Migu Culture Technology Co.,Ltd | China
Zeyu Peng is an accomplished Graphics AI Algorithm Engineer based in Shenzhen, Guangdong, with a strong academic foundation in Mathematics and Applied Mathematics, earning an MSc from Wuhan University (2015–2018) specializing in Optimization Theory, Algorithms, and Applications, and a BSc from Central South University (2011–2015). Professionally, Zeyu has extensive experience in AI-driven graphics and operations research, currently contributing to Migu Culture Technology as a Graphics AI Algorithm Engineer since 2021, and previously serving as an Algorithm Engineer at China Southern Airlines IT Research Institute (2018–2021). His notable projects include developing a Speech-to-3D Facial Animation Generation engine, creating multi-style performance rule sets, and implementing both combination optimization and diffusion-based facial animation synthesis frameworks for digital humans. He has also engineered a Co-Speech Gesture Generation system leveraging language models, inverse kinematics, and diffusion models to produce contextually appropriate gestures synchronized with speech. Additionally, Zeyu has worked on FlowMatch and FastPitch-based AI voice conversion models for speech and singing applications, and contributed to large-scale aviation optimization projects such as flight pairing and crew scheduling, involving graph-based algorithms, branch-and-bound frameworks, and operations research algorithm libraries. He has developed a Facial Detection SDK with MTCNN for autonomous airport check-in systems and a luggage damage detection system utilizing object detection and classification techniques. Zeyu’s technical expertise spans C++, Python, Rust, and frameworks like PyTorch and ONNX, complemented by proficiency in LaTeX, Maya, Blender, and certifications including CET-6 and TOEFL. His research contributions are evidenced by publications in high-impact venues such as The Journal of Supercomputing and CGI2025, along with multiple patents in facial animation synthesis, video generation, and digital human animation technologies. Overall, Zeyu demonstrates a rare combination of deep theoretical knowledge, practical algorithmic implementation, and impactful contributions to AI-driven virtual reality, digital human animation, and optimization systems, establishing him as a leading engineer and innovator in graphics AI and virtual human technologies.
Profile: Google Scholar
Featured Publications
Peng, Z., Ma, D., & Lv, X. (2025, July). A two stage co-speech gesture generation method. Oral Presentation at CGI 2025.
Peng, Z., Li, H., & Wang, S. (2025, April). FaceImitate: Speech-driven 3D facial animation synthesis from imitation. The Journal of Supercomputing.
Peng, Z., & Wang, S. (2024, October 11). A video synthesis method, device, storage medium, and program product based on a classification model [Patent No. CN118764575A]. Migu Culture Technology Co., Ltd.; China Mobile Communications Group Co., Ltd.
Peng, Z., Wang, S., & Li, H. (2024, July 9). Facial motion prediction methods, devices, media, and computer program products [Patent No. CN118314614A]. Migu Culture Technology Co., Ltd.; China Mobile Communications Group Co., Ltd.