South Korea’s Electronics and Telecommunications Research Institute (ETRI) has developed advanced AI technology capable of generating hyper-realistic avatars that can speak and interact naturally using only a single portrait photo. This breakthrough is positioned as a next-generation interface for intuitive communication between humans and machines, particularly in the context of fully autonomous vehicles.

Unlike conventional AI assistants, which typically follow scripted commands with minimal facial movement, ETRI’s avatars feature detailed facial expressions and precise mouth synchronization. This enables the avatars to engage in lifelike, conversational interactions, transforming the experience of in-vehicle communication. The avatars can simulate realistic communication not only with drivers but also with pedestrians, supporting more natural and human-centered dialogue.

The foundation of this technology is a proprietary algorithm that focuses on learning and synthesizing only the facial areas relevant to speech, such as the lips, teeth, chin, and surrounding skin. By avoiding the processing of unrelated facial features, the system achieves highly accurate lip-syncing and enhanced expression, capturing subtle facial movements like skin wrinkles during speech.

Demonstrations of the technology have received international recognition for both visual fidelity and synchronization precision, including presentations at leading conferences like CVPR and AAAI.

The potential applications extend far beyond autonomous vehicles. ETRI envisions widespread use across industries such as digital kiosks, banking, customer service, news broadcasting, and advertising. The avatars can serve as digital representatives or assistants, potentially reducing human labor in various service sectors while enhancing user experience.

The Mobility User Experience Research Section at ETRI, which leads this development, is also working on AI-based driver interface systems. These systems aim to assess driver and pedestrian emotional states, fatigue, and concentration levels, contributing to safer and more responsive mobility solutions.

The new AI avatar framework, listed on the ETRI Technology Transfer site as “A Framework for Photorealistic Talking Face Generation,” is positioned for commercialization. The institute is actively pursuing technology transfer opportunities to accelerate its adoption across industries.

This innovation reflects a broader push toward inclusive and intelligent mobility experiences. By providing a more natural and accessible interface, the technology aims to enhance digital literacy and usability, especially for elderly or marginalized users, ensuring that future smart mobility services are inclusive and widely accessible.

By Impact Lab