Morpheus: A Neural-driven Animatronic Face with
Hybrid Actuation and Diverse Emotion Control

RSS 2025

Software

Zongzheng Zhang^1,2* · Jiawen Yang^1* · Ziqiao Peng¹ ·
Meng Yang⁴ · Jianzhu Ma¹, Lin Cheng⁵ · Huazhe Xu³ . Hang Zhao³ and Hao Zhao^1,2

¹ Institute for AI Industry Research (AIR), Tsinghua University, ² Beijing Academy of Artificial Intelligence (BAAI),
³ Institute for Interdisciplinary Information Sciences(IIIS), Tsinghua University,
⁴ MGI Tech, Shenzhen, China, ⁵ Beihang University
_{(* indicates equal contribution)}
RSS official | Project Page | arXiv | Hardware | Software

Environment

• Linux
• Python 3.10
• Pytorch 2.3.1
• CUDA 12.1
• Blender 3.4.1
• ffmpeg 4.4.1

Clone the repo:

git clone https://github.com/ZZongzheng0918/Morpheus-Software.git
cd Morpheus-Software

Create conda environment:

conda create -n morpheus python=3.10
conda activate morpheus
pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu121
pip install -r requirements.txt

Demo

Download Blender and put it in this directory.

wget https://ftp.nluug.nl/pub/graphics/blender/release/Blender3.4/blender-3.4.1-linux-x64.tar.xz
tar -xf blender-3.4.1-linux-x64.tar.xz
mv blender-3.4.1-linux-x64 blender && rm blender-3.4.1-linux-x64.tar.xz

Download the pretrained models from model.pth. Put the pretrained models under pretrain_model folder. Put the audio under aduio folder and run

python demo.py --wav_path "./audio/disgust.wav"

The generated animation will be saved in result folder.

License and Acknowledgements

This source code is licensed under the MIT liscence found in the LICENSE file in the root directory of this repository.

Citation

If you find this project useful, feel free to cite our work!

@article{zhang2025morpheus,
  title={Morpheus: A Neural-driven Animatronic Face with Hybrid Actuation and Diverse Emotion Control},
  author={Zhang, Zongzheng and Yang, Jiawen and Peng, Ziqiao and Yang, Meng and Ma, Jianzhu and Cheng, Lin and Xu, Huazhe and Zhao, Hang and Zhao, Hao},
  journal={arXiv preprint arXiv:2507.16645},
  year={2025}
}

Acknowledgement

Here are some great resources we benefit:

• EmoTalk for codebase
• Faceformer for training pipeline
• EVP for training dataloader
• Speech-driven-expressions for rendering
• Wav2Vec2 Content and Wav2Vec2 Emotion for audio encoder
• Head Template for visualization.

Thanks to John Hable for sharing his head template under the CC0 license, which is very helpful for us to visualize the results.