• Our model is trained in the space of face coefficients. Therefore, you need to prepare your data in the form of coefficients. You may refer to Spectre or DiffPoseTalk for reference.

bash train.sh

Inference

bash test.sh

---

Citation

@ARTICLE{AdaMesh2025,
  author={Chen, Liyang and Bao, Weihong and Lei, Shun and Tang, Boshi and Wu, Zhiyong and Kang, Shiyin and Huang, Haozhi and Meng, Helen},
  journal={IEEE Transactions on Multimedia}, 
  title={AdaMesh: Personalized Facial Expressions and Head Poses for Adaptive Speech-Driven 3D Facial Animation}, 
  year={2025},
  volume={27},
  pages={3598-3609},
  doi={10.1109/TMM.2025.3535287}
}