Microsoft has filed a patent application for technology that could alter spoken words during livestreams and other live video broadcasts while digitally changing speakers' lip movements to match the replacement audio.

The system, described in a patent application published October 29, 2026, would use generative Super Intelligence technology to identify selected words in a live audio recording, replace them with different wording, and modify the corresponding video frames. The goal would be to make the edited speech appear natural, even though the speaker never actually said the replacement words.

The technology could be customized according to individual viewer profiles, meaning two people watching the same broadcast could potentially hear different versions of what was said. For example, the system could remove profanity, offensive language, filler words, or phrases considered inappropriate for a particular audience. It could also correct mispronunciations, replace misspoken words, and translate speech into another language while generating matching lip movements.

Microsoft's patent describes applications extending beyond livestreams, including lectures, sporting events, award ceremonies, television programs, and other live presentations. The proposed system could introduce a delay of several seconds to process and modify the audio and video before delivering the altered broadcast to viewers.

The patent also describes retaining the original, unmodified recordings as a reference for what speakers actually said. That feature could be important if edited versions are later questioned or compared with the original broadcast.

However, the filing does not mean Microsoft is preparing to release the technology as a commercial product. It outlines a proposed system and its potential applications, and the company has not announced plans to deploy it.