The visual tells that used to give digital humans away are disappearing. NVIDIA introduced RTX Neural Shaders at CES 2025 and gave a GDC 2025 talk titled 'Crossing the Uncanny Valley with RTX Neural Face Rendering,' and has since shipped neural rendering tools and Audio2Face-3D for Unreal Engine developers, demoed on a real-time conversational MetaHuman. In parallel, 3D Gaussian splatting has become the dominant academic method for real-time, animatable photorealistic avatars.
The 'almost' is doing real work. NVIDIA's own framing is about narrowing the valley, not eliminating it, and researchers still wrestle with fine detail, lighting and motion under interaction. But the trajectory — photoreal faces rendered live, driven by audio — is clear and shipping into developer tools.
Real-time photorealism is the enabling layer for believable digital humans, from customer-service avatars to celebrity IP. As rendering closes the last visual gaps, the bottleneck shifts from 'does it look real?' to 'does it behave and sound real?' — moving the frontier from graphics to behavior.
Key Facts
- NVIDIA RTX Neural Shaders introduced at CES 2025
- GDC 2025 talk: 'Crossing the Uncanny Valley with RTX Neural Face Rendering'
- Audio2Face-3D and neural rendering shipped for Unreal Engine
- 3D Gaussian splatting: the dominant real-time avatar method
- Gap narrowing, not eliminated (fine detail, lighting, interaction)
Frequently Asked
Is the uncanny valley solved?
Not fully. NVIDIA's own framing is about narrowing the valley, and researchers still struggle with fine detail, lighting and motion under interaction — but real-time photoreal faces are shipping into developer tools.
What technology is closing the gap?
NVIDIA's RTX neural face rendering and Audio2Face-3D for Unreal Engine, plus 3D Gaussian splatting, the dominant academic method for real-time animatable photorealistic avatars.
Why does real-time rendering matter?
It is the enabling layer for believable digital humans; as visuals get solved, the frontier shifts from how avatars look to how they behave and sound.