Nvidia

NVIDIA's cloud gaming patent addresses a persistent performance bottleneck by reimagining how shader compilation is handled across player bases. Rather than requiring each device to independently compile shaders for the same game, the system stores compiled results in a shared cloud repository. When one player's GPU processes a particular shader configuration, that work becomes instantly available to every future player encountering the same scenario, identified through state-data hashing that ensures compatibility across different Hardware setups.

Two AI and machine learning patents explore different applications of neural networks in content generation. The first focuses on character and physics animation, training networks on motion data to output force-based control signals rather than direct positional outputs. This approach aims to produce animation that adapts more naturally to varied situations compared to earlier methods that struggled with brittleness and limited movement repertoires. The second patent tackles automated sports and event commentary by constructing a pipeline that interprets video frames and feeds structured prompts into a language model, which then generates natural language narration for defined time intervals. This framework replaces the manual scripting and rigid templates that previous automated commentary systems relied upon, allowing the language model to produce flexible descriptions of on-screen action.