Microsoft Patents AI That Can Alter Live Speech and Lip Movements in Real Time

Microsoft has filed a patent application for an artificial intelligence system designed to modify what a person says during a live video stream, with both the audio and the speaker's lip movements adjusted so that the change is difficult to detect. According to the filing, the edit would take roughly one to ten seconds to complete.

The application, titled “Near Real-Time AI Enhancement of Video or AR Presentations of Live Events” and published under US application number 20260341309, describes a system that replaces spoken words and then redraws the speaker's mouth to match the new audio. The document suggests the tool could remove filler words such as “um,” slang, mumbled or mispronounced words, and offensive language.

The filing also indicates that substitutions could be tailored to individual viewers based on profile details such as age, location and language. A separate claim covers live translation paired with lip-sync, allowing a presentation to appear as though the speaker is talking in the viewer's language.

Because the changes would be applied while the video is streaming, audiences would not necessarily know what was originally said. That capability raises concerns about misinformation and fraud, and the filing reportedly contains only one dependent claim that preserves the original recording.

It is important to note that this is a patent application, not a released product. Microsoft has not indicated whether it intends to bring the technology to market. The company has been expanding its AI efforts across several areas, including a major Copilot revamp and AI tools for medical diagnosis.

The publication of the filing adds to a growing debate over how AI-generated edits should be disclosed, particularly in live broadcasts where viewers have no way to verify the source material. How regulators and platforms respond to such tools could shape the standards for transparency in streaming and online presentations.