Microsoft

Microsoft filed 2 Audio patents that both tackle the challenge of making games accessible to players with sound sensitivities. The first describes a system that uses generative AI to identify and transform specific sound classes in any dynamic Audio stream, operating at runtime without requiring developers to modify individual games. This approach goes beyond traditional volume sliders by letting players semantically target particular types of sounds. The second patent covers a similar real-time transformation capability but emphasizes user-defined training, where players can teach the system to recognize their personal triggers using Audio samples, with a feedback loop that refines the model when it misses or incorrectly processes sounds. The trained models can move with players across devices through a global account.

The single AI & Machine Learning patent describes a dual-layer system for game creation and evolution. GPT models assist designers during the development phase by helping construct narrative elements, while a separate generative layer responds to player behavior during live gameplay. The system monitors implicit engagement signals rather than explicit surveys, using this data to automatically trigger content changes and inform future game versions. Players shape the experience through their actions without accessing traditional authoring tools, creating a closed feedback loop between behavior and content generation.