Tencent filed 22 patent applications in H1 2026 across UI/UX (9), Game Engines (5), AI & Machine Learning (4), Networking (2), Game Mechanics (1), and Graphics (1).
The UI/UX patents cover interface improvements like split-screen mobile gaming, picture-in-picture game windows, dual field-of-view displays for MOBAs, smart ball-passing controls for sports games, and pet-capture threshold indicators for RPGs. Game_engine applications include resource-efficient instant replay systems, dynamic server tick rates, physics-based projectile throwing with stamina-linked trajectory visualization, and GPU shader-based texture decoding. The AI & Machine Learning patents describe systems for automated game level verification, AI-powered garment customization, reinforcement learning for NPC combat behaviors, and intelligent aim-assist that tracks relevant targets, while Networking patents address ISP-optimized connection routing and voiceprint-based cooperative actions.
The 5 Game Engines patents span a wide range of in-game systems, from replay mechanics to combat interactions. One application describes a replay system for mobile games that pulls character models, textures, and animations directly from resources already loaded during a battle, avoiding the need to duplicate assets and making real-time event playback viable on Hardware that couldn't previously support it. Another patent covers a crowd-control mechanic where the duration of a time-stop or stun effect isn't fixed at cast time but instead extends as the player continues landing attacks, with energy accumulation tied to combo consistency and timing precision. A third application addresses aim-assist in multi-enemy scenarios, using contact condition analysis to identify a single primary target among several moving enemies and applying speed-based tracking to that target alone, rather than splitting attention across the group. The remaining 2 patents venture into environment design and social gameplay: one describes a system where players can physically reposition Platformss during a live match, reshaping the arena's spatial layout mid-fight, while the other layers hidden-identity faction mechanics onto real-time combat so that each interaction simultaneously reveals information about both players' secret roles.
The lone Game Mechanics patent focuses on the connection between a character's physical condition and how a thrown projectile behaves. Rather than displaying a static arc, the system adjusts the trajectory visualization in real time based on the character's current stamina level, so players can see how fatigue will affect a throw before they commit to it.
Nine UI/UX patents address a broad set of interface challenges across multiple game genres. A picture-in-picture application renders the game in a floating, resized window during idle moments and layers only the most time-sensitive information on top, so players can attend to other tasks without losing track of what matters in-session. A separate patent for MOBA games introduces on-demand display of a second field-of-view derived from a character's pose, giving players concurrent visual perspectives without requiring them to reposition the camera. Another MOBA-focused application automates the targeting and distance resolution during flash-skill combos, removing the need to manually aim the flash while also managing the initiating skill. For open-world and MMO contexts, a path-recording system lets players mark quest locations during normal play and automatically converts those markers into shareable navigation routes that others can follow with a single tap. The split-screen application creates dedicated sub-interfaces for a zoomed game view and persistent controls, allowing continuous gameplay alongside other mobile apps simultaneously. A sports game patent replaces directional swipe inputs for passing with a set of thumb-reachable buttons, each linked to a specific teammate through on-screen identifiers, so target selection no longer depends on precise screen positioning. An RPG-oriented patent adds visual threshold markers to status bars during creature encounters, showing players exactly when a capture attempt will succeed rather than leaving them to guess based on energy levels alone. The voiceprint cooperation patent embeds a player's biometric Audio signature into in-game cooperation requests, allowing the system to identify the requester and tailor cooperative responses without requiring live voice communication. Finally, a battle royale patent describes a dual-control interface where a single player manages both a deployed aerial drone and their ground character at the same time through distinct, concurrently displayed UI elements.
The 2 Networking patents each target a different layer of multiplayer infrastructure. One decouples the location of a game's hosting server from the network entry point players use to connect, allowing the system to route each player through the ISP peering path with the lowest latency while still giving operators freedom to choose hosting locations based on cost or compute needs. The other application adjusts server tick rates dynamically across the course of a match, scaling refresh frequency up during high-intensity combat phases and pulling it back during quieter periods, rather than holding a constant rate throughout the session.
Four AI & Machine Learning patents cover generation, verification, and behavior systems. One application lets players input text keywords alongside an existing garment's technical maps (including diffuse, normal, and material data) to generate new clothing designs that remain compatible with the game engine's rendering requirements. A second patent describes an NPC control architecture that separates individual character logic from group-level state management, with the two layers communicating through status parameters to produce behaviors that respond dynamically to player actions. A third application replaces manual playtesting of user-generated levels with an AI-driven system that tests all navigation routes simultaneously, categorizes specific error types such as unreachable jump distances or impossible landing points, and offers automated geometry adjustments to fix them. The fourth uses reinforcement learning combined with LSTM networks to train NPCs for combat scenarios, processing both real-time state data and historical context to produce decision-making that adapts across different fight conditions without requiring manual behavior tree configuration.
The single Graphics patent takes aim at texture loading performance by using GPU shader work groups to decode compressed texture blocks in parallel, assigning one work group per block so all blocks process simultaneously rather than sequentially. This removes the CPU from the decode pipeline entirely, reducing load times without adding CPU overhead.
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.