AI & Machine Learning

Sony received 2 patents this month. The first tackles sports video editing by using an AI system that tracks objects during gameplay, identifies noteworthy moments, and assigns each an excitement score through what Sony calls a "point excitement matrix." This scoring mechanism allows the system to automatically assemble highlight reels based on emotional intensity rather than simply detecting when certain events occur. The second patent describes a machine learning approach that transforms household devices into extensions of the gaming experience, adjusting lights, fans, and thermostats in response to what happens on screen. The system goes beyond basic trigger responses by using multi-modal feature extraction and a state vector model that learns from user preferences, with a latency compensation component that anticipates upcoming game events to prepare device changes in advance.

EA's 2 patents address production challenges in game development. One covers a text-to-speech system that gives developers fine-grained control over how non-player characters speak by using reference audio to shape emotional tone and prosody. The architecture separates the content of speech from its stylistic qualities into distinct feature embeddings, which are then merged during the decoding phase, allowing for expressive vocal performances without requiring multi-stage processing. The other patent applies machine learning to motion capture workflows, automatically predicting the positions of markers that become obscured during recording sessions. This removes the need for animators to manually reconstruct missing data or schedule additional capture sessions, streamlining a traditionally labor-heavy part of character animation.