Sony

The 8 AI & Machine Learning patents span dialogue systems, content generation, and automated development tools. One application describes NPCs that understand player speech and respond with dialogue synchronized to actual game state, creating a feedback loop between language models and the game engine. Another converts gameplay telemetry into personalized storybooks, podcasts, and highlight reels by chaining two generative models together. A third watches players struggle through difficult sections and generates personalized practice scenarios with visual controller guides timed to help build muscle memory. The system that converts text and game data into haptic feedback uses machine learning to generate tactile signals dynamically rather than relying on pre-programmed responses. One filing describes fusing camera, IMU, and microphone data through a unified ML pipeline to interpret player intent without a controller. Another application auto-detects performance issues like frame drops during game execution, diagnoses them with ML models, and attempts automated fixes before escalating to human developers. A route-based storytelling system feeds real-time navigation data into an LLM to generate location-aware interactive narratives for car passengers, displayed on vehicle windows. The final patent in this category uses ML to reverse-engineer controller inputs purely from gameplay video, overlaying button presses for viewers learning from recorded footage.

Two Hardware patents repurpose physical devices as gaming peripherals. One coordinates a drone swarm to project game visuals onto outdoor surfaces, with each drone acting as a repositionable display unit that collectively renders a complete image. The other creates a bidirectional API between game software and parked vehicles, allowing games to control headlights, horns, seats, and other automotive components as haptic feedback mechanisms.

Both UI/UX patents address player attention and communication. The muscle memory trainer analyzes gameplay struggles to generate practice scenarios that mirror difficulty without spoilers, pairing them with real-time visual controller input guides. The VR communication system displays visual cues to others when players are too immersed to interrupt, queuing incoming messages and delivering them based on gameplay intensity and message severity.

Two Audio patents adapt music to gameplay dynamically. One uses AI to detect gameplay tempo from in-game events and syncs music accordingly, both in real-time and during post-production video editing. The other morphs portions of personal playlists to match the original game score's style during specific activities, creating a hybrid soundtrack that adapts moment-to-moment.

The single VR & AR patent allocates image processing power based on player gaze and gestures, processing sensor data at different rates depending on inferred intent. This reduces compute load while improving responsiveness in areas where players focus, using a database of gameplay objects combined with live tracking to create contextual awareness.

Two Monetization patents turn loading screens into content delivery opportunities. Both detect wait-state triggers during gameplay sessions and substitute ads, trailers, or mini-games into these intervals. The approach differs from static loading screens by treating wait time as a programmable slot that can be filled with contextually relevant content at the Platforms level.

The cloud gaming patent combines bug detection with highlight generation through automated video capture. When aggregate player failure data reveals problem areas, the system triggers targeted capture requests to client machines for remote diagnosis and patching, while simultaneously using the same infrastructure to create highlight reels from extraordinary moments.

Platform patents address authentication and social discovery. One uses the controller as an authentication node, detecting when a phone is nearby and auto-logging players into their console without manual credential entry. The other analyzes past gameplay sessions to identify players with overlapping schedules and compatible gaming habits, generating friend recommendations based on probabilistic temporal overlap rather than relying on manual searches.