Microsoft is developing AI that can correct speech right during a stream

Microsoft has patented a new AI technology for Xbox that will be able to alter people's speech directly during broadcasts.

Microsoft has filed a patent for technology capable of editing speech during video broadcasts using AI. The solution could be useful when watching gaming events, esports tournaments, and livestreams related to Xbox.

The system would analyze audio and then use AI to replace or remove individual words. At the same time, the video would be altered so that the speaker's lip movements match the new speech. The processed broadcast would play with a delay of approximately 1 to 10 seconds. Moreover, the algorithm could take into account the viewer's profile, including their age, language, and location.

The patent describes several use cases for the technology. AI could remove profanity, correct slips of the tongue and stuttering, as well as replace obscure expressions. In particular, it highlights the ability to translate speech into another language with corresponding lip sync adjustments. This would allow watching broadcasts in a foreign language without the usual mismatch between dubbing and video.

The technology is potentially capable of running directly on smartphones, tablets, cameras, and augmented reality glasses. However, the document does not confirm that Microsoft intends to implement it in Xbox consoles or gaming services.

The patent application was filed in April 2025 and published on October 29, 2026. Notably, in their current form, the claims have already received an examiner's rejection. Therefore, the patent registration itself does not mean that the technology will ever hit the market.