Nvidia

The single Graphics patent tackles the challenge of generating additional frames between rendered frames to make gameplay feel smoother without introducing the lag that would ruin competitive play. Where existing frame generation techniques typically insert one interpolated frame between each pair of rendered frames, this approach uses deep neural networks to produce multiple interpolated frames per rendered frame while keeping latency low enough for fast-paced gaming and VR applications.

Four AI and machine learning patents span character animation, visual customization, video synthesis, and quality assurance. One application introduces a training method that lets virtual characters move with lifelike precision while still adapting to new scenarios, combining motion tracking's accuracy with distribution matching's flexibility in a single reinforcement learning framework. Another allows game streamers to describe visual effects in plain language and receive dynamically generated shader code that applies those effects in real time, bypassing the need for preset filter menus. A third patent separates the appearance of objects from their motion patterns during video generation, giving creators independent control over what things look like and how they move within AI-generated footage. The fourth replaces manual gameplay testing with vision-language models that watch video and listen to Audio to spot bugs automatically, using embedding comparisons to focus processing power on frames where meaningful changes occur.

Nvidia's Audio patent connects camera-based head tracking to spatial sound generation, creating 3D Audio that responds to where a player is looking. The system achieves this with lower computational overhead than traditional head-tracking Audio approaches, making immersive spatial sound practical for resource-constrained environments like VR headsets and gaming PCs.