NVIDIA Open-Sources Audio2Face: Generative AI for Lifelike Character Animation is Now Free

NVIDIA has announced the open-sourcing of its Audio2Face model and accompanying SDK, enabling developers to integrate lifelike character animation into games and 3D applications, with a comprehensive open-source training framework for fine-tuning or customisation, according to NVIDIA.

Project R2X, first showcased at CES 2025, uses generative AI to transform speech into facial expressions and lip movements, delivering accurate lip-sync and expressive emotion for in-game dialogues, customer service bots, and real-time virtual interactions. The release aims to lower the barrier to entry by providing tools that can be paired with other generative AI capabilities, broadening the ecosystem for digital humans across gaming, media, and enterprise applications.

The technology has already seen adoption among studios such as Codemasters, GSC Game World, NetEase, and Perfect World, with independent vendors including Convai, Inworld AI, Reallusion, Streamlabs and UneeQ leveraging Audio2Face for immersive virtual experiences. The open-sourcing effort is framed as a step toward standardisation and democratisation of digital-human technologies.