AI & Machine Learning

Sony received 10 patents this month, with several addressing how to keep players engaged when returning to games after breaks. One system automatically generates personalized video recaps by analyzing player progress through the console's Universal Data System, combining existing content to create "previously on" summaries without requiring developers to manually author content for every story path. Another takes this further by determining whether a player quit due to difficulty, boredom, or external factors, then constructs adaptive tutorials ranging from skill refreshers to full training programs that address specific weaknesses. A third patent describes adaptive difficulty that learns a player's behavioral style, rewards it, then challenges them to break out of their comfort zone, using emotional state detection to prevent frustration from turning into disengagement. On the content creation side, Sony filed a patent for an AI-powered game scene generation system where designer-authored keyframes act as hard constraints on generative models, ensuring AI-produced content remains stylistically consistent without frame-by-frame human authorship. Another addresses NPC behavior through a single agent trained over a continuous distribution of reward weights, enabling designers to dial in personality styles via sliders at inference without retraining separate models. The company also filed multiple patents around conversational AI, including NPCs that listen to real player conversations in co-op games and generate contextually aware dialogue using large language models trained on actual gameplay chat, with responses that expire if no longer relevant. A related system operates in dual modes, both refining player communication in real-time and generating dynamic NPC responses timed to specific game events and proximity-based interactions. Sony's remaining patents cover crowd-controlled gaming that uses camera-based gesture recognition to let thousands of stadium fans collectively control video games on massive displays, and a notification system that delivers real-world texts and emails through in-game NPCs at contextually appropriate moments using two specialized AI engines for timing decisions and NPC behavior execution.

Microsoft filed 2 patents, both addressing voice transformation for multiplayer role-playing games. The first combines real-time voice conversion with dynamic character state modeling that adjusts vocal output based on in-game conditions like running or being underwater while preserving the speaker's natural speech patterns, updating synthesis every frame based on game state data. The second uses computer vision to analyze player and character physical attributes, then applies similarity matching algorithms and real-time voice transformation to automatically select and blend appropriate voice characteristics that match character appearance while preserving privacy.

Bandai Namco received 1 patent for an AI-driven avatar creation system that combines passive user behavioral and profile data already stored in a user management system with active multi-modal inputs like text, image, and voice to drive an automated character generation process. The system can produce multiple character variants simultaneously for user selection, eliminating manual customization complexity while achieving higher preference accuracy.

EA filed 1 patent for an AI-powered audio assistant that processes gameplay state data to generate real-time contextual descriptions for visually impaired players, including character movements, object distances, directions, and environmental context that text-to-speech screen readers cannot provide.