This month's AI & Machine Learning category includes 13 filed patent applications from 8 companies: Sony (5), Nvidia (2), Atombeam Technologies (1), EA (1), Tencent (1), AT&T (1), Triumph Labs (1), and Adeia (1).
The patents cover AI systems that personalize and automate gaming experiences, including Sony's technologies for customizing voice pitch in text-to-speech dialog, personalizing trophy appearances, generating transitional gameplay content, and modifying in-game audio based on player preferences. Nvidia filed patents for cost-efficient image insertion in chatbot responses and AI-driven 3D mesh generation for game assets. Other applications describe systems for dynamic content filtering (AT&T), character-to-task assignment optimization (EA), skill-based matchmaking adjustment, context-aware gameplay assistance, persistent cognitive AI teammates (Triumph Labs), automated level verification for user-generated content, and relationship-based NPC personality systems.
Sony received 5 patents covering personalization and automation across multiple gaming touchpoints. The text-to-speech pitch control system divides pitch variations into discrete categories that can be manipulated through neural networks, allowing character voices to be adjusted without bringing actors back to the studio. A separate patent describes adaptive trophy designs that change based on how players actually earned them, reflecting individual playstyles rather than showing identical achievements to everyone. The audio personalization system learns which in-game sounds annoy or fatigue players, then modifies or replaces them in real-time while preserving sounds the developer wants to protect. Another application addresses gameplay assistance by analyzing video frames to predict what players might be confused about and generating answers before they even ask, removing the need to pause and search for help. The final patent tackles highlight creation by using multiple AI models in sequence to generate smooth transitional footage between key gameplay moments, eliminating manual editing.
AT&T filed 1 patent for a content filtering system that modifies gaming and media experiences on the fly based on who is watching. Rather than blocking access to mature content entirely, the system adapts visuals, audio, and even gameplay mechanics in real-time, adjusting the severity of what appears on screen according to viewer age and individual preferences.
Electronic Arts received 1 patent for an AI-driven assignment algorithm that matches multiple characters to multiple locations or tasks simultaneously. The ring-based approach solves coordination problems across entire teams at once rather than calculating paths for each character independently, making complex multi-character scenarios computationally feasible.
Triumph Labs filed 1 patent describing a matchmaking system that adjusts skill-based pairing criteria when player populations drop. The system predicts when wait times will spike and preemptively loosens matching thresholds to keep queues moving, balancing competitive integrity against the need to get players into matches quickly.
Nvidia received 2 patents addressing AI-generated content for games and virtual environments. One application describes a method for chatbots and NPCs to include relevant images in text responses without running expensive multimodal AI models, instead pre-indexing text-image pairs and retrieving them through vector similarity searches. The other patent covers an auto-regressive auto-encoder that generates 3D meshes with over 8000 faces, using a tokenization method that compresses geometry into compact representations while maintaining the quality artists expect.
Atombeam Technologies filed 1 patent for a cognitive AI system designed to function as a persistent teammate that continues thinking even when not actively prompted. The architecture includes sleep cycles for memory consolidation, maintains cognitive state across system restarts, and generates thoughts autonomously, moving beyond traditional prompt-and-response interactions toward AI that behaves more like a human community member who remembers past experiences.
Tencent received 1 patent for an automated verification system that tests user-generated game levels for playability errors. The AI uses pathfinding algorithms to check whether all areas are reachable, identifies specific problems like jumps that are too far or platforms that are too high, and offers automatic geometry adjustments to fix navigation issues without requiring manual playtesting.
Adeia filed 1 patent describing an NPC personality system that uses relationship graphs and large language models to make background characters respond dynamically to player actions. The system propagates personality traits and knowledge through a network of character relationships, constraining what each NPC knows based on their connection strength to the player and their position in the story to prevent them from revealing information they shouldn't have access to.