Microsoft · Filed Feb 5, 2025 · Published Aug 6, 2026 · verified — real USPTO data

Microsoft Files Patent Letting Gamers Swap Out Any In-Game Sound With AI

Hate the sound of gunshots in games but love everything else? Microsoft has filed a patent for a system that lets you point at any sound, show the AI a few examples, and have it replaced across every game you play.

Microsoft Patent: AI That Rewrites Game Audio in Real Time — figure from US 2026/0224991 A1
Figure from the official USPTO publication.
See all 12 drawings from this filing ↓
Publication number US 2026/0224991 A1
Applicant Microsoft Technology Licensing, LLC
Filing date Feb 5, 2025
Publication date Aug 6, 2026
Inventors Tero Juhani PATANA, Erik Roberto GARCIA, Shuoqi Scott WANG
CPC classification 463/35
Grant likelihood Medium
Examiner WILLIAMS, ROSS A (Art Unit 3715)
Status Docketed New Case - Ready for Examination (Mar 18, 2025)
Document 20 claims

How Microsoft's per-sound game audio filter would work

Imagine you're playing a game and one particular sound drives you up the wall, maybe it's the shriek of a specific enemy, the crack of a gun, or a sound effect that triggers a headache. Right now, your only options are to lower the overall volume or live with it. Microsoft's newly filed patent describes a system that would let you fix that with pinpoint accuracy.

Here's how it would work for you: you'd open a settings panel and give the system a few examples of the sound you want to change. You could also tell it how to change it, whether that means replacing it with something softer, removing it entirely, or swapping it for a completely different sound. An AI model then learns what that sound category looks like and automatically intercepts it in any game you launch.

The system is designed to keep learning. If the AI misses an instance or catches something it shouldn't, you can flag it, and the model adjusts. The idea is a personal audio profile that gets better the more you use it, without touching the game's code at all.

How the generative AI model learns and replaces sounds live

The patent describes a generative AI audio pipeline that sits between the game's raw audio output and your speakers or headset. Instead of modifying the game itself, it operates on the live audio stream, intercepting and rewriting specific sounds on the fly.

The user interaction works in a few steps:

  • You provide one or more audio samples of the sound category you want to change (for example, clips of gunfire across different games).
  • You specify a transformation type, which could mean silencing the sound, replacing it with an alternative, or altering its character.
  • The gaming device trains a generative AI model on those samples so it can recognize that sound class wherever it appears in a dynamic audio stream.
  • When a game starts, the system generates a prompt (an instruction packet) for the AI, telling it exactly what to listen for and what to do with it.
  • The AI's modified output becomes the actual audio you hear, replacing the original stream entirely.

The patent also describes a feedback loop: as you hear the transformed audio, you can flag errors, which refine the model over time. The setting applies globally, meaning every game the device runs gets filtered without any per-game configuration.

We find one patent like this every day. Get the best of each week in your inbox, free →

What this means for accessibility and game audio control

For players with auditory sensitivities, PTSD triggers, or hearing impairments, this is a meaningful accessibility tool. Today, most games offer volume sliders and a handful of presets. A system that can surgically remove or replace a specific sound class, across all games, without needing developer support, would fill a gap that game studios rarely prioritize.

Beyond accessibility, this kind of system has obvious appeal for content creators who stream games and want to avoid copyright-flagged sound effects, or for parents who want to soften the audio violence of a game without switching the game off entirely. Because the technology works at the device level rather than inside the game, it could apply to any title, including older games that will never receive an accessibility patch.

Editorial take

This is a genuinely practical idea. Audio accessibility in gaming has been neglected for years, and a device-level AI filter would help people that studio-side settings menus consistently ignore. The feedback-tuning loop is the detail that makes it credible rather than just clever: without that, the model would make too many mistakes to be usable.

There are more where this came from

We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.

The drawings

12 drawing sheets from US 2026/0224991 A1 · click any drawing to enlarge

Patent filing page

Source. Full patent text and figures from the official USPTO publication PDF.

Editorial commentary on a publicly published patent application. Not legal advice.