Sony filed 42 patents this quarter across Streaming (1), Audio (6), Platforms (7), AI & Machine Learning (13), VR & AR (2), Cloud Gaming (6), UI/UX (1), Hardware (4), Networking (1), and Monetization (1).
The AI & Machine Learning category covers systems that adapt gameplay based on eye-tracking data, generate personalized coaching and practice minigames, auto-create avatars from body movements, and enable players to delegate sessions to customizable bot agents. Cloud_gaming patents describe predictive frame rendering, split-client architectures where phones control games displayed on TVs, and neural object compression for asset delivery. Platform filings include cross-metaverse population sharing to address empty virtual worlds, entrance announcement systems for notable users, hot-patching to update games mid-session, and matchmaking that learns player style preferences.
A single Streaming patent describes automated narration for gameplay clips, using AI to generate dynamic commentary tied to in-game events captured in recorded video. The system eliminates manual editing by producing commentary-enriched highlight reels at the system level rather than as a post-production task.
Six Audio patents span positioning, communication, personalization, and gameplay integration. One uses spread-spectrum sound signals and reflected waves to track headsets and smartphones indoors, distinguishing direct from reflected waves via unique spreading codes to improve positioning accuracy for immersive Audio. Another automatically reconfigures headphone communication channels based on in-game events, muting or filtering voice chat without game software intervention, and supports physical NFC attachments on headphones to unlock game powers. A third generates dynamic boot sounds using machine learning models trained on time, location, weather, and user listening history instead of fixed startup chimes. A fourth detects when hands obstruct speakers on handheld devices and reroutes important Audio to unobstructed channels in real time. The fifth auto-selects songs from personal music libraries to score gameplay clips, matching tempo and mood through attribute-to-property matching. The sixth treats Audio as a structured input, driving NPC behavior and asset interactions through sentiment analysis of spoken keywords and beat-synchronized game logic.
Platform filings total 7 and address metaverse mechanics, parental controls, and distribution infrastructure. Two tackle virtual world population issues: one triggers special performances when notable users enter a space to create social hierarchy, while another broadcasts live activity from one metaverse into another to drive user migration between Platformss. Three parental control patents shift from time-based limits to gameplay-aware restrictions, including activity-count enforcement that respects natural stopping points, geofencing that restricts interaction to players within a defined radius, and context-aware systems that evaluate mid-boss or online match status before applying cutoffs. A hot-patching system injects compiled code updates into live game binaries without interrupting play. An automated asset generation system emulates consoles to discover and create marketing key art and preview materials, identifying the correct emulator from software package format and iteratively satisfying Platforms content criteria.
Thirteen AI and ML patents form the largest category, covering adaptation, accessibility, matchmaking, and content generation. One uses eye-tracking as an engagement inference engine, adjusting Audio and visuals based on detected focus regions rather than just cursor control. Another infers physical limitations from actuation patterns and uses AI to augment inputs in real time, creating adaptive compensation without manual accessibility configuration. A third clones real players as AI companions by extracting behavioral attributes from recorded gameplay to construct NPCs that mirror human playstyles. Cross-Platforms frame adaptation enhances visibility of critical assets when Streaming from console to mobile by understanding which elements matter to gameplay. One system mines aggregated play data to quantify which activities correlate with trophy and item unlocks, providing likelihood measures derived from population behavior. Matchmaking combines implicit behavioral data with explicit preference declarations and performs complementary skill matching rather than pure skill parity. Personalized trailer generation constructs individualized videos by selecting content sections based on console usage data. A Platforms-level accessibility engine transforms visuals and Audio in real time using trained AI that identifies element-output relationships including occluded elements, adapting all associated outputs simultaneously. Gameplay video search uses AI-generated embeddings of controller input speed, completion times, and device settings as indexing mechanisms for semantic queries. Resource allocation predicts player skill level from behavior to proactively assign computing power before game state generation. A coaching system analyzes controller signals and gameplay video jointly to generate real-time instructions and opponent-specific practice minigames. An AI agent framework allows players to delegate sessions to bots with whitelist and blacklist controls over permitted actions. Character creation combines motion-captured body movements for animation generation with natural language descriptions for generative AI visual creation, producing behavioral and visual personalization simultaneously.
Two VR and AR patents address avatar portability and population density. The first extends style transfer to full 3D geometry and texture adaptation, enabling cross-metaverse avatar portability by co-adapting shape and appearance using Platforms-specific visual feature databases. The second imports avatar representations from separate metaverse Platformss and displays them in visually distinct form to create cross-Platforms social presence when user counts drop.
Cloud gaming receives 6 patents focused on latency, asset management, and distributed architecture. Predictive debugging pre-executes all possible next debug steps speculatively and caches outputs, converting latency-bound remote operations into cache lookups. Haptic feedback adjusts controller resistance based on network quality, proactively signaling degraded connections to limit player inputs to those most likely to transmit reliably, while server-side displacement complementation and timeout-based reset close perceptual gaps. Asset decomposition enables extraction, local editing, and reintegration of sub-components into live cloud assets without rebuilding entire packages. Predictive frame rendering combines controller input history with motion vector data to speculatively pre-render partial frames client-side, reconciling them with server data upon arrival. Neural object representations convert large assets into compressed neural proxies via virtual camera capture pipelines, preserving fidelity for client-side rendering without transmitting full geometry and texture data. Split-client architecture separates input capture on phones from video rendering on TVs, allowing distinct devices to handle control and display independently with all processing server-side.
A single UI and UX patent ties annotations to game state data rather than timestamps, enabling contextual delivery during live gameplay when actual conditions match annotated moments. Tips, commentary, or highlights created from one session surface dynamically in subsequent sessions regardless of pacing differences.
Four Hardware patents cover controllers, peripherals, and biometric authentication. One relocates wireless host functions from consoles into peripherals already positioned near users, reducing signal path length and interference by adapting antenna characteristics based on peripheral type and spatial relationships. A modular controller features a rotating dial and touchscreen display as a magnetically repositionable module, enabling personalized ergonomic layouts and auto-pause-on-notification driven by embedded processors. EMG-based biometric authentication uses muscle signals as passive, continuous login mechanisms that capture dynamic movement signatures unique to each person's neuromuscular system. Wearable headphone attachments carrying RFID-style identifiers unlock game modifications and customizations via attach-and-play, with headphones serving as authenticated intermediaries between physical tokens and running game instances.
Networking gets 1 patent describing multi-host wireless hub migration that distributes connectivity across multiple wired hosts and uses ML models trained on gameplay feature data to proactively migrate peripheral connections before performance degrades. The system routes VR headsets and controllers through the nearest wired dongle using player movement, positional data, and input patterns as migration triggers rather than generic network handoff metrics.
The Monetization category contains 1 patent where game controllers serve as active relay devices for payment data, initiating proximity detection when purchase events occur on consoles and transferring payment information from tapped phones without manual entry. The controller creates a closed-loop, event-driven payment handoff rather than requiring direct console-to-phone pairing.
All data sourced from USPTO patent filings. Google Patents may take several weeks to index recent publications. If a link is unavailable, search for the patent number at USPTO Patent Public Search.