NVIDIA Audio2Face is a foundation application that uses generative AI to instantly create expressive facial animations for 3D characters from just an audio source. It simplifies the process of animating a character's facial features to match any voice-over track, making it useful for games, films, real-time digital assistants, and more.

Audio2Face is part of NVIDIA Omniverse, a real-time collaboration platform that allows creators to work together on complex projects using GPU-accelerated rendering and AI. The app analyzes an audio input and uses a pre-trained deep neural network to drive the 3D vertices of a character mesh, creating facial animations in real-time.

Key features include the ability to use live or recorded audio, retarget animations to any 3D human or human-like face, run multiple instances to animate multiple characters, control emotional expressions, and export animations to tools like Blender and Unreal Engine. Audio2Face streamlines facial animation workflows and enables realistic lip-syncing and facial expressions synchronized with audio.