Sony received 20 granted patents this month across 10 categories: UI/UX (2), Hardware (2), Audio (3), Graphics (1), Cloud Gaming (4), Platforms (1), AI & Machine Learning (4), VR & AR (1), Streaming (1), and Game Mechanics (1).
The Cloud Gaming and AI & Machine Learning patents describe systems for latency reduction, session recovery, replay-based jump-in, AI players that mimic user style, and minigame generation from recorded gameplay. UI/UX and Hardware patents cover adaptive feature discovery, controller-based authentication, and button lighting that responds to in-game context. Other filings address dynamic Graphics quality switching, keyword-driven music generation, Audio codec optimization, real-world object fingerprinting for digital twins, VR collision avoidance, viewer interaction in livestreams, and camera-based posture monitoring.
Two UI and UX patents describe systems that respond to player behavior in real time. The first watches what a user does on a console, then decides when and how to surface relevant feature hints based on situational awareness and a personalized profile, updating itself from whether the user acted on the suggestion. The second ties controller lighting to individual buttons, letting the game illuminate specific inputs according to their role at each moment of play, with control signals sent directly from the game apparatus.
Two Hardware patents address controller authentication and console integration. One patent describes a controller that orchestrates its own login handshake by pinging a nearby phone, receiving an acknowledgment, fetching the player's credential, and answering the console without requiring the user to type. The other describes an endpoint management bridge that mirrors PCIe endpoints so a PS5-class APU board can sit in a host server slot for development, storage sharing, and frame capture while keeping the console's PCIe domain isolated from the host's fabric.
Three Audio patents cover music generation, sound effect tooling, and CPU-adaptive decoding. The first has the game console emit keywords describing player actions and game state, which a separate music generator turns into a live, personalized soundtrack that updates with each keyword. The second derives constraints such as duration, material, and speed from an animation, then auto-randomizes pitch, timing, and volume to produce fitted sound effect variants without manual tuning. The third pre-sorts Audio packet components by magnitude so the receiver can drop the quietest components and save CPU, scaling Audio resolution to workload like level of detail in Graphics.
A single Graphics patent describes quality switching triggered by signals from the application itself, via a mode switch message, frame analysis, or Audio cues, instead of relying on frame time measurements as in dynamic resolution scaling. Because the trigger is semantic, the system can render a paused detailed screen at high quality and adjust multiple rendering parameters together rather than only resolution.
Four cloud gaming patents address latency, session recovery, and replay-based navigation. One shrinks large keyframes in game streams by blurring, downscaling, or rotating them before encoding, then lets the client sharpen, upscale, or reorient to trade picture quality for lower latency on constrained networks. Another keeps a session alive when the primary connection fails, pivoting to a low-bandwidth side channel that sends template-based event messages to the player's phone, accepts commands from it, and can run autoplay to move the avatar away from danger. A third coordinates server flip-time and early scan-out with client display scheduling to trim one-way latency, deriving scan-out speed from the chipset's maximum pixel clock and the client's display image size. The fourth enables jump-in by replaying stored inputs from a snapshot to reconstruct play state in a fresh game instance up to a chosen moment, then handing over live control at that point instead of only seeking within a video.
A single Platforms patent scans a real-world object to build a distinct physical fingerprint, validates ownership by matching it to a specific instance, and unlocks a corresponding digital twin for a metaverse avatar or gaming system. The mechanism filters wear-and-tear anomalies and normalizes virtual attributes to mirror the real object, letting unique or older items be tied to a metaverse identity without relying on a point-of-sale system that only works for new purchases.
Four AI and machine learning patents describe systems that learn from player behavior and generate playable content. One filters chat messages for relevance to gameplay within a selected time window, extracts keywords weighted by prominence, and presents tap-through summaries tied to in-game events so a user who stepped away can catch up on what friends discussed. Another converts recorded gameplay into playable minigame fragments by analyzing game state and input activity to find a region of interest, selecting valid start and end boundaries, then recommending fragments based on the user's time and goals while updating the model from multimodal feedback. A third trains an AI player from a user's inputs and the game states those inputs caused, lets it play in the user's style, and hands control back at a chosen point rather than running generic autoplay or bot behavior. The fourth reads a real viewer's image or speech through a camera and microphone, then repaints virtual humans and pets in the metaverse with matching apparel, skin color, emotion, voice, and behavior, propagating the customization to other displays according to the real-world person's status.
A single VR patent addresses free-roam safety by feeding real-world presence of non-headset people into the world construction step, so the virtual scene shows an object at the corresponding position and players do not collide with visitors they cannot see. A stay-area variant renders a whole region as an object when someone enters a preset zone, with entry and exit reported by a sensor or input device.
One Streaming patent maps viewer cursor control commands and sentiment reactions into game-specific visual and haptic outputs on the player's display, with the ability to promote those inputs into game actions, lock out cursors during cutscenes, rotate viewer permissions, alter playout when reaction thresholds are met, and award trophies or viewer rewards. The system ties viewer input to the live game execution context and its current event or state rather than only overlaying chat.
A single game mechanics patent uses console or headset cameras to learn a player's baseline sitting position from their own standing image, spots shoulder lean or hunch, then slows or restricts gameplay until the user sits up. The approach relies on relative change from a personal baseline rather than struggling with different body sizes and proportions as prior posture feedback devices did.
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.